Wireless communication methods, user equipment, and base station
Patent Information
- Application Number
- PCT/CN2025/084808
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-03-25
- Publication Date
- 2026-10-01
Smart Images

Figure CN2025084808_01102026_PF_FP_ABST
Abstract
Description
WIRELESS COMMUNICATION METHODS, USER EQUIPMENT, AND BASE STATIONTECHNICAL FIELD
[0001] The present disclosure relates to the field of communication systems, and more particularly, to wireless communication methods, a user equipment (UE) , and a base station.BACKGROUND
[0002] In wireless communication systems, channel state information (CSI) is important for improving network performance. With the introduction of artificial intelligence (AI) technology, AI-based CSI prediction, beam management, and CSI compression have made efficient performance monitoring and reporting mechanisms important to ensuring model reliability and accuracy. A user equipment (UE) needs to report performance monitoring results; however, under resource constraints, some CSI reports may not be completed. When multiple AI models need to run simultaneously, the management of processing units and storage units becomes important to ensure CSI processing efficiency and prevent reporting failures. Addressing at least one of these issues is important for enhancing AI-based CSI processing capabilities and maintaining network performance.SUMMARY
[0003] An object of the present disclosure is to propose wireless communication methods, a user equipment (UE) , and a base station, which can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0004] In a first aspect of the present disclosure, a wireless communication method performed by a user equipment (UE) includes: reporting, to a base station, a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting, receiving a CSI reporting configuration information from the base station, receiving channel measurement signals from the base station, reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting, and receiving a life cycle management (LCM) indication information from the base station.
[0005] In a second aspect of the present disclosure, a wireless communication method performed by a user equipment (UE) , includes: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.
[0006] In a third aspect of the present disclosure, a wireless communication method performed by a user equipment (UE) , includes: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; reporting, to the base station, performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and receiving a life cycle management (LCM) indication information from the base station; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.
[0007] In a fourth aspect of the present disclosure, a wireless communication method performed by a user equipment (UE) , includes: receiving a monitoring indication information from a base station; and reporting, to the base station, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.
[0008] In a fifth aspect of the present disclosure, a wireless communication method performed by a user equipment (UE) includes: receiving a channel state information (CSI) reporting configuration information from the base station; receiving channel measurement signals from the base station; reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information; receiving monitoring measurement signals from the base station; and reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.
[0009] In a sixth aspect of the present disclosure, a wireless communication method performed by a base station, comprising: receiving, from a user equipment (UE) , a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting; transmitting, to the UE, a CSI reporting configuration information; transmitting, to the UE, channel measurement signals; receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting; and transmitting, to the UE, a life cycle management (LCM) indication information.
[0010] In a seventh aspect of the present disclosure, a wireless communication method performed by a base station, comprising: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.
[0011] In an eighth aspect of the present disclosure, a wireless communication method performed by a base station, comprising: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and transmitting, to the UE, a life cycle management (LCM) indication information; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.
[0012] In a ninth aspect of the present disclosure, a wireless communication method performed by a base station, comprising: transmitting, a user equipment (UE) , a monitoring indication information; and receiving, from the UE, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.
[0013] In a tenth aspect of the present disclosure, a wireless communication method performed by a base station, comprising: transmitting, to a user equipment (UE) , a channel state information (CSI) reporting configuration information; transmitting, to the UE, channel measurement signals; receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; transmitting, to the UE, monitoring measurement signals; and receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.
[0014] In an eleventh aspect of the present disclosure, a UE includes a memory, a transceiver, and a processor coupled to the memory and the transceiver. The UE is configured to perform the above method.
[0015] In a twelfth aspect of the present disclosure, a base station includes a memory, a transceiver, and a processor coupled to the memory and the transceiver. The base station is configured to perform the above method.
[0016] In a thirteenth aspect of the present disclosure, a non-transitory machine-readable storage medium has stored thereon instructions that, when executed by a computer, cause the computer to perform the above method.
[0017] In a fourteenth aspect of the present disclosure, a chip includes a processor, configured to call and run a computer program stored in a memory, to cause a device in which the chip is installed to execute the above method.
[0018] In a fifteenth aspect of the present disclosure, a computer readable storage medium, in which a computer program is stored, causes a computer to execute the above method.
[0019] In a sixteenth aspect of the present disclosure, a computer program product includes a computer program, and the computer program causes a computer to execute the above method.
[0020] In a seventeenth aspect of the present disclosure, a computer program causes a computer to execute the above method.BRIEF DESCRIPTION OF DRAWINGS
[0021] In order to illustrate the embodiments of the present disclosure or related art more clearly, the following figures will be described in the embodiments are briefly introduced. It is obvious that the drawings are merely some embodiments of the present disclosure, a person having ordinary skill in this field can obtain other figures according to these figures without paying the premise.
[0022] FIG. 1A is a block diagram of an artificial intelligence (AI) general framework for wireless communication according to an embodiment of the present disclosure.
[0023] FIG. 1B is a block diagram of one or more user equipments (UEs) and a base station of communication in a communication network system according to an embodiment of the present disclosure.
[0024] FIG. 2A is a flowchart illustrating a wireless communication method performed by a UE according to an embodiment of the present disclosure.
[0025] FIG. 2B is a flowchart illustrating a wireless communication method performed by a base station according to an embodiment of the present disclosure.
[0026] FIG. 2C is a flowchart illustrating a wireless communication method according to an embodiment of the present disclosure.
[0027] FIG. 3A is a flowchart illustrating a wireless communication method performed by a UE according to an embodiment of the present disclosure.
[0028] FIG. 3B is a flowchart illustrating a wireless communication method performed by a base station according to an embodiment of the present disclosure.
[0029] FIG. 4A is a flowchart illustrating a wireless communication method performed by a UE according to an embodiment of the present disclosure.
[0030] FIG. 4B is a flowchart illustrating a wireless communication method performed by a base station according to an embodiment of the present disclosure.
[0031] FIG. 5A is a flowchart illustrating a wireless communication method performed by a UE according to an embodiment of the present disclosure.
[0032] FIG. 5B is a flowchart illustrating a wireless communication method performed by a base station according to an embodiment of the present disclosure.
[0033] FIG. 6A is a flowchart illustrating a wireless communication method performed by a UE according to an embodiment of the present disclosure.
[0034] FIG. 6B is a flowchart illustrating a wireless communication method performed by a base station according to an embodiment of the present disclosure.
[0035] FIG. 7 is a block diagram of an example of a computing device according to an embodiment of the present disclosure.
[0036] FIG. 8 is a block diagram of a communication system according to an embodiment of the present disclosure.DETAILED DESCRIPTION OF EMBODIMENTS
[0037] Embodiments of the present disclosure are described in detail with the technical matters, structural features, achieved objects, and effects with reference to the accompanying drawings as follows. Specifically, the terminologies in the embodiments of the present disclosure are merely for describing the purpose of the certain embodiment, but not to limit the disclosure.
[0038] The technical solutions of the embodiments of the present disclosure can be applied to various communication systems, such as a global system of mobile communication (GSM) system, a code division multiple access (CDMA) system, a wideband code division multiple access (WCDMA) system, a general packet radio service (GPRS) , a long term evolution (LTE) system, a LTE frequency division duplex (FDD) system, a LTE time division duplex (TDD) system, an advanced long term evolution (LTE-A) system, a new radio (NR) system, an evolution system of a NR system, a LTE-based access to unlicensed spectrum (LTE-U) system, a NR-based access to unlicensed spectrum (NR-U) system, an universal mobile telecommunication system (UMTS) , a global interoperability for microwave access (WiMAX) communication system, wireless local area networks (WLAN) , wireless fidelity (Wi-Fi) , a future 5th generation (5G) system (may also be called a new radio (NR) system) , a 6G system, a 7G system, or other communication systems, etc.
[0039] Optionally, a network such as a base station mentioned in the embodiments of the present application can provide a communication coverage for a specific geographic area and can communicate with a user equipment (UE) located in the coverage area. Optionally, the base station may be a gNB, a base transceiver station (BTS) in the GSM or in the CDMA system, or may be a NodeB (NB) in the WCDMA system, or may be an evolutional Node B (eNB or eNodeB) in the LTE system, or a radio controller in a cloud radio access network (CRAN) .
[0040] A user equipment (UE) may refer to an access terminal, a subscriber unit, a subscriber station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent, or a user device. The access terminal may be a cellular radio telephone, a cordless telephone, a session initiation protocol (SIP) telephone, a wireless local loop (WLL) station, a personal digital assistant (PDA) , a handheld device with wireless communication functions, a computing device, other processing devices coupled with a wireless modem, an in-vehicle device, a wearable device, a terminal device in a future 5G network, a terminal device in a future evolved public land mobile network (PLMN) , etc.
[0041] Optionally, the communication system in the embodiment of the present application may be applied to an unlicensed spectrum, where the unlicensed spectrum may also be considered as a shared spectrum; or the communication system in the embodiment of the present application may also be applied to a licensed spectrum, where the licensed spectrum can also be considered an unshared spectrum.
[0042] “Pre-determinded” can refer to “pre-configured” , “network configured” , “preset” , “pre-defined” , or “pre-defined rules” . “Preset” , “pre-defined” , or “pre-defined rules” may be achieved by pre-storing corresponding codes, tables, or other manners for indicating relevant information in devices (e.g., including a UE and a network device) . The specific implementation is not limited in the present disclosure. For example, “preset” and “pre-defined” may refer to those defined in a protocol. It is also to be understood that in the disclosure, “protocol” may refer to a standard protocol in the field of communication, which may include, for example, an LTE protocol, NR protocol and relevant protocol applied in the future communication system, which is not limited in the present disclosure.
[0043] In a multiple-input multiple-output (MIMO) communication system, channel state information (CSI) measurement is important for both time division duplex (TDD) and frequency division duplex (FDD) communication systems. In a TDD system, a network can obtain an uplink (UL) channel through channel measurement and then derive a downlink (DL) channel based on the channel reciprocity between UL and DL. However, in FDD systems, since channel reciprocity does not hold, a precoding matrix for the DL channel is obtained through UE feedback. Additionally, in TDD systems, considering coverage limitations of UL channel measurement, the precoding matrix can also be obtained through UE feedback. Based on this background, the first edition of the 3rd generation partnership project (3GPP) standard studied CSI feedback, which has undergone multiple iterations and evolutions in subsequent versions.
[0044] With the rapid development of artificial intelligence / machine learning (AI / ML) , 3GPP has begun studying AI-based CSI prediction and / or compression, AI-based positioning, and AI-based beam management. Meanwhile, the specification work for AI-based CSI prediction, AI-based positioning, and AI-based beam management began in Release 19. Since some issues remain unresolved, AI-based CSI compression requires further study.
[0045] Based on the discussion and conclusions, a general framework for wireless communication systems is presented as follows: Based on the Rel-19 research, the following conclusions have been reached. For example, a legacy feedback mechanism of “typeII-Doppler-r18” can be used as a starting point for model inference in AI-based CSI prediction using a UE-side model. For performance monitoring, three types of procedures have been proposed by some companies: Type 1: A UE calculates a performance monitoring output and reports it to a network (NW) , which then determines whether to provide feedback to a non-AI CSI prediction, such as the “Rel-18 Doppler codebook. ” Type 2: The UE reports a ground-truth CSI to the NW, and the NW calculates the performance metric and determines whether to provide feedback to the non-AI CSI prediction. Type 3: The UE reports the performance monitoring metric to the NW, and the NW determines whether to provide feedback to the non-AI CSI prediction.
[0046] FIG. 1A is a block diagram of an artificial intelligence (AI) general framework for wireless communication according to an embodiment of the present disclosure. FIG. 1A illustrates that, in some embodiments, the framework includes functional modules: data collection, model training, model management, model storage, and inference. Moreover, the following conclusions have been reached: the objectives of model monitoring include at least model inactivity / deactivation, model switching, model updates, and functionality feedback.
[0047] Based on the Rel-19 research, the following conclusions have been reached. For example, the legacy feedback mechanism of "TypeII-Doppler-r18" can be used as a starting point for model inference in AI-based CSI prediction using UE-side models. For performance monitoring, three types of procedures have been proposed by some companies:
[0048] Type 1: The UE calculates the performance metric (s) . The UE reports performance monitoring output that facilitates the functionality fallback decision at the network. The details of the performance monitoring output can be further defined. The NW may configure a threshold criterion to facilitate UE-side performance monitoring (if needed) . The NW makes decisions regarding the functionality fallback operation (fallback mechanism to legacy CSI reporting) .
[0049] Type 2: The UE reports predicted CSI and / or the corresponding ground truth. The NW calculates the performance metrics. The NW makes decisions regarding the functionality fallback operation (fallback mechanism to legacy CSI reporting) .
[0050] Type 3: The UE calculates the performance metric (s) . The UE reports the performance metric (s) to the NW. The NW makes decisions regarding the functionality fallback operation (fallback mechanism to legacy CSI reporting) .
[0051] Furthermore, there are several technologies related to eType II codebook, Rel-18 eType II doppler codebook, CSI processing unit (CPU) occupancy rules, and priority rules for CSI reports, as illustrated below.
[0052] eType II codebook:
[0053] The existing Rel-16 eType-II codebook applies a three-level codebook framework where W1∈CP×2L indicates a spatial domain basis matrix, P denotes the number of antenna ports, and L is the number of spatial domain basis vectors chosen in a single polarization direction. is the projection coefficient achieved by project precoding matrix in spatial domain and frequency domain bases matrix, Mv indicates the number of frequency domain basis vectors associate with layer v; denotes the frequency domain bases matrix, and N3 is the number of subbands of PMI. The dimensions of the matrices in the above codebook framework are partially or fully determined by the NW, such as L, Mv, and other parameters. The NW determines the number of spatial domain bases, the number of frequency domain bases, and the control factor for the number of non-zero coefficients, as shown in the following Table 1 and Table 2.
[0054] Table 1: Codebook parameter configurations for L, β and pυ.
[0055] Table 2: Parameters for eType II codebook:
[0056] R indicates the number of PMI subbands in the CQI sub-band; β is used to control the maximum number of non-zero coefficients. For example, the maximum number of non-zero coefficients reported by the UE for the first layer can be expressed as Then, the location of the non-zero coefficient is indicated through bitmap, and the length of bitmap can be expressed as 2LMv. In addition, for Rel-16 eType-II codebook, both the spatial domain bases and frequency domain bases are constructed using orthogonal DFT vectors. For example, spatial domain bases matrix W1 is a block diagonal matrix w∈CP / 2×L, and the L column vectors of w is selected from the orthogonal DFT vector set with dimension P / 2. Moreover, the Mv column vectors of Wf are selected from the orthogonal DFT vector set with dimension N3.
[0057] Rel-18 eType II Doppler codebook:
[0058] The existing Rel-18 eType-II Doppler codebook applies a three-level codebook framework, given by where W1∈CP×2L indicates a spatial domain bases matrix, P denotes the number of antenna ports, and L is the number of spatial domain basis vectors chosen in a single polarization direction. is the projection coefficient achieved by project precoding matrix in spatial domain and frequency domain bases matrix, Mv indicates the number of frequency domain basis vectors associate with layer v; denotes the frequency domain bases matrix, and N3 is the number of subbands of PMI. The dimensions of the matrices in the above codebook framework are either partially or fully determined by the NW, such as L, Mv, and other parameters. The NW specifies the number of spatial domain bases, the number of frequency domain bases, and the control factor for the number of non-zero coefficients, as shown in the following Table 3.
[0059] Table 3: Codebook parameter configurations for L, β and pυ:
[0060] R indicates the number of PMI subbands in the CQI sub-band; β is used to control the maximum number of non-zero coefficients. For example, the maximum number of non-zero coefficients reported by the UE for the first layer can be expressed as The location of the non-zero coefficients is indicated through a bitmap, and the length of the bitmap can be expressed as 2LMv. Additionally, in the Rel-16 eType-II codebook, both the spatial domain bases and frequency domain bases are constructed using orthogonal DFT vectors. For example, the spatial domain basis matrix W1 is a block diagonal matrix given by: w∈CP / 2×L, and the L column vectors of w are selected from the orthogonal DFT vector set with a dimension of P / 2. Moreover, the Mv column vectors of Wf are selected from the orthogonal DFT vector set with a dimension of N3.
[0061] In the Rel-18 eType-II Doppler codebook, the parameter N4 is introduced to indicate the dimension of the time-domain channel, where N4∈ {1, 2, 4, 8} , and the dimension is Q after time-domain compression. Subject to UE capability, a UE configured with a CSI-ReportConfig that includes the higher-layer parameter N4 and has reportQuantity set to 'cri-RI-PMI-CQI' is assumed to support UE-side CSI prediction. The reported PMI indicates predicted precoder matrices associated with N4 consecutive slot intervals, each with a duration of d slots, where the value of N4∈ {1, 2, 4, 8} is configured by the higher-layer parameter N4. If the UE is configured with an aperiodic CSI-RS resource set for channel measurement, the value, in number of slots, of the time unit d∈ {1, m} is configured by higher layer parameter d, where m is defined in 3GPP technical specification 38.214 Clause 5.2.1.4.1. If the UE is configured with a periodic or semi-persistent CSI-RS resource set for channel measurement, the value of d is equal to the periodicity of the CSI-RS resource. The earliest of the N4 slot intervals starts at slot l=n+δ, where n is the uplink slot in which the CSI is reported and the slot offset δ∈ {-nCSI_ref, 0, 1, 2} is configured by higher layer parameter delta, where nCSI_ref defined in 3GPP technical specification 38.214 Clause 5.2.2.5 and the value δ=-nCSI_ref can be configured subject to UE capability.
[0062] For N4=1, the UE is expected to report a predicted PMI for slot interval [l, l+d-1] and the slot offset value δ=-nCSI_ref can be configured only for d>1. A UE can be configured with N4=1 if the higher layer parameter codebookType is set to 'typeII-Doppler-r18' , or 'typeII-Doppler-PortSelection-r18' . The reported CQI is associated with slot l and the reported PMI.
[0063] For N4>1, the UE is expected to report a PMI which indicates predicted precoder matrices associated with slot intervals [l+j·d, l+ (j+1) ·d-1] , for j=0, …, N4-1. A UE can be configured with N4>1 if the higher layer parameter codebookType is set to 'typeII-Doppler-r18' . The UE is configured by higher layer parameter TDCQI to report X∈ {1, 2} CQIs for each subband in the CSI reporting band, if cqi-FormatIndicator is set to 'subbandCQI' , or X∈ {1, 2} CQIs for the entire CSI reporting band, if cqi-FormatIndicator is set to 'widebandCQI' . For X=2, the second CQI includes a 4-bit wideband CQI index and, if subband CQI reporting is configured, a 2-bit subband CQI index, calculated independently from the first CQI, as described in 3GPP technical specification 38.214 Clause 5.2.2.1, and the two CQIs are reported in the same CSI report.
[0064] CPU occupancy rules:
[0065] The UE indicates the number of supported simultaneous CSI calculations NCPU with parameter simultaneousCSI-ReportsPerCC or [simultaneousCSI-SubReportsPerCC-r18] in a component carrier, and simultaneousCSI-ReportsAllCC or [simultaneousCSI-SubReportsAllCC-r18] across all component carriers. If UE is configured with at least one CSI report setting with sub-configuration in a component carrier, UE shall use parameter [simultaneousCSI-SubReportsPerCC-r18] in the component carrier; otherwise, UE shall use simultaneousCSI-ReportsPerCC in the component carrier. If UE is configured with at least one CSI reporting setting with sub-configuration in any component carrier, UE shall use [simultaneousCSI-SubReportsAllCC-r18] ; otherwise, UE shall use simultaneousCSI-ReportsAllCC. If a UE supports NCPU simultaneous CSI calculations it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-L unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-L CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to the UE is not required to update the N-M requested CSI reports with lowest priority (according to 3GPP technical specification 38.214 Clause 5.2.5) , where 0≤M≤N is the largest value such that holds.
[0066] Priority rules for CSI reports:
[0067] CSI reports are associated with a priority value PriiCSI (y, k, c, s) =2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s, where: y=0 for aperiodic CSI reports to be carried on PUSCH y=1 for semi-persistent CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUCCH and y=3 for periodic CSI reports to be carried on PUCCH; k=0 for CSI reports carrying L1-RSRP or L1-SINR and k=1 for CSI reports not carrying L1-RSRP or L1-SINR; c is the serving cell index and Ncells is the value of the higher layer parameter maxNrofServingCells; for a CSI report configured with LTM-CSI-ReportConfig, c is the serving cell index value where the report configuration is configured; s is the reportConfigID and Msis the value of the higher layer parameter maxNrofCSI-ReportConfigurations; for a CSI report configured with LTM-CSI-ReportConfig, s is the LTM-CSI-ReportConfigID and Ms is the value of the higher layer parameter maxNrofLTM-CSI-ReportConfigurations. A first CSI report is considered to have priority over a second CSI report if the associated value (PriiCSI (y, k, c, s) ) of the first report is lower than that of the second report.
[0068] Based on the above background, some embodiments of the present disclosure further consider at least one issue related to performance monitoring result reporting, CSI reporting priority, CSI processing time, channel state information (CSI) memory unit (CMU) occupancy rules, and other related aspects. The at least one issue is illustrated below.
[0069] Issue 1: Trigger method of performance monitoring:
[0070] Based on the conclusions of the previous meeting, performance monitoring result reporting can be periodic, semi-persistent, aperiodic, and / or event-driven. However, considering the computational complexity of the UE, some examples consider that focusing solely on the reporting mechanism is insufficient, as the UE must calculate performance metrics for event-driven reporting. In fact, if the generalization capability of the model is sufficiently robust, continuous performance monitoring may become unnecessary. Moreover, model inference accuracy is affected by the channel environment, making fixed periodic monitoring result reports potentially unnecessary. Therefore, in the present disclosure, some embodiments design methods to reduce UE processing complexity, such as trigger mechanisms and reporting mechanisms for performance monitoring. The specific method design is detailed in Embodiment 1.
[0071] Issue 2: Performance monitoring result reporting:
[0072] For AI-based CSI prediction, the main performance monitoring metrics include a normalized mean squared error (NMSE) and a squared generalized cosine similarity (SGCS) . However, the specific content to be reported by the UE based on these performance metrics still requires further clarification in the standards. Based on the current progress of the standards, the UE may report performance monitoring results based on NMSE and / or SGCS, or it may report the comparison results of NMSE and / or SGCS against predefined thresholds. Therefore, regarding the aforementioned performance monitoring metrics, some embodiments primarily consider issues such as the specific content to be reported and the mapping rules for the reported content. Specific solutions can be found in Embodiments 3 and 4.
[0073] Issue 3: Performance monitoring result reporting priority:
[0074] For AI-based CSI prediction, beam management, CSI compression, and / or CSI prediction and compression, performance monitoring is required for each use case to ensure the reliability of AI models. Each use case may also require the UE to report performance monitoring results. Considering that performance monitoring results can be regarded as a new type of reporting quantity, the reporting priority for different reporting quantities requires further study. In some examples, the issue of reporting priority is raised, and the specific method is provided in Embodiment 2.
[0075] Issue 4: CSI processing timeline:
[0076] For AI-based use cases, such as AI-based CSI prediction, AI-based beam management (BM) , AI-based CSI compression, AI-based CSI prediction and compression, or other use cases that require the UE to report CSI or another quantity using physical uplink shared channel (PUSCH) or physical uplink control channel (PUCCH) , there are specific timeline considerations. From a timeline perspective, one key difference between AI / ML-based processing and non-AI / ML-based processing is model deployment. The model needs to be deployed to the AI / ML processing memory before performing inference-this procedure can be understood as functionality activation. Since multiple models supported by the UE share the AI / ML processing memory, keeping a specific model stored in the AI / ML processing memory at all times would lead to inefficient use of storage. Therefore, the model can be flushed from memory after each inference occasion is completed, this procedure can be understood as functionality deactivation. Based on the discussion above, when triggering aperiodic or semi-persistent CSI reporting for AI-based use cases, both the network (NW) and the UE need to consider the processing time required for functionality activation or deactivation. If the functionality activation time occurs later than the CSI-RS transmission time, as shown in the figure below, the CSI reporting may not function properly based on the existing standard. Therefore, some examples raise this issue, and in Embodiment 5, at least one method to resolve this problem is proposed.
[0077] Issue 5: CPU occupancy rules:
[0078] For AI-based CSI reports, such as CSI prediction, beam management, CSI compression, and CSI prediction and compression, the UE is required to report performance monitoring results to ensure the accuracy of model inference. Moreover, CSI reporting for model inference and performance monitoring can be reported separately. However, if inference result reporting is unsuccessful at a given time, performance monitoring results cannot be generated. Consequently, the UE is unable to update the requested CSI report for performance monitoring. Instead, the UE can update subsequent requested CSI reports with lower priority than the CSI report for performance monitoring. However, based on the existing CPU occupancy rules, the above situation is not supported. Therefore, some examples raise this issue, and in Embodiment 6, at least one method to resolve this problem is proposed.
[0079] Issue 6: CPU and CMU (CSI memory unit) occupancy rules:
[0080] To support simultaneous CSI reports, the UE may need to deploy multiple models or functionalities in the AI / ML processing memory, each of which corresponds to a specific memory storage requirement. If the number of models or functionalities running simultaneously exceeds the overall available memory storage capacity, some CSI reports must be canceled. Moreover, based on the existing standard, each CSI report requires the allocation of certain CPU resources. If the remaining CPU resources cannot support the CSI reports, the UE will be unable to update the requested CSI report. Based on the discussion above, CSI reporting depends on two factors: CPU and CMU. To ensure a consistent understanding of CSI updating behavior between the gNB and the UE, introducing a mechanism for aligning memory occupancy and CPU occupancy can be considered. Therefore, some examples raise this issue, and in Embodiment 7, at least one method to resolve this problem is proposed.
[0081] In some embodiments of the present disclosure, AI-based CSI prediction and / or compression technology is considered for MIMO communication systems, relating to the CSI prediction method, CSI feedback method, model data collection, model or functionality performance monitoring, and model-based or functionality-based life cycle management (LCM) , among other aspects.
[0082] Before introducing the specific content of some embodiments in the present disclosure, some concepts and terms mentioned in some embodiments of the present disclosure, which are applicable to the entire present disclosure, are explained as follows.
[0083] The performance monitoring result can refer to the performance monitoring output, the performance monitoring metric, or both. Performance monitoring can involve monitoring the performance of a model or a functionality. Note that the performance monitoring output can be obtained either through a comparison between the ground truth and the predicted CSI, or through a comparison between a performance metric and a certain threshold. In some examples, it does not exclude the possibility that the performance monitoring result is the ground truth CSI.
[0084] Moreover, in some embodiments of the present disclosure, some methods apply to all AI-based CSI reports, such as the process timeline, the CPU utilization rule, and the CMU utilization rule. Note that AI-based CSI reports can be applied to AI-based CSI prediction, AI-based beam management, AI-based CSI compression, AI-based CSI prediction and compression, and other related aspects.
[0085] Additionally, in this patent disclosure document, CSI can include multiple types of reporting quantities. For example, CSI reports may include: CSI reports carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) , CSI reports not carrying L1-RSRP or L1-SINR, CSI reports carrying performance monitoring results, CSI reports not carrying performance monitoring results, CSI reports carrying performance monitoring results along with L1-RSRP and L1-SINR, CSI reports carrying performance monitoring results along with other CSI parameters excluding L1-RSRP and L1-SINR, CSI reports carrying performance monitoring results for L1-RSRP and L1-SINR reporting, and / or CSI reports carrying performance monitoring results for other CSI parameters excluding L1-RSRP and L1-SINR.
[0086] Moreover, CSI reporting can be triggered by the network (NW) or initiated by the UE, and the CSI reporting container can be PUSCH, PUCCH, or MAC CE. Additionally, CSI reports that do not carry L1-RSRP or L1-SINR may include the following reporting quantities: 'cri-RI-i1' , 'cri-RI-i1-CQI' , 'cri-RI-CQI' , 'cri-RSRP' , 'ssb-Index-RSRP' , 'cri-RSRP-Index' , 'ssb-Index-RSRP-Index' , 'cri-SINR' , 'ssb-Index-SINR' , 'cri-SINR-Index' , 'ssb-Index-SINR-Index' , 'tdcp' , 'cri-RI-PMI-CQI' , 'cri-RI-LI-PMI-CQI' , and / or other similar reporting quantities. Note that the predicted CSI mentioned in some embodiments of the present disclosure can refer to either the predicted actual channel or the predicted precoding matrix information. In principle, both the actual channel and precoding matrix information can be based on RB granularity, multiple RB granularity, or subband granularity.
[0087] The performance monitoring results in some embodiments of the present disclosure primarily consider NMSE, SGCS, or GCS between the actual measured CSI and the predicted CSI. However, the reporting scheme for monitoring content is equally applicable to the evaluation of other performance indicators.
[0088] Furthermore, not all steps in the flowchart of air interface interactions mentioned in some embodiments of the present disclosure are mandatory, nor do they represent all possible interaction processes between the network and the terminal. They merely illustrate some potential air interface interaction processes considered in in some embodiments of the present disclosure. Additionally, the sequence of different signaling is not fixed, and there is no strict order between different signaling messages.
[0089] In summary, at least one solution mentioned in some embodiments of the present disclosure is applicable to the aforementioned scenarios and are suitable for NR and subsequent communication systems, such as 6G and 7G. The at least one solution is also applicable to WiFi or other similar communication systems.
[0090] The at least one main inventive point in some embodiments of the present disclosure is as follows: For performance monitoring, UE group-based indication information is designed to activate or deactivate performance monitoring result calculation. Meanwhile, a dynamic adjustment mechanism for the monitoring window length and reporting periodicity is introduced in this innovation. Additionally, reporting a smaller portion of the performance monitoring results within the monitoring window is considered to reduce feedback overhead. Furthermore, a new parameter is introduced and combined with an existing parameter to jointly control CSI reporting priority. Considering that functionality activation requires some time, activation time is introduced in the CSI reporting reference times Zref and Z'ref. Further, given the relationship between performance monitoring results and model inference results, when model inference is unavailable or not reported, the UE updates CSI reports with lower priority compared to performance monitoring results reporting.
[0091] Through some embodiments of the present disclosure, the UE can support AI-based CSI prediction using a UE-side model to enhance CSI prediction accuracy and further improve system capacity. Meanwhile, some embodiments also achieve at least one of the following technical effects:
[0092] (1) Reducing computational complexity: Some embodiments may allow the NW to instruct the UE to compute the performance metric, thereby reducing the computational complexity for the UE. With this method, the UE is not required to compute the performance metric continuously for periodic, semi-persistent, or event-driven reporting. Meanwhile, dynamically adjusting the length of the monitoring window further helps reduce the UE’s computational complexity. This technical effect may correspond to Embodiment 1.
[0093] (2) Reducing reporting overhead: Some embodiments may allow the UE to report only the minimum performance monitoring results, e.g., SGCS or NMSE, within the monitoring window, thereby reducing the reporting overhead of the UE. This technical effect may correspond to Embodiments 3 and / or 4.
[0094] (3) Enhancing system robustness: Some embodiments may allow the UE to report event-driven performance monitoring results with higher priority than periodic or semi-persistent predicted CSI reporting. With this method, the NW can determine whether to fall back to the legacy Rel-18 Doppler codebook in a timely manner. This helps maintain system stability and robustness. This technical effect may correspond to Embodiment 2.
[0095] (4) Improving resource utilization: Some embodiments may allow the UE to skip updating the requested CSI reports for performance monitoring when the corresponding CSI reports for model inference are unavailable or not reported. Instead, the UE can update subsequent requested CSI reports with lower priority if sufficient CPU resources are available. The CPU occupancy rule described may help improve UE resource utilization. This technical effect may correspond to Embodiments 6 and / or 7.
[0096] (5) Reducing control signaling overhead: Some embodiments may allow the UE to indicate a group of UEs to activate or deactivate performance monitoring result calculations, particularly for event-driven monitoring result reporting. This indication method may help reduce control signaling overhead. This technical effect may correspond to Embodiment 1.
[0097] FIG. 1B illustrates that, in some embodiments, one or more user equipments (UEs) 10 and a base station 20 such as a base station (e.g., next generation NodeB (gNB) or eNB) of communication in a communication network system 30 (e.g., an NR system) according to an embodiment of the present disclosure are provided. The communication network system 30 includes the one or more UEs 10 and the base station 20. The one or more UEs 10 may include a memory 12, a transceiver 13, and a processor 11 coupled to the memory 12 and the transceiver 13. The base station 20 may include a memory 22, a transceiver 23, and a processor 21 coupled to the memory 22 and the transceiver 23. The processor 11 or 21 may be configured to implement proposed functions, procedures and / or methods described in this description. Layers of radio interface protocol may be implemented in the processor 11 or 21. The memory 12 or 22 is operatively coupled with the processor 11 or 21 and stores a variety of information to operate the processor 11 or 21. The transceiver 13 or 23 is operatively coupled with the processor 11 or 21, and the transceiver 13 or 23 transmits and / or receives a radio signal.
[0098] The processor 11 or 21 may include application-specific integrated circuit (ASIC) , other chipset, logic circuit and / or data processing device. The memory 12 or 22 may include read-only memory (ROM) , random access memory (RAM) , flash memory, memory card, storage medium and / or other storage device. The transceiver 13 or 23 may include baseband circuitry to process radio frequency signals. When the embodiments are implemented in software, the techniques described herein can be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The modules can be stored in the memory 12 or 22 and executed by the processor 11 or 21. The memory 12 or 22 can be implemented within the processor 11 or 21 or external to the processor 11 or 21 in which case those can be communicatively coupled to the processor 11 or 21 via various means as is known in the art.
[0099] In some embodiments, the UE 10 is configured to perform: reporting, to a base station, a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting, receiving a CSI reporting configuration information from the base station, receiving channel measurement signals from the base station, reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting, and receiving a life cycle management (LCM) indication information from the base station.
[0100] In some embodiments, the UE 10 is configured to perform: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.
[0101] In some embodiments, the UE 10 is configured to perform: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; reporting, to the base station, performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and receiving a life cycle management (LCM) indication information from the base station; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.
[0102] In some embodiments, the UE 10 is configured to perform: receiving a monitoring indication information from a base station; and reporting, to the base station, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.
[0103] In some embodiments, the UE 10 is configured to perform: receiving a channel state information (CSI) reporting configuration information from the base station; receiving channel measurement signals from the base station; reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information; receiving monitoring measurement signals from the base station; and reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.
[0104] This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0105] In some embodiments, the base station 20 is configured to perform: receiving, from a user equipment (UE) , a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting; transmitting, to the UE, a CSI reporting configuration information; transmitting, to the UE, channel measurement signals; receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting; and transmitting, to the UE, a life cycle management (LCM) indication information.
[0106] In some embodiments, the base station 20 is configured to perform: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.
[0107] In some embodiments, the base station 20 is configured to perform: determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and transmitting, to the UE, a life cycle management (LCM) indication information; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.
[0108] In some embodiments, the base station 20 is configured to perform: transmitting, a user equipment (UE) , a monitoring indication information; and receiving, from the UE, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.
[0109] In some embodiments, the base station 20 is configured to perform: transmitting, to a user equipment (UE) , a channel state information (CSI) reporting configuration information; transmitting, to the UE, channel measurement signals; receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; transmitting, to the UE, monitoring measurement signals; and receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.
[0110] This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0111] In the above, some examples describe a method for a user equipment (UE) to perform AI-based channel state information (CSI) reporting, including reporting support for processors and CSI memory units (CMUs) , receiving CSI reporting and monitoring configurations, conducting CSI measurements and predictions, reporting performance monitoring results based on predefined priorities, and adapting reporting strategies according to different life cycle management (LCM) stages to enable smarter CSI reporting and efficient system resource management. This method enhances CSI prediction accuracy and system capacity while reducing computational complexity and signaling overhead, improving system robustness and resource utilization.
[0112] FIG. 2A is an example of a wireless communication method 200A performed by a UE according to an embodiment of the present disclosure. The wireless communication method 200A performed by a UE is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 200A performed by a UE using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 200A performed by a UE includes: an operation 202A, reporting, to a base station, a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting, an operation 204A, receiving a CSI reporting configuration information from the base station, an operation 206A, receiving channel measurement signals from the base station, an operation 208A, reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information, an operation 210A, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting, and an operation 212A, receiving a life cycle management (LCM) indication information from the base station. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0113] FIG. 2B is an example of a wireless communication method 200B performed by a base station according to an embodiment of the present disclosure. The wireless communication method 200B performed by a base station is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 200B performed by a base station using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 200B performed by a base station includes: an operation 202B, receiving, from a user equipment (UE) , a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting; an operation 204B, transmitting, to the UE, a CSI reporting configuration information; an operation 206B, transmitting, to the UE, channel measurement signals; an operation 208B, receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; an operation 210B, receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting; and an operation 212B, transmitting, to the UE, a life cycle management (LCM) indication information. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0114] In some examples, the processor comprises at least one CSI processing unit for AI-based CSI reporting. In some examples, the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression. In some examples, the wireless communication method further comprises receiving a downlink data information from the base station. In some examples, the wireless communication method further comprises determining predefined rules supporting AI-based CSI reporting, wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CMU usage.
[0115] FIG. 2A and FIG. 2B respectively illustrate example wireless communication methods performed by a user equipment (UE) and a base station, which together outline the end-to-end interaction procedure for AI-based Channel State Information (CSI) reporting. These figures define a series of operations, such as capability reporting, configuration reception, signal measurement, CSI prediction and reporting, performance monitoring, and life cycle management (LCM) indication exchanges between the UE and the base station. FIG. 2C builds upon this framework by providing a more detailed and implementation-focused example of the AI-based CSI reporting process. In particular, FIG. 2C elaborates on the signaling procedures, priority rules, processing time alignment, and resource constraints (e.g., CPU and CMU usage) , offering a refined and comprehensive view of the internal processing and reporting mechanisms involved in AI / ML-enabled CSI reporting.
[0116] FIG. 2C is an example of a wireless communication method according to an embodiment of the present disclosure. The wireless communication method is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method using any suitably configured hardware and / or software. FIG. 2C illustrates that, in some embodiments, the reporting process for AI-based CSI prediction is similar to the PMI prediction and CSI reporting process based on the eType-II codebook in 3GPP Rel-18. The primary difference between the two lies in whether AI is used for CSI prediction. However, AI-based CSI prediction also involves a series of operations, such as data collection, inference, model monitoring, and LCM management. In this regard, some embodiments primarily consider at least one of the following: the relevant signaling and procedures when data collection is based on network-side instructions, the air interface signaling interactions involved in model monitoring, the related aspects of LCM management, performing AI-based CSI prediction and reporting the CSI. The signaling interaction process between the base station and the UE, including at least one innovative point, is at least as shown in FIG. 2C below:
[0117] FIG. 2C illustrates that, in some embodiments, the wireless communication method includes at least one of the following operations:
[0118] Operation 1: Predefine the relevant rules supporting AI-based CSI prediction functionality, including: the reporting priority of the AI model's reporting content at different LCM stages, the specified CSI processing time corresponding to CSI reporting at different stages of the AI model, and CPU usage and / or CMU usage. In some embodiments, the specific details are as follows: Reporting priority of LCM stage reporting content. There is no distinction between AI-based CSI reporting and traditional CSI reporting.
[0119] Reporting Priority Handling Mechanism in the LCM Stage: In some embodiments, a unified mechanism is proposed for determining the reporting priority of LCM (Lifecycle Management) stage content, including AI-based CSI reports, traditional CSI reports, and performance monitoring results. Importantly, this invention does not distinguish between AI-based and traditional CSI reporting but instead extends the existing priority calculation formula to handle various types of reports appropriately. To achieve this goal, at least one solution such as Solution 1 and Solution 2 are proposed.
[0120] Reporting priority of LCM stage reporting content: Solution 1: By introducing parameter x and combining it with the existing parameter k within the priority calculation formula PriCSI (y, k, c, s) , the reporting of performance monitoring results, event-triggered monitoring results, and CSI reporting can be differentiated. For example, k=0 and x=1 represent CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR. Additionally, setting the coefficient factor l≥8 ensures that the priority of CSI reporting is higher than that of performance monitoring result reporting. The specific method can be referred to in Method 1 of Embodiment 2.
[0121] Reporting priority of LCM stage reporting content: Solution 2: The parameter k within the priority calculation formula PriCSI (y, k, c, s) can be expressed by at least one of k∈ {0, 1, 2, ..., 10} , and each k value can express at least one of the following cases: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR; CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR; Event-triggered CSI reports carrying L1-RSRP or L1-SINR; Additionally, the coefficient l in PriCSI (y, k, c, s) is kmax+1 and kmax indicates the maximum value of k. The specific method can refer to Method 3 in Embodiment 2.
[0122] In the above reporting priority handling mechanism in the LCM stage, some examples are that regardless of whether CSI reports are AI-based or not, the system can manage the timing and importance of different report types effectively by introducing or extending priority calculation parameters. This allows for flexible and efficient handling of resource scheduling and LCM decisions when multiple sources of CSI and monitoring information are present.
[0123] The reporting priority handling mechanism in the LCM (Lifecycle Management) stage constitutes an independent invention point. This mechanism focuses on determining and managing the priority of different types of reporting content, such as AI-based CSI reporting, traditional CSI reporting, and performance monitoring result reporting, within the LCM stage. This invention is not limited to a specific implementation but provides a flexible framework that can be adapted to various reporting configurations and system capabilities. For example, it allows for differentiated handling of CSI reports that carry performance monitoring results or event-triggered information, while maintaining consistency across AI and non-AI-based functionalities. The proposed priority handling logic can be implemented by introducing additional parameters or extending existing ones within the priority calculation formula, enabling fine-grained control of reporting behavior. The design ensures seamless coordination between different reporting types and optimizes network responsiveness. Importantly, this invention point may be applied independently, or in combination with other embodiments or invention points, such as AI-based CSI compression, switching mechanisms between reporting types, or model monitoring frameworks. The described mechanisms are not limited to a specific signaling format or standard, and may be adapted to future enhancements, thereby offering broad applicability and deployment flexibility across different system architectures and standard releases.
[0124] CSI processing time: By simultaneously adding the functionality activation time ΔT to both CSI reporting reference times Zref and Z'ref, sufficient time for CSI reporting during aperiodic CSI reporting is ensured. Meanwhile, by stipulating that the offset X for the delivery of aperiodic CSI-RS resources during aperiodic CSI reporting be greater than a certain value Y, it ensures that the AI-based functionality can complete its activation before channel measurement, thereby enabling model inference operations. The specific method can be referred to in Embodiment 5.
[0125] The mechanism for configuring CSI processing time can constitute an independent invention point in this disclosure. It addresses the need to ensure sufficient time for processing AI-based CSI reporting in aperiodic scenarios, particularly when model activation and inference are involved. This invention point can be applied independently or in combination with other mechanisms, such as AI-based CSI compression, switching mechanisms, or performance monitoring frameworks. It is not limited to any specific signaling format, standard specification, frequency band, or CSI report type, and can be adapted for both current and future communication systems (e.g., 5G, 6G, 7G, etc. ) The described configuration logic supports flexible deployment across devices with different processing capabilities and scales well with increasing AI integration in communication protocols.
[0126] CPU utilization rules:
[0127] CPU utilization rules: Solution 1: If the CSI reporting for performance monitoring corresponds to model inference results that are unavailable or not reported, the UE is not required to update the requested CSI reports for performance monitoring. Instead, if sufficient CPU resources are available, the UE is required to update the subsequent requested CSI report, which has a lower priority than the CSI reports for performance monitoring (according to 3GPP TS 38.214 Clause 5.2.5) . The specific method can be referred to in Embodiment 6.
[0128] CPU utilization rules: Solution 2: If the overall available memory for all AI models, functionalities, or features is considered a separate UE capability, then updating the requested CSI reports depends on two factors: CPU and CMU (CSI Memory Unit) . This means that the UE is only required to update the first M requested CSI reports with the highest priority (according to 3GPP TS 38.214 Clause 5.2.5) , where M=min {M1, M2} , M1 represents the first M1 CSI reports with the highest priority that the CPU of the UE can support (that may represent the maximum number of highest-priority CSI reports that the UE's CPU can support) , and M2 represents the first M2 CSI reports with the highest priority that the CMU of the UE can support (that may represent the maximum number of highest-priority CSI reports that the UE's CMU can support) . More details can be found in Method 1 of Embodiment 7.
[0129] The proposed CPU utilization rules mechanism can constitute an independent invention point. Its core purpose is to optimize the update and reporting behavior of CSI (Channel State Information) under resource-constrained conditions by intelligently managing computation resources, particularly in scenarios involving AI model inference and performance monitoring. This invention point can be implemented independently as a standalone resource management mechanism or combined with other invention points, such as AI-based CSI compression, report switching strategies, or LCM decision mechanisms. It is not limited by platform, standard, or report type, and can be extended or adapted to fit different UE capabilities, network requirements, or standard evolutions. The proposed mechanism is highly suitable and scalable for intelligent communication systems, especially those operating with multi-model AI inference and dynamic CSI reporting in 5G, 6G, or 7G architectures.
[0130] Operation 2: The UE reporting supports the capability of AI-based CSI prediction, including information such as whether the overall CPU usage is accounted for separately or jointly between legacy CSI reporting and AI / ML-based CSI reporting. Additionally, performance metrics for AI-based use cases, such as NMSE and / or SGCS, are supported.
[0131] Operation 3: The UE receives configuration information, which primarily includes measurement reference signal configuration, CSI reporting configuration, and model or function monitoring result reporting configuration.
[0132] Operation 4: The UE receives measurement reference signals, and the results obtained from these measurements are primarily used as inputs for the model. The UE performs inference based on the received measurement reference signals to obtain the predicted CSI and then compresses the predicted CSI for reporting.
[0133] Operation 5: The UE reports CSI by compressing and transmitting the predicted CSI.
[0134] Operation 6: The UE receives monitoring indication information, which is used to instruct the terminal to perform monitoring result calculations for event-driven monitoring result reporting. The indicated information includes at least one of the following items:
[0135] A 1-bit field indicating whether to trigger performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, nothing happens. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, nothing happens.
[0136] A 1-bit field indicating whether to trigger or deactivate performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, performance monitoring is deactivated. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, performance monitoring is deactivated.
[0137] A field specifying the duration of performance monitoring.
[0138] A G-bit field indicating that at least one UE group requires performance monitoring. Each bit specifies whether the associated UE group requires performance monitoring. Each bit is associated with one UE group. The specific method can be referred to in Embodiment 1.
[0139] The design and usage of indicated information to trigger or control performance monitoring constitutes an independent invention point. This invention addresses the need for flexible and efficient signaling from the network to the UE, enabling dynamic control of event-driven monitoring result reporting. This invention proposes that the indicated information received by the UE in the form of a monitoring indication message can include various control fields. These fields allow the network to selectively initiate, deactivate, or configure performance monitoring based on system needs or operational conditions. This invention point can be implemented independently to enhance UE-side flexibility in performance monitoring activation and control. It can also be combined with other invention points, such as event-triggered CSI reporting, model management, or performance-based LCM mechanisms. The design is not limited by specific protocol layers, bit positions, or signaling formats, and can be easily adapted to future 5G, 6G, or 7G protocol evolution. The bit-level control granularity ensures low signaling overhead, making it well-suited for massive device deployment scenarios or energy-sensitive applications.
[0140] Operation 7: The UE receives monitoring measurement signals, which are primarily used for CSI measurement. The measured CSI serves as the ground truth for model or function monitoring.
[0141] Operation 8: The UE reports monitoring results, which are primarily used by the network side to make decisions regarding the life cycle management (LCM) of the model or functionality. The main monitoring report content can be presented in at least one of the following forms:
[0142] Monitoring report content: Solution 1: The SGCS is calculated across all sub-bands and / or all layers for a single predicted CSI time occasion, and the M monitoring results with the smallest SGCS values of the predicted CSI within the monitoring window are reported as CSI Part 1. The reported content can be mapped according to the time-domain order of the predicted CSI. When differential reporting is used, priority is given to mapping the position information of the predicted CSI corresponding to the maximum or minimum SGCS value, as well as the quantized information of the maximum or minimum SGCS value. The remaining information can be mapped according to the time-domain order of the predicted CSI. The specific method can be referred to in Embodiment 3.
[0143] Monitoring report content: Solution 2: The SGCS is calculated across all sub-bands and / or all layers for a single predicted CSI time occasion, and the comparison results of the M monitoring results with the smallest SGCS values of the predicted CSI within the monitoring window against a threshold are reported as CSI Part 1. The reported content can be mapped according to the time-domain order of the predicted CSI. The specific method can be referred to in Embodiment 4.
[0144] The design of the monitoring report content can constitute an independent invention point, focusing on how the UE composes and reports performance monitoring results to support life cycle management (LCM) decisions at the network side. This mechanism enables the network to assess model / functionality behavior over time and make informed decisions regarding activation, replacement, or deactivation of AI-based functionalities. This invention defines multiple flexible methods for generating and structuring monitoring report content, based on SGCS (Statistical Gap-based Comparison Score) values derived from predicted CSI data within a monitoring window. This invention point can be independently implemented as part of a flexible performance monitoring framework for AI model or functionality management. It may also be combined with other invention points, such as differential quantization schemes, reporting priority mechanisms, or model switching logic. The mechanism is not limited by specific thresholds, mapping orders, or SGCS calculation methods, and supports customization based on operator policy, UE capability, or model complexity. The reporting structure is designed to minimize signaling overhead while maintaining sufficient granularity for model evaluation, making it highly suitable for next-generation 5G, 6G, or 7G systems with AI-based prediction and feedback loops.
[0145] Operation 9: The UE receives LCM indication information, which is used to instruct the UE to perform corresponding LCM operations, primarily including functionality fallback operations. The specific signaling bearer methods are as follows: The fallback operation can be explicitly or implicitly indicated through the deactivation of performance monitoring result reporting via DCI or MAC CE messages. The specific method can be referred to in Embodiments 3 and 8.
[0146] Operation 10: The UE receives downlink data information, i.e., the UE receives downlink data sent by the network after precoding is applied to the data based on the CSI reported by the UE.
[0147] FIG. 2C presents a detailed example of a wireless communication method for AI-based CSI prediction, expanding upon the foundational processes outlined in FIG. 2A and FIG. 2B. It describes a comprehensive sequence that includes data collection, model inference, model monitoring, and life cycle management (LCM) . The method defines priority calculation rules, timing alignment for aperiodic CSI reporting, CPU / CMU resource constraints, performance metrics (such as NMSE and SGCS) , reporting configurations, and compression strategies. Specific operations include receiving measurement reference signals, performing AI model inference and compressed reporting, executing event-triggered monitoring based on monitoring indications, receiving monitoring measurement signals, reporting model performance results, and executing fallback operations based on LCM instructions. Moreover, the method supports flexible sequencing and combinations of operations, enhancing the scalability and implementation adaptability of the AI-driven CSI reporting framework. This method provides a comprehensive and flexible AI-driven CSI reporting framework that significantly improves prediction accuracy, resource efficiency, and strengthens model monitoring and life cycle management capabilities.
[0148] It is to be understood that at least one of the operations illustrated in FIG. 2A to FIG. 2C may be performed independently or in combination with other operations, either sequentially or concurrently, depending on implementation preferences. Furthermore, individual operations or solutions described may be selectively activated, modified, or omitted to suit different deployment scenarios or device capabilities. This modular and flexible design ensures that the disclosed embodiments can be adapted to diverse system requirements and evolving AI-based communication technologies, thereby extending the scope and applicability of the present disclosure.
[0149] The reporting priority of LCM stage reporting content:
[0150] FIG. 3A is an example of a wireless communication method 300A performed by a UE according to an embodiment of the present disclosure. The wireless communication method 300A performed by a UE is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 300A performed by a UE using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 300A performed by a UE includes: an operation 302A, determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and an operation 304A, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0151] FIG. 3B is an example of a wireless communication method 300B performed by a base station according to an embodiment of the present disclosure. The wireless communication method 300B performed by a base station is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 300B performed by a base station using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 300B performed by a base station includes: an operation 302B, determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; and an operation 304B, receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0152] In some examples, the predefined rules supporting AI-based CSI reporting comprise at least one of the following: the priority of the CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage. In some examples, a priority of the CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting. In some examples, the processor comprises at least one CSI processing unit for AI-based CSI reporting. In some examples, the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression. In some examples, the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting. In some examples, the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the performance monitoring results. In some examples, the coefficient factor is equal to a maximum value of the first parameter plus 1. In some examples, values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following: for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ; for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR; for CSI report reporting carrying the L1-RSRP or the L1-SINR; for CSI reporting not carrying the L1-RSRP or the L1-SINR; for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR; for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR; for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR.
[0153] The wireless communication methods 300A and 300B illustrated in FIG. 3A and FIG. 3B respectively provide a high-level overview of operations performed by a UE and a base station in support of AI-based CSI reporting. These methods establish a foundation for prioritizing CSI reporting based on predefined rules, including life cycle management (LCM) stage priorities and differentiated CSI types. To further clarify how such predefined rules can be implemented, the following provides more specific examples and parameter configurations that demonstrate practical application strategies. It should be noted that at least one operation depicted in FIG. 3A or FIG. 3B, or in the exemplary methods described herein, can be independently executed or combined with other operations based on implementation requirements. This flexibility enables a broader protection scope and supports diverse system capabilities and deployment scenarios. The examples below define in greater detail how parameters such as priority indicators, CSI processing time, processor and memory resource usage can be utilized to dynamically adjust reporting behavior. By integrating such configurations into the priority calculation logic (e.g., using PriCSI (y, k, c, s) ) , the system can achieve more accurate scheduling, resource efficiency, and seamless support for both AI-based and traditional CSI reporting mechanisms.
[0154] For examples, predefine the relevant rules supporting AI-based CSI prediction functionality, including: the reporting priority of the AI model's reporting content at different LCM stages, the specified CSI processing time corresponding to CSI reporting at different stages of the AI model, and CPU usage and / or CMU usage. In some embodiments, the specific details are as follows: Reporting priority of LCM stage reporting content. There is no distinction between AI-based CSI reporting and traditional CSI reporting.
[0155] Solution 1: By introducing parameter x and combining it with the existing parameter k within the priority calculation formula PriCSI (y, k, c, s) , the reporting of performance monitoring results, event-triggered monitoring results, and CSI reporting can be differentiated. For example, k=0 and x=1 represent CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR. Additionally, setting the coefficient factor l≥8 ensures that the priority of CSI reporting is higher than that of performance monitoring result reporting. The specific method can be referred to in Method 1 of Embodiment 2.
[0156] Solution 2: The parameter k within the priority calculation formula PriCSI (y, k, c, s) can be expressed by at least one of k∈ {0, 1, 2, ..., 10} , and each k value can express at least one of the following cases: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR; CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR; Event-triggered CSI reports carrying L1-RSRP or L1-SINR; Additionally, the coefficient l in PriCSI (y, k, c, s) is kmax+1 and kmax indicates the maximum value of k. The specific method can refer to Method 3 in Embodiment 2.
[0157] The configuration of reporting priority at different Life Cycle Management (LCM) stages, as illustrated in FIG. 3A and FIG. 3B, constitutes an independent invention point. This invention introduces a flexible priority mechanism that applies to both AI-based and non-AI-based CSI reporting, as well as performance monitoring result reporting. By integrating predefined parameters-such as additional priority indicators (e.g., parameter x) , CSI processing time, and resource usage constraints (CPU / CMU) -into the existing CSI priority calculation formula, the system enables accurate differentiation and scheduling of reporting types across LCM stages. The invention supports various configurations, such as using coefficient factors to rank CSI and performance monitoring reports, and mapping different values of parameter k to specific report types. This priority handling logic can be implemented independently or in combination with other modules such as AI-based prediction, beam management, or compression. It is not limited by signaling format, CSI type, or standard version, and can be flexibly adapted to diverse deployment scenarios in 5G, 6G or 7G systems.
[0158] CSI processing time:
[0159] FIG. 4A is an example of a wireless communication method 400A performed by a UE according to an embodiment of the present disclosure. The wireless communication method 400A performed by a UE is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 400A performed by a UE using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 400A performed by a UE includes: an operation 402A, determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; an operation 404A, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and an operation 406A, receiving a life cycle management (LCM) indication information from the base station; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0160] FIG. 4B is an example of a wireless communication method 400B performed by a base station according to an embodiment of the present disclosure. The wireless communication method 400B performed by a base station is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 400B performed by a base station using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 400B performed by a base station includes: an operation 402B, determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; an operation 404B, receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; and an operation 406B, transmitting, to the UE, a life cycle management (LCM) indication information; wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0161] In some examples, the processing time is managed by at least one of the following: adding an activation time to a first reference time and a second reference time for CSI reporting; stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value. In some examples, the processor usage and / or the CMU usage is managed by at least one of the following: if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported and / or if sufficient processing resources are available, the UE updates a requested CSI report with a lower priority than the performance monitoring results. In some examples, the UE determines the CSI reporting based on at least one of the following: calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting; prioritizing CSI reporting based on a CMU availability and / or a CPU availability.
[0162] In some examples, for AI-based CSI prediction, if a number of CSI-RS resources is larger than 12, an occupy CPU ∈ {8, 12, 16, 18, 20} ; if the number of CSI-RS resources is less than 12, OCPU=Y1·K, where Y1∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by a UE capability indication, K is a parameter; if N4>1, OCPU=max (Y2·N4, 4) and Y2∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by the UE capability indication. In some examples, the UE switches between an AI-based CSI reporting and a non-AI-based CSI reporting via a downlink control information (DCI) or a medium access control-control element (MAC CE) message; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the DCI or the MAC CE message to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting. In some examples, an AI-based CSI reporting fallback or deactivate time point is located before a first non-AI-based CSI reporting. In some examples, the MAC CE message or the DCI comprises at least one of the following: a field used to indicate an applicable CSI reporting configuration ID; a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting; a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.
[0163] The wireless communication methods 400A and 400B illustrated in FIG. 4A and FIG. 4B respectively describe operations performed by a user equipment (UE) and a base station to support AI-based CSI reporting, including the use of predefined rules, performance monitoring, and LCM (Life Cycle Management) signaling. These figures provide a generalized process flow applicable to both network and device sides and highlight how AI-enhanced reporting functionalities can be coordinated through signaling and intelligent resource management. To further illustrate the practical application of these concepts, the following sections present more detailed and concrete examples regarding processing time adjustments, CPU and CMU usage control, fallback mechanisms, and signaling configurations such as MAC CE and DCI. It should be noted that at least one operation described in FIG. 4A, FIG. 4B, or in the exemplary embodiments below, may be implemented independently or in conjunction with others, allowing for a flexible deployment strategy adaptable to specific network capabilities and UE configurations. This design ensures broader applicability and scalability across varying system conditions while maintaining a low overhead and high efficiency for CSI reporting and AI functionality activation.
[0164] Some examples are illustrated below. CSI processing time: By simultaneously adding the functionality activation time ΔT to both CSI reporting reference times Zref and Z'ref, sufficient time for CSI reporting during aperiodic CSI reporting is ensured. Meanwhile, by stipulating that the offset X for the delivery of aperiodic CSI-RS resources during aperiodic CSI reporting be greater than a certain value Y, it ensures that the AI-based functionality can complete its activation before channel measurement, thereby enabling model inference operations. The specific method can be referred to in Embodiment 5.
[0165] The CSI processing time configuration mechanism illustrated in FIG. 4A and FIG. 4B constitutes an independent invention point, aiming to ensure that AI functionalities can be successfully activated and complete model inference during aperiodic CSI reporting through flexible timing control strategies. Specifically, this mechanism includes: (1) simultaneously adding a functionality activation time ΔT to both CSI reporting reference time points, to guarantee sufficient processing time for the UE; and (2) stipulating that the offset X for the delivery of aperiodic CSI-RS resources during aperiodic CSI reporting must be greater than a predetermined threshold Y, ensuring that AI functionalities are fully activated before channel measurement begins. This design can be implemented independently or in combination with other control signaling mechanisms such as MAC CE or DCI, improving the accuracy of scheduling and the efficiency of resource management for AI-based CSI reporting. The mechanism is not limited by specific report formats, scheduling methods, or standard versions and can be flexibly adapted based on UE capabilities, network load, or AI functionality type. It offers high scalability and general applicability, making it suitable for AI-enhanced CSI reporting scenarios in 5G, 6G, or 7G systems.
[0166] CPU utilization rules:
[0167] Solution 1: If the CSI reporting for performance monitoring corresponds to model inference results that are unavailable or not reported, the UE is not required to update the requested CSI reports for performance monitoring. Instead, if sufficient CPU resources are available, the UE is required to update the subsequent requested CSI report, which has a lower priority than the CSI reports for performance monitoring (according to 3GPP TS 38.214 Clause 5.2.5) . The specific method can be referred to in Embodiment 6.
[0168] Solution 2: If the overall available memory for all AI models, functionalities, or features is considered a separate UE capability, then updating the requested CSI reports depends on two factors: CPU and CMU (CSI Memory Unit) . This means that the UE is only required to update the first M requested CSI reports with the highest priority (according to 3GPP TS 38.214 Clause 5.2.5) , where M=min {M1, M2} , M1 represents the first M1 CSI reports with the highest priority that the CPU of the UE can support (that may represent the maximum number of highest-priority CSI reports that the UE's CPU can support) , and M2 represents the first M2 CSI reports with the highest priority that the CMU of the UE can support (that may represent the maximum number of highest-priority CSI reports that the UE's CMU can support) . More details can be found in Method 1 of Embodiment 7.
[0169] The CPU utilization rules illustrated in FIG. 4A and FIG. 4B represent an independent invention point, providing a flexible mechanism for managing CSI reporting behavior based on the availability of computing and memory resources at the UE side. This invention addresses practical challenges in AI-based CSI reporting by introducing two resource-aware strategies: (1) if model inference results are unavailable, the UE is not required to update performance monitoring-related CSI reports, and instead, may update lower-priority CSI reports only if CPU resources are sufficient; and (2) if the total available memory for AI models and functionalities is treated as a separate capability, the UE updates only the first M highest-priority CSI reports, where M = min {M1, M2} , with M1 and M2 representing the limits imposed by CPU and CSI Memory Unit (CMU) resources, respectively. These approaches ensure efficient resource allocation and priority-based scheduling under varying system loads. The mechanism can be independently implemented or combined with other LCM, signaling, or scheduling methods, and is adaptable to diverse UE hardware configurations and network deployments, offering strong flexibility and scalability across AI-enhanced 5G, 6G, or 7G systems.
[0170] Indicated information:
[0171] FIG. 5A is an example of a wireless communication method 500A performed by a UE according to an embodiment of the present disclosure. The wireless communication method 500A performed by a UE is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 500A performed by a UE using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 500A performed by a UE includes: an operation 502A, receiving a monitoring indication information from a base station; and an operation 504A, reporting, to the base station, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0172] FIG. 5B is an example of a wireless communication method 500B performed by a base station according to an embodiment of the present disclosure. The wireless communication method 500B performed by a base station is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 500B performed by a base station using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 500B performed by a base station includes: an operation 502B, transmitting, a user equipment (UE) , a monitoring indication information; and an operation 504B, receiving, from the UE, performance monitoring results for AI-based CSI reporting; wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0173] In some examples, the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.
[0174] The wireless communication methods 500A and 500B illustrated in FIG. 5A and FIG. 5B respectively outline a general framework for enabling performance monitoring result calculation and reporting between a UE and a base station. These methods are designed to support AI-based CSI reporting by allowing the network to control when and how the UE performs monitoring operations through monitoring indication information. This signaling may include simple 1-bit triggers, group-based configurations, or time-bound instructions. To further demonstrate how such mechanisms can be practically implemented, the following provides more specific examples that detail different configurations of the monitoring indication information, including trigger conditions, duration settings, and group-based control logic. It should be emphasized that at least one of the operations shown in FIG. 5A, FIG. 5B, or the examples that follow may be performed independently or in combination with others depending on system capabilities, deployment scenarios, and network policies. This modularity ensures greater implementation flexibility and allows the invention to adapt seamlessly to diverse use cases, thereby expanding the protection scope and enhancing robustness under various communication conditions.
[0175] For example, the UE receives monitoring indication information, which is used to instruct the terminal to perform monitoring result calculations for event-driven monitoring result reporting. The indicated information includes at least one of the following items:
[0176] A 1-bit field indicating whether to trigger performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, nothing happens. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, nothing happens.
[0177] A 1-bit field indicating whether to trigger or deactivate performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, performance monitoring is deactivated. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, performance monitoring is deactivated.
[0178] A field specifying the duration of performance monitoring.
[0179] A G-bit field indicating that at least one UE group requires performance monitoring. Each bit specifies whether the associated UE group requires performance monitoring. Each bit is associated with one UE group. The specific method can be referred to in Embodiment 1.
[0180] The indicated information mechanism described in FIG. 5A and FIG. 5B constitutes an independent invention point, offering a flexible signaling framework that enables the network to dynamically control performance monitoring result calculations at the UE side. This mechanism utilizes predefined signaling fields-including simple 1-bit triggers, monitoring duration fields, and G-bit group-specific indicators-to instruct the UE when and how to perform event-driven performance monitoring in support of AI-based CSI reporting. For example, a 1-bit field may activate or deactivate monitoring based on its value, while a G-bit field allows selective activation for specific UE groups. These configurations can be carried via DCI or MAC CE messages and applied at the individual UE, UE group, or cell level, offering scalability and adaptability across deployment scenarios. The modular design ensures that each signaling element can be implemented independently or in combination with others, providing a highly customizable solution that enhances CSI reporting efficiency, reduces signaling overhead, and supports intelligent monitoring operations in 5G, 6G, or 7G networks.
[0181] Monitoring report content:
[0182] FIG. 6A is an example of a wireless communication method 600A performed by a UE according to an embodiment of the present disclosure. The wireless communication method 600A performed by a UE is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 600A performed by a UE using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 600A performed by a UE includes: an operation 602A, receiving a channel state information (CSI) reporting configuration information from the base station; an operation 604A, receiving channel measurement signals from the base station; an operation 606A, reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information; an operation 608A, receiving monitoring measurement signals from the base station; and an operation 610A, reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0183] FIG. 6B is an example of a wireless communication method 600B performed by a base station according to an embodiment of the present disclosure. The wireless communication method 600B performed by a base station is configured to implement some embodiments of the disclosure. Some embodiments of the disclosure may be implemented into the wireless communication method 600B performed by a base station using any suitably configured hardware and / or software. In some embodiments, the wireless communication method 600B performed by a base station includes: an operation 602B, transmitting, to a user equipment (UE) , a channel state information (CSI) reporting configuration information; an operation 604B, transmitting, to the UE, channel measurement signals; an operation 606B, receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information; an operation 608B, transmitting, to the UE, monitoring measurement signals; and an operation 610B, receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting. This can solve issues in the prior art and other issues, enhance a CSI prediction accuracy, improve a system capacity, reduce a computation complexity, reduce a reporting overhead, enhance a system robust, improve resource utilization, and / or reduce a control signaling overhead, etc.
[0184] In some examples, the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression. In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the UE triggers reporting of performance monitoring results. In some examples, a reporting period and / or a monitoring window length for performance monitoring results are configurable via a downlink control information (DCI) , a medium access control-control element (MAC CE) message, or a radio resource control (RRC) message; and / or a reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message. In some examples, the UE calculates a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and selects M performance monitoring results from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1. In some examples, M is configured by the base station or predefined. In some examples, the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting.
[0185] In some examples, when the UE uses differential reporting in the performance monitoring results, the UE uses a priority to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value. In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following: periodic reporting; semi-persistent reporting; aperiodic reporting; event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold. In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises: reporting, to the base station, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following: a bitmap representation; a combined digital representation; a reported CSI index. In some examples, the UE calculates a SGCS or NMSE across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and reports, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1. In some examples, the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.
[0186] The wireless communication methods 600A and 600B illustrated in FIG. 6A and FIG. 6B respectively describe the overall signaling and operation flows performed by a user equipment (UE) and a base station for AI-based CSI reporting and performance monitoring. These methods define how the CSI reporting configuration, channel and monitoring measurement signals, and corresponding CSI or performance monitoring results are exchanged between the UE and the network. Particularly, the methods highlight the application of a predefined priority mechanism to ensure that CSI reporting takes precedence over performance monitoring, thereby optimizing system responsiveness and processing efficiency. To provide a more comprehensive understanding of how such priority-based reporting is implemented, the following examples offer detailed mechanisms and metrics, such as SGCS (Squared Generalized Cosine Similarity) and NMSE (Normalized Mean Squared Error) , for selecting and structuring monitoring results. These examples demonstrate specific procedures for event-triggered reporting, comparison against thresholds, differential reporting techniques, and the use of bitmap or index-based reporting formats. Notably, at least one operation depicted in FIG. 6A, FIG. 6B, or the exemplary methods below may be carried out independently or in any suitable combination, offering flexibility to adapt the solution across diverse network deployments, UE capabilities, and service requirements. This modular and extensible design not only enhances implementation adaptability but also ensures broad protection coverage under varying usage scenarios.
[0187] For example, the UE reports monitoring results, which are primarily used by the network side to make decisions regarding the life cycle management (LCM) of the model or functionality. The main monitoring report content can be presented in at least one of the following forms:
[0188] Solution 1: The SGCS is calculated across all sub-bands and / or all layers for a single predicted CSI time occasion, and the M monitoring results with the smallest SGCS values of the predicted CSI within the monitoring window are reported as CSI Part 1. The reported content can be mapped according to the time-domain order of the predicted CSI. When differential reporting is used, priority is given to mapping the position information of the predicted CSI corresponding to the maximum or minimum SGCS value, as well as the quantized information of the maximum or minimum SGCS value. The remaining information can be mapped according to the time-domain order of the predicted CSI. The specific method can be referred to in Embodiment 3.
[0189] Solution 2: The SGCS is calculated across all sub-bands and / or all layers for a single predicted CSI time occasion, and the comparison results of the M monitoring results with the smallest SGCS values of the predicted CSI within the monitoring window against a threshold are reported as CSI Part 1. The reported content can be mapped according to the time-domain order of the predicted CSI. The specific method can be referred to in Embodiment 4.
[0190] The monitoring report content mechanism described in FIG. 6A and FIG. 6B represents an independent invention point, offering a flexible and scalable approach to structuring and transmitting performance monitoring results for AI-based CSI reporting. This invention enables the UE to selectively report the most representative performance metrics-such as SGCS (Squared Generalized Cosine Similarity) or NMSE (Normalized Mean Squared Error) -based on predefined rules and priority configurations. Specifically, the UE may report: (1) the M monitoring results with the smallest SGCS or NMSE values within a monitoring window (Solution 1) , or (2) their comparison against a predefined threshold (Solution 2) . These results can be reported using time-domain mapping, differential reporting techniques, or compact formats such as bitmaps or indices, thus reducing signaling overhead and improving resource efficiency. The mechanism supports multiple triggering modes (periodic, semi-persistent, aperiodic, or event-driven) and can adapt to different network configurations or UE capabilities. Moreover, the CSI and monitoring report prioritization ensures that critical reporting is not delayed or dropped under constrained resources. This modular design allows each element to be deployed independently or in combination with others, providing broad protection scope and flexible applicability across diverse 5G, 6G, or 7G communication scenarios.
[0191] In some embodiments of the present application, Embodiment 1, Embodiment 2, Embodiment 3, Embodiment 4, Embodiment 5, Embodiment 6, Embodiment 7, Embodiment 8, and / or Embodiment 9 may be implemented in combination with one another. The Embodiment 1, Embodiment 2, Embodiment 3, Embodiment 4, Embodiment 5, Embodiment 6, Embodiment 7, Embodiment 8, and / or Embodiment 9 may also be implemented independently. In some embodiments of the present application, the solutions described in multiple embodiments may be implemented either in combination or independently.
[0192] Embodiment 1: Trigger method of performance monitoring.
[0193] In some examples, the wireless communication method further comprises: receiving a monitoring indication information from the base station, wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following: a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation; a field indicating a duration of performance monitoring result calculation; a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group. In some examples, the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.
[0194] In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the UE triggers reporting of performance monitoring results. In some examples, a reporting period and / or a monitoring window length for performance monitoring results are configurable via a DCI, a MAC CE message, or a radio resource control (RRC) message; and / or a reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message.
[0195] The content above outlines the fundamental principles of performance monitoring triggering in AI-based CSI reporting, where the UE receives monitoring indication information from the base station and responds based on predefined reporting priorities and triggering conditions. It describes basic methods such as passive or active 1-bit triggers, duration indications, and UE group-specific monitoring controls. The following section provides more concrete implementation examples that build upon these principles. It expands the design by considering the computational burden on the UE and proposes refined mechanisms for triggering and reporting performance monitoring results based on network-indicated events and dynamically configurable parameters.
[0196] Importantly, the operations illustrated in this embodiment, whether related to event-driven triggering, periodic / semi-persistent / aperiodic configuration, or monitoring duration control, can be performed individually or in combination, in sequential or parallel fashion. This modular and flexible approach ensures that the present disclosure supports a wide range of deployment scenarios and UE capabilities, allowing for broader protection scope and practical adaptability in real-world network conditions.
[0197] Based on the conclusions of the previous meeting, performance monitoring result reporting can be periodic, semi-persistent, aperiodic, and / or event-driven. However, considering the computational complexity of the UE, we believe that focusing solely on the reporting mechanism is insufficient, as the UE must calculate performance metrics for all types of reporting: periodic, semi-persistent, aperiodic, and event-driven.
[0198] Assume that performance monitoring result reporting is event-driven. If a defined event is satisfied, it triggers the UE to report the performance monitoring results. Clearly, before such a trigger occurs, the UE must continuously calculate the performance monitoring metrics. In fact, if the generalization capability of the model is sufficiently robust, continuous performance monitoring may become unnecessary.
[0199] Therefore, in the present disclosure, some examples consider how to design a trigger mechanism for performance monitoring. Some examples provide at least one of the following methods:
[0200] Method 1: Network-indicated triggering via DCI or MAC CE.
[0201] The network indicates to the UE that it may perform performance monitoring via DCI or MAC CE signaling. The indication information in the DCI or MAC CE may include at least one of the following items:
[0202] 1-bit trigger indication (passive deactivation) : A 1-bit field is used to indicate whether to trigger performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, no action is taken. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, no action is taken.
[0203] 1-bit trigger indication (active deactivation) : A 1-bit field is used to indicate whether to trigger or deactivate performance monitoring. In some examples, if the bit is 1, performance monitoring is activated, and / or if the bit is 0, performance monitoring is deactivated. In another example, if the bit is 0, performance monitoring is activated, and / or if the bit is 1, performance monitoring is deactivated.
[0204] Monitoring duration indication: A field that specifies the duration of performance monitoring. For example, X bits may be used to: indicate the number of predicted CSIs to be monitored, and / or indicate the number of monitoring windows. The length of a monitoring window may be either configured by RRC or predefined. For example, a window could include N≥1 predicted CSIs. predicted CSIs. If the RRC configures M candidate values {k1, k2, ..., kM} for the number of predicted CSIs or windows, then X is equal to where which represents the number of bits required to encode M candidates.
[0205] UE group-based monitoring indication: A G-bit field is used to indicate that at least one UE group requires performance monitoring. Each bit specifies whether the associated UE group requires performance monitoring. Each bit is associated with one UE group. Each bit corresponds to a specific UE group. In some examples, a bit value of 1 indicates that the associated group requires performance monitoring. In another example, a bit value of 0 indicates that the associated group requires performance monitoring.
[0206] Note: The DCI or MAC CE information in this method can be implemented in one of the following ways: By introducing a new DCI format or MAC CE format, or by reusing an existing DCI format or MAC CE format and adding or reusing specific fields, as mentioned above.
[0207] For example, if a new DCI format is introduced, the DCI can be scrambled using MEI-RNTI, a specific RNTI used to distinguish DCI messages for triggering performance monitoring. The size of the new DCI format is indicated by the higher-layer parameter payloadSizeDCI-X-Y. The number of information bits in format X_Y must be equal to or less than the payload size of format X-Y. If the number of information bits in format X-Y is less than the format size, the remaining bits are reserved.
[0208] Clearly, based on the above indication, a fixed period of time can be ensured for performance monitoring. The reporting of performance monitoring results can then be event-driven. For example, if one or more performance monitoring results for predicted CSI are below or above a certain threshold, the UE reports the performance monitoring results.
[0209] Furthermore, the DCI or MAC CE information in this method can be UE-specific, UE group-specific, or cell-level. In some examples, if the DCI or MAC CE is applied at the cell level. In some examples, the bit in the 1-bit trigger indication (passive deactivation) is set to 1, this indicates that all applicable UEs under the cell will perform performance monitoring. In another example, the bit in the 1-bit trigger indication (passive deactivation) is set to 0, this indicates that all applicable UEs under the cell will perform performance monitoring.
[0210] Method 2: Dynamically configuring the reporting periodicity and / or the length of the monitoring window for performance monitoring.
[0211] Performance monitoring result reporting can be periodic, semi-persistent, or aperiodic. Based on the existing CSI framework, the reporting periodicity of periodic or semi-persistent CSI reporting is typically configured by RRC. However, considering the generalization capability of AI models, the accuracy of CSI prediction may vary depending on channel conditions. Therefore, in this method, for performance monitoring result reporting, the reporting periodicity and / or the length of the monitoring window are dynamically configurable via DCI or MAC CE. For periodic performance monitoring result reporting, the length of the monitoring window and / or the reporting period are configured by the RRC. These configurations can be specified in CSI-ReportConfig, CodebookParameters, or CodebookConfig. Note: The above configuration method also applies to semi-persistent or aperiodic performance monitoring result reporting.
[0212] For semi-persistent performance monitoring result reporting:
[0213] When using PUCCH: The reporting periodicity and / or the length of the monitoring window are indicated via MAC CE. This can be achieved by adding specific fields to the legacy MAC CE used for triggering semi-persistent CSI reporting.
[0214] When using PUSCH: The reporting periodicity and / or the length of the monitoring window are indicated via DCI. This can be achieved by adding specific fields to the legacy DCI used for triggering semi-persistent CSI reporting.
[0215] For example, the RRC configures M candidate reporting periods and / or K candidate monitoring window lengths, which can be specified in CSI-ReportConfig, CodebookParameters, or CodebookConfig. When triggering semi-persistent performance monitoring result reporting, at least one additional field is added to the legacy MAC CE or DCI to indicate the selected reporting period and / or monitoring window length.
[0216] For instance, bits and / or bits can be used to indicate the selected reporting period and / or monitoring window length. Additionally, the reporting period and monitoring window length can be jointly indicated using a single field. For example, the UE uses bits to indicate an index corresponding to a predefined set of configurations {mx, kx} , where each configuration specifies the values of the reporting period and monitoring window length, as shown in the following Table 4. Note that it is possible for mx to equal my and kx to equal ky.
[0217] Table 4: Reporting period and monitoring window length combination.
[0218] For aperiodic performance monitoring result reporting, at least one monitoring window length is configured by the RRC, and this configuration can be specified in CSI-ReportConfig, CodebookParameters, or CodebookConfig. If the RRC configures M candidate values {k1, k2, ..., kM} for monitoring window lengths, the UE can indicate one of these values using bits in the DCI.
[0219] Note: In this scheme, the monitoring window length can be expressed as the number of CSI measurement occasions or CSI prediction occasions within the monitoring window. The end of the monitoring window is defined as either the CSI reference resource nCSI_ref, or the last CSI measurement occasion before nCSI_ref. For example, for performance monitoring with periodic and / or semi-persistent CSI-RS resources, the NW can configure the number of CSI measurement occasions or CSI prediction occasions to express the monitoring window length. Additionally, the interval between adjacent CSI measurement occasions or CSI prediction occasions can be a single period, a multiple of the periodic interval, or multiple periodic intervals.
[0220] Note that in this scheme, M∈ [1, 1024] , K∈ [1, 1024] , and N∈ [1, 24] . Moreover, mx∈ [1, 1024] and kx∈ [1, 1024] , and usually kx≤mx although this is not mandatory. These parameter values can be predefined, and the unit of the reporting period can be in slots, milliseconds (ms) , seconds (s) , minutes (min) , hours, or days. Additionally, the specific values of the reporting period and / or the monitoring window length can also be configured by the RRC, for example, in CSI-ReportConfig, CodebookParameters, or CodebookConfig.
[0221] Note: Performance monitoring results can be reported based on either a single CSI prediction occasion or multiple CSI prediction occasions. Subject to UE capability, a UE configured with a CSI-ReportConfig that includes the higher-layer parameter window-length and has reportQuantity set to performance-metric is assumed to support performance monitoring result reporting. For clarity and convenience, WL is used to represent the parameter window-length. The reported performance monitoring results are based on WL consecutive slot intervals, each with a duration of d slots, where the value of WL∈ [1, 1024] is configured by the higher-layer parameter window-length.
[0222] For aperiodic performance monitoring result reporting, a UE configured with a CSI-ReportConfig that includes the higher-layer parameter window-length and has reportQuantity set to performance-metric is expected to be configured with K∈ [1, 20] aperiodic CSI-RS resources in the resource set for performance monitoring. For an aperiodic CSI-RS resource set used for performance monitoring, the K CSI-RS resources are triggered by the same triggering instance, and the separation between two consecutive CSI-RS resources is d∈ [1, 20] slots, which is configured by a higher-layer parameter in the NZP-CSI-RS-ResourceSet. The K aperiodic CSI-RS resources are transmitted in the order of the CSI-RS resource IDs configured in the CSI-RS resource set. The UE shall assume that the antenna ports with the same port index across the K aperiodic CSI-RS resources are considered the same. If interference measurement is performed on CSI-IM, only one resource is configured in the corresponding csi-IM-ResourceSet.
[0223] If the UE is configured with a periodic or semi-persistent CSI-RS resource set for performance monitoring, the value of d is equal to the periodicity of the CSI-RS resource. The end of the WL slot intervals can be either the first slot before nCSI_ref, or the first CSI measurement occasion before nCSI_ref, where nCSI_ref represents the location of the CSI reference resource for performance monitoring result reporting. Note that kx can be equal to WL.
[0224] This embodiment discusses performance monitoring result reporting, which can be periodic, semi-persistent, aperiodic, and / or event-driven. Due to the computational complexity of User Equipment (UE) , merely focusing on the reporting mechanism is insufficient, as the UE must continuously calculate performance metrics. At least one of the following methods are proposed to optimize the triggering and reporting mechanisms: Method 1: Network-indicated triggering via DCI or MAC CE: The network signals the UE to activate performance monitoring using DCI or MAC CE, which includes trigger indications, monitoring duration, and group-based monitoring. The method ensures a fixed monitoring period and enables event-driven reporting when performance monitoring results exceed a threshold. Method 2: Dynamic configuration of reporting periodicity and monitoring window length: The reporting periodicity and monitoring window length can be dynamically adjusted via DCI or MAC CE, adapting to channel conditions. Applies to periodic, semi-persistent, and aperiodic reporting scenarios. Allows flexible resource allocation by enabling the UE to skip or delay reporting based on available CPU and memory resources. The scheme also supports configuring monitoring windows, CSI measurement occasions, and adaptation for aperiodic CSI-RS resource sets, ensuring efficient performance monitoring while minimizing computational and reporting overhead. Enhances efficiency by optimizing resource usage and reducing unnecessary computations, ensuring accurate AI-based CSI prediction while maintaining system stability and reporting flexibility.
[0225] Embodiment 2: Performance monitoring result reporting priority.
[0226] In some examples, the priority of the CSI reporting at the different LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting. In some examples, the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting. In some examples, the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the CSI reporting. In some examples, the coefficient factor is equal to a maximum value of the first parameter plus 1.
[0227] In some examples, values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following: for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ; for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR; for CSI report reporting carrying the L1-RSRP or the L1-SINR; for CSI reporting not carrying the L1-RSRP or the L1-SINR; for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR; for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR; for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR. In some examples, the processing time is managed by at least one of the following: adding an activation time to a first reference time and a second reference time for CSI reporting; stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value.
[0228] To further support performance monitoring requirements for AI-based CSI prediction, beam management, CSI compression, and prediction, this embodiment extends the discussion on how to prioritize different types of CSI reporting. The content above has outlined a preliminary structure for prioritizing CSI reporting using the first parameter, second parameter, or a coefficient factor, including conditions that differentiate AI-based and non-AI-based CSI reports. To ensure efficient allocation of system resources, avoid redundant reporting, and prioritize high-value performance information, the following section provides more concrete examples of how to define and calculate reporting priorities.
[0229] These detailed examples include priority definitions based on whether L1-RSRP or L1-SINR is carried, whether performance monitoring results are included, and whether the report is event-triggered. The methods proposed can be applied independently or in combination, depending on network configuration, UE capabilities, or specific deployment needs, thereby enabling broader and more flexible protection scope. Additionally, the computation formulas and associated parameter settings in each method are highly scalable and can effectively support future standard evolution and diverse implementation strategies. The following describes specific examples and algorithmic illustrations of these mechanisms.
[0230] For AI-based CSI prediction, beam management, CSI compression, and / or CSI prediction and compression, performance monitoring is required for each use case to ensure the reliability of the AI models. Each use case may also require the UE to report performance monitoring results. Considering that performance monitoring results can be regarded as a new type of reporting quantity, the reporting priority among different reporting quantities requires further study. In the present disclosure, some examples provide at least one of the following schemes, and the concrete implementation may apply at least one of the following:
[0231] For two overlapping PUSCHs, the priority rules in this clause apply to physical channels with the same priority index, as specified in Clause 9 of [6, TS 38.213] , if either: the UE is not configured with enableSTx2PofmDCI, or the UE is configured with the higher-layer parameter PDCCH-Config that contains two different values of coresetPoolIndex in ControlResourceSet, and the UE is configured with enableSTx2PofmDCI, and the two overlapping PUSCHs are associated with the same value of coresetPoolIndex.
[0232] Method 1: The reporting priority of performance monitoring results is higher than that of CSI reporting. CSI reporting does not distinguish between AI-based CSI reporting and non-AI CSI reporting. CSI reports are associated with a priority value, which can be determined as follows:
[0233] PriCSI (x, y, k, c, s) =l·Ncells·Ms·x+2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. y=0 for aperiodic CSI reports to be carried on PUSCH. The formula applies to aperiodic CSI reports carried on PUSCH, semi-persistent CSI reports carried on PUSCH or PUCCH, and periodic CSI reports carried on PUCCH. y=1 is for semi-persistent CSI reports to be carried on PUSCH, y=2 is for semi-persistent CSI reports to be carried on PUCCH, and y=3 is for periodic CSI reports to be carried on PUCCH; c is the serving cell index, and Ncells is the value of the higher layer parameter maxNrofServingCells; s is the reportConfigID, and Msis the value of the higher layer parameter maxNrofCSI-ReportConfigurations.
[0234] l is a constant value, and l≥8. For PriCSI (x, y, k, c, s) as mentioned above, the value of k and x can express at least one of the following cases:
[0235] Case 1: The values of k and x represent at least one of the following ways:
[0236] k=0 and x = 0 for CSI reports carrying L1-RSRP or L1-SINR; k=1 and x = 0 for CSI reports not carrying L1-RSRP or L1-SINR; k=0 and x=1 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x=1 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR.
[0237] Case 2: The values of k and x represent at least one of the following ways: k=0 and x = 0 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x = 0 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR; k=0 and x=1 for CSI reports carrying L1-RSRP or L1-SINR; k=1 and x=1 for CSI reports not carrying L1-RSRP or L1-SINR.
[0238] Note that, for Case 1 and Case 2, the performance monitoring results reporting can either be based on event-driven reporting or not based on event-driven reporting. Moreover, in order to reduce the reporting overhead and latency, the performance monitoring results reporting and / or the CSI reports can be based on the event-driven.
[0239] Case 3: The values of k and x represent at least one of the following ways:
[0240] k=0 and x = 0 for Event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x = 0 for Event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR; k=0 and x=1 for CSI reports carrying L1-RSRP or L1-SINR; k=1 and x=1 for CSI reports not carrying L1-RSRP or L1-SINR; k=0 and x = 2 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x = 2 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR.
[0241] Case 4: The values of k and x represent at least one of the following ways: k=0 and x = 0 for CSI reports carrying L1-RSRP or L1-SINR; k=1 and x = 0 for CSI reports not carrying L1-RSRP or L1-SINR; k=0 and x=1 for event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x=1 for event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR; k=0 and x = 2 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x = 2 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR.
[0242] Case 5: The values of k and x represent at least one of the following ways: k=0 and x = 0 for event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x = 0 for event-triggered CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR; k=0 and x=1 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports carrying on L1-RSRP or L1-SINR; k=1 and x=1 for CSI reports carrying the performance monitoring results, where the performance monitoring results are for CSI reports not carrying on L1-RSRP or L1-SINR; k=0 and x = 2 for CSI reports not carrying L1-RSRP or L1-SINR; k=1 and x = 2 for CSI reports carrying L1-RSRP or L1-SINR.
[0243] Method 2: Distinguish the priority value for different CSI reporting quantities, such as: CSI reports carrying L1-RSRP or L1-SINR, CSI reports not carrying L1-RSRP or L1-SINR, CSI reports carrying performance monitoring results, where the results are associated with CSI reports carrying L1-RSRP or L1-SINR, CSI reports carrying performance monitoring results, where the results are associated with CSI reports not carrying L1-RSRP or L1-SINR, and so on. CSI reporting does not differentiate between AI-based and non-AI-based CSI reporting.
[0244] CSI reports are associated with a priority value, which can be determined as follows: PriCSI (y, k, c, s) = 2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. y=0 for aperiodic CSI reports to be carried on PUSCH. y=1 for semi-persistent CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUCCH, and y=3 for periodic CSI reports to be carried on PUCCH; c is the serving cell index and Ncells is the value of the higher layer parameter maxNrofServingCells; s is the reportConfigID and Ms is the value of the higher layer parameter maxNrofCSI-ReportConfigurations. In some examples, the smaller the calculated parameter value, the higher the priority.
[0245] Moreover, based on PriCSI (y, k, c, s) in this method, the value of k can be expressed by at least one of k∈ {0, 1, 2, ..., 10} , and each k value can express at least one of the following cases:
[0246] Case 1: CSI reports carrying L1-RSRP or L1-SINR.
[0247] Case 2: CSI reports not carrying L1-RSRP or L1-SINR.
[0248] Case 3: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR.
[0249] Case 4: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0250] Case 5: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports.
[0251] Case 6: Event-triggered CSI reports carrying L1-RSRP or L1-SINR.
[0252] Case 7: Event-triggered CSI reports not carrying L1-RSRP or L1-SINR.
[0253] Case 8: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports.
[0254] Case 9: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0255] Case 10: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR.
[0256] For example, based on the k values mentioned above, at least one of the following cases are provided:
[0257] Case 1: k=0 for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports not carrying L1-RSRP or L1-SINR, k=2 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR, and k=3 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0258] Case 2: k=0 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR, k=2 for CSI reports carrying L1-RSRP or L1-SINR, and k=3 for CSI reports not-carrying L1-RSRP or L1-SINR.
[0259] Case 3: k=0 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports, k=1 for CSI reports carrying L1-RSRP or L1-SINR, and k=2 for CSI reports not-carrying L1-RSRP or L1-SINR.
[0260] Case 4: k=0 for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports not-carrying L1-RSRP or L1-SINR, and k=2 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports.
[0261] Moreover, the priority value for CSI reporting also can be determined as follows: PriCSI (x, y, k, c, s) =l·Ncells·Ms·x+ 2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. x=0 for CSI reports not carrying performance monitoring results, x=1 for CSI reports carrying performance monitoring results; l is a constant value, and l=6, where kmax is the maximum value of k. For PriCSI (x, y, k, c, s) , the other parameters of value are the same as mentioned above. Note that, for event-triggered performance monitoring results reporting, the reporting priority value can be determined as PriCSI (y, k, c, s) when k=2 or k=3, and y = 0.
[0262] Method 3: Distinguish the priority value for different CSI reporting quantities, such as CSI reports carrying L1-RSRP or L1-SINR, CSI reports not carrying L1-RSRP or L1-SINR, CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR, CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR, and so on. The CSI reporting does not differentiate between AI-based CSI reporting and non-AI based CSI reporting.
[0263] CSI reports are associated with a priority value, which can be determined as follows: PriCSI (y, k, c, s) = l·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. y=0 for aperiodic CSI reports to be carried on PUSCH, y=1 for semi-persistent CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUCCH, and y=3 for periodic CSI reports to be carried on PUCCH; c is the serving cell index and Ncellsis the value of the higher layer parameter maxNrofServingCells; s is the reportConfigID and Msis the value of the higher layer parameter maxNrofCSI-ReportConfigurations.
[0264] Moreover, based on PriCSI (y, k, c, s) in this method, the value of k can be expressed by at least one of k∈ {0, 1, 2, ..., 10} , and each k value can express at least one of the following cases:
[0265] Case 1: CSI reports carrying L1-RSRP or L1-SINR.
[0266] Case 2: CSI reports not carrying L1-RSRP or L1-SINR.
[0267] Case 3: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR.
[0268] Case 4: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0269] Case 5: CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports.
[0270] Case 6: Event-triggered CSI reports carrying L1-RSRP or L1-SINR.
[0271] Case 7: Event-triggered CSI reports not carrying L1-RSRP or L1-SINR.
[0272] Case 8: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports.
[0273] Case 9: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0274] Case 10: Event-triggered CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR.
[0275] Additionally, the value of l=kmax+1 and kmax indicates the maximum value of k. For example, based on the k values mentioned above, at least one of the following cases is outlined:
[0276] Case 1: k=0 for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports not carrying L1-RSRP or L1-SINR, k=2 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR, and k=3 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR; In this case, l= 4.
[0277] Case 2: k=0 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR, k=2 for CSI reports carrying L1-RSRP or L1-SINR, and k=3 for CSI reports not-carrying L1-RSRP or L1-SINR; In this case, l= 4.
[0278] Case 3: k=0 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports, k=1 for CSI reports carrying L1-RSRP or L1-SINR, and k=2 for CSI reports not-carrying L1-RSRP or L1-SINR; In this case, l= 3.
[0279] Case 4: k=0 for CSI reports carrying L1-RSRP or L1-SINR, k=1 for CSI reports not-carrying L1-RSRP or L1-SINR, and k=2 for CSI reports carrying performance monitoring results, where the performance monitoring results are for CSI reports; In this case, l= 3.
[0280] Method 4: Distinguish the priority value for event-driven CSI reports. The CSI reporting does not differentiate between AI-based CSI reporting and non-AI-based CSI reporting. In this method, different y values are defined for event-driven report quantities. Then, the value of y can be expressed by at least one of y∈ {0, 1, 2, ..., 8} , and each y value can express at least one of the following cases:
[0281] Case 1: Aperiodic CSI reports are to be carried on PUSCH.
[0282] Case 2: Event-driven CSI reports are to be carried on PUSCH, and the PUSCH resource is scheduled by DCI or active by DCI, e.g., Type 2 PUSCH resource.
[0283] Case 3: Event-driven CSI reports are to be carried on PUCCH.
[0284] Case 4: Semi-persistent CSI reports are to be carried on PUSCH.
[0285] Case 5: Semi-persistent CSI reports are to be carried on PUCCH.
[0286] Case 6: Periodic CSI reports are to be carried on PUCCH.
[0287] Case 7: Aperiodic CSI reports are to be carried on PUSCH, Event-driven CSI reports are to be carried on PUSCH, or Event-driven CSI reports are to be carried on PUCCH.
[0288] Case 8: Aperiodic CSI reports are to be carried on PUSCH or Event-driven CSI reports are to be carried on PUSCH.
[0289] Case 9: Aperiodic CSI reports are to be carried on PUSCH or Event-driven CSI reports are to be carried on PUCCH.
[0290] Case 10: Event-driven aperiodic CSI reports are to be carried on PUSCH.
[0291] Case 11: Event-driven aperiodic CSI reports are to be carried on PUCCH.
[0292] Case 12: Event-driven semi-persistent CSI reports are to be carried on PUSCH.
[0293] Case 13: Event-driven semi-persistent CSI reports are to be carried on PUCCH.
[0294] Case 14: Event-driven semi-persistent CSI reports are to be carried on PUSCH, Semi-persistent CSI reports to be carried on PUSCH.
[0295] Case 15: Event-driven Semi-persistent CSI reports are to be carried on PUCCH, or Semi-persistent CSI reports are to be carried on PUCCH.
[0296] Case 16: Event-driven CSI reports are to be carried on PUSCH, and the PUSCH resource is configured by RRC or active by DCI, e.g., Type 2 PUSCH resource.
[0297] Case 17: Event-driven CSI reports are to be carried on PUSCH.
[0298] Case 18: Event-driven CSI reports are to be carried on PUCCH, and the PUCCH resource is configured by RRC and / or indicated / triggered by DCI / MAC CE.
[0299] Case 19: Event-driven CSI reports are to be carried on PUCCH, and the PUCCH resource is configured by RRC, selected by UE, and / or indicated / triggered by DCI / MAC CE.
[0300] For example, based on the k values mentioned above, at least one of the following cases are outlined:
[0301] Case 1: y=0 for aperiodic CSI reports to be carried on PUSCH or event-driven CSI reports to be carried on PUSCH / PUCCH, y=1 for semi-persistent CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUCCH, and y=3 for periodic CSI reports to be carried on PUCCH.
[0302] Case 2: y=0 for event-driven CSI reports to be carried on PUSCH / PUCCH, y=1 for aperiodic CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUSCH, y=3 for semi-persistent CSI reports to be carried on PUCCH, and y=4 for periodic CSI reports to be carried on PUCCH.
[0303] Case 3: y=0 for event-driven CSI reports to be carried on PUSCH / PUCCH, y=1 for aperiodic CSI reports to be carried on PUSCH, y=2 for semi-persistent CSI reports to be carried on PUSCH, y=3 for semi-persistent CSI reports to be carried on PUCCH, and y=4 for periodic CSI reports to be carried on PUCCH.
[0304] Note that, in this method, the event-driven CSI reports can: (1) carry L1-RSRP or L1-SINR; (2) not carry L1-RSRP or L1-SINR; (3) carry performance monitoring results; (4) carry performance monitoring results, where the performance monitoring results are for CSI reports carrying L1-RSRP or L1-SINR; and / or (5) carry performance monitoring results, where the performance monitoring results are for CSI reports not carrying L1-RSRP or L1-SINR.
[0305] Additionally, CSI reports are associated with a priority value, and the computation of priority value PriCSI can express at least one of the follows:
[0306] Item 1: Based on Method 1, PriCSI can be expressed as: PriCSI (x, y, k, c, s) =l·Ncells·Ms·x+2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. l is a constant value, and l≥2ymax+kmax+1, where ymax indicates maximum value of y, and kmax is the maximum value of k. Note that the definition of other parameters is the same as in Method 1, and l·Ncells·Ms·x is not necessary; it depends on whether the performance monitoring results are the candidate CSI report quantity.
[0307] Item 2: Based on Method 2, PriCSI can be expressed as: PriCSI (x, y, k, c, s) =l·Ncells·Ms·x+ 2·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. l is a constant value, and l≥2ymax, where ymax indicates maximum value of y. Note that the definition of other parameters is the same as in Method 2, and l·Ncells· Ms·x is not necessary; it depends on whether the performance monitoring results are the candidate CSI report quantity.
[0308] Item 3: Based on Method 3, PriCSI can be expressed as: PriCSI (y, k, c, s) = l·Ncells·Ms·y+Ncells·Ms·k+Ms·c+s. Note that the definition of other parameters is the same as in Method 3.
[0309] Method 5: If the UE reports the ground truth for performance monitoring, then the CSI report for performance monitoring has either a lower or higher priority compared to all CSI report (s) that are not intended for performance monitoring. This applies regardless of the value of PriCSI in the event of a collision with CSI report (s) that are not for performance monitoring. Note that the ground truth may include L1-RSRP or L1-SINR, or it may not include L1-RSRP or L1-SINR.
[0310] Moreover, if the CSI report (s) for performance monitoring and the CSI report (s) that are not for performance monitoring have the same PriCSI value, then the CSI report (s) that are not for performance monitoring are assigned either a lower or higher priority. Note that the formula for calculating thePriCSI value under this method can be any of the previously described methods in this embodiment, or it can reuse the legacy formulaPriCSI (y, k, c, s) based on 3GPP 38.214 Clause 5.2.5.
[0311] Note that the aforementioned methods are based on the assumption that CSI reporting does not differentiate between AI-based and non-AI-based CSI reporting. However, based on the priority calculation formulas mentioned above, if two CSI reporting instances on PUSCH with the same priority value overlap in the time domain, and one is AI-based while the other is non-AI-based, then the AI-based CSI reporting has a higher priority.
[0312] Note that the performance monitoring output can be obtained through a comparison between the ground truth and the predicted CSI, or through the comparison between a performance metric and a certain threshold. This does not exclude the possibility that the performance monitoring result is the ground truth CSI. For example, in AI-based CSI prediction, the UE can report the ground truth CSI to the network. The network can then calculate the performance metric based on the ground truth and predicted CSI, and make a decision regarding a functionality fallback operation. In the above methods, the CSI reports carry performance monitoring results, and those results may be the ground truth CSI.
[0313] Moreover, in the above method, event-triggered CSI reports carrying L1-RSRP or L1-SINR can be obtained by UE initial beam management. Event-triggered CSI reports not carrying L1-RSRP or L1-SINR can be obtained by UE initial CSI reporting. If the event-triggered CSI report (s) and legacy CSI report (s) have the same PriCSI value, then each of the event-triggered CSI report (s) have a lower or higher priority.
[0314] This embodiment focuses on prioritizing performance monitoring results reporting for AI-based CSI prediction, beam management, CSI compression, and other related functions. The key challenge addressed is the relative priority of different reporting quantities, including CSI reports carrying L1-RSRP / L1-SINR, performance monitoring results, and event-driven reports. To tackle this, at least one of the following methods are proposed: Method 1: assigning higher priority to performance monitoring results: Establishes a formula for prioritizing different CSI reports. Defines cases for CSI reports with and without performance monitoring results. Method 2: differentiating priority for various CSI reporting quantities: Distinguishes reports based on whether they contain L1-RSRP, L1-SINR, or performance monitoring results. Assigns priority values using a structured formula. Method 3: categorizing event-driven CSI reports separately: Defines a range of scenarios for event-driven CSI reports and their impact on prioritization. Method 4: Introducing event-specific priority adjustments: Establishes a hierarchy of priority values for different event-triggered reports. Overall, these methods ensure efficient resource allocation, minimize reporting overhead, and prioritize crucial network measurements to enhance AI-based CSI prediction accuracy. Improves system performance by prioritizing critical CSI reports, ensuring reliable AI model performance, and reducing unnecessary reporting overhead. Method 5 outlines priority rules for CSI reporting in the context of performance monitoring, regardless of whether the reports are AI-based or non-AI-based. CSI reports used for performance monitoring may be assigned higher or lower priority than those not used for monitoring, depending on specific conditions, including collisions and identical priority values. The performance monitoring output can be derived from comparisons between ground truth and predicted CSI, or between performance metrics and thresholds, with the ground truth CSI itself potentially serving as the performance monitoring result. In AI-based scenarios, the UE may report ground truth CSI, enabling the network to assess prediction accuracy and decide on fallback operations accordingly.
[0315] Embodiment 3: Performance monitoring results reporting (type 3) :
[0316] In some examples, the UE calculates a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and selects M performance monitoring results from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1. In some examples, M is configured by the base station or predefined. In some examples, the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting. In some examples, when the UE uses differential reporting in the performance monitoring results, the UE uses a priority to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value. In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following: periodic reporting; semi-persistent reporting; aperiodic reporting; event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold. In some examples, reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises: reporting, to the base station, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following: a bitmap representation; a combined digital representation; a reported CSI index.
[0317] The following paragraph connects the content above with the subsequent section, indicating that the latter provides more concrete implementation examples: The preceding content outlines the foundational principles of Performance Monitoring Results Reporting (Type 3) in AI-based CSI prediction. Specifically, the UE calculates SGCS or NMSE across sub-bands and / or layers and selects M monitoring results with the smallest values within a monitoring window to be reported as the first part of the AI-based CSI report. It explains how the performance monitoring results are mapped and reported based on predefined reporting priorities, time-domain ordering, and differential reporting methods. Reporting can be periodic, semi-persistent, aperiodic, or event-triggered. The following section provides more detailed implementation examples, describing how the UE uses its local AI model for CSI prediction and calculates NMSE and SGCS as performance metrics. Various cases are introduced for computing these metrics based on time occasions, layers, and sub-bands. Additionally, multiple reporting methods are described, including absolute and differential quantization schemes.
[0318] These examples detail how the reporting quantity, referred to as "performance-metric" , is structured using mapping tables, showing how SGCS values and their location indices are encoded. The section also elaborates on the configuration and reuse of CSI-RS resources for performance monitoring, whether periodic, semi-persistent, or aperiodic, and how these resources may overlap with or differ from those used for model inference. Based on the reported performance, the network (NW) can instruct the UE to fall back to the legacy Rel-18 Doppler codebook using either implicit or explicit signaling. It is important to note that the operations described in this embodiment, such as performance metric calculation, quantization, mapping, reporting, and fallback mechanisms, can be executed independently or in combination. This modular and flexible design broadens the protection scope and enhances adaptability across different deployment scenarios and UE capabilities, enabling efficient resource utilization and dynamic performance optimization.
[0319] For AI-based CSI prediction using a UE-side model, the performance metrics for performance monitoring can include NMSE and / or SGCS, which are derived from the predicted CSI and its corresponding ground-truth CSI. Based on ongoing discussions in the current standard, the UE reports the performance metrics for performance monitoring to the network. The network then determines whether to fall back to the legacy Rel-18 Doppler codebook scheme.
[0320] Based on the aforementioned discussion, the NMSE or SGCS can be calculated using at least one of the following methods:
[0321] Case 1: Across all sub-bands and all layers for a single predicted CSI time occasion.
[0322] Case 2: Across all sub-bands for each layer in a single predicted CSI time occasion.
[0323] Case 3: Across all sub-bands, all layers, and all predicted CSI time occasions within the monitoring window.
[0324] Case 4: Across all sub-bands and all predicted CSI time occasions within the monitoring window for each layer. Note that the sub-band can be one RB or multiple RBs.
[0325] This embodiment specifies the reporting contents of the UE for performance monitoring based on at least one of the different methods used to calculate NMSE or SGCS.
[0326] Case 1: the SGCS calculates across all sub-bands and all layers for a single predicted CSI time occasion. In this case, the specific implementation can apply at least one of the following approaches:
[0327] Method 1: The UE reports SGCS or SGCS differential quantization values for each predicted CSI associated with the ground truth within the monitoring window. Each SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . The value of X depends on the elements in the quantization table. If the quantization table includes Y elements, then each SGCS can be expressed using bits. For example, if Y = 8, then X= 3. Note that the difference between adjacent values in the quantization table can be either equal or unequal.
[0328] Moreover, the reporting of the SGCS quantization value can be based on a differential quantization method. For example, the reported quantization value for each SGCS can be the difference between the SGCS value and the maximum or minimum value in the quantization table. Additionally, the reported SGCS quantization value can also be the difference between the current SGCS value and the maximum or minimum SGCS value within the monitoring window. If this differential quantization method is used, the UE is required to report the maximum or minimum SGCS quantization value along with the location of the SGCS value associated with the predicted CSI. Note that the number of bits used for the maximum or minimum SGCS quantization value may differ from those used for the differential SGCS quantization value. The location information of the maximum or minimum SGCS value can be reported using a bitmap, a combined digital representation, or a predicted CSI index. Additionally, the quantization value selection can be either up quantization or down quantization. The maximum or minimum SGCS value is the result of quantization based on a quantization table, which falls within the range [0, 1] . The quantization table can be either a uniform or non-uniform quantization table, such as Table 5.2.2.2.3-2 in 3GPP 38.214.
[0329] For this method, the reporting quantity can be defined as a "performance-metric, " which is regarded as a type of CSI, and the CSI feedback consists of a single part. Note that the term "performance-metric" is just a placeholder and can be replaced with another name. The mapping rules for the reporting quantity "performance-metric" can be expressed as follows:
[0330] Assume that the maximum index for the contents in the reporting quantity ‘performance-metric’ is N, and each index value corresponds to at least one of the following items: The SGCS quantization value or SGCS differential quantization value of the k-th predicted CSI and k≤N; The maximum SGCS quantization value; The location index of the maximum or minimum SGCS value.
[0331] For example, the reporting quantity "performance-metric" is mapped according to the time order of the predicted CSI, as shown in Table 5.
[0332] Table 5: Reporting quantity:
[0333] Moreover, the reporting quantity "performance-metric" is mapped according to the time order of the predicted CSI, excluding the maximum or minimum SGCS value, as shown in Table 6.
[0334] Table 6: Reporting quantity.
[0335] Method 2: The UE reports M minimum SGCS values or minimum SGCS differential quantization values for each predicted CSI associated with the ground truth within the monitoring window. Each SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . Moreover, the value of M can be predefined, configured via RRC, MAC CE, or DCI, or selected by the UE. In addition, M may be indicated by MAC CE or DCI from a dataset that is either predefined or configured by RRC. Similar to Method 1, if M SGCS values are quantized based on a quantization table, the UE may also be required to report the location of the M SGCS values corresponding to predicted CSI within the monitoring window. Note: location information reporting is not mandatory. If the SGCS quantization value is reported based on a differential quantization method-for example, where the reported quantization value for each SGCS is the difference between the SGCS value and the maximum or minimum value in the quantization table-the UE may also need to report the location information for the M SGCS values in the monitoring window. Again, this location information is optional. If the reported SGCS quantization value is the difference between the current SGCS value and the maximum or minimum SGCS value among the M values, the UE may also be required to report the location information of the M SGCS values corresponding to predicted CSI in the monitoring window, and / or the location of the maximum or minimum SGCS value, which can be placed at the beginning of the UCI or MAC CE. If location information for all M SGCS values is not reported, the UE may still be required to report the location of the predicted CSI associated with the maximum or minimum SGCS value in the monitoring window. For example, this location can be indicated using a bitmap, a combined digital representation, or a predicted CSI index. Similarly, the location information of predicted CSI associated with the M SGCS values can also be reported using a bitmap, combined digital representation, or predicted CSI index. Note: The number of bits used for the maximum or minimum SGCS quantization value can differ from the number of bits used for the differential SGCS quantization values. Note that this method applies to scenarios where the monitoring window includes a large number of predicted CSI instances, such as when the number of predicted CSIs in the monitoring window exceeds 4 or 8. If the number of predicted CSIs is less than or equal to 4 or 8, all SGCS values are required to be reported, and the M value does not need to be configured.
[0336] Similar to Method 1, the reporting quantity can be defined as ‘performance-metric’ , which consists of a single part, e.g., CSI Part 1. The mapping rules for the reporting quantity ‘performance-metric’ can be described as follows: Assume that the maximum index for the contents in the reporting quantity ‘performance-metric’ is N, and each index value corresponds to at least one of the following items: The location index of M SGCS values within the monitoring window. The M SGCS quantization values or SGCS differential quantization values within the monitoring window. The location indication of the maximum or minimum SGCS value within the monitoring window. The location indication of the M–1 SGCS values within the monitoring window, excluding the maximum or minimum SGCS value. The maximum or minimum SGCS quantization value among the M SGCS values. The M–1 SGCS differential quantization values within the monitoring window, excluding the maximum or minimum SGCS value.
[0337] For example, the reporting quantity ‘performance-metric’ is mapped according to the time order of the predicted CSI, as shown in Table 7.
[0338] Table 7: Reporting quantity.
[0339] Moreover, the reporting quantity ‘performance-metric’ is mapped according to the time order of the predicted CSI, excluding the maximum or minimum SGCS value, as shown in Table 8.
[0340] Table 8: Reporting quantity.
[0341] Case 2: the SGCS calculates across all sub-bands for each layer in a single predicted CSI time occasion.
[0342] In this case, the specific implementation can apply at least one of the following approaches:
[0343] Method 1: The UE reports SGCS or SGCS differential quantization values for each layer of each predicted CSI within the monitoring window. Each SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . The value of X depends on the elements in the quantization table. If the quantization table includes Y elements, then each SGCS can be expressed using bits. For example, if Y = 8, then X= 3. Note that the difference between adjacent values in the quantization table can be either equal or unequal.
[0344] Moreover, the reporting of SGCS quantization values can be based on a differential quantization method. For example, the reported quantization value for each SGCS can be the difference between the SGCS value and the maximum or minimum value in the quantization table. Additionally, the reported SGCS quantization value can also be the difference between the current SGCS value and the maximum or minimum SGCS value within the monitoring window. When using this differential quantization method, the UE is required to report the maximum or minimum SGCS quantization value and the location information of the SGCS value associated with the predicted CSI. Note that the number of bits used for the maximum or minimum SGCS quantization value can differ from those used for the differential SGCS quantization values. The location of the maximum or minimum SGCS value can be reported using a bitmap, a combined digital representation, or the predicted CSI index. It is also noted that the quantization value selection can be either upward (Up) quantization or downward (Down) quantization.
[0345] For this method, the reporting quantity can be defined as a "performance-metric, " which is considered a type of CSI, and the CSI feedback consists of a single part. Note that the term "performance-metric" is just a placeholder and can be replaced with another name. The mapping rules for the reporting quantity "performance-metric" can be described as follows: Firstly, based on the aforementioned methods, the reporting content may include at least one of the following items: The SGCS quantization values or SGCS differential quantization values of the predicted CSI, mapped in time order. For each predicted CSI, the SGCS quantization values or SGCS differential quantization values for each layer are mapped in ascending order based on the layer number. The SGCS differential quantization values of the predicted CSI, mapped in time order. For each predicted CSI, the SGCS differential quantization values for each layer are mapped in ascending order based on the layer number. Note that the SGCS differential quantization values do not include the maximum or minimum SGCS quantization value. The maximum or minimum SGCS quantization value. The location index of the maximum or minimum SGCS value.
[0346] For example, if the SGCS quantization values or SGCS differential quantization values are not obtained based on the maximum or minimum SGCS value, then the reporting quantity "performance-metric" is mapped as shown in Table 9.
[0347] Table 9: Reporting quantity.
[0348] Moreover, if the SGCS differential quantization values are obtained based on the maximum or minimum SGCS value, then the reporting quantity "performance-metric" is mapped as shown in Table 10.
[0349] Table 10: Reporting quantity.
[0350] Method 2: The UE reports M minimum SGCS or minimum SGCS differential quantization values for each layer of each predicted CSI within the monitoring window. Each SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . Moreover, the value of M may be predefined, configured by RRC, MAC CE, or DCI, or selected by the UE. In addition, the value of M can be indicated via MAC CE or DCI from a dataset that is either predefined or configured by the RRC. Similar to Method 1, if SGCS quantization is based on a quantization table, the UE may also be required to report the location of the SGCS values for the predicted CSI within the monitoring window. Note that reporting the location information is not mandatory. If the SGCS quantization value is reported using a differential quantization method. For example, if each SGCS value is reported as the difference between the actual SGCS value and the maximum or minimum value in the quantization table. The UE may also need to report the location information of the SGCS values for the predicted CSI in the monitoring window. Again, reporting the location information is optional. If the reported SGCS quantization value is the difference between the current SGCS value and the maximum or minimum SGCS value among the M × L SGCS values (where L is the number of layers) , the UE may also be required to report the location information of the SGCS values for the predicted CSI in the monitoring window. In this case, the location information of the maximum or minimum SGCS value can be mapped at the beginning of the UCI or MAC CE. If location information for the SGCS of the predicted CSI is not reported, the UE may instead be required to report the layer index and the location information of the maximum or minimum SGCS value in the monitoring window.
[0351] For example, the location information of the predicted CSI associated with the maximum or minimum SGCS value can be reported using a bitmap, a combined digital representation, or the predicted CSI index. The layer index can be expressed using Y=log2 (Rmax) bits, whereRmax is the maximum Rank value. Note that the number of bits used for the maximum or minimum SGCS quantization value may differ from those used for the differential SGCS quantization value. Similar to Method 1, the reporting quantity can be defined as a ‘performance-metric’ , which consists of a single part (e.g., CSI Part 1) . The mapping rules for the reporting quantity ‘performance-metric’ can be expressed as follows:
[0352] Firstly, based on the aforementioned methods, the reporting content may include at least one of the following items: The location information of the SGCS values within the monitoring window. The SGCS quantization values or SGCS differential quantization values of the predicted CSI, mapped in time order. For each predicted CSI, the SGCS quantization values or SGCS differential quantization values for each layer are mapped in ascending order based on the layer number. The SGCS differential quantization values of the predicted CSI, mapped in time order. For each predicted CSI, the SGCS differential quantization values for each layer are mapped in ascending order based on the layer number. Note: The SGCS differential quantization values do not include the maximum or minimum SGCS quantization value. The maximum or minimum SGCS quantization value. The location information of the maximum or minimum SGCS value. This location information includes two parts: (1) the location of the predicted CSI associated with the maximum (or minimum) SGCS value within the monitoring window, and (2) the corresponding layer index information.
[0353] For example, if the SGCS quantization values or SGCS differential quantization values are not obtained based on the maximum or minimum SGCS value, then the reporting quantity 'performance-metric' is mapped as shown in Table 11.
[0354] Table 11: Reporting quantity.
[0355] Moreover, if the SGCS differential quantization values are obtained based on the maximum or minimum SGCS value, then the reporting quantity 'performance-metric' is mapped as shown in Table 12.
[0356] Table 12: Reporting quantity.
[0357] Note that this method applies to scenarios where the monitoring window includes a large number of predicted CSI instances, such as when the number of predicted CSIs in the monitoring window exceeds 4 or 8. If the number of predicted CSIs is less than or equal to 4 or 8, all SGCS values are required to be reported, and the M value does not need to be configured. Moreover, the above method also applies to reporting all comparison results for the predicted CSIs in the monitoring window, meaning there are M predicted CSIs. If all SGCS values for the predicted CSIs in the monitoring window are reported, the location information for each SGCS is not required.
[0358] Case 3: the SGCS calculates across all sub-bands, all layer, and all predicted CSI time occasions within the monitoring window.
[0359] In this case, the specific implementation can apply at least one of the following approaches:
[0360] Method 1: The UE reports the SGCS quantization value to NW, and SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . The value of X depends on the elements in the quantization table. If the quantization table includes Y elements, then each SGCS can be expressed using bits. For example, if Y = 8, then X= 3. Note that the difference between adjacent values in the quantization table can be either equal or unequal.
[0361] Case 4: the SGCS calculates across all sub-bands and all predicted CSI time occasions within the monitoring window for each layer.
[0362] In this case, the specific implementation can apply at least one of the following approaches:
[0363] Method 1: The UE reports the SGCS quantization value for each layer to NW, and SGCS quantization value occupies X bits, where X can be one of {1, 2, 3, 4, 5, 6, 7} . The value of X depends on the elements in the quantization table. If the quantization table includes Y elements, then each SGCS can be expressed using bits. For example, if Y = 8, then X= 3. Note that the difference between adjacent values in the quantization table can be either equal or unequal.
[0364] Moreover, the reporting of the SGCS quantization value can be based on a differential quantization method. For example, the reported quantization value for each SGCS can be the difference between the SGCS value and the maximum or minimum value in the quantization table. Additionally, the reported SGCS quantization value can be defined as the difference between the maximum or minimum SGCS value and the other SGCS values. When using this differential quantization method, the UE is required to report the maximum or minimum SGCS quantization value along with its corresponding layer index. Note that the number of bits used for representing the maximum or minimum SGCS quantization value may differ from those used for the differential SGCS quantization values. The layer indication information for the maximum or minimum SGCS value can be expressed using a bitmap, a combined digital representation, or a layer index. It should also be noted that the quantization value selection can be either up quantization or down quantization. In this method, the reporting quantity can be defined as a performance metric, which is considered a type of CSI. The CSI feedback consists of a single component. The mapping rules for the reporting quantity referred to as the performance metric can be defined as follows:
[0365] Assume that the maximum index for the contents in the reporting quantity performance metric is N, and each index value corresponds to at least one of the following items: The SGCS quantization value or the differential quantization value of each layer; the layer indication information of the maximum or minimum SGCS value; the maximum or minimum SGCS quantization value; the SGCS quantization values for the other layers, excluding the layer with the maximum or minimum SGCS value.
[0366] For example, the reporting quantity performance metric is mapped according to the layer order of the predicted CSI, as shown in Table 13.
[0367] Table 13: Reporting quantity.
[0368] Moreover, the reporting quantity performance metric is mapped according to the layer order of the predicted CSI, excluding the layer with the maximum or minimum SGCS value, as shown in Table 14.
[0369] Table 14: Reporting quantity.
[0370] Note that in this embodiment, performance monitoring result reporting can be periodic, semi-persistent, or aperiodic. The performance monitoring window can be configured by the network (NW) . For periodic or semi-persistent CSI reporting used for model inference, the CSI-RS resource can also be periodic or semi-persistent and may be reused for performance monitoring. Moreover, the NW can configure specific CSI-RS resources for performance monitoring. For example: For model inference using a periodic CSI-RS resource, the CSI-RS resource for performance monitoring can be periodic, semi-persistent, or aperiodic. For model inference using a semi-persistent CSI-RS resource, the CSI-RS resource for performance monitoring can be semi-persistent or aperiodic. For model inference using an aperiodic CSI-RS resource, the CSI-RS resource for performance monitoring can be aperiodic. Additionally, the CSI-RS resources for model inference and performance monitoring can be configured within the same resource set or in different resource sets. Based on the above reporting content, the NW may determine whether to fall back to the legacy Rel-18 Doppler codebook scheme. The NW can instruct the UE to fall back to the legacy Rel-18 Doppler codebook using either an explicit or implicit method. For example, the NW can deactivate performance monitoring result reporting via DCI or MAC CE. If the UE receives deactivation information, it will consider falling back to the legacy Rel-18 Doppler codebook. Alternatively, a 1-bit indication can be used to explicitly signal whether to fall back to the legacy Rel-18 Doppler codebook.
[0371] Note that the above-mentioned methods also apply to the performance monitoring metric NMSE, since the range of NMSE is [0, +∞) . Therefore, when applying the differential quantization method, the maximum NMSE value is normalized to 1, and it does not need to be reported. Moreover, the differential quantization table can refer to Table 5.2.2.2.3-2 in 3GPP 38.214. Additionally, for the methods mentioned above, the UE may also report timestamp information, which can indicate the time location of the start or end of the predicted CSI in the monitoring window, such as the frame number and slot number.
[0372] This embodiment describes Performance Monitoring Results Reporting (Type 3) for AI-based CSI prediction using UE-side models. It defines how performance metrics-such as NMSE and SGCS-are calculated, quantized, and reported to the network (NW) . Multiple methods and cases are presented for reporting these metrics, depending on the time occasion, layers, and sub-bands involved. The report content can be in absolute or differential form, and quantization is adaptable based on various conditions. The reporting quantity, termed as performance-metric, is structured with mapping tables showing how information like SGCS values and their location indices are ordered. Moreover, CSI-RS resources used for performance monitoring can be periodic, semi-persistent, or aperiodic, and may overlap or differ from those used for model inference. Based on the reported performance, the NW can instruct the UE to fall back to the legacy Rel-18 Doppler codebook, either implicitly or explicitly. This method provides a flexible, efficient, and adaptive mechanism for monitoring AI-based CSI prediction accuracy, enabling optimized feedback and dynamic fallback to legacy schemes based on real-time performance.
[0373] Embodiment 4: Performance monitoring results reporting (type 1) :
[0374] In some examples, the UE calculates a SGCS or NMSE across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and reports, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1. In some examples, the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.
[0375] The above section outlines a simplified threshold-based performance monitoring result reporting mechanism (Type 1) , where the UE calculates SGCS or NMSE and compares the M smallest values against a predefined threshold. Instead of reporting full metric values, the UE only reports whether the metrics exceed the threshold, thereby significantly reducing reporting overhead. This threshold-based method improves efficiency while still providing meaningful performance insight to the network. The following section provides more concrete examples and implementations based on this concept, detailing how different calculation granularities, such as across layers, time occasions, and sub-bands, can be applied in practice.
[0376] It should be noted that in this embodiment, the various operations-such as calculating SGCS / NMSE, comparing to thresholds, and reporting either bitmaps or counts-can be performed independently or in combination depending on configuration. This modular structure offers a highly flexible design, allowing the solution to adapt to diverse deployment scenarios, network preferences, and UE capabilities. Such flexibility extends the protection scope of the present disclosure while enabling optimized trade-offs between reporting accuracy and signaling efficiency.
[0377] For AI-based CSI prediction using UE-side models during performance monitoring, it is necessary to evaluate the similarity between the predicted CSI and the corresponding measured CSI using certain performance metrics, such as NMSE or SGCS. However, directly reporting NMSE or SGCS results may lead to relatively high reporting overhead and may not provide significant benefit for network-side decision-making. Based on conclusions from the Rel-19 discussions, the network may configure corresponding threshold criteria to assist with UE performance monitoring.
[0378] Based on embodiment 3, the NMSE or SGCS can be calculated apply at least one of the following ways:
[0379] Case 1: Across all sub-bands and all layers for a single predicted CSI time occasion.
[0380] Case 2: Across all sub-bands for each layer in a single predicted CSI time occasion.
[0381] Case 3: Across all sub-bands, all layers, and all predicted CSI time occasions within the monitoring window.
[0382] Case 4: Across all sub-bands and all predicted CSI time occasions within the monitoring window for each layer. Note that the sub-band can be one RB or multiple RBs.
[0383] This embodiment specifies the reporting contents of the UE for performance monitoring based on the different methods used to calculate NMSE or SGCS.
[0384] Case 1: the SGCS or NMSE calculates across all sub-bands and all layers for a single predicted CSI time occasion.
[0385] In this case, the specific implementation can apply at least one of the following approaches:
[0386] Method 1: The UE reports M comparison results between the M minimum SGCS or NMSE values and the threshold within the monitoring window, using 1 bit to indicate each comparison result. If the bit is 1 or 0, it means that SGCS or NMSE value is larger than the threshold.
[0387] Method 2: The UE reports the number of SGCS or NMSE values that are lower than the threshold within the M minimum SGCS or NMSE values, or within the monitoring window. The bits overhead depend on the M or the number of predicted CSI within the monitoring window, it can be expressed as or where Y is the number of predicted CSI or measurement CSI within the monitoring window.
[0388] Case 2: the SGCS or NMSE calculates across all sub-bands for each layer in a single predicted CSI time occasion. In this case, the specific implementation can apply at least one of the following approaches:
[0389] Method 1: The UE reports M × L comparison results between the M minimum SGCS or NMSE values and the threshold for each layer within the monitoring window, using 1 bit to indicate each comparison result, where L is the number of layers. A bit value of 1 indicates that the SGCS or NMSE value is greater than the threshold. Similar to Embodiment 3, the reporting quantity can be defined as a performance metric, which consists of a single part, e.g., CSI Part 1. The mapping rules for the reporting quantity performance metric are illustrated in Table 15.
[0390] Table 15: Reporting quantity.
[0391] Method 2: The UE reports the number of SGCS or NMSE values that are lower than the threshold within the M minimum SGCS or NMSE values, or within the monitoring window for each layer. The bit overhead depends on the number of layers (L) , as well as M or the number of predicted CSIs within the monitoring window. It can be expressed as or where Y represents the number of predicted CSIs or measurement CSIs within the monitoring window.
[0392] Similar to Embodiment 3, the reporting quantity can be defined as a performance metric, which consists of a single part, e.g., CSI Part 1. The mapping rules for the reporting quantity performance metric are described as shown in Table 16.
[0393] Table 16: Reporting quantity.
[0394] Case 3: The SGCS or NMSE is calculated across all sub-bands, all layers, and all predicted CSI time occasions within the monitoring window.
[0395] In this case, the specific implementation can apply at least one of the following approaches:
[0396] Method 1: The UE reports the comparison result between the SGCS or NMSE value and a threshold, using 1 bit to indicate the comparison result. A bit value of 1 (or 0) indicates that the SGCS or NMSE value is greater than the threshold.
[0397] Case 4: The SGCS is calculated across all sub-bands and all predicted CSI time occasions within the monitoring window, for each layer.
[0398] In this case, the specific implementation can apply at least one of the following approaches:
[0399] Method 1: The UE reports the comparison result between the SGCS or NMSE value and the threshold for each layer. The total bit overhead is L, where L is the number of layers. A bit value of 1 (or 0) indicates that the SGCS or NMSE value is greater than the threshold.
[0400] Note: For the methods described above, the value of M is configured by RRC, MAC CE, or DCI. The M value may have multiple candidate options and can be either predefined or configured by RRC, with the final selection determined by MAC CE or DCI. The above methods apply to scenarios where the monitoring window includes a large number of predicted CSI instances, such as when the number of predicted CSIs in the monitoring window exceeds 4 or 8. If the number of predicted CSIs is less than or equal to 4 or 8, all SGCS values are required to be reported, and the M value does not need to be configured. Moreover, the above methods also apply to reporting all comparison results for the predicted CSI in the monitoring window, meaning there are M predicted CSI instances. If all comparison results for the predicted CSI in the monitoring window are reported, the location information for each SGCS is not required. Additionally, for the methods mentioned above, the UE may also report timestamp information, which can indicate the time location of the start or end of the predicted CSI in the monitoring window, such as the frame number and slot number. Additionally, performance monitoring may support both SGCS and NMSE as performance metrics. The specific performance metric used by the UE can be configured via RRC, MAC CE, or DCI. Furthermore, both SGCS and NMSE can be configured by RRC, and the UE selects the appropriate one based on indications from MAC CE or DCI. Meanwhile, based on the reported performance metrics, the network (NW) may determine whether to fall back to the legacy Rel-18 Doppler codebook scheme. The NW can instruct the UE to fall back using either an explicit or implicit method. For example, the NW may deactivate performance monitoring result reporting via DCI or MAC CE. Upon receiving such deactivation information, the UE considers falling back to the legacy Rel-18 Doppler codebook. Alternatively, a 1-bit indication can be included to explicitly signal whether to fall back to the legacy Rel-18 Doppler codebook.
[0401] This embodiment introduces Performance Monitoring Results Reporting (Type 1) for AI-based CSI prediction using UE-side models. It proposes threshold-based methods to reduce the reporting overhead compared to directly reporting NMSE or SGCS values. Four calculation cases are defined (based on time, layer, and frequency granularity) , each with reporting methods such as bitmaps or count-based feedback. The UE may report either comparison results or the number of values below the threshold, with bit overhead optimized based on the number of layers or prediction windows. The metric used (SGCS or NMSE) and the reporting granularity are configurable by RRC, MAC CE, or DCI. Based on the UE's reported performance, the network can decide to fall back to the legacy Rel-18 Doppler codebook, using explicit or implicit signaling. This embodiment offers a low-overhead, threshold-based feedback mechanism for AI-based CSI prediction, enabling efficient network-side decisions while minimizing reporting complexity.
[0402] Embodiment 5: The process timeline for all AI-based CSI reports:
[0403] For AI-based use cases, such as AI-based CSI prediction, AI-based beam management (BM) , AI-based CSI compression, AI-based CSI prediction and compression, or other scenarios where the UE is required to report CSI or another quantity via PUSCH or PUCCH, one key difference between AI / ML-based processing and non-AI / ML-based processing, from a timeline perspective, is model deployment. Before inference can be performed, the model must be deployed to the AI / ML processing memory, a procedure that can be understood as functionality activation. Since multiple models supported by the UE may share the AI / ML processing memory, storing a specific model in memory at all times would result in inefficient use of storage resources. Therefore, the model can be removed from memory after each inference occasion is completed. This can be understood as functionality deactivation. Based on the discussion above, when triggering aperiodic or semi-persistent CSI reporting for AI-based use cases, it is required that the model input be ready prior to functionality activation. To address this issue, this embodiment provides at least one of the following methods:
[0404] Method 1: For aperiodic CSI reporting, distinguishing the functionality activation or inactivation state. Different functionality states correspond to different ZAI-ref and Z'AI-ref (n) .
[0405] When triggering aperiodic CSI reporting, if the functionality is active, then ZAI-ref = Zref and Z'AI-ref (n) =Z'ref (n) . If the functionality is inactive, then ZAI-ref = Zref+ΔT and Z'AI-ref (n) =Z'ref (n) +ΔT, where ΔT represents the time required for functionality activation. For different numerologies (μ) , the value of ΔT may have different or equal values.
[0406] When the CSI request field in a DCI triggers CSI report (s) on PUSCH, the UE shall provide a valid CSI report for the n-th triggered report. For AI-based use cases, this is required only if the first uplink symbol carrying the corresponding CSI report (s) , including the effect of timing advance, starts no earlier than symbol ZAI-ref, Similarly, for the n-th CSI report in AI-based use cases, the first uplink symbol carrying the report, including the effect of timing advance, shall start no earlier than symbol Z'AI-ref (n) , where ZAI-ref or Zref is defined as the next uplink symbol with its CP starting Tproc, CSI= (Z) (2048+144) ·κ2-μ·TC+Tswitch after the end of the last symbol of the PDCCH triggering the CSI report (s) , and where Z'AI-ref (n) or Z'ref (n) , is defined as the next uplink symbol with its CP starting T′proc, CSI= (Z′) (2048+144) ·κ2-μ·TC after the end of the latest symbol, in time, among the following resources used for the n-th triggered CSI report: Aperiodic CSI-RS for channel measurements, aperiodic CSI-IM used for interference measurements, and / or aperiodic NZP CSI-RS for interference measurement.
[0407] Moreover, Z, Z' and μ are defined as: and where M is the number of updated CSI report (s) according to Clause 5.2.1.6, (Z (m) , Z′ (m) ) corresponds to the m-th updated CSI report and is defined as (Z2+14 (K-1) m, Z′2) , with (Z2, Z′2) of table 5.4-2, if the CSI report is configured with N4=1, codebookType is set to 'typeII-Doppler-r18'or 'typeII-Doppler-PortSelection-r18' and the corresponding NZP-CSI-RS-ResourceSet for channel measurement is aperiodic with K CSI-RS resources, or (Z2+w, Z′2) , with (Z2, Z′2) of table 5.4-2, if the CSI report is configured with N4=1, codebookType is set to 'typeII-Doppler-r18' or 'typeII-Doppler-PortSelection-r18' and the corresponding NZP-CSI-RS-ResourceSet for channel measurement is periodic or semi-persistent with a single CSI-RS resource, or (Z2+14 (K-1) m, Z′2) or (Z2+14 (K-1) m+Z′2, 2Z′2) , according to UE reported capability, with (Z2, Z′2) of table 5.4-2, if the CSI report is configured with N4>1, codebookType is set to 'typeII-Doppler-r18' and the corresponding NZP-CSI-RS-ResourceSet for channel measurement is aperiodic with K CSI-RS resources, or (Z2+w, Z′2) or (Z2+w+Z′2, 2Z′2) , according to UE reported capability, with (Z2, Z′2) of table 5.4-2, if the CSI report is configured with N4>1, codebookType is set to 'typeII-Doppler-r18' and the corresponding NZP-CSI-RS-ResourceSet for channel measurement is periodic or semi-persistent with a single CSI-RS resource. Z2 is the same as 3GPP 38.214 Table 5.4-2: CSI computation delay requirement 2. Moreover, in this method, Z2 and Z′2 can also replace by Z4 and Z′4, where Z4 and Z′4 are defined use to express the CSI computation delay requirement for AI-based use cases. Note that Z4 and Z′4 can include the functionality activated time or exclude the functionality activated time.
[0408] Method 2: Configure a sufficiently large offset X between the slot containing the DCI that triggers a set of aperiodic NZP CSI-RS resources and the slot in which the CSI-RS resource set is transmitted.
[0409] For aperiodic and semi-persistent CSI reporting on PUSCH, to ensure that the model input is ready before functionality activation, the configuration of offset X canbe sufficiently large.
[0410] For example, in the case of aperiodic CSI-RS resources and aperiodic CSI reporting, offset X represents the number of slots between the slot containing the DCI that triggers the aperiodic NZP CSI-RS resources and the slot in which the CSI-RS resource set is transmitted. For aperiodicTriggeringOffset, at least one of the following values are defined as follows: Value 0 corresponds to 0 slots. Value 1 corresponds to 1 slot. Value 2 corresponds to 2 slots. Value 3 corresponds to 3 slots. Value 4 corresponds to 4 slots. Value 5 corresponds to 16 slots. Value 6 corresponds to 24 slots. Therefore, for AI-based use cases utilizing aperiodic CSI-RS resources, the UE can be configured with an offset X that is greater than or equal to a predefined value Y. In other words, the UE is not expected to be configured with an offset X less than Y slots. If the existing standard does not support configuration for Y slots, an extended offset X configuration can be introduced. X can be configured with at least one value in the range X∈ [5, 128] , Y is predefined. Y can be predefined at least one value from the range [5, 128] . Note that the at least one of the above methods described above are not mutually exclusive and can be used in combination.
[0411] This embodiment outlines the process timeline for AI-based CSI reports, emphasizing the additional time required for model deployment (functionality activation) before inference and model removal (functionality deactivation) after inference. To ensure timely and valid CSI reporting via PUSCH or PUCCH, particularly for aperiodic or semi-persistent CSI reporting, at least one of the following methods are proposed: Method 1 defines distinct timing references (ZAI-ref and Z'AI-ref (n) depending on whether the model is already loaded or not, with detailed equations accounting for CSI computation delays and various configuration scenarios based on 3GPP specifications. Method 2 involves configuring a sufficiently large offset X between the CSI-triggering DCI and the actual CSI-RS resource transmission, ensuring the model has enough time to prepare for inference. The at least one of the following methods aims to guarantee that the model input is ready before activation and can be used independently or in combination for enhanced reliability. This embodiment ensures timely and efficient CSI reporting in AI-based systems by aligning model deployment timelines with resource scheduling, thus enhancing compatibility with existing and future 5G / NR standards.
[0412] Embodiment 6: CPU utilization rule for all AI-based CSI reports.
[0413] In some examples, wherein the processor usage and / or the CMU usage is managed by at least one of the following: if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported, and / or if sufficient processing resources are available, the UE updates a requested CSI report with a lower priority than the performance monitoring results.
[0414] Building upon the general rule for CPU utilization in AI-based CSI reporting, this embodiment introduces several concrete examples that illustrate how processing resources can be efficiently allocated under varying system configurations and constraints. These examples provide practical implementations of the proposed CPU management strategy, taking into account different architectural assumptions-whether the CPU is shared, partially separated, or fully separated across legacy and AI / ML-based functionalities. Each method reflects a scenario-specific application of the rule, ensuring flexibility and scalability. Notably, at least one operation or configuration illustrated in these examples may be independently implemented, or may be combined with other methods in an adaptive manner, depending on system needs. This modular approach enables a broader and more adaptable protection scope, allowing the techniques to remain effective across diverse device capabilities and deployment scenarios.
[0415] CSI reporting for AI-based CSI prediction can adopt the same reporting method as the traditional Rel-18 Doppler codebook. However, due to differences in processing complexity between AI-predicted CSI and traditionally calculated CSI, the number of CSI processing units corresponding to the reporting of AI-predicted CSI needs to be redefined. Similarly, for the reporting of model monitoring, CSI processing units also need to be defined.
[0416] One issue is whether the overall CPU can be shared or separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, as well as among different AI / ML features or functionalities (e.g., beam management, CSI prediction, and CSI compression) . At least one of the following candidate options can be considered: Option 1: The overall CPU is shared between legacy CSI reporting and AI / ML-based CSI reporting, and also shared among CSI report-related AI / ML functionalities. Option 2: The overall CPU is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, but shared among CSI report-related AI / ML functionalities. Option 3: The overall CPU is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, and also separately accounted for among CSI report-related AI / ML functionalities.
[0417] Assume that the overall CPU usage is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, and is shared among AI / ML functionalities related to CSI reporting. Another key issue is determining whether the CPU cost for performance monitoring result reporting for AI-based use cases or features can be attributed to legacy CSI reporting or AI / ML-based CSI reporting. Moreover, if the performance monitoring result reporting is attributed to legacy CSI reporting, the CPU allocated for AI / ML-based CSI reporting becomes limited, while the CPU for legacy CSI reporting remains unrestricted. In this situation, continuous reporting of performance monitoring results may lead to a waste of computational resources.
[0418] Based on the problems mentioned above, this embodiment provides some methods to address them. The specific methods can perform at least one of the following:
[0419] Before introducing the methods, the definitions of some parameters are provided as follows:
[0420] For option 1, NCPU represent that a UE supports NCPU simultaneous CSI calculations it is said to have NCPU CSI processing units for processing CSI reports.
[0421] For option 2, NCPUrepresents that a UE supports NCPUsimultaneous legacy CSI or non-AI / ML-based CSI calculations. It is stated that the UE has NCPU CSI processing units for processing legacy CSI reports or non-AI / ML-based CSI reports. Meanwhile, NAI-CPU represents that a UE supports NAI-CPU simultaneous AI / ML-based CSI calculations. It is stated that the UE has NAI-CPU CSI processing units for processing AI / ML-based CSI reports. Note that, for option 2, the CPU cost of the performance monitoring results reporting is attributed to NCPU and / or NAI-CPU.
[0422] For option 3, NCPUrepresents that a UE supports NCPUsimultaneous legacy CSI or non-AI / ML-based CSI calculations. It is stated that the UE has NCPU CSI processing units for processing legacy CSI reports or non-AI / ML-based CSI reports. Meanwhile, NAI-CPU-X represents that UE supports the functionality, feature, functionality group, or feature group AI-based X, and the functionality of feature can be beaming management, CSI prediction, CSI compression, CSI prediction and compression, or CSI prediction and CSI compression.
[0423] Method 1: Assuming that the overall CPU is shared, as in Option 1, if the CSI reporting for performance monitoring corresponds to model inference results that are unavailable or not being reported, the UE shall not update the requested CSI reports for performance monitoring. Instead, if sufficient CPU resources are available, the UE shall update the next requested CSI report that has a lower priority (as defined in 3GPP 38.214 Clause 5.2.5) than the CSI reports for performance monitoring.
[0424] If a UE supports NCPU simultaneous CSI calculations it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-L unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-L CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to
[0425] If the CSI reporting for performance monitoring is the M-th CSI report, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority than the M-th CSI report for performance monitoring, where 0≤K≤N-M-1 is the largest value such that holds.
[0426] Method 2: Assuming that the overall CPU is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, as in Option 2, and that the CSI reporting for performance monitoring is attributed to AI / ML-based CSI reporting. In this situation, if the CSI reporting for performance monitoring corresponds to model inference results that are unavailable or not being reported, the UE shall not update the requested CSI reports for performance monitoring. Instead, if sufficient CPU resources are available, the UE shall update the next requested CSI report that has a lower priority (as defined in 3GPP 38.214 Clause 5.2.5) than the CSI reports for performance monitoring.
[0427] If a UE supports NCPU-AI simultaneous CSI calculations it is said to have NCPU-AI CSI processing units for processing CSI reports. If P CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-AI-P unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-AI-P CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to
[0428] If the CSI reporting for performance monitoring is the M-th CSI report, where 0≤M≤N, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority than the M-th CSI report for performance monitoring, where 0≤K≤N-M-1 is the largest value such that holds.
[0429] Method 3: Assuming that the overall CPU is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, as in Option 2, and that the CSI reporting for performance monitoring is attributed to legacy CSI reporting. In this situation, if the CSI reporting for performance monitoring corresponds to model inference results that are unavailable or not being reported, the UE shall not update the requested CSI reports for performance monitoring. Instead, if sufficient CPU resources are available, the UE shall update the next requested CSI report that has a lower priority (as defined in 3GPP 38.214 Clause 5.2.5) than the CSI reports for performance monitoring.
[0430] If a UE supports NCPU simultaneous CSI calculations for legacy CSI reporting it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-L unoccupied CPUs.
[0431] If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-L CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to
[0432] If the CSI reporting for performance monitoring is the M-th CSI report, where 0≤M≤N, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority than the M-th CSI report for performance monitoring, where 0≤K≤N-M-1 is the largest value such that holds.
[0433] Method 4: Assuming that the overall CPU is separately accounted for between legacy CSI reporting and AI / ML-based CSI reporting, as in Option 2, and that the CSI reporting for performance monitoring is attributed to legacy CSI reporting and / or AI / ML-based CSI reporting. The CSI reports for performance monitoring are given priority in utilizing the CPU allocated for AI-based CSI reporting or legacy CSI reporting. If the CPU resources for AI-based CSI reporting or legacy CSI reporting are limited, the CSI reports for performance monitoring will occupy the available CPU resources allocated to either legacy CSI reporting or AI-based CSI reporting.
[0434] If a UE supports NCPU simultaneous CSI calculations for legacy CSI reporting it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-L unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-L CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to
[0435] If a UE supports NCPU-AI simultaneous CSI calculations for AI-based CSI reporting it is said to have NCPU-AI CSI processing units for processing CSI reports. If P CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-AI-P unoccupied CPUs. If G CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-AI-P CPUs are unoccupied, where each CSI report g=0, …, G-1 corresponds to
[0436] Regardless of which part of the CPU is occupied by the CSI reporting for performance monitoring. If the CSI reporting for performance monitoring is the M-th CSI report, where 0≤M≤N or 0≤M≤G, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority than the M-th CSI report for performance monitoring, where 0≤K≤N-M-1 or 0≤K≤G-M-1 is the largest value such that or holds.
[0437] Method 5: Assuming that the overall CPU is separately counted between legacy CSI reporting and AI / ML-based CSI reporting, as in Option 3, and the CSI reporting for performance monitoring is attributed to either legacy CSI reporting or the corresponding AI / ML-based CSI reporting, depending on the specific feature or functionality.
[0438] If a UE supports NCPU simultaneous CSI calculations for legacy CSI reporting it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-L unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-L CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to
[0439] If a UE supports NCPU-AI-X simultaneous CSI calculations for legacy CSI reporting it is said to have NCPU CSI processing units for processing CSI reports. If L CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-AI-X-L unoccupied CPUs. If G CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-AI-X-L CPUs are unoccupied, where each CSI report g=0, …, G-1 corresponds to
[0440] Case 1: the CSI reporting for performance monitoring is attributed to legacy CSI reporting.
[0441] For case 1, if the CSI reporting for performance monitoring is the M-th CSI report, where 0≤M≤N, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority (according to 3GPP 38.214 Clause 5.2.5) than the M-th CSI report for performance monitoring, where 0≤K≤N-M-1 is the largest value such that holds.
[0442] Case 2: the CSI reporting for performance monitoring is attributed to the corresponding AI / ML-based CSI reporting, depending on the specific feature, functionality, feature group, or functionality group.
[0443] For case 2, if the CSI reporting for performance monitoring is the M-th CSI report, where 0≤M≤G, and it corresponds to the model inference results that are not available or not reporting. Then the UE does not update the M-th requested CSI report for performance monitoring. Instead, the UE is required to update the subsequent K requested CSI reports, which have a lower priority (according to 3GPP 38.214 Clause 5.2.5) than the M-th CSI report for performance monitoring, where 0≤K≤G-M-1is the largest value such that holds.
[0444] Note that the above methods only consider cases where single-model or single-functionality inference results are unavailable or not reported. Additionally, these methods can also be extended to handle scenarios where continuous or discontinuous multi-model or multi-functionality inference results are unavailable or not reported.
[0445] Method 6: Assuming that for AI-based CSI prediction, the number of predicted CSI occasions, N4, can be any of {1, 2, 4, 8, 12, 16, 20} . For aperiodic measurement and reporting, the number of aperiodic CSI-RS resources, K, can be any of {2, 4, 8, 12, 16, 20} . Then, the number of CPUs occupied by CSI reporting can be determined using the following methods:
[0446] If the corresponding CSI-RS Resource Set for channel measurement is aperiodic and configured with K CSI-RS resources, OCPU-AI∈ {8, 12, 16, 18, 20} for K>12, OCPU-AI=8 for K=12 and OCPU-AI=Y1·K for K<12, where Y1∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by the UE capability indication.
[0447] If the corresponding CSI-RS resource set for channel measurement is periodic or semi-persistent and configured with a single CSI-RS resource, OCPU-AI=4 for N4=1 and OCPU-AI=max (Y2·N4, 4) for N4>1, where the value of N4 is configured by the higher layer parameter N4, and Y2∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by the UE capability indication.
[0448] Based on the aforementioned methods, for the UE, the CPU occupancy for both AI-based and non-AI-based CSI reporting may be based on the same unit or satisfy a certain ratio. For example, a single CPU for AI-based CSI reporting may be X times that for non-AI-based CSI reporting, where 0 < X < 1024. Moreover, for the UE, whether the CPU is shared or separately allocated for both AI-based and non-AI-based CSI reporting depends on the UE capability. For example, a specific field may be added to indicate whether the CPU is shared or separately allocated for both AI-based and non-AI-based CSI reporting in UE capability reporting.
[0449] Moreover, for a periodic or semi-persistent CSI-RS resource in a CSI-RS resource set for channel measurement linked to a CSI-ReportConfig configured with the higher layer parameter codebookType set to 'AI-based CSI predication' , the CSI-RS resource and the CSI-RS ports within the CSI-RS resource are counted KP times, where the value of KP∈ {1, 2, 4, 6, 8, 10, 12} is indicated by the UE capability.
[0450] This embodiment proposes detailed CPU utilization rules for AI-based CSI reporting, addressing how processing resources can be allocated across legacy and AI / ML-based CSI tasks. It first outlines three CPU configuration options (shared, partially separated, or fully separated across functionalities) and introduces definitions such as NCPU, NCPU-AI, and NCPU-AI-X to represent the number of CSI processing units supported by the UE. Based on these configurations, these methods are proposed to handle cases where AI inference results are unavailable or not reported. For example, some methods define CPU occupancy for AI-based CSI reporting based on CSI-RS type, resource count, and UE capability. CPU use may differ between AI and non-AI reporting, with allocation (shared or separate) indicated by the UE. These methods ensure that lower-priority CSI reports can still be processed efficiently by reassigning available CPU units, depending on the architecture. The embodiment also includes mathematical constraints to guide CPU allocation and supports flexible attribution of CPU cost depending on whether performance monitoring is considered a legacy or AI / ML-based task. Lastly, the methods can be extended to cover complex multi-model or multi-functionality scenarios, ensuring robustness. This embodiment ensures efficient and flexible CPU resource management for AI-based and legacy CSI reporting, minimizing processing waste while maintaining system performance-even when AI inference results are unavailable.
[0451] Embodiment 7: The CPU and CMU utilization rule for all AI-based CSI reports.
[0452] In some examples, the UE determines the CSI reporting based on at least one of the following: calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting; prioritizing CSI reporting based on a CMU availability and / or a CPU availability.
[0453] In this embodiment, prioritizing CSI reporting based on CMU (Channel Measurement Unit) availability and / or CPU availability serves as the foundational rule for efficient resource management. To ensure optimal performance and adaptability under dynamic processing loads, the system intelligently determines whether and how to proceed with CSI report generation depending on the availability of computational resources. The following are more specific examples illustrating how this prioritization can be implemented under various CPU and CMU allocation strategies. These examples not only address practical execution paths for different system architectures but also show how reporting decisions can be flexibly adjusted based on real-time resource conditions. At least one of the described operations in these examples can be applied independently or in combination with others, depending on implementation needs. This design ensures broader applicability and scalability across diverse UE capabilities, enabling more robust and efficient CSI reporting even in resource-constrained environments.
[0454] To support simultaneous CSI reports, the UE may need to deploy multiple models or functionalities in the AI / ML processing memory, each requiring a dedicated memory allocation. If the number of models or functionalities expected to run simultaneously exceeds the available memory capacity, some CSI reports must be canceled. To ensure consistent understanding of CSI update behavior between the gNB and the UE, it may be necessary to introduce a memory occupancy alignment mechanism. Under this mechanism, the UE is not required to update an AI / ML-based CSI report if the available memory cannot accommodate its required memory footprint.
[0455] Based on the above reasons, if a UE supports NCMU simultaneous CSI calculations it is said to have NCMU CSI memory units for processing CSI reports. If L CMUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCMU-L unoccupied CMUs. If N CSI reports start occupying their respective CMUs on the same OFDM symbol on which NCMU-L CMUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to the UE is not required to update the N-M requested CSI reports with lowest priority, where 0≤M≤N is the largest value such that holds.
[0456] Moreover, based on the embodiment 6, assume that the overall CPU is separately counted between legacy CSI reporting and AI / ML-based CSI reporting, and shared among CSI report related AI / ML functionalities. Based on the assumption, NAI-CPU represents that a UE supports NAI-CPU simultaneous AI / ML-based CSI calculations. It is stated that the UE has NAI-CPU CSI processing units for processing AI / ML-based CSI reports. If P CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-AI-P unoccupied CPUs. If N CSI reports start occupying their respective CPUs on the same OFDM symbol on which NCPU-AI-P CPUs are unoccupied, where each CSI report n=0, …, N-1 corresponds to the UE is not required to update the NCPU-AI-P requested CSI reports with lowest priority, where 0≤Q≤N is the largest value such that holds.
[0457] It is clear that updating the requested CSI reports depends on two factors: whether the CPU is limited and whether the CMU is limited. Therefore, to align the understanding of CSI update behavior between the gNB and the UE, it may be necessary to introduce a memory occupancy and CPU occupancy alignment mechanism.
[0458] Based on the issues mentioned above, this embodiment provides several methods to address them. The specific methods may perform at least one of the following operations:
[0459] Assume that P CPUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCPU-AI-P unoccupied CPUs, and L CMUs are occupied for calculation of CSI reports in a given OFDM symbol, the UE has NCMU-L unoccupied CMUs. If N CSI reports start occupying their respective CPUs and CMUs on the same OFDM symbol on which NCPU-AI-P CPUs and NCMU-L are unoccupied, where each CSI report n=0, …, N-1 corresponds to and respectively.
[0460] Method 1: The UE is not required to update the N-M requested CSI reports with lowest priority (according to 3GPP 38.214 Clause 5.2.5) , where M=min {M1, M2} , 0≤M1≤N is the largest value such that holds, and 0≤M2≤N is the largest value such that holds. That means the UE only requires to update the first M requested CSI reports with the highest priority (according to 3GPP 38.214 Clause 5.2.5) .
[0461] Method 2: Assume that 0≤M2≤N is the largest value such that holds, and 0≤K2≤N-M2-x is the largest value such that holds. This means that the memory allocated for the (M2+1) -th to (M2+x) -th CSI reports is limited. If the memory required for the (M2+1) -th to (M2+x) -th CSI reports is not considered, the remaining memory can support the requirements of lower-priority (according to 3GPP 38.214 Clause 5.2.5) CSI reports compared to the (M2+1) -th to (M2+x) -th CSI reports.
[0462] Meanwhile, 0 ≤K1≤N-M2-x is the largest value such that holds.
[0463] Based on the above assumptions, the UE is not required to update (M2+1) -th to (M2+x) -th CSI reports, instead the UE update the subsequent K requested CSI reports that is with lower priority than (M2+1) -th to (M2+x) -th CSI report, where K=min {K1, K2} .
[0464] Method 3: Assume that 0≤M1≤N is the largest value such that holds, and 0≤K1≤N-M1-y is the largest value such that holds. This means that the CPU allocated for the (M1+1) -th to (M1+y) -th CSI reports is limited. If the CPU required for the (M1+1) -th to (M1+y) -th CSI reports is not considered, the remaining CPU can support the requirements of lower-priority (according to 3GPP 38.214 Clause 5.2.5) CSI reports compared to the (M1+1) -th to (M1+y) -th CSI reports.
[0465] Meanwhile, 0 ≤K2≤N-M1-y is the largest value such that holds.
[0466] Based on the above assumptions, the UE is not required to update (M1+1) -th to (M1+y) -th CSI reports, instead the UE update the subsequent K requested CSI reports with lower priority than (M1+1) -th to (M1+y) -th CSI report, where K=min {K1, K2} .
[0467] Method 4: Considering that each AI functionality requires a certain amount of memory, the UE is required to report its memory storage capability. This method introduces three parameters to represent the UE capability: (1) the maximum number of functionalities; (2) the maximum number of models for each functionality: uploaded or activated; and (3) the total number of models: uploaded or activated. Note that the above capabilities may be defined per band, per cell, per BWP, or across all bands.
[0468] This embodiment focuses on defining the combined CPU and CMU (CSI memory unit) utilization rules for AI-based CSI reporting. It begins by highlighting the challenge of limited AI / ML processing memory, which may prevent multiple AI models or functionalities from running simultaneously. To resolve potential resource contention, this embodiment introduces a memory and CPU occupancy alignment mechanism to determine whether specific CSI reports can be updated. It defines parameters such as NCMU, NCPU-AI, and individual resource consumption per CSI report (e.g., and ) , and proposes three methods to prioritize CSI reporting based on available resources. Each method includes mathematical conditions to ensure that only high-priority CSI reports are updated, while lower-priority ones are skipped when memory or processing resources are insufficient. Some methods define UE memory capability reporting based on the number of AI functionalities and models, with scope specified per band, cell, BWP, or across all bands. These methods also accommodate dynamic scenarios where certain CSI reports are skipped to allow lower-priority ones to proceed if resource thresholds permit. This embodiment enables intelligent resource allocation for CSI reporting by prioritizing updates based on real-time CPU and memory availability, ensuring efficient system performance under resource constraints.
[0469] Embodiment 8: Switching mechanism between AI-based CSI reports and Legacy CSI reports.
[0470] In some examples, the UE switches between an AI-based CSI reporting and a non-AI-based CSI reporting via a MAC CE message or a DCI; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the MAC CE message or the DCI to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting. In some examples, the MAC CE message or the DCI comprises at least one of the following: a field used to indicate an applicable CSI reporting configuration ID; a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting; a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.
[0471] To provide a concrete implementation of the switching mechanism between AI-based and non-AI-based CSI reporting as outlined above, the following examples demonstrate specific signaling procedures, configuration structures, and operational flows under various system settings. These examples elaborate how such switching can be achieved through MAC CE or DCI signaling, and how CSI-ReportConfig IDs, codebook types, and performance monitoring are managed during transitions. Each case reflects a different level of configurability, ranging from shared CSI-ReportConfig with codebook enumeration to completely separate configurations with bi-directional associations, allowing for a highly adaptable and scalable solution. Importantly, the methods and operations illustrated herein may be applied independently or in any suitable combination, depending on deployment needs or network implementation strategies. This modular and extensible design ensures that the embodiment remains compatible with a wide range of device capabilities and standardization requirements, while also maximizing flexibility for dynamic system behavior and performance optimization.
[0472] Based on the ongoing discussions in current standards, the network determines whether to perform a fallback operation based on the reported performance monitoring results. However, the specific switching mechanism remains undefined. Therefore, this embodiment provides at least one of the following methods:
[0473] Case 1: Both AI-based CSI reports and non-AI-based CSI reports for the same functionality share a single CSI-ReportConfig, which supports both AI-based and non-AI-based codebooks. For example, within the CSI-ReportConfig, the codebook type is configured using an enumeration that includes both AI-based and non-AI-based codebooks.
[0474] Moreover, performance monitoring for AI-based CSI reports is required, and the UE must report the performance monitoring results. To associate the quantities of both the AI-based CSI reports and the performance monitoring result reporting, the CSI-ReportConfig can be used to establish a relationship between them. For example, the CSI-ReportConfig ID for performance monitoring result reporting can be configured within the CSI-ReportConfig for AI-based CSI reports. Similarly, the CSI-ReportConfig ID for AI-based CSI reports can be configured within the CSI-ReportConfig for performance monitoring result reporting. Note that this configuration can be either unidirectional or bidirectional.
[0475] Method 1: For periodic CSI reporting, support for AI-based or non-AI-based codebooks can be configured by the RRC. This configuration can be achieved, for example, by using a 1-bit indicator to specify which type of codebook to apply or by predefining the codebook type to be used.
[0476] For semi-persistent or aperiodic CSI reporting, support for AI-based or non-AI-based codebooks can be configured by MAC CE or DCI. This configuration can also be achieved, for example, by using a 1-bit indicator to specify the type of codebook or by predefining the codebook type.
[0477] Based on the above configuration, switching between AI-based and non-AI-based codebooks can be indicated by MAC CE or DCI. The MAC CE or DCI information includes at least one of the following items: A field indicating the applied codebook type; A field indicating the CSI-ReportConfig ID; Serving Cell ID; BWP ID; Time domain resource assignment; A field indicating whether to activate or deactivate performance monitoring result reporting.
[0478] Note that during the switch, if no specific CSI report parameter configuration is provided, it means the CSI reporting parameters prior to the switch are reused. For example, when switching from AI-based CSI prediction to the Rel-18 Doppler codebook, if no new parameters are configured, the parameters for AI-based CSI prediction will continue to apply to the Rel-18 Doppler codebook.
[0479] If the codebook type switches from AI-based to non-AI-based, performance monitoring result reporting for AI-based CSI reports is also deactivated. Conversely, if the codebook type switches from non-AI-based to AI-based, performance monitoring result reporting is activated. Additionally, activation or deactivation of performance monitoring result reporting can also be indicated by a single bit. In some examples, if the bit is set to 1, it indicates that performance monitoring result reporting is also enabled when AI-based CSI reporting is activated. In another example, if the bit is set to 0, it indicates that performance monitoring result reporting is also enabled when AI-based CSI reporting is activated.
[0480] Case 2: AI-based CSI reports and non-AI-based CSI reports for the same functionality are configured with different CSI-ReportConfig IDs, respectively. Additionally, within the CSI-ReportConfig for AI-based CSI reports, at least one associated CSI-ReportConfig ID for non-AI-based CSI reports is configured. Similarly, within the CSI-ReportConfig for non-AI-based CSI reports, at least one associated CSI-ReportConfig ID for AI-based CSI reports is configured. Note that the association between the CSI-ReportConfig for AI-based CSI reports and the CSI-ReportConfig for non-AI-based CSI reports can be unidirectional. For example, under the CSI-ReportConfig for AI-based CSI reports, two associated CSI-ReportConfig IDs for non-AI-based CSI reports may be configured: one for semi-persistent CSI reporting and the other for aperiodic CSI reporting.
[0481] Note that for periodic CSI reporting, the current CSI-ReportConfig may be associated with a CSI-ReportConfig for semi-persistent and / or aperiodic CSI reporting. Similarly, for semi-persistent CSI reporting, the current CSI-ReportConfig may be associated with a CSI-ReportConfig for periodic and / or aperiodic CSI reporting.
[0482] Moreover, performance monitoring for AI-based CSI reports is required, and the UE must report the performance monitoring results. To establish an association between the AI-based CSI reports and performance monitoring result reporting, CSI-ReportConfig can be used. For example, the CSI-ReportConfig ID for performance monitoring result reporting can be configured under the CSI-ReportConfig for AI-based CSI reports. Likewise, the CSI-ReportConfig ID for AI-based CSI reports can be configured under the CSI-ReportConfig for performance monitoring result reporting. Note that this configuration can be either unidirectional or bidirectional.
[0483] Method 2: Based on the above configuration of Case 2, the switch between AI-based CSI reports and non-AI based CSI reports can be indicated through MAC CE or DCI.
[0484] For periodic, semi-persistent, or aperiodic CSI reporting, the MAC CE or DCI information includes at least one of the following items:
[0485] A field used to indicate the CSI-ReportConfig ID for the current CSI reports, similar to the CSI request field in the legacy DCI used for triggering CSI reporting or the Sifield in the legacy MAC CE used for triggering semi-persistent CSI reporting. For example, the field indicates that the CSI-ReportConfig for AI-based CSI reports or non-AI based CSI reports is ID-X.
[0486] A field used to indicate whether to switch the CSI-ReportConfig for the current CSI reports to the associated CSI-ReportConfig. The bit length of this field depends on the number of associated CSI-ReportConfig instances. If the number of associated CSI-ReportConfig instances is N, the bit length of this field is In some examples, if this field is 1 bit and the bit is set to 1, it indicates that the CSI reports can be switched to another associated CSI-ReportConfig. In another example, if this field is 1 bit and the bit is set to 0, it indicates that the CSI reports can be switched to another associated CSI-ReportConfig.
[0487] A field used to indicate the time-domain resource assignment for CSI reports. It is similar with the time domain resource assignment field in the legacy DCI used for triggering CSI reporting.
[0488] A field used to indicate the index of the selected periodicity and offset value for CSI reporting. For example, under the CSI-ReportConfig for non-AI-based CSI reports or AI-based CSI reports, if multiple candidate periodicity and offset values are configured, this field can indicate the index of the selected periodicity and offset value.
[0489] Based on the above, there is an association between the CSI-ReportConfig for AI-based CSI reports and the CSI-ReportConfig for performance monitoring. When switching from AI-based CSI reports to non-AI-based CSI reports, the performance monitoring result reporting for AI-based CSI reports is also deactivated. Conversely, when switching from non-AI-based CSI reports to AI-based CSI reports, the performance monitoring result reporting for AI-based CSI reports is activated. Moreover, the activation or deactivation of performance monitoring result reporting can also be indicated by a single bit. In some examples, if the bit is set to 1, it indicates that when AI-based CSI reporting is activated, the performance monitoring result reporting for AI-based CSI reports is also activated. In another example, if the bit is set to 0, it indicates that when AI-based CSI reporting is activated, the performance monitoring result reporting for AI-based CSI reports is also activated.
[0490] Method 3: For case 2, the legacy semi-persistent and aperiodic CSI reporting trigger mechanisms are reused with some enhancements. Additionally, the switching between AI-based CSI reports and non-AI-based CSI reports can be indicated via an enhanced legacy MAC CE signal or a DCI signal, which is used for triggering semi-persistent or aperiodic CSI reporting.
[0491] For the periodic or semi-persistent CSI reports, the switching between AI-based codebooks and non-AI based codebooks can be indicated by MAC CE or DCI.
[0492] Assume that the CSI-ReportConfig ID of the current AI-based periodic or semi-persistent CSI reporting is associated with a CSI-ReportConfig for non-AI-based semi-persistent CSI reporting. That is the CSI-ReportConfig ID of the current AI-based periodic or semi-persistent CSI reporting is configured under the CSI-ReportConfig for non-AI-based semi-persistent CSI reporting. If it wants to switch from AI-based periodic or semi-persistent CSI reporting to non-AI-based semi-persistent CSI reporting, the MAC CE information can be expressed as at least one of the following in Table 17:
[0493] Table 17: MAC CE information.
[0494] S: This field indicates whether to activate or deactivate the CSI reporting, where the CSI-ReportConfig ID of the CSI reporting is associated with the CSI-ReportConfig of the deactivate or activated semi-persistent CSI reporting, which corresponds to Si. Note that the above location of the field S in MAC CE is just an example, and it can also be located in other locations.
[0495] The other field in this MAC CE signal is the same as the legacy MAC CE used to inactivate or deactivate semi-persistent CSI reports.
[0496] For example, if using the MAC CE signal to trigger the non-AI-based semi-persistent CSI reporting, which correspond to CSI-ReportConfigId related to S0. In some examples, if the field S is set to 1, the associated AI-based periodic or semi-persistent CSI reporting is deactivated. In another example, if the field S is set to 0, the associated AI-based periodic or semi-persistent CSI reporting is deactivated. The CSI-ReportConfig ID for the AI-based periodic CSI reporting is associated with the CSI-ReportConfig for the non-AI-based semi-persistent CSI reporting.
[0497] Note that the legacy MAC CE is only used to inactivate or deactivate semi-persistent CSI reports on PUCCH. If it intends to switch periodic AI-based CSI reports to semi-persistent non-AI-based CSI reports, it requires adding a field to indicate the CSI-ReportConfig ID for the periodic AI-based CSI reports. Then the MAC CE information can be expressed as follows in Table 18:
[0498] Table 18: MAC CE information.
[0499] S: This field indicates whether the MAC CE signal is used to trigger the switch between AI-based CSI reports and non-AI-based CSI reports. In some examples, if this field is set to 1, it means that the MAC CE signal is used for this purpose; otherwise, it is not. In another example, if this field is set to 0, it means that the MAC CE signal is used for this purpose; otherwise, it is not. Note that the above location of the field S in MAC CE is just an example, and it can also be located in other locations.
[0500] CSI-ReportConfig ID: This field indicates the CSI-ReportConfig ID for the periodic CSI reports. In some examples, if the field S is set to 1, it means that the periodic CSI reports need to be deactivated. In another example, if the field S is set to 0, it means that the periodic CSI reports need to be deactivated. Note that the bit length of this field depends on the number of CSI-ReportConfig IDs configured for periodic AI-based CSI reports, non-AI-based CSI reports, or both.
[0501] The other field in this MAC CE signal is the same as the legacy MAC CE used to inactivate or deactivate semi-persistent CSI reports.
[0502] Moreover, if it wants to switch AI-based periodic or semi-persistent CSI reporting to non-AI-based semi-persistent or aperiodic CSI reporting, the DCI signal can be expressed as follows:
[0503] Reused the DCI format 0_1 or DCI format 0_2 to activate semi-persistent or aperiodic CSI reporting, DCI format 0_1 and DCI format 0_2 contains a CSI request field which indicates the semi-persistent or aperiodic CSI trigger state to activate or deactivate. Meanwhile, add a field “Associated CSI request” or reuse the remaining bits in DCI format 0_1 and DCI format 0_2 to indicate whether to deactivate periodic or semi-persistent CSI reporting. The CSI-ReportConfig ID for AI-based periodic or semi-persistent CSI reporting is configured under the CSI-ReportConfig for non-AI-based semi-persistent or aperiodic CSI reporting. In some examples, if the bit length of the field is 1 and the bit is set to 1, then the AI-based periodic or semi-persistent CSI reporting is deactivated. In another example, if the bit length of the field is 1 and the bit is set to 0, then the AI-based periodic or semi-persistent CSI reporting is deactivated.
[0504] For example, if using the DCI signal to trigger the non-AI-based semi-persistent or aperiodic CSI reporting, which corresponds to CSI-ReportConfigId related to CSI request. In some examples, if the field "Associated CSI request" is set to 1, then the associated AI-based periodic or semi-persistent CSI reporting is deactivated. In another example, if the field "Associated CSI request" is set to 0, then the associated AI-based periodic or semi-persistent CSI reporting is deactivated. The CSI-ReportConfig ID of the AI-based periodic or semi-persistent CSI reporting is associated with the CSI-ReportConfig of the non-AI-based semi-persistent or aperiodic CSI reporting.
[0505] Based on the aforementioned, there is an association between the CSI-ReportConfig for AI-based CSI reports and the CSI-ReportConfig for performance monitoring. When switching from AI-based CSI reports to non-AI-based CSI reports, the performance monitoring result reporting for AI-based CSI reports is also deactivated. Conversely, if switching from non-AI-based CSI reports to AI-based CSI reports, the performance monitoring result reporting for AI-based CSI reports is also activated. Moreover, the activation or deactivation of performance monitoring result reporting can also be indicated by a single bit. In some examples, if the bit is set to 1, it indicates that performance monitoring result reporting for AI-based CSI reports is also activated when AI-based CSI reports are enabled. In another example, if the bit is set to 0, it indicates that performance monitoring result reporting for AI-based CSI reports is also activated when AI-based CSI reports are enabled.
[0506] Additionally, the field “Associated CSI request” can also be used for indicating the CSI-ReportConfigId of the deactivated or activated CSI reporting.
[0507] Note that this method also applies when switching from non-AI-based periodic or semi-persistent CSI reporting to AI-based semi-persistent or aperiodic CSI reporting.
[0508] Case 3: Both AI-based and non-AI-based CSI reports for the same functionality are configured with different CSI-ReportConfig IDs. However, the measurement resources can be reused between AI-based and non-AI-based CSI reports. This means that when switching between the two, the CSI-RS resource configuration remains unchanged. Moreover, the periodic or semi-persistent CSI-RS measurement resources used for AI-based CSI reports can also be used for performance monitoring. Therefore, the CSI-ReportConfigs for AI-based CSI reporting, non-AI-based CSI reporting, and performance monitoring can be associated with the same CSI-RS resource or resource set.
[0509] For this case the switching between AI-based CSI reports and non-AI-based CSI reports can use at least one of the following methods:
[0510] Method 4: For Case 3, the legacy semi-persistent and aperiodic CSI reporting trigger mechanisms are reused with certain enhancements. Additionally, the switching between AI-based and non-AI-based CSI reports can be indicated via an enhanced legacy MAC CE signal or a DCI signal, which is used to trigger semi-persistent or aperiodic CSI reporting.
[0511] For periodic or semi-persistent CSI reports, the switching between AI-based and non-AI-based codebooks can be indicated by a MAC CE or DCI.
[0512] For example, if switching from AI-based periodic or semi-persistent CSI reporting to non-AI-based semi-persistent CSI reporting, the MAC CE signal can be represented in Table 19 as at least one of the following:
[0513] Table 19: MAC CE information.
[0514] S: This field indicates whether to activate or deactivate the CSI reporting, where the CSI-ReportConfig of the CSI reporting and the deactivated or activated semi-persistent CSI reporting, which corresponds to Si, are associated with the same CSI-RS resource or resource set. Note that the above location of the field S in MAC CE is just an example, and it can also be located in other locations. Moreover, the CSI reporting can be AI-based periodic / semi-persistent CSI reporting, performance monitoring result reporting, or a combination of the AI-based reporting and the performance monitoring result reporting.
[0515] The other field in this MAC CE signal is the same as the legacy MAC CE used to inactivate or deactivate semi-persistent CSI reports.
[0516] For example, if the MAC CE signal is used to trigger the non-AI-based semi-persistent CSI reporting, which corresponds to CSI-ReportConfigId related to S0, and the field of S is set to 1, then AI-based periodic or semi-persistent CSI reporting will be deactivated. In this case, the AI-based periodic or semi-persistent CSI reporting and non-AI-based semi-persistent CSI reporting are associated with the same CSI-RS resource or resource set.
[0517] Moreover, if switching from AI-based periodic or semi-persistent CSI reporting to non-AI-based semi-persistent or aperiodic CSI reporting, the DCI signal can be expressed as follows:
[0518] DCI format 0_1 or DCI format 0_2 can be reused to activate or deactivate semi-persistent or aperiodic CSI reporting. These formats contain a CSI request field that indicates the trigger state-activation or deactivation-of semi-persistent or aperiodic CSI reporting. In addition, this method introduces a new field called 'Associated CSI request' or reuses the remaining bits in DCI format 0_1 and 0_2 to indicate whether to activate or deactivate periodic or semi-persistent CSI reporting. The associated CSI reporting and the activated or deactivated semi-persistent or aperiodic CSI reporting share the same CSI-RS resource or resource set. For example, the field may be 1 bit in length: if the bit is set to 1, the AI-based periodic or semi-persistent CSI reporting is activated or deactivated. Furthermore, the CSI reporting may refer to AI-based periodic / semi-persistent CSI reporting, performance monitoring result reporting, or a combination of both.
[0519] Additionally, the field “Associated CSI request” can also be used for indicating the CSI-ReportConfigId of the deactivated or activated CSI reporting.
[0520] Note that this method also applies to switching non-AI-based periodic or semi-persistent CSI reporting to AI-based semi-persistent or aperiodic CSI reporting.
[0521] Case 4: Both AI-based and non-AI-based CSI reports for the same functionality are configured with different CSI-ReportConfig IDs. However, the measurement resources can be reused for both types of CSI reports, meaning that when switching between AI-based and non-AI-based CSI reports, the CSI-RS resource configuration remains unchanged. Moreover, performance monitoring is required for AI-based CSI reports, and the UE is expected to report the performance monitoring results. To associate the quantities between AI-based CSI reports and performance monitoring result reporting, the CSI-ReportConfig can be used to establish a relationship between them. For example, the CSI-ReportConfig ID for performance monitoring result reporting can be configured under the CSI-ReportConfig for AI-based CSI reports, and vice versa. This configuration can be either unidirectional or bidirectional.
[0522] For case 4, the switching between AI-based CSI reports and non-AI-based CSI reports can use at least one of the following methods:
[0523] Method 5: For case 4, the legacy semi-persistent and aperiodic CSI reporting trigger mechanisms are reused with some enhancements. Additionally, the switching between AI-based CSI reports and non-AI-based CSI reports can be indicated via an enhanced legacy MAC CE signal or a DCI signal, which is used for triggering semi-persistent or aperiodic CSI reporting.
[0524] For the periodic or semi-persistent CSI reports, the switching between AI-based codebooks and non-AI based codebooks can be indicated by MAC CE or DCI.
[0525] For example, if it wants to switch from AI-based periodic or semi-persistent CSI reporting to non-AI-based semi-persistent CSI reporting, the MAC CE signal can be expressed as at least one of the following:
[0526] Table 20: MAC CE information.
[0527] S: This field indicates whether to activate or deactivate the CSI reporting, where the CSI-ReportConfig of the CSI reporting and the deactivated or activated semi-persistent CSI reporting, which corresponds to Si, are associated with the same CSI-RS resource or resource set. Note that the above location of the field S in MAC CE is just an example, and it can also be located in other locations.
[0528] Moreover, the CSI-ReportConfig for AI-based CSI reports and the CSI-ReportConfig for performance monitoring result reports are either mutually associated or unilaterally associated. If AI-based CSI reports is activated or deactivated, then the associated performance monitoring result reporting is activated or deactivated.
[0529] The other field in this MAC CE signal is the same as the legacy MAC CE used to inactivate or deactivate semi-persistent CSI reports.
[0530] Additionally, the above switch effect can also be indicated by DCI as the method 4. Note that in the aforementioned methods, AI-based CSI reports and non-AI-based CSI reports perform the same functions, such as CSI prediction, Beam management, CSI compression, and CSI prediction and compression. For example, AI-based CSI reports, non-AI-based CSI reports, and CSI reporting for performance monitoring can be AI-based CSI prediction, Rel-18 Doppler codebook, and performance monitoring reporting for AI-based CSI prediction.
[0531] This embodiment outlines various methods to enable switching between AI-based and non-AI-based CSI reports in 5G networks, 6G networks, 7G networks, or other networks, particularly focusing on mechanisms involving shared or separate CSI-ReportConfig IDs. It includes configurations for periodic, semi-persistent, and aperiodic CSI reporting, utilizing RRC, MAC CE, and DCI signaling. Some cases describe scenarios where AI-based and non-AI-based CSI reports share the same measurement resources but are configured with different CSI-ReportConfig IDs. Switching between the two can be managed via enhanced MAC CE or DCI signals, with additional fields like "Associated CSI request" used to control activation / deactivation. These methods also support performance monitoring for AI-based CSI, allowing association between CSI reporting and monitoring via configurable relationships. AI and non-AI-based reports serve the same functions, such as CSI prediction, beam management, and compression. The embodiment introduces flexible switching mechanisms, enabling seamless fallback based on performance monitoring results. It also details how performance monitoring reporting aligns with CSI reporting transitions, using single-bit indicators and associations between different configurations to maintain reporting continuity and resource efficiency. This mechanism enables dynamic and efficient switching between AI-based and legacy CSI reporting while maintaining accurate performance tracking and minimizing signaling overhead.
[0532] Embodiment #9: Switching time point:
[0533] As mentioned in Embodiment 8, the switching between AI-based CSI reporting and non-AI-based CSI reporting can be indicated by a MAC CE or DCI. However, the effective implementation of the DCI or MAC CE signal requires some time. Therefore, this embodiment discusses the switching time point during the transition between AI-based and non-AI-based CSI reporting, and provides at least one of the following methods:
[0534] Method 1: Assume that if a MAC CE is used to indicate the switching from AI-based CSI reporting to non-AI-based CSI reporting, the MAC CE message serves a dual purpose: it activates the semi-persistent non-AI-based CSI reporting on one hand, and deactivates the periodic or semi-persistent AI-based CSI reporting on the other hand.
[0535] Based on the legacy MAC CE activation signal, use MAC CE to activate semi-persistent reporting on PUCCH, the UE would transmit a PUCCH with HARQ-ACK information in slot n corresponding to the PDSCH carrying the activation command, the indicated semi-persistent reporting setting can be applied starting from the first slot that is after slot where μ is the SCS configuration for the PUCCH.
[0536] To avoid a break in CSI feedback, the periodic or semi-persistent CSI reporting can be deactivated at least at one of the following time points:
[0537] (1) The slot where the NW transmits the MAC CE message.
[0538] (2) Slot n, where the UE transmit a PUCCH with HARQ-ACK information corresponding to the PDSCH carrying the activation command.
[0539] (3) slot
[0540] (4) slot where the first slot after slot
[0541] (5) The slot preceding the first semi-persistent CSI reporting.
[0542] (6) The slot where the first semi-persistent CSI reporting occurs.
[0543] Method 2: Assume that if a DCI is used to indicate the switching from AI-based CSI reporting to non-AI-based CSI reporting, the DCI message serves a dual purpose: it activates the semi-persistent or aperiodic non-AI-based CSI reporting on one hand, and deactivates the periodic or semi-persistent AI-based CSI reporting on the other hand.
[0544] Based on the legacy DCI activation signal, use DCI to activate semi-persistent or aperiodic reporting on PUSCH, the UE would transmit the first CSI reporting at the slot where and are the SCS for PUSCH and PUCCH, respectively, K2 is time offset, the unit is slot, which is based on the SCS of PUSCH.
[0545] To avoid a break in CSI feedback, the periodic or semi-persistent CSI reporting can be deactivated at least at one of the following time points:
[0546] (1) Slot n, where the NW transmits the DCI message.
[0547] (2) Slot Ks, where the UE transmit the first CSI reporting.
[0548] (3) Slot Ks-1, where the slot preceding the first semi-persistent CSI reporting.
[0549] This focuses on determining the appropriate switching time point when transitioning between AI-based and non-AI-based CSI reporting, as signaled by MAC CE or DCI. Since the activation or deactivation via MAC CE or DCI requires processing time, this embodiment proposes detailed methods to avoid interruption in CSI feedback. For MAC CE-triggered switching, several candidate deactivation time points are proposed around the activation slot and corresponding PUCCH transmissions. Similarly, for DCI-triggered switching, deactivation timing is aligned with the first CSI report slot or the slot just before it. These strategies ensure seamless continuity of CSI reporting during transition. This embodiment ensures smooth switching between AI-based and non-AI-based CSI reporting by aligning activation and deactivation timings, thereby avoiding feedback interruption. Each method described in this embodiment can be implemented independently or in combination with other methods within this embodiment or with other embodiments, allowing flexible deployment based on system requirements and UE capabilities.
[0550] Commercial interests for some embodiments are as follows. 1. Solve issues in the prior art and other issues. 2. Enhance a CSI prediction accuracy. 3. Improve a system capacity. 4. Reduce a computation complexity. 5. Reduce a reporting overhead. 6. Enhance a system robust. 7. Improve resource utilization. 8. Reduce a control signaling overhead. 9. Provide a good communication performance. 10. Provide high reliability. Some embodiments can incorporate the AI / ML based solution into the current protocols seamlessly with good backward compatibility. Some embodiments of the present disclosure can be used in many applications. Some embodiments of the present disclosure are used by chipset vendors, video system development vendors, automakers including cars, trains, trucks, buses, bicycles, moto-bikes, helmets, and etc., drones (unmanned aerial vehicles) , smartphone makers, communication devices for public safety use, AR / VR / MR device maker for example gaming, conference / seminar, education purposes. Some embodiments of the present disclosure are a combination of “techniques / processes” that can be adopted in video standards to create an end product. Some embodiments of the present disclosure propose technical mechanisms. The at least one proposed solution, method, system, and apparatus of some embodiments of the present disclosure may be used for current and / or new / future standards regarding communication systems such as a UE, a base station, and / or a communication system. Compatible products follow at least one proposed solution, method, system, and apparatus of some embodiments of the present disclosure. The proposed solution, method, system, and apparatus are widely used in a UE, a base station, and / or a communication system. With the implementation of the at least one proposed solution, method, system, and apparatus of some embodiments of the present disclosure, at least one modification to wireless communication methods and apparatus are considered for standardizing.
[0551] FIG. 7 is an example of a computing device 1100 according to an embodiment of the present disclosure. Any suitable computing device can be used for performing the operations described herein. For example, FIG. 7 illustrates an example of the computing device 1100 that can implement some embodiments of f using any suitably configured hardware and / or software. In some embodiments, the computing device 1100 can include a processor 1112 that is communicatively coupled to a memory 1114 and that executes computer-executable program code and / or accesses information stored in the memory 1114. The processor 1112 may include a microprocessor, an application-specific integrated circuit ( “ASIC” ) , a state machine, or other processing device. The processor 1112 can include any of a number of processing devices, including one. Such a processor can include or may be in communication with a computer-readable medium storing instructions that, when executed by the processor 1112, cause the processor to perform the operations described herein.
[0552] The memory 1114 can include any suitable non-transitory computer-readable medium. The computer-readable medium can include any electronic, optical, magnetic, or other storage device capable of providing a processor with computer-readable instructions or other program code. Non-limiting examples of a computer-readable medium include a magnetic disk, a memory chip, a read-only memory (ROM) , a random access memory (RAM) , an application specific integrated circuit (ASIC) , a configured processor, optical storage, magnetic tape or other magnetic storage, or any other medium from which a computer processor can read instructions. The instructions may include processor-specific instructions generated by a compiler and / or an interpreter from code written in any suitable computer-programming language, including, for example, C, C++, C#, visual basic, java, python, perl, javascript, and actionscript.
[0553] The computing device 1100 can also include a bus 1116. The bus 1116 can communicatively couple one or more components of the computing device 1100. The computing device 1100 can also include a number of external or internal devices such as input or output devices. For example, the computing device 1100 is illustrated with an input / output ( “I / O” ) interface 1118 that can receive input from one or more input devices 1120 or provide output to one or more output devices 1122. The one or more input devices 1120 and one or more output devices 1122 can be communicatively coupled to the I / O interface 1118. The communicative coupling can be implemented via any suitable manner (e.g., a connection via a printed circuit board, connection via a cable, communication via wireless transmissions, etc. ) . Non-limiting examples of input devices 1120 include a touch screen (e g., one or more cameras for imaging a touch area or pressure sensors for detecting pressure changes caused by a touch) , a mouse, a keyboard, or any other device that can be used to generate input events in response to physical actions by a user of a computing device. Non-limiting examples of output devices 1122 include a liquid crystal display (LCD) screen, an external monitor, a speaker, or any other device that can be used to display or otherwise present outputs generated by a computing device.
[0554] The computing device 1100 can execute program code that configures the processor 1112 to perform one or more of the operations described above with respect to some embodiments of FIG. 1 to FIG. 6B. The program code may be resident in the memory 1114 or any suitable computer-readable medium and may be executed by the processor 1112 or any other suitable processor.
[0555] The computing device 1100 can also include at least one network interface device 1124. The network interface device 1124 can include any device or group of devices suitable for establishing a wired or wireless data connection to one or more data networks 1128. Non limiting examples of the network interface device 1124 include an Ethernet network adapter, a modem, and / or the like. The computing device 1100 can transmit messages as electronic or optical signals via the network interface device 1124.
[0556] FIG. 8 is a block diagram of an example of a communication system 1200 according to an embodiment of the present disclosure. Embodiments described herein may be implemented into the communication system 1200 using any suitably configured hardware and / or software. FIG. 8 illustrates the communication system 1200 including a radio frequency (RF) circuitry 1210, a baseband circuitry 1220, an application circuitry 1230, a memory / storage 1240, a display 1250, a camera 1260, a sensor 1270, and an input / output (I / O) interface 1280, coupled with each other at least as illustrated.
[0557] The application circuitry 1230 may include a circuitry such as, but not limited to, one or more single-core or multi-core processors. The processors may include any combination of general-purpose processors and dedicated processors, such as graphics processors, application processors. The processors may be coupled with the memory / storage and configured to execute instructions stored in the memory / storage to enable various applications and / or operating systems running on the system. The communication system 1200 can execute program code that configures the application circuitry 1230 to perform one or more of the operations described above with respect to some embodiments of FIG. 1 to FIG. 6B. The program code may be resident in the application circuitry 1230 or any suitable computer-readable medium and may be executed by the application circuitry 1230 or any other suitable processor.
[0558] The baseband circuitry 1220 may include circuitry such as, but not limited to, one or more single-core or multi-core processors. The processors may include a baseband processor. The baseband circuitry may handle various radio control functions that may enable communication with one or more radio networks via the RF circuitry. The radio control functions may include, but are not limited to, signal modulation, encoding, decoding, radio frequency shifting, etc. In some embodiments, the baseband circuitry may provide for communication compatible with one or more radio technologies. For example, in some embodiments, the baseband circuitry may support communication with an evolved universal terrestrial radio access network (EUTRAN) and / or other wireless metropolitan area networks (WMAN) , a wireless local area network (WLAN) , a wireless personal area network (WPAN) . Embodiments in which the baseband circuitry is configured to support radio communications of more than one wireless protocol may be referred to as multi-mode baseband circuitry.
[0559] In various embodiments, the baseband circuitry 1220 may include circuitry to operate with signals that are not strictly considered as being in a baseband frequency. For example, in some embodiments, baseband circuitry may include circuitry to operate with signals having an intermediate frequency, which is between a baseband frequency and a radio frequency. The RF circuitry 1210 may enable communication with wireless networks using modulated electromagnetic radiation through a non-solid medium. In various embodiments, the RF circuitry may include switches, filters, amplifiers, etc. to facilitate the communication with the wireless network. In various embodiments, the RF circuitry 1210 may include circuitry to operate with signals that are not strictly considered as being in a radio frequency. For example, in some embodiments, RF circuitry may include circuitry to operate with signals having an intermediate frequency, which is between a baseband frequency and a radio frequency.
[0560] In various embodiments, the transmitter circuitry, control circuitry, or receiver circuitry discussed above with respect to some embodiments of FIG. 1 to FIG. 6B may be embodied in whole or in part in one or more of the RF circuitry, the baseband circuitry, and / or the application circuitry. As used herein, “circuitry” may refer to, be part of, or include an application specific integrated circuit (ASIC) , an electronic circuit, a processor (shared, dedicated, or group) , and / or a memory (shared, dedicated, or group) that execute one or more software or firmware programs, a combinational logic circuit, and / or other suitable hardware components that provide the described functionality. In some embodiments, the electronic device circuitry may be implemented in, or functions associated with the circuitry may be implemented by, one or more software or firmware modules. In some embodiments, some or all of the constituent components of the baseband circuitry, the application circuitry, and / or the memory / storage may be implemented together on a system on a chip (SOC) . The memory / storage 1240 may be used to load and store data and / or instructions, for example, for system. The memory / storage for one embodiment may include any combination of suitable volatile memory, such as dynamic random access memory (DRAM) ) , and / or non-volatile memory, such as flash memory.
[0561] In various embodiments, the I / O interface 1280 may include one or more user interfaces designed to enable user interaction with the system and / or peripheral component interfaces designed to enable peripheral component interaction with the system. User interfaces may include, but are not limited to a physical keyboard or keypad, a touchpad, a speaker, a microphone, etc. Peripheral component interfaces may include, but are not limited to, a non-volatile memory port, a universal serial bus (USB) port, an audio jack, and a power supply interface. In various embodiments, the sensor 1270 may include one or more sensing devices to determine environmental conditions and / or location information related to the system. In some embodiments, the sensors may include, but are not limited to, a gyro sensor, an accelerometer, a proximity sensor, an ambient light sensor, and a positioning unit. The positioning unit may also be part of, or interact with, the baseband circuitry and / or RF circuitry to communicate with components of a positioning network, e.g., a global positioning system (GPS) satellite.
[0562] In various embodiments, the display 1250 may include a display, such as a liquid crystal display and a touch screen display. In various embodiments, the communication system 1200 may be a mobile computing device such as, but not limited to, a laptop computing device, a tablet computing device, a netbook, an ultrabook, a smartphone, an AR / VR glasses, etc. In various embodiments, system may have more or less components, and / or different architectures. Where appropriate, methods described herein may be implemented as a computer program. The computer program may be stored on a storage medium, such as a non-transitory storage medium.
[0563] A person having ordinary skill in the art understands that each of the units, algorithm, and steps described and disclosed in the embodiments of the present disclosure are realized using electronic hardware or combinations of software for computers and electronic hardware. Whether the functions run in hardware or software depends on the condition of application and design requirement for a technical plan. A person having ordinary skill in the art can use different ways to realize the function for each specific application while such realizations should not go beyond the scope of the present disclosure. It is understood by a person having ordinary skill in the art that he / she can refer to the working processes of the system, device, and unit in the above-mentioned embodiment since the working processes of the above-mentioned system, device, and unit are basically the same. For easy description and simplicity, these working processes will not be detailed.
[0564] It is understood that the disclosed system, device, and method in the embodiments of the present disclosure can be realized with other ways. The above-mentioned embodiments are exemplary only. The division of the units is merely based on logical functions while other divisions exist in realization. It is possible that a plurality of units or components are combined or integrated in another system. It is also possible that some characteristics are omitted or skipped. On the other hand, the displayed or discussed mutual coupling, direct coupling, or communicative coupling operate through some ports, devices, or units whether indirectly or communicatively by ways of electrical, mechanical, or other kinds of forms.
[0565] The units as separating components for explanation are or are not physically separated. The units for display are or are not physical units, that is, located in one place or distributed on a plurality of network units. Some or all of the units are used according to the purposes of the embodiments. Moreover, each of the functional units in each of the embodiments can be integrated in one processing unit, physically independent, or integrated in one processing unit with two or more than two units. If the software function unit is realized and used and sold as a product, it can be stored in a readable storage medium in a computer. Based on this understanding, the technical plan proposed by the present disclosure can be importantly or partially realized as the form of a software product. Or, one part of the technical plan beneficial to the conventional technology can be realized as the form of a software product. The software product in the computer is stored in a storage medium, including a plurality of commands for a computational device (such as a personal computer, a server, or a network device) to run all or some of the steps disclosed by the embodiments of the present disclosure. The storage medium includes a USB disk, a mobile hard disk, a read-only memory (ROM) , a random access memory (RAM) , a floppy disk, or other kinds of media capable of storing program codes.
[0566] While the present disclosure has been described in connection with what is considered the most practical and preferred embodiments, it is understood that the present disclosure is not limited to the disclosed embodiments but is intended to cover various arrangements made without departing from the scope of the broadest interpretation of the appended claims.
Claims
1.A wireless communication method performed by a user equipment (UE) , comprising:reporting, to a base station, a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting;receiving a CSI reporting configuration information from the base station;receiving channel measurement signals from the base station;reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information;reporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting; andreceiving a life cycle management (LCM) indication information from the base station.2.The wireless communication method according to claim 1, wherein the processor comprises at least one CSI processing unit for AI-based CSI reporting.3.The wireless communication method according to claim 1 or 2, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.4.The wireless communication method according to any one of claims 1 to 3, further comprising:receiving a monitoring indication information from the base station, wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following:a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation;a field indicating a duration of performance monitoring result calculation;a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.5.The wireless communication method according to claim 4, wherein the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.6.The wireless communication method according to any one of claims 1 to 5, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the UE triggers reporting of performance monitoring results.7.The wireless communication method according to any one of claims 1 to 6, wherein a reporting period and / or a monitoring window length for performance monitoring results are configurable via a DCI, a MAC CE message, or a radio resource control (RRC) message; and / ora reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message.8.The wireless communication method according to any one of claims 1 to 7, wherein the UE calculates a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and selects M performance monitoring results from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1.9.The wireless communication method according to claim 8, wherein M is configured by the base station or predefined.10.The wireless communication method according to claim 8 or 9, wherein the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting.11.The wireless communication method according to claim 9 or 10, wherein when the UE uses differential reporting in the performance monitoring results, the UE uses a priority to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value.12.The wireless communication method according to any one of claims 1 to 11, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following:periodic reporting;semi-persistent reporting;aperiodic reporting;event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold.13.The wireless communication method according to any one of claims 1 to 12, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises:reporting, to the base station, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following:a bitmap representation;a combined digital representation;a reported CSI index.14.The wireless communication method according to any one of claims 1 to 13, wherein the UE calculates a SGCS or NMSE across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and reports, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1.15.The wireless communication method according to claim 14, wherein the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.16.The wireless communication method according to any one of claims 1 to 15, further comprising:receiving a downlink data information from the base station.17.The wireless communication method according to any one of claims 1 to 16, further comprising determining predefined rules supporting AI-based CSI reporting, wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following:a priority of a CSI reporting at different LCM stages;a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model;a processor usage and / or a CMU usage.18.The wireless communication method according to claim 17, wherein the priority of the CSI reporting at the different LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.19.The wireless communication method according to claim 18, wherein the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting.20.The wireless communication method according to claim 18 or 19, wherein the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the CSI reporting.21.The wireless communication method according to claim 20, wherein the coefficient factor is equal to a maximum value of the first parameter plus 1.22.The wireless communication method according to any one of claims 18 to 21, wherein values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following:for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ;for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI report reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR.23.The wireless communication method according to any one of claims 17 to 22, wherein the processing time is managed by at least one of the following:adding an activation time to a first reference time and a second reference time for CSI reporting;stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value.24.The wireless communication method according to any one of claims 17 to 23, wherein the processor usage and / or the CMU usage is managed by at least one of the following:if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported, and / or if sufficient processing resources are available, the UE updates a requested CSI report with a lower priority than the performance monitoring results.25.The wireless communication method according to any one of claims 17 to 24, wherein the UE determines the CSI reporting based on at least one of the following:calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting;prioritizing CSI reporting based on a CMU availability and / or a CPU availability.26.The wireless communication method according to any one of claims 17 to 25, wherein the UE switches between an AI-based CSI reporting and a non-AI-based CSI reporting via a MAC CE message or a DCI; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the MAC CE message or the DCI to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting.27.The wireless communication method according to claim 26, wherein the MAC CE message or the DCI comprises at least one of the following:a field used to indicate an applicable CSI reporting configuration ID;a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting;a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.28.A wireless communication method performed by a user equipment (UE) , comprising:determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; andreporting, to the base station, performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.29.The wireless communication method according to claim 28, wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following:the priority of the CSI reporting at different LCM stages;a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model;a processor usage and / or a CSI memory unit (CMU) usage.30.The wireless communication method according to claim 28 or 29, wherein a priority of the CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.31.The wireless communication method according to any one of claims 28 to 30, wherein the processor comprises at least one CSI processing unit for AI-based CSI reporting.32.The wireless communication method according to any one of claims 28 to 31, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.33.The wireless communication method according to any one of claims 28 to 32, wherein the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting.34.The wireless communication method according to according to any one of claims 28 to 33, wherein the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the performance monitoring results.35.The wireless communication method according to claim 34, wherein the coefficient factor is equal to a maximum value of the first parameter plus 1.36.The wireless communication method according to any one of claims 28 to 35, wherein values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following:for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ;for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI report reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR.37.A wireless communication method performed by a user equipment (UE) , comprising:determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting;reporting, to the base station, performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; andreceiving a life cycle management (LCM) indication information from the base station;wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.38.The wireless communication method according to claim 37, wherein the processing time is managed by at least one of the following:adding an activation time to a first reference time and a second reference time for CSI reporting;stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value.39.The wireless communication method according to claim 37 or 38, wherein the processor usage and / or the CMU usage is managed by at least one of the following:if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported and / or if sufficient processing resources are available, the UE updates a requested CSI report with a lower priority than the performance monitoring results.40.The wireless communication method according to any one of claims 37 to 39, wherein the UE determines the CSI reporting based on at least one of the following:calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting;prioritizing CSI reporting based on a CMU availability and / or a CPU availability.41.The wireless communication method according to any one of claims 37 to 40, wherein for AI-based CSI prediction, if a number of CSI-RS resources is larger than 12, an occupy CPU ∈ {8, 12, 16, 18, 20} ; if the number of CSI-RS resources is less than 12, OCPU=Y1·K, where Y1∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by a UE capability indication, K is a parameter; if N4>1, OCPU=max (Y2·N4, 4) and Y2∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by the UE capability indication.42.The wireless communication method according to any one of claims 37 to 41, wherein the UE switches between an AI-based CSI reporting and a non-AI-based CSI reporting via a downlink control information (DCI) or a medium access control-control element (MAC CE) message; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the DCI or the MAC CE message to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting.43.The wireless communication method according to any one of claims 37 to 42, wherein an AI-based CSI reporting fallback or deactivate time point is located before a first non-AI-based CSI reporting.44.The wireless communication method according to any one of claims 37 to 43, wherein the MAC CE message or the DCI comprises at least one of the following:a field used to indicate an applicable CSI reporting configuration ID;a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting;a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.45.A wireless communication method performed by a user equipment (UE) , comprising:receiving a monitoring indication information from a base station; andreporting, to the base station, performance monitoring results for AI-based CSI reporting;wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following:a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation;a field indicating a duration of performance monitoring result calculation;a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.46.The wireless communication method according to claim 45, wherein the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.47.A wireless communication method performed by a user equipment (UE) , comprising:receiving a channel state information (CSI) reporting configuration information from the base station;receiving channel measurement signals from the base station;reporting, to the base station, CSI measurement and / or prediction results based on the CSI reporting configuration information;receiving monitoring measurement signals from the base station; andreporting, to the base station, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.48.The wireless communication method according to claim 47, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.49.The wireless communication method according to claim 47 or 48, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the UE triggers reporting of performance monitoring results.50.The wireless communication method according to any one of claims 47 to 49, wherein a reporting period and / or a monitoring window length for performance monitoring results are configurable via a downlink control information (DCI) , a medium access control-control element (MAC CE) message, or a radio resource control (RRC) message; and / ora reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message.51.The wireless communication method according to any one of claims 47 to 50, wherein the UE calculates a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and selects M performance monitoring results from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1.52.The wireless communication method according to claim 51, wherein M is configured by the base station or predefined.53.The wireless communication method according to claim 51 or 52, wherein the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting.54.The wireless communication method according to claim 52 or 53, wherein when the UE uses differential reporting in the performance monitoring results, the UE uses a priority to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value.55.The wireless communication method according to any one of claims 47 to 54, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following:periodic reporting;semi-persistent reporting;aperiodic reporting;event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold.56.The wireless communication method according to any one of claims 47 to 55, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises:reporting, to the base station, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following:a bitmap representation;a combined digital representation;a reported CSI index.57.The wireless communication method according to any one of claims 47 to 56, wherein the UE calculates a SGCS or NMSE across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and reports, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1.58.The wireless communication method according to claim 57, wherein the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.59.A wireless communication method performed by a base station, comprising:receiving, from a user equipment (UE) , a support of a number of processors and / or channel state information (CSI) memory units (CMUs) for artificial intelligence (AI) -based CSI reporting;transmitting, to the UE, a CSI reporting configuration information;transmitting, to the UE, channel measurement signals;receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information;receiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting; andtransmitting, to the UE, a life cycle management (LCM) indication information.60.The wireless communication method according to claim 59, wherein the processor comprises at least one CSI processing unit for AI-based CSI reporting.61.The wireless communication method according to claim 59 or 60, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.62.The wireless communication method according to any one of claims 59 to 61, further comprising:transmitting, to the UE, a monitoring indication information, wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following:a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation;a field indicating a duration of performance monitoring result calculation;a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.63.The wireless communication method according to claim 62, wherein the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.64.The wireless communication method according to any one of claims 59 to 63, wherein receiving, from the UE, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the base station receives performance monitoring results from the UE.65.The wireless communication method according to any one of claims 59 to 64, wherein a reporting period and / or a monitoring window length for performance monitoring results are configurable via a DCI, a MAC CE message, or a radio resource control (RRC) message; and / ora reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message.66.The wireless communication method according to any one of claims 59 to 65, wherein a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) is calculated across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and M performance monitoring results are selectd from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1.67.The wireless communication method according to claim 66, wherein M is configured by the base station or predefined.68.The wireless communication method according to claim 66 or 67, wherein the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting.69.The wireless communication method according to claim 67 or 68, wherein when differential reporting is applied in the performance monitoring results, a priority is applied to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value.70.The wireless communication method according to any one of claims 59 to 69, wherein receiving, from the UE, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following:periodic reporting;semi-persistent reporting;aperiodic reporting;event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold.71.The wireless communication method according to any one of claims 59 to 70, wherein receiving, from the UE, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises:reporting, to the base station, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following:a bitmap representation;a combined digital representation;a reported CSI index.72.The wireless communication method according to any one of claims 59 to 71, wherein a SGCS or NMSE is calculated across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and is reported, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1.73.The wireless communication method according to claim 72, wherein the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.74.The wireless communication method according to any one of claims 59 to 73, further comprising:transmitting, to the UE, a downlink data information.75.The wireless communication method according to any one of claims 59 to 74, further comprising determining predefined rules supporting AI-based CSI reporting, wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following:a priority of a CSI reporting at different LCM stages;a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model;a processor usage and / or a CMU usage.76.The wireless communication method according to claim 75, wherein the priority of the CSI reporting at the different LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.77.The wireless communication method according to claim 76, wherein the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting.78.The wireless communication method according to claim 76 or 77, wherein the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the CSI reporting.79.The wireless communication method according to claim 78, wherein the coefficient factor is equal to a maximum value of the first parameter plus 1.80.The wireless communication method according to any one of claims 76 to 79, wherein values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following:for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ;for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI report reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR.81.The wireless communication method according to any one of claims 75 to 80, wherein the processing time is managed by at least one of the following:adding an activation time to a first reference time and a second reference time for CSI reporting;stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value.82.The wireless communication method according to any one of claims 75 to 81, wherein the processor usage and / or the CMU usage is managed by at least one of the following:if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported, and / or if sufficient processing resources are available, a requested CSI report with a lower priority than the performance monitoring results is updated.83.The wireless communication method according to any one of claims 75 to 82, wherein the CSI reporting is determined based on at least one of the following:calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting;prioritizing CSI reporting based on a CMU availability and / or a CPU availability.84.The wireless communication method according to any one of claims 75 to 83, wherein an AI-based CSI reporting and a non-AI-based CSI reporting is switched between via a MAC CE message or a DCI; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the MAC CE message or the DCI to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting.85.The wireless communication method according to claim 84, wherein the MAC CE message or the DCI comprises at least one of the following:a field used to indicate an applicable CSI reporting configuration ID;a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting;a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.86.A wireless communication method performed by a base station, comprising:determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting; andreceiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on a priority of a CSI reporting at different life cycle management (LCM) stages, wherein the priority of the CSI reporting at the LCM stages comprises at least a first parameter and / or a second parameter to differentiate types of the CSI reporting.87.The wireless communication method according to claim 86, wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following:the priority of the CSI reporting at different LCM stages;a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model;a processor usage and / or a CSI memory unit (CMU) usage.88.The wireless communication method according to claim 86 or 87, wherein a priority of the CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.89.The wireless communication method according to any one of claims 86 to 88, wherein the processor comprises at least one CSI processing unit for AI-based CSI reporting.90.The wireless communication method according to any one of claims 86 to 89, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.91.The wireless communication method according to any one of claims 86 to 90, wherein the types of the CSI reporting comprise an AI-based CSI reporting and a non-AI CSI reporting.92.The wireless communication method according to according to any one of claims 86 to 91, wherein the priority of the CSI reporting at the different LCM stages comprises a coefficient factor; at least one of the first parameter, the second parameter, and the coefficient factor is used to determine a reporting priority of the CSI reporting and / or a reporting priority of the performance monitoring results.93.The wireless communication method according to claim 92, wherein the coefficient factor is equal to a maximum value of the first parameter plus 1.94.The wireless communication method according to any one of claims 86 to 93, wherein values of the first parameter and / or the second parameter in the CSI reporting are defined for at least one of the following:for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying a layer 1 reference signal received power (L1-RSRP) or a layer 1 signal to interference and noise ratio (L1-SINR) ;for event-triggered CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI report reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting not carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting carrying the L1-RSRP or the L1-SINR;for CSI reporting carrying performance monitoring results, where the performance monitoring results are for CSI reporting not carrying the L1-RSRP or the L1-SINR;for event-triggered CSI reporting carrying the L1-RSRP or the L1-SINR.95.A wireless communication method performed by a base station, comprising:determining predefined rules supporting artificial intelligence (AI) -based channel state information (CSI) reporting;receiving, from a user equipment (UE) , performance monitoring results for AI-based CSI reporting based on the predefined rules supporting AI-based CSI reporting; andtransmitting, to the UE, a life cycle management (LCM) indication information;wherein the predefined rules supporting AI-based CSI reporting comprise at least one of the following: a priority of a CSI reporting at different LCM stages; a processing time corresponding to the CSI reporting at the different LCM stages of an AI / machine learning (ML) model; a processor usage and / or a CSI memory unit (CMU) usage.96.The wireless communication method according to claim 95, wherein the processing time is managed by at least one of the following:adding an activation time to a first reference time and a second reference time for CSI reporting;stipulating that an offset for delivery of aperiodic channel state information reference signal (CSI-RS) resources during an aperiodic CSI reporting is greater than a predetermined value.97.The wireless communication method according to claim 95 or 96, wherein the processor usage and / or the CMU usage is managed by at least one of the following:if the CSI reporting for performance monitoring corresponds to model inference results that are not available or not reported and / or if sufficient processing resources are available, the UE updates a requested CSI report with a lower priority than the performance monitoring results.98.The wireless communication method according to any one of claims 95 to 97, wherein the CSI reporting is determined based on at least one of the following:calculating a number of unoccupied CMUs and determining whether the unoccupied CMUs are sufficient to support the CSI reporting;prioritizing CSI reporting based on a CMU availability and / or a CPU availability.99.The wireless communication method according to any one of claims 95 to 98, wherein for AI-based CSI prediction, if a number of CSI-RS resources is larger than 12, an occupy CPU ∈ {8, 12, 16, 18, 20} ; if the number of CSI-RS resources is less than 12, OCPU=Y1·K, where Y1∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by a UE capability indication, K is a parameter; if N4>1, OCPU=max (Y2·N4, 4) and Y2∈ {1 / 4, 1 / 3, 1 / 2, 1, 2, 3, 4, 5, 6} is reported by the UE capability indication.100.The wireless communication method according to any one of claims 95 to 99, wherein an AI-based CSI reporting and a non-AI-based CSI reporting is switched between via a downlink control information (DCI) or a medium access control-control element (MAC CE) message; and / or the LCM indication information is used to indicate the UE to perform an LCM operation, the LCM operation comprises a fallback operation, and the fallback operation is indicated through the DCI or the MAC CE message to activate or deactivate non-AI-based CSI reporting and / or deactivate AI-based CSI reporting.101.The wireless communication method according to any one of claims 95 to 100, wherein an AI-based CSI reporting fallback or deactivate time point is located before a first non-AI-based CSI reporting.102.The wireless communication method according to any one of claims 95 to 101, wherein the MAC CE message or the DCI comprises at least one of the following:a field used to indicate an applicable CSI reporting configuration ID;a field used to indicate a type of codebook for AI-based CSI reporting or non-AI-based CSI reporting;a field used to indicate activate or deactivate the CSI reporting, wherein the CSI reporting comprises at least one of following information: the performance monitoring results, AI-based CSI measurement results, AI-based CSI prediction results, non-AI-based CSI measurement results, non-AI-based CSI prediction results.103.A wireless communication method performed by a base station, comprising:transmitting, a user equipment (UE) , a monitoring indication information; andreceiving, from the UE, performance monitoring results for AI-based CSI reporting;wherein the monitoring indication information is used to indicate the UE to perform performance monitoring result calculation and comprises at least one of the following:a 1-bit field indicating whether to trigger the performance monitoring result calculation, where a first value of the 1-bit field activates the performance monitoring result calculation, and a second value of the 1-bit field results in no action or second value of the 1-bit field deactivates the performance monitoring result calculation;a field indicating a duration of performance monitoring result calculation;a G-bit field indicating whether at least one of G UE groups requires the performance monitoring result calculation, where each bit of the G-bit field indicates whether an associated UE group requires the performance monitoring result calculation, and each bit the G-bit field corresponds to the associated UE group.104.The wireless communication method according to claim 103, wherein the monitoring indication information is carried by a downlink control information (DCI) or a medium access control-control element (MAC CE) message, and the DCI or the MAC CE message is UE-specific, UE group-specific, or cell-level.105.A wireless communication method performed by a base station, comprising:transmitting, to a user equipment (UE) , a channel state information (CSI) reporting configuration information; transmitting, to the UE, channel measurement signals;receiving, from the UE, CSI measurement and / or prediction results based on the CSI reporting configuration information;transmitting, to the UE, monitoring measurement signals; andreceiving, from the UE, performance monitoring results for AI-based CSI reporting based on a predefined reporting priority, wherein a priority of a CSI reporting is higher than a priority of a performance monitoring reporting for AI-based CSI reporting.106.The wireless communication method according to claim 105, wherein the CSI reporting of the AI-based CSI reporting comprises an AI-based CSI prediction, an AI-based beam management, an AI-based CSI compression, or an AI-based CSI prediction and compression.107.The wireless communication method according to claim 105 or 106, wherein reporting, to the base station, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is event-driven, and when one or more performance monitoring results for the AI-based CSI reporting are less than or greater than a preset threshold, the UE triggers reporting of performance monitoring results.108.The wireless communication method according to any one of claims 105 to 107, wherein a reporting period and / or a monitoring window length for performance monitoring results are configurable via a downlink control information (DCI) , a medium access control-control element (MAC CE) message, or a radio resource control (RRC) message; and / ora reporting period for a periodic reporting or a semi-persistent reporting for performance monitoring results is configured by the RRC message.109.The wireless communication method according to any one of claims 105 to 108, wherein a squared generalized cosine similarity (SGCS) or a normalized mean squared error (NMSE) is calculated across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and M performance monitoring results are selected from the performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window to report as a first part of the AI-based CSI reporting, where M is greater than or equal to 1.110.The wireless communication method according to claim 109, wherein M is configured by the base station or predefined.111.The wireless communication method according to claim 109 or 110, wherein the M monitoring results with the smallest SGCS or NMSE values of the AI-based CSI reporting are mapped according to a time-domain order of the AI-based CSI reporting.112.The wireless communication method according to claim 110 or 111, wherein when differential reporting is applied in the performance monitoring results, a priority is applied to map a position information of the AI-based CSI reporting corresponding to a maximum or minimum SGCS or NMSE value.113.The wireless communication method according to any one of claims 105 to 112, wherein receiving, from the UE, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority is based on at least one of the following:periodic reporting;semi-persistent reporting;aperiodic reporting;event-triggered reporting when performance monitoring metrics of the performance monitoring results exceed a predefined threshold.114.The wireless communication method according to any one of claims 105 to 113, wherein receiving, from the UE, the performance monitoring results for AI-based CSI reporting based on the predefined reporting priority comprises:receiving, from the UE, a location information of SGCS or NMSE values associated with a reported CSI for AI-based CSI reporting using at least one of the following:a bitmap representation;a combined digital representation;a reported CSI index.115.The wireless communication method according to any one of claims 105 to 114, wherein a SGCS or NMSE is calculated across sub-bands and / or layers for the AI-based CSI reporting in one time occasion, and is reported, as a first part of the AI-based CSI reporting, comparison results of M performance monitoring results with smallest SGCS or NMSE values of the AI-based CSI reporting within a monitoring window against a threshold, where M is greater than or equal to 1.116.The wireless communication method according to claim 115, wherein the comparison results are mapped according to a time-domain order of the AI-based CSI reporting.117.A user equipment (UE) , comprising:a memory;a transceiver; anda processor coupled to the memory and the transceiver;wherein the UE is configured to perform any one of claims 1 to 58.118.A base station, comprising:a memory;a transceiver; anda processor coupled to the memory and the transceiver;wherein the network is configured to perform any one of claims 59 to 116.