Channel state information processing method based on AI / ML and wireless communication equipment

Abstract

The embodiment of the invention provides a channel state information (CSI) processing method based on artificial intelligence (AI) / machine learning (ML). And the base station receives the AI / ML-based CSI processing capability reported by user equipment (UE). The base station determines an AI / ML-based CSI measurement configuration based on the AI / ML-based CSI processing capability. The CSI measurement configuration based on the AI / ML comprises one or more of the following items: configuration of CSI resource scheduling, configuration of CSI reporting and configuration of CSI measurement triggering.

Description

Technical Field

[0001] This application relates to the field of communication systems, and more particularly to a channel state information (CSI) processing method and wireless communication device based on artificial intelligence (AI) / machine learning (ML). Background Technology

[0002] 3GPP RAN1 TR38.843 comprehensively discusses the integration of AI and ML into the NR air interface. These techniques are used to enhance CSI feedback, thereby improving the accuracy of time-domain CSI prediction, beam management, and even positioning. Furthermore, the application of AI / ML methods in areas such as load balancing and Radio Resource Management (RRM) algorithms has been discussed in RAN3 / RAN2. These standardized research efforts consistently demonstrate significant performance improvements compared to non-AI / ML processing.

[0003] Technical issues CSI (Computer-Assisted Signal Processing) plays a crucial role in supporting Multiple-Input Multiple-Output (MIMO) and Radio Resource Management (RRM) algorithms, significantly contributing to system throughput and spectral efficiency. The introduction of AI is no exception, potentially reshaping CSI design. For example, AI-related information (including prediction accuracy) needs to be integrated into the traditional CSI reporting process, while also considering the time required for CSI measurement and computation. In traditional communication systems, consistency in CSI measurement and reporting behavior between the base station and user equipment (UE) is typically ensured by defining the CSI processing unit (CPU) and the latency requirements for different types of CSI computations. Summary of the Invention

[0004] The purpose of this disclosure is to provide a wireless communication device (e.g., a user equipment (UE) or a base station) and a channel state information (CSI) processing method based on artificial intelligence (AI) / machine learning (ML).

[0005] In a first aspect, one embodiment of the present invention provides an AI / ML-based CSI processing method, executed by a base station, comprising: The ability to receive CSI reports from UEs based on AI / ML; and The AI / ML-based CSI measurement configuration is determined based on the AI / ML-based CSI processing capability, wherein the AI / ML-based CSI measurement configuration includes one or more of the following: CSI resource scheduling configuration, CSI reporting configuration, and CSI measurement triggering configuration.

[0006] In a second aspect, one embodiment of the present invention provides a base station including a processor configured to invoke and run a computer program stored in a memory to cause a device in which the processor resides to perform the disclosed method.

[0007] In a third aspect, one embodiment of the present invention provides an AI / ML-based CSI processing method, executed by a UE, comprising: The UE reports its CSI processing capabilities based on AI / ML; Receives AI / ML-based CSI measurement configuration from the base station, wherein the AI / ML-based CSI measurement configuration is based on the AI / ML-based CSI processing capability; and The first operation is performed according to the AI / ML-based CSI measurement configuration; The AI / ML-based CSI measurement configuration includes one or more of the following: CSI resource scheduling configuration, CSI reporting configuration, and CSI measurement triggering configuration.

[0008] In a fourth aspect, one embodiment of the present invention provides a user equipment (UE) including a processor configured to invoke and run a computer program stored in a memory to cause the device in which the processor resides to perform the disclosed method.

[0009] In a fifth aspect, one embodiment of the present invention provides an AI / ML-based CSI processing method, which can be executed in a UE, comprising: Receive AI / ML-based CSI measurement configuration and AI / ML-based CSI reference signal; Wherein, the time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time; and The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability.

[0010] In a sixth aspect, one embodiment of the present invention provides a user equipment (UE) including a processor configured to invoke and run a computer program stored in a memory to cause the device in which the processor resides to perform the disclosed method.

[0011] In a seventh aspect, one embodiment of the present invention provides an AI / ML-based CSI processing method, executed by a base station, comprising: The base station sends an AI / ML-based CSI measurement configuration and an AI / ML-based CSI reference signal to enable the UE to perform CSI measurements based on the AI / ML-based CSI measurement configuration and the AI / ML-based CSI reference signal. The time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time. The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability. Receive the results reported by the CSI.

[0012] In an eighth aspect, one embodiment of the present invention provides a base station including a processor configured to invoke and run a computer program stored in a memory to cause a device in which the processor resides to perform the disclosed method.

[0013] The disclosed method can be implemented in a chip. The chip may include a processor configured to invoke and run a computer program stored in memory to cause the device in which the chip resides to perform the disclosed method.

[0014] The disclosed method can be programmed as computer-executable instructions stored in a non-transitory computer-readable medium. When the non-transitory computer-readable medium is loaded into a computer, it instructs the computer's processor to execute the disclosed method.

[0015] The non-transitory computer-readable medium may include at least one of the following: hard disk, read-only optical disk (CD-ROM), optical storage device, magnetic storage device, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and flash memory.

[0016] The disclosed method can be programmed into a computer program product that causes a computer to execute the disclosed method.

[0017] The disclosed method can be programmed into a computer program that causes a computer to execute the disclosed method. Attached Figure Description

[0018] To more clearly illustrate the embodiments or related technologies of this disclosure, the accompanying drawings involved in the embodiments are briefly described below. Obviously, the drawings are only some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0019] Figure 1 The diagram illustrates the CSI processing guidelines in the NR system and the CSI measurement and calculation time.

[0020] Figure 2 A schematic diagram of a wireless communication system including user equipment (UE), base stations, and network entities is shown.

[0021] Figure 3 A schematic diagram of an AI / ML functional framework system for implementing model management methods based on machine learning (ML) models is shown.

[0022] Figure 4 A schematic diagram of the overall solution of the disclosed method is shown.

[0023] Figure 5 A schematic diagram of one embodiment of the disclosed method is shown.

[0024] Figure 6 A schematic diagram of another embodiment of the disclosed method is shown.

[0025] Figure 7 The diagram illustrates the timing sequence of CSI reporting, AI / ML model activation, and AI / ML model inference in one example.

[0026] Figure 8 A schematic diagram of the CSI reporting, AI / ML model activation, and AI / ML model inference timing is shown in another example.

[0027] Figure 9 A schematic diagram of a wireless communication system according to an embodiment of the present disclosure is shown. Detailed Implementation

[0028] The embodiments of this disclosure will be described in detail with reference to the accompanying drawings, covering aspects such as technical content, structural features, implementation objectives, and effects. It should be noted that the terminology used in the embodiments of this disclosure is only used to describe the purpose of specific embodiments and is not intended to limit this disclosure.

[0029] The abbreviations used in this specification are as follows: Table 1

[0030] Embodiments of this disclosure relate to artificial intelligence (AI) and machine learning (ML) in the new radio (NR) air interface, and address AI / ML-specific CSI processing and computational latency issues, aiming to enable AI models to operate normally in the network while achieving extremely low signaling interaction between the gNB and the UE.

[0031] This disclosure presents a novel framework for CSI measurement using AI / ML models. The framework covers reference signal configuration, CSI measurement, and CSI reporting. This approach offers several advantages, including reduced signaling overhead, good scalability, efficient resource allocation based on different model requirements, and the unification of single-sided and two-sided model processing into a single workflow by introducing an AI model identifier (ID).

[0032] The embodiments of this disclosure provide: A set of CSI processing guidelines for AI / ML models, including defining CPU N for AI / ML models. CPU,AI / ML and the CPU O occupied CPU,AI / ML And the CSI processing criteria for the UE when AI / ML-based and non-AI / ML-based CSI processing coexist on the UE.

[0033] A set of CSI measurement timings and configurations specifically for AI / ML model inference, including T proc,CSI,AI / ML and T' proc,CSI,AI / ML This covers aspects such as activation time and inference time of AI / ML models, as well as modifications to the existing CSI measurement time estimation process.

[0034] Define Z Ref,AI / ML The cyclic prefix (CP) starts at time T after the last symbol of the PDCCH that triggered the CSI report ends. proc,CSI,AI / ML =(Z)(2048+144) κ2 -μ T c +T switch The next upline symbol. Define Z' Ref,AI / ML The CP start time is T' after the last symbol in the following items ends in time. proc,CSI,AI / ML =(Z')(2048+144) κ2 -μ The next uplink symbol of T_c: Aperiodic CSI-RS resource for channel measurements, Aperiodic CSI-IM resource for interference measurements, and Aperiodic NZP CSI-RS resource for interference measurements. When the Aperiodic CSI-RS is used for channel measurements triggered for the nth CSI report, Z, Z', and T... switch Defined in 3GPP TS38.214.

[0035] The T proc,CSI,AI / ML The time difference between the last symbol of the PDCCH that triggers the CSI report and the next uplink symbol after the end of the last symbol of that PDCCH; the T' proc,CSI,AI / MLThe time difference between the last symbol of the reference signal used for CSI measurement and reporting and the next uplink symbol after the last symbol of the reference signal ends.

[0036] The definitions that may be used in this specification include: CSI Processing Unit (CPU): This term refers to the system's ability to perform CSI measurements and calculations simultaneously.

[0037] Global AI / ML Model ID: This identifier is used to identify an AI / ML model. It can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, or an AI / ML management platform-specific unique ID.

[0038] Logical AI / ML Model ID: This identifier is used to identify the AI / ML model used in the network and has a certain mapping relationship with the global ID. The logical ID can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, a cell-specific ID, a link-specific ID, a TA-specific ID, a CU-specific ID, a DU-specific ID, a UPF-specific ID, an AMF-specific ID, an RRC-specific ID, or a network slice-specific ID. "Cell-specific" means that each AI / ML model has a unique ID within a specific cell. Similarly, "link-specific," "TA-specific," "CU-specific," "DU-specific," "UPF-specific," "AMF-specific," and "network slice-specific" represent unique IDs for AI / ML models within a specific context in the network.

[0039] AI / ML model related information: This information includes one or more of the following: global AI / ML model ID, provider, scenario, features, functions, version, accuracy, RRC descriptor, and AI / ML model descriptor.

[0040] Scene: The scenes of AI / ML models can include, but are not limited to, indoor, outdoor, flight, underwater, and movement speed.

[0041] Features: AI / ML model features may include, but are not limited to, CSI compression, CSI prediction, beam management, positioning, handover, and radio resource management.

[0042] Functionality: The functions of the AI / ML model may include, but are not limited to, CSI compression, CSI prediction, beam management, localization, handover, radio resource management, and / or the usage methods of the AI / ML model. These usage methods may include monitoring, inference, and / or training.

[0043] RRC Descriptor: The RRC descriptor includes reference signal configuration requirements and CSI measurement requirements for AI / ML models, which may include at least one of the following: reference signal time requirements (e.g., period, time step, frequency hopping mechanism, etc.), reference signal frequency requirements (e.g., frequency resources, frequency hopping mechanism, etc.), number of reference signal antenna ports, and CSI content (e.g., channel matrix, eigenvector, etc.).

[0044] AI / ML Model Descriptor: This term represents a detailed description of the attributes of an AI / ML model. This description may include, but is not limited to, AI / ML type, accuracy, input data requirements, output data, monitoring methods, input data distribution, and output data distribution.

[0045] For the sake of simplicity, the terms AI / ML model, AI model, ML model, and model are used interchangeably in this specification.

[0046] 1. CSI handling guidelines for non-AI / ML applications See Figure 1 The UE (e.g., UE 10) indicates the number N of simultaneous CSI measurements and calculations it supports via the parameter simultaneousCSI-ReportsPerCC in one component carrier and the parameter simultaneousCSI-ReportsAllCC across all component carriers. CPU If the UE supports N CPU If multiple CSI measurements and calculations are performed simultaneously, then it is considered to have N. CPU There are L CSI processing units (CPUs) used for processing CSI reports. If, in a given OFDM symbol, L CPUs are used for CSI report calculations, then the UE has N CPU There are L unoccupied CPUs. If, on the same OFDM symbol, N CSIs report starting to occupy their respective CPUs, and at this time there are N... CPU There are L unoccupied CPUs, of which the nth CSI reports occupied CPU. There are 1 CPU, where n is from 0 to N. 1. A variable of integers that changes, i.e., n=0,…,N 1. Then the UE does not need to update the lowest priority N. M requested CSI reports, where 0 ≤ M ≤ N. This allows for... Maximize, while satisfying The UE should not be configured to have more than N. CPU A non-periodic CSI trigger state is reported. The processing of CSI reports has been defined to occupy a certain number of CPUs across multiple symbols.

[0047] 2. CSI measurement and computation latency for non-AI / ML applications After receiving the CSIReportConfig from the downlink control information (DCI), the UE (e.g., UE 10) begins the CSI measurement and calculation process, ultimately obtaining the CSI measurement / calculation results to report to the gNB (e.g., gNB 20) on the PUCCH or PUSCH. The entire process can be divided into four parts: DCI decoding, beam switching (optional), CSI measurement / calculation, and transmission preparation. When the CSI request field in the DCI triggers a CSI report on the PUSCH, the UE should provide a valid CSI report for the nth triggered CSI report. — If the first uplink symbol used to carry the corresponding CSI report (including the effects of timing advance) is not earlier than the start of symbol Zref; and — If the first uplink symbol used to carry the nth CSI report (including the effects of timing advance) does not begin earlier than symbol Z'ref(n); Zref is defined as the next uplink symbol starting at the beginning of the CP after the last symbol of the PDCCH that triggered the CSI report ends. Z'ref(n) is defined as the next uplink symbol after the last symbol of the CSI-RS resource in time, at which its CP begins. .

[0048] On the UE side, since the requirements of AI / ML models for CSI processing units (CPU) are completely different from those of non-AI / ML models during operation, it is necessary to redefine a set of CSI processing criteria applicable to AI / ML models as well as CSI measurement and calculation latency.

[0049] 1. The requirements for AI CPUs differ significantly from those of traditional CPUs. When AI / ML models are applied to CSI measurement processes (including CSI prediction, channel information feedback, beam management, and localization), it becomes clear that the requirements of AI / ML models in terms of computing units, memory, storage space, and other hardware resources differ significantly from those of traditional CSI measurement and calculation methods that do not employ AI / ML (i.e., non-AI / ML CSI measurement / calculation). Therefore, the CPU usage and CPU occupancy metrics originally defined for non-AI / ML models are insufficient to assess the resource utilization of AI / ML models. Thus, it is necessary to redefine the relevant parameters for AI / ML models.

[0050] 2. For AI / ML, it is necessary to reconstruct CSI measurement and computation time estimation. Unlike non-AI / ML algorithms used for CSI measurement and computation, AI / ML model activation (e.g., loading the AI / ML model into the UE's memory (RAM)) is a necessary step before inference and requires a certain amount of time. This model activation process should only begin after the UE receives the CSI-RS. Therefore, if the UE initiates AI / ML model inference immediately upon the start of conventional CSI measurement / computation as described in Section 5.4 of TS 38.214 (where CSI measurement and computation begin after receiving the CSI-RS), the overall time for CSI measurement and computation using AI / ML model inference will be prolonged due to the required AI / ML model activation time.

[0051] Different AI capabilities can lead to different inference times, rendering traditional fixed CSI measurements and computation time definitions inapplicable. Furthermore, even with the same AI functionality, using different AI algorithms (such as RNNs and CNNs) can introduce differences in computational latency.

[0052] See Figure 2 A communication system including UE 10a, base station 20a, base station 20b and network entity device 30 performs the disclosed method according to an embodiment of the present disclosure. Figure 2 For illustrative purposes only and not as a limitation, this system may include additional UEs, base stations (BSs), and core network (CN) entities. Connections between devices and between device components are indicated by lines and arrows in the diagram. UE 10a may include processor 11a, memory 12a, and transceiver 13a. Base station 20a may include processor 21a, memory 22a, and transceiver 23a. Base station 20b may include processor 21b, memory 22b, and transceiver 23b. Network entity device 30 may include processor 31, memory 32, and transceiver 33. Each processor 11a, 21a, 21b, and 31 may be configured to implement the functions, procedures, and / or methods described herein. The various layers of the radio interface protocol may be implemented in processors 11a, 21a, 21b, and 31. Each memory 12a, 22a, 22b, and 32 is used to operatively store various programs and information for use by the respective processor. Each transceiver 13a, 23a, 23b, and 33 is operatively coupled to a corresponding processor and is used to transmit and / or receive wireless signals. Each base station 20a and 20b can be one of an eNB, gNB, or other wireless nodes.

[0053] Each processor 11a, 21a, 21b, and 31 may include a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), other chipsets, logic circuits, and / or data processing devices. Each memory 12a, 22a, 22b, and 32 may include read-only memory (ROM), random access memory (RAM), flash memory, memory cards, storage media, and / or other storage devices. Each transceiver 13a, 23a, 23b, and 33 may include baseband circuitry and radio frequency (RF) circuitry for processing radio frequency signals. When the embodiments are implemented in software, the techniques described herein can be implemented through modules, processes, functions, entities, etc., which perform the functions described herein. These modules may be stored in memory and executed by the processor. The memory may be implemented internally or externally to the processor and may be communicatively connected to the processor in various ways known in the art.

[0054] Network entity device 30 may be a node in the core network (CN). The CN may include an LTE core network or a 5GC, which may include User Plane Function (UPF), Session Management Function (SMF), Access and Mobility Management Function (AMF), Unified Data Management (UDM), Policy Control Function (PCF), Control Plane / User Plane Separation (CUPS), Authentication Server (AUSF), Network Slice Selection Function (NSSF), and Network Exposure Function (NEF).

[0055] See Figure 3 A system 100 for CSI processing based on AI / ML includes a data acquisition unit 101, a model training unit 102, an actor 103, and a model inference unit 104. It should be noted that... Figure 3 This method is not necessarily limited to this example. It is applicable to any machine learning-based design. Its general steps include data acquisition and / or model training and / or model inference and / or execution.

[0056] Data acquisition unit 101 is a function that provides input data to model training unit 102 and model inference unit 104. Data preparation related to AI / ML algorithms (such as data preprocessing, cleaning, formatting, and transformation) is not performed in data acquisition unit 101. Examples of input data may include measurements from the UE or different network entities, feedback from actuator 103, and outputs from the AI / ML model. Training data is the data required as input to model training unit 102. Inference data is the data required as input to model inference unit 104.

[0057] The model training unit 102 is a function that performs ML model training, validation, and testing. If necessary, the model training unit 102 is also responsible for data preparation (e.g., data preprocessing, cleaning, formatting, and transformation) based on the training data provided by the data acquisition unit 101. Model deployment / updates occur between the model training unit 102 and the model inference unit 104, including deploying or updating the AI / ML model (e.g., the trained ML model 105a or 105b) to the model inference unit 104. The model training unit 102 uses the data unit as training data to train the ML model 105a and generates the trained ML model 105b from the ML model 105a.

[0058] The model inference unit 104 is a function that provides AI / ML model inference output (e.g., prediction or decision). The AI / ML model inference output is the output of the ML model 105b. If necessary, the model inference unit 104 is also responsible for data preparation (e.g., data preprocessing, cleaning, formatting, and transformation) based on the inference data provided by the data acquisition unit 101. The output shown between units 103 and 104 in the figure is the AI / ML model inference output generated by the model inference unit 104.

[0059] The executor 103 is a function that receives output from the model inference unit 104 and triggers or executes corresponding actions. The executor 103 can trigger actions against other entities or itself. The feedback between the executor 103 and the data acquisition unit 101 is information that may be used to generate training data, inference data, or performance feedback.

[0060] Figure 4 A flowchart illustrating the overall solution for AI / ML-based CSI processing is provided. See also... Figure 4 The overall solution of the disclosed method is as follows.

[0061] Step 1: The UE (e.g., UE 10) sends the parameters simultaneousCSI-ReportsPerCC and simultaneousCSI-ReportsAllCC to the gNB (e.g., gNB 20). These two parameters can be used to calculate the number of available CSI processing units (CPUs) of the UE, denoted as N. CPU The N CPU It is part of the UE capability.

[0062] Step 2: The gNB provides CSI measurement configuration based on the UE's capabilities and performs CSI reporting processing.

[0063] Step 3: After receiving the CSI measurement request, the UE decodes the received message. Upon successful decoding, the UE activates the AI / ML model used for CSI measurement / computation inference.

[0064] Step 4: Upon receiving the CSI-RS, the UE can perform CSI measurement / calculation through AI / ML model inference. Within a specified time, the UE generates a CSI report and sends it to the gNB.

[0065] See Figure 5 The UE 10 in the specification may include one of UE 10a; the gNB 20 in the specification may include base station 20a or 20b. It should be noted that although the following description uses a gNB as an example of a base station, the disclosed method can also be applied to other types of base stations, such as eNBs or base stations for 5G and later systems. The transmission of uplink (UL) control signals or data can be a transmission operation from the UE to the base station; the transmission of downlink (DL) control signals or data can be a transmission operation from the base station to the UE. The disclosed method is described in detail below. UE 10 and the base station (e.g., gNB 20) perform an AI / ML-based CSI processing method.

[0066] UE 10 reports its AI / ML-based CSI processing capabilities (step S001). In one embodiment, the UE's AI / ML-based CSI processing capabilities include one or more of the following: The number of AI / ML-based CSI processing units (CPUs) in the UE; The number of AI / ML-based CSI processing processes for the UE; The number of AI / ML-based CSI storage units in the UE; The number of AI / ML-based CSI computing units in the UE; and The number of AI / ML-based CSI power consumption units in the UE.

[0067] In one embodiment, the UE's AI / ML-based CSI processing capability is determined based on one or more of the following: Subcarrier spacing (SCS); CSI parameter information based on AI / ML; and Features of CSI models based on AI / ML.

[0068] gNB 20 receives the AI / ML-based CSI processing capability (step S002).

[0069] gNB 20 determines the AI / ML-based CSI measurement configuration 402 based on the AI / ML-based CSI processing capability, wherein the AI / ML-based CSI measurement configuration includes one or more of the following: CSI resource scheduling configuration, CSI reporting configuration, and CSI measurement triggering configuration (step S003).

[0070] UE 10 receives AI / ML-based CSI measurement configuration 402 from gNB 20, wherein the configuration is based on the AI / ML-based CSI processing capability (step S004).

[0071] UE 10 performs a first operation (step S006) based on the AI / ML-based CSI measurement configuration 402.

[0072] When determining the AI / ML-based CSI measurement configuration based on the AI / ML-based CSI processing capabilities, UE 10 and / or gNB 20 maintain one or more of the following relationships: The AI / ML-based CSI measurement configuration occupies less than or equal to the AI / ML-based CSI processing capacity. The weighted value of the AI / ML-based CSI processing resources occupied by the AI / ML-based CSI measurement configuration is less than or equal to the AI / ML-based CSI processing capacity; or The sum of the AI / ML-based CSI processing resources used by the AI / ML-based CSI measurement configuration and the CSI processing resources used by the non-AI / ML-based CSI measurement is less than or equal to the UE's processing capacity.

[0073] In one embodiment, the AI / ML-based CSI processing resources include one or more of the following: The number of AI / ML-based CSI processing units (CPUs) of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI processing processes of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI storage units occupied by the AI / ML-based CSI measurement configuration in the UE; The number of AI / ML-based CSI calculation units in the UE occupied by the AI / ML-based CSI measurement configuration; and The number of AI / ML-based CSI power consumption units of the UE occupied by the AI / ML-based CSI measurement configuration.

[0074] In one embodiment, the AI / ML-based CSI processing resources are determined based on one or more of the following: Subcarrier spacing (SCS); CSI parameter information based on AI / ML; and CSI AI / ML model features.

[0075] In one embodiment, the UE's AI / ML-based CSI processing resources are obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the UE's AI / ML-based CSI processing resources occupied by it are pre-configured or pre-defined by the base station; and The base station receives a message from the UE, wherein the message includes one or more of the following: UE's AI / ML-based CSI processing resources; Instructions indicating that AI / ML-based CSI processing resources and non-AI / ML-based CSI processing resources are separated from each other.

[0076] Alternatively, the UE's AI / ML-based CSI processing resources may be obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the UE's AI / ML-based CSI processing resources occupied by it are pre-configured or pre-defined by the base station; and The base station receives a message from the UE, wherein the message includes one or more of the following: UE's AI / ML-based CSI processing resources; Indication information indicating that the AI / ML-based CSI processing resources and the non-AI / ML-based CSI processing resources belong to the same common CSI processing unit (CPU) resource pool.

[0077] In some embodiments of this disclosure, when the AI / ML-based CSI measurement configuration exceeds the UE's AI / ML-based CSI processing capability, the UE discards the AI / ML-based CSI measurement configuration that exceeds the processing capability; when the configuration is less than or equal to the UE's AI / ML-based CSI processing capability, the UE performs the first operation according to the AI / ML-based CSI measurement configuration.

[0078] In some embodiments of this disclosure, when the weighted sum calculated from the AI / ML-based CSI measurement configuration and its weights exceeds the UE's AI / ML-based CSI processing capability, the UE discards the AI / ML-based CSI measurement configuration that exceeds the processing capability; when the weighted sum is less than or equal to the UE's AI / ML-based CSI processing capability, the UE performs the first operation according to the AI / ML-based CSI measurement configuration.

[0079] See Figure 6 In one embodiment of the disclosed method, UE 10 sends an uplink message to transmit CPU usage information for CSI processing of AI / ML.

[0080] The gNB 20 sends an AI / ML-based CSI measurement configuration 402 and an AI / ML-based CSI reference signal 403 to enable the UE to perform CSI measurements based on the AI / ML-based CSI measurement configuration and the AI / ML-based CSI reference signal. The time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time. The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability (step S011).

[0081] UE 10 receives AI / ML-based CSI measurement configuration 402 and AI / ML-based CSI reference signal 403 (step S012). The time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time. The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability.

[0082] UE 10 performs CSI reporting based on the AI / ML-based CSI measurement configuration 402 and the AI / ML-based CSI reference signal 403 to send CSI reporting result 404 (step S013).

[0083] gNB 20 receives the results reported by CSI (e.g., CSI report 404) (step S014).

[0084] In some embodiments of this disclosure, the first processing time includes one or more of the following: The time required to decode the AI / ML-based CSI measurement configuration; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; and The duration required for UE antenna switching.

[0085] In some embodiments of this disclosure, the second processing time includes one or more of the following: Used for part or all of the time required to activate AI / ML models; The time required for AI / ML model switching; and The time required for AI / ML model inference.

[0086] In some embodiments of this disclosure, the first processing time and / or the second processing time are determined by one or more of the following: CSI measurement data based on AI / ML; AI / ML-based CSI measurement features or functions; CSI measurement time based on AI / ML; AI / ML model parameters used for CSI measurements; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; and Subcarrier spacing (SCS).

[0087] In some embodiments of this disclosure, the first processing time and / or the second processing time are obtained by one or more of the following: Predefined or preconfigured; Empirical values ​​derived from statistics on the time consumed in executing the AI / ML-based CSI measurement configuration; and The uplink message carries the values ​​of the first processing time and / or the second processing time, wherein the uplink message is transmitted in UE capability reporting, scheduling request, physical uplink shared channel (PUSCH) or physical uplink control channel (PUCCH).

[0088] In some embodiments of this disclosure, the second processing time includes a portion or all of the time required for AI / ML model activation; If the AI / ML model activation begins after receiving the AI / ML-based CSI reference signal, the second processing time includes the total duration required for AI / ML model activation. If the AI / ML model activation begins before receiving the AI / ML-based CSI reference signal and completes after receiving the AI / ML-based CSI reference signal, then the second processing time includes a portion of the time required for AI / ML model activation; and If the AI / ML model activation is completed before the AI / ML-based CSI reference signal is received, the second processing time does not include the time required for AI / ML model activation.

[0089] The following provides a detailed explanation of AI / ML-based virtualized CPUs and their processing principles.

[0090] During the process of performing CSI measurement and calculation using AI / ML models on the UE side, the CSI measurement requirements configured by the base station for the UE, or the CSI measurement requirements configured by the UE itself, or the CSI measurement update requirements (including CSI-RS resource configuration, CSI reporting configuration, CSI measurement configuration, CSI triggering, etc.), should be determined based on at least one of the following parameters: The number of AI / ML processes supported simultaneously is denoted as N. CPU,AI / ML This parameter indicates that the UE has N CPU,AI / ML One dedicated AI / ML CSI processing unit (CPU) is used to process AI / ML-based CSI reports, where N CPU,AI / ML It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0091] The number of CPUs used by the nth AI / ML-based CSI report is denoted as . Where n is from 0 to N AI / ML 1. A variable integer: This parameter represents the UE's occupancy in the nth AI / ML-based CSI report. CPUs, of which It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0092] For example, if in a given OFDM symbol, there is L AI / ML If one dedicated AI / ML CPU is used for AI / ML-based CSI reporting calculations, then the UE has N CPU,AI / ML L AI / ML There are N unused AI / ML dedicated CPUs. AI / ML Several AI / ML-based CSI reports begin occupying their respective dedicated AI / ML CPUs on the same OFDM symbol, while at this time there are N CPU,AI / ML L AI / ML There are 10 unused AI / ML dedicated CPUs, where each AI / ML-based CSI report n=0,…,N AI / ML 1 corresponds to the occupation If there are 1 CPU, then the UE does not need to update the lowest priority N. AI / ML M AI / ML One requested AI / ML-based CSI report (M AI / ML <L AI / ML ), where 0≤M AI / ML ≤N AI / ML And MAI / ML To meet The maximum value. The UE should not be configured to have more than N. CPU,AI / ML The reported configuration includes aperiodic CSI trigger states that use AI / ML model inference. In other words, the CPU usage of AI / ML-based CSI processing satisfies the following formula:

[0093] CSI processing may include either or both of CSI measurements and CSI calculations. Additionally, parameters... It can be determined based on one or more factors, such as CSI measurement content, AI / ML functions, AI / ML models, etc. The specific method depends on the implementation. The value can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB20.

[0094] 1. Option A: Predefined by wireless communication system CPU usage Examples are shown in the table below. Related to CSI measurement content, AI / ML functions, and AI / ML models. O_k is a predefined... The value represents the dedicated AI / ML CPU consumed, which may be affected by processing resources, storage resources, computing power, and power consumption. The variable k is an integer representing the column index. Class A / B / C represent different types of CSI-ReportConfig. For example, Class A indicates that reportQuantity is set to "none," Class B indicates it is set to "RSRP" or "SINR," and Class C indicates CSI measurements and calculations including codebook types. Type A / B / C represent different AI / ML functions. For example, Type A represents AI / ML for CSI measurements, Type B represents AI / ML for beamforming measurements, and Type C represents AI / ML for localization. Model A / B / C represent different AI / ML models. For example, Model A represents an RNN-based AI / ML model, Model B represents a CNN-based AI / ML model, and similarly, Model C represents an AI / ML model based on a certain neural network, etc.

[0095] Table 2: Examples of Value Determination

[0096] The following example illustrates the parameters How to determine O based solely on CSI measurement content?k The value of is given in this example, which illustrates three scenarios based on different types of CSI-ReportConfig.

[0097] Case 1: When the type of CSI-ReportConfig is Class A, ; Scenario 2: When the type of CSI-ReportConfig is Class B, ; Scenario 3: When the CSI-ReportConfig type is Class C, the CSI measurement / calculation metrics are relatively complex, and the measurement task is heavy. Therefore, the AI / ML model used for CSI measurement / calculation also has high complexity and resource overhead. Therefore, the parameters... The value is set to the maximum value of the UE's capability, that is... .

[0098] 2. Option B: Instructed by UE to gNB In option B, each AI / ML-based CSI report... Obtained through UE reporting. After receiving the information reported by the UE, the gNB, provided that the total resources used for AI / ML model inference do not exceed the UE's AI / ML model inference capability, begins to determine the AI / ML-based CSI measurement configuration allocated to the UE, namely:

[0099] To enable the base station to accurately identify the UE's AI / ML CPU usage under different conditions (i.e., different measurements, AI / ML functions, AI / ML models), thereby maximizing the UE's computing power and CSI feedback capability, the information fed back by the UE to the base station should include one or more of the following parameters: AI / ML model-related information used for AI / ML-based CSI and the CPU usage. The information fed back by the UE is model-specific metadata and can be included in UE capability reporting, uplink control / data channels (e.g., PUSCH / PUCCH), MAC layer control units, RRC layer signaling, or other appropriate control mechanisms in the radio protocol stack.

[0100] AI / ML model related information: The AI / ML model related information includes AI / ML model attribute description information, which is used to identify and distinguish the different AI / ML models used by the UE or gNB in ​​CSI measurement and calculation.

[0101] CPU usage of AI / ML-based CSI The CPU usage for the nth AI / ML-based CSI measurement and computation can vary depending on the different CSI, and / or AI / ML features / functions, and / or AI / ML models. Furthermore, the amount of resources used... It may include at least one of the following parameters: , , or , representing the processing resources, storage resources, computing power, or power consumption of the nth CSI measurement, calculation, or report, respectively. Alternatively, it represents the amount of resources used. Parameters can also be included. , , or The weighted sum. Parameters , , or The definition is explained in Example 2.

[0102] In Example 1, parameter N CPU,AI / ML This parameter, N, represents the number of AI / ML processes supported simultaneously. It is a virtualization parameter derived from and influenced by various physical factors, including processing power, memory, computing capacity, and energy consumption. Only one parameter, N, is used. CPU,AI / ML This may not fully reflect the UE's ability to simultaneously process AI / ML-based CSI measurements and calculations. In this embodiment, the CSI measurement requirements configured by the base station for the UE, or the CSI measurement requirements configured by the UE itself, or the CSI measurement update requirements (including CSI-RS resource configuration, CSI reporting configuration, CSI measurement configuration, CSI triggering, etc.), should be determined based on at least one of the following parameters: The amount of processing resources supported for AI / ML-based CSI measurements, Nproc: represents the processing resources that the UE can currently use for AI / ML model computation, where the parameter Nproc can be indicated by the UE or defined by system parameters.

[0103] The amount of storage resources Nmem supported for AI / ML-based CSI measurements: represents the available RAM resources on the UE that can be used for AI / ML model inference, where the parameter Nmem can be indicated by the UE or defined by system parameters.

[0104] The amount of computing power supported for AI / ML-based CSI measurements, Ncomp: represents the available computing resources on the UE that can be used for AI / ML model inference, where the parameter Ncomp can be indicated by the UE or defined by system parameters.

[0105] The amount of power consumption Npower supported for AI / ML-based CSI measurements: represents the maximum power consumption available on the UE for AI / ML model inference, where the parameter Npower can be indicated by the UE or defined by system parameters.

[0106] Therefore, the nth AI / ML-based CSI report (n=0, 1, …, N) AI / ML 1) The processing resources, storage resources, computing power, and power consumption occupied should be defined as follows: The amount of processing resources used by the nth AI / ML-based CSI report Where n is from 0 to N AI / ML 1. A changing integer variable: representing the UE's occupancy in the nth AI / ML-based CSI report. One processing resource, of which It can be indicated by the UE or defined by system parameters.

[0107] The amount of storage resources used by the nth AI / ML-based CSI report Where n is from 0 to N AI / ML 1. A changing integer variable: representing the UE's occupancy in the nth AI / ML-based CSI report. One storage resource, of which It can be indicated by the UE or defined by system parameters.

[0108] The amount of computing power used by the nth AI / ML-based CSI report Where n is from 0 to N AI / ML 1. A changing integer variable: representing the UE's occupancy in the nth AI / ML-based CSI report. One computing power, of which It can be indicated by the UE or defined by system parameters.

[0109] The power consumption of the nth AI / ML-based CSI report Where n is from 0 to N AI / ML 1. A changing integer variable: representing the UE's occupancy in the nth AI / ML-based CSI report. One power consumption, of which It can be indicated by the UE or defined by system parameters.

[0110] For example, after receiving the above four parameters, gNB 20 determines the AI / ML-based CSI reporting configuration (i.e., CSI reporting configuration) to be assigned to the UE based on an assessment of the UE's capabilities, wherein the assessment satisfies at least one of the following conditions:

[0111] In addition, parameters , , , It can be determined based on one or more factors, such as CSI measurement content, AI / ML model, AI / ML functionality, etc. The specific method depends on the implementation. , , and One or more values ​​in the range can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to the gNB20. Furthermore, , , and Some of the values ​​can be obtained from predefined system parameters, while others can be indicated by the UE 10 through a reporting message sent to gNB 20.

[0112] 1. Option A: , , and One or more of them are predefined by the wireless communication system The following table provides an example, in which , , and One or more of these are predefined by the wireless communication system. In this table, P k M k C k and O k These represent predefined consumption values ​​for processing resources, storage resources, computing power, and power consumption, respectively, where k is an integer ranging from 1 to 7. Class A / B / C indicates the type of CSI-ReportConfig, Type A / B / C indicates the AI / ML function type, and Model A / B indicates the AI / ML model type, with the same meaning as described in Example 1.

[0113] Table 3

[0114] The following example illustrates the parameters , , , How to determine P based solely on CSI measurement content? k M k C k O k The value of is given in this example, which illustrates three scenarios based on different types of CSI-ReportConfig.

[0115] Case 1: When the type of CSI-ReportConfig is Class A, ( , , , = (0, 0, 0, 0); Case 2: When the type of CSI-ReportConfig is Class B, ( , , , = (1, 1, 1, 1); Scenario 3: When the CSI-ReportConfig type is Class C, the CSI measurement / calculation metrics are relatively complex, and the measurement task is heavy. Therefore, the AI / ML model used for CSI measurement / calculation also has high complexity and resource overhead. Therefore, the values ​​of each parameter are set to the maximum value of the UE capability, i.e. =Nproc, =Nmem, =Ncomp, and =Npower.

[0116] 2. Option B: , , and One or more of them are indicated by the UE to the gNB These parameters can be transmitted via UE capability reporting, MAC-CE, or RRC messages carried by PUSCH / PUCCH, and must include at least one of the following: AI / ML model related information: The AI / ML model related information includes AI / ML model attribute description information, which is used to identify an AI / ML model. This identifier can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, or an AI / ML management platform-specific unique ID.

[0117] Number of processing processes This parameter describes the number of processing steps consumed by the AI / ML model.

[0118] Storage resources occupied This parameter describes the RAM resources consumed during AI / ML model inference.

[0119] Computing power consumed This parameter describes the computational power consumed during the AI / ML model inference process.

[0120] Power consumption This parameter describes the power consumption during the AI / ML model inference process.

[0121] Example 2: AI / ML Dedicated CPU and Processing Criteria Based on Weighted Parameters of One or More Physical Information In Example 2, the parameters represent the UE capabilities used to support AI / ML-based CSI measurements. , , and One or more of these requests are sent from the UE (e.g., UE 10) to the gNB 20, which may result in significant air interface overhead. In this embodiment, the CSI measurement requirements configured by the base station for the UE, or the CSI measurement requirements configured by the UE itself, or the CSI measurement update requirements (including CSI-RS resource configuration, CSI reporting configuration, CSI measurement configuration, CSI triggering, etc.), should be determined based on at least one of the following parameters: The number N of weighted AI / ML dedicated CSI processing units supported for both CSI measurement and calculation. ALL : The parameter N ALL Indicates that the UE has N ALL One AI / ML-dedicated CSI processing unit is used to process AI / ML-based CSI reports, where N ALL It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0122] The priority (weight value) of processing resources used by the nth AI / ML-based CSI report. , where n ranges from 0 to N_(AI / ML) 1. Changing integer variables: This priority level This indicates the importance of processing resources when evaluating the UE's capabilities for AI / ML-based CSI measurements and computing, where It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0123] The priority (weight value) of storage resources used by the nth AI / ML-based CSI report. Where n is from 0 to N AI / ML 1. Changing integer variables: This priority level This indicates the importance of storage resources when evaluating a UE's capabilities for AI / ML-based CSI measurements and computing. It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0124] The priority (weight value) of computing power used by the nth AI / ML-based CSI report. Where n is from 0 to N AI / ML 1. Changing integer variables: This priority level This indicates the importance of computing power when evaluating a UE's ability to perform AI / ML-based CSI measurements and calculations. It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0125] The power consumption priority (weight value) of the nth AI / ML-based CSI report. Where n is from 0 to N AI / ML 1. Changing integer variables: This priority level This indicates the importance of power consumption when evaluating the UE's capabilities for AI / ML-based CSI measurements and computing. It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0126] For example, priority parameters (weight values) can be used. , , as well as The parameters are calculated as follows:

[0127] Wherein, parameter N AI / ML Nproc, Nmem, Ncomp, and Npower are defined in Example 2.

[0128] At the same time, the UE should not be configured to have more than N ALL A reported configuration of a non-periodic CSI trigger state that uses AI / ML model inference.

[0129]

[0130] In addition, parameters , , and It can be determined based on one or more factors, such as CSI measurement content, AI / ML model, AI / ML functionality, etc. The specific method depends on the implementation. , , and One or more values ​​in the range can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to the gNB20. Furthermore, , , and Some of the values ​​can be obtained from predefined system parameters, while others can be indicated by the UE 10 through a reporting message sent to gNB 20.

[0131] 1. Option A: , , and One or more of them are predefined by the wireless communication system The following table provides an example, in which , , and Predefined by the wireless communication system. In this table, P proc k P mem k P comp k and P power k These represent predefined values ​​for the priority (weight) of processing resources, storage resources, computing power, and power consumption, respectively. Variable k is an integer representing the column index. Class A / B / C represents the type of CSI-ReportConfig, Type A / B / C represents the AI / ML function type, and Model A / B represents the AI / ML model type; their meanings are the same as described in Example 1.

[0132] Table 4

[0133] The following example illustrates the parameters , , and How to determine based solely on CSI measurement content? P proc k, P mem k, Pcomp k, P power k The value of i. This example illustrates three cases based on different types of CSI-ReportConfig.

[0134] Case 1: When the type of CSI-ReportConfig is Class A, ( , , , = (0, 0, 0, 0); Case 2: When the type of CSI-ReportConfig is Class B, ( , , , = (0.25, 0.25, 0.25, 0.25); Scenario 3: When the CSI-ReportConfig type is Class C, the CSI measurement / calculation metrics are relatively complex, and the measurement task is heavy. Therefore, the AI / ML models used for CSI measurement / calculation also have high complexity and resource overhead. Thus, storage resources and computing power take precedence over processing resources and power consumption, i.e. ( , , , =(0.15, 0.35, 0.35, 0.15).

[0135] 2. Option B: , , and One or more of them are indicated by the UE to the gNB.

[0136] Each AI / ML-based CSI report , , and One or more of these parameters are obtained through reporting by the UE (e.g., UE 10). After receiving the above four parameters, gNB 20 determines the AI / ML-based CSI reporting configuration (i.e., CSI reporting configuration) to be assigned to the UE (e.g., UE 10) based on an assessment of the UE's capabilities, wherein the assessment satisfies at least one of the following conditions:

[0137] These parameters can be transmitted via UE capability reporting, MAC-CE, or RRC messages carried by PUSCH / PUCCH, and must include at least one of the following: AI / ML model related information: The AI / ML model related information includes AI / ML model attribute description information, which is used to identify an AI / ML model. This identifier can be a globally unique ID, a PLMN-specific unique ID, an operator-specific unique ID, or an AI / ML management platform-specific unique ID.

[0138] Process resource priorities This parameter describes the priority of the processing resources consumed by the AI / ML model.

[0139] Storage resource priority This parameter describes the priority of RAM resources consumed during AI / ML model inference.

[0140] Priority of consuming computing power This parameter describes the priority of computing power consumed during AI / ML model inference.

[0141] Power consumption priority This parameter describes the priority of power consumption during AI / ML model inference.

[0142] Example 3: Compatibility of AI / ML-based and non-AI / ML-based CSI measurements and calculations To ensure CSI measurement compatibility between AI / ML and non-AI / ML models, the UE should be able to support both types of models simultaneously. It should be noted that traditional solutions and other compatible solutions developed or expanded in the future can also be incorporated into the embodiments of this disclosure. This is because in some scenarios, the AI / ML model may switch or fall back to the non-AI / ML model. Therefore, the UE should have the ability to perform CSI measurements using both AI / ML and non-AI / ML models simultaneously within the same device. Thus, the UE (e.g., UE 10) can send signaling to the base station to notify the base station how to configure non-AI / ML CSI measurements and AI / ML-based CSI measurements based on the UE's capabilities. The UE can simultaneously send N_CPU corresponding to the non-AI / ML CSI measurement capability and N_CPU corresponding to the AI / ML CSI measurement capability to the base station. CPU,AI / MLAI / ML-based CSI measurements and non-AI / ML-based CSI measurements can run simultaneously or they can not share the same hardware and system resources. For the scenario of running AI / ML and non-AI / ML models for CSI measurements on the same UE, there are two scenarios: one where the AI / ML model runs on a separate microprocessor unit (MPU), and the other where the AI / ML and non-AI / ML models run on a shared MPU. The UE can indicate the following information: AI / ML model related information: The AI / ML model related information includes AI / ML model attribute description information, which is used to identify the AI / ML model used by the UE or gNB for CSI measurement and calculation.

[0143] CSI Computing Resource Occupation Scheme Indication Information: This indication information is used to indicate whether AI / ML-based and non-AI / ML-based CSI measurements and computations occupy a separate CPU resource pool or share the same CPU resource pool, where CPU can be computing processing resources, storage resources, or other resources.

[0144] Relationship information is used to indicate the relationship between the CPUs used for AI / ML-based CSI measurements and calculations and the CPUs used for non-AI / ML-based CSI measurements and calculations: the CPUs may include the total number of CPUs and / or the CPUs used for a single CSI calculation, reporting or measurement.

[0145] The CSI computing resource occupancy scheme indication information can be implicitly or explicitly transmitted by the UE (e.g., UE 10) to the base station (e.g., gNB 20) via signaling. This indication information can be implicitly represented through relational information or explicitly represented through direct indication. A specific implementation example is as follows: (1) Option 1: Determined by relational information. When the relational information is a valid value, it means that AI / ML-based CSI measurements and calculations and non-AI / ML-based CSI measurements and calculations share the same CPU resource pool; when the relational information is an invalid value, it means that the two use independent CPU resource pools. The invalid value can be represented as NULL, maximum value, minimum value or other alternative values.

[0146] Furthermore, relational information can describe the relationship between the CPUs used for CSI measurements and computations based on AI / ML and those not based on AI / ML. For example, this relationship can be represented as follows (but is not limited to): The proportionality coefficient Kn is used to determine the CPU consumption relationship between AI / ML processing and non-AI / ML processing in the nth AI / ML-based CSI report, where n is from 0 to N. AI / ML 1. A variable integer: This scaling factor Kn represents the ratio of CPU used by the AI / ML model to that used by the non-AI / ML model. Kn can be obtained from predefined system parameters or indicated by the UE 10 through a reporting message sent to gNB 20. This conversion relationship can be expressed as follows: .

[0147] The CPU consumption difference Dn between AI / ML processing and non-AI / ML processing in the nth AI / ML-based CSI report is used, where n is from 0 to N. AI / ML 1. A variable integer: This difference, Dn, represents the difference in CPU usage between the AI / ML model and the non-AI / ML model. Dn can be obtained from predefined system parameters or indicated by the UE 10 through a reporting message sent to gNB 20. This conversion relationship can be represented as follows: .

[0148] (2) Option 2: The UE (e.g., UE 10) needs to send a new indication message to the gNB (e.g., gNB 20) to indicate whether the AI / ML model and the non-AI / ML model use separate CPU resource pools or share the same CPU resource pool. For example, the indication message can be in binary form, where 0 indicates separate and 1 indicates shared.

[0149] After determining whether the AI / ML model and the non-AI / ML model use independent or shared MPU resources, and the relationship between the CPUs in CSI measurement and calculation, the gNB (e.g., gNB 20) can allocate a CSI measurement configuration (including CSI-RS resources, CSI reporting, CSI triggering, etc.) that matches the capabilities of the UE (e.g., UE 10) based on the following two scenarios: (1) Case 1: AI / ML-based and non-AI / ML-based CSI measurements and calculations use separate CPU resource pools.

[0150] Each CPU resource pool may include hardware and system resources. In this case, the hardware and system resources used when performing CSI measurements or calculations using AI / ML models and non-AI / ML models are independent of each other and do not affect each other. Therefore, according to the technical solution in Embodiment 1, when the base station (e.g., gNB 20) issues CSI measurement configurations for AI / ML models and non-AI / ML models to the UE (e.g., UE 10), it should ensure that the configurations meet the corresponding UE capabilities, as follows: Wherein, parameter N represents the number of CSI reporting tasks implemented using a non-AI / ML model, parameter N AI / MLThis indicates the number of CSI reporting tasks implemented using AI / ML model inference. The number of CSI reporting tasks can be configured in the CSI reporting configuration.

[0151] In this case, the UE (e.g., UE 10) does not need to send additional signaling to instruct the gNB on how to configure CSI measurements and calculations based on AI / ML and non-AI / ML models, because the configuration can be directly determined by the UE's capabilities.

[0152] (2) Case 2: AI / ML-based and non-AI / ML-based CSI measurements and calculations share the same CPU resource pool.

[0153] In this scenario, if the hardware and system resources consumed when running AI / ML models and non-AI / ML models on the same microprocessor unit (MPU) are shared, then when the CSI reporting configuration issued by the gNB for non-AI / ML models and AI / ML models, it will occupy N_CPU and N_ML respectively. CPU,AI / ML When the resource limit is reached, it is highly likely that the CPU used by the UE to perform the corresponding CSI measurement and calculation will directly exceed its capacity, resulting in some CSI reporting configurations being unable to be processed and reducing the success rate of CSI feedback.

[0154] Therefore, to avoid the above situation, the resource consumption of AI / ML models and non-AI / ML models cannot be considered independently; instead, the conversion relationship between their CPU consumption needs to be considered. This conversion relationship allows the CPU consumption of one type of CSI measurement and calculation to be converted into the CPU consumption of another type of CSI measurement and calculation for unified evaluation. CSI measurements and calculations include both AI / ML-based and non-AI / ML-based types.

[0155] The conversion relationship between CPU consumption based on AI / ML models and non-AI / ML models in CSI measurement and computation should be determined by at least one of the parameters Kn and Dn. For example, the total CPU consumption needs to satisfy the following formula:

[0156] In another example, the total CPU consumption needs to satisfy the following formula:

[0157] Example 4: Time consumption of CSI measurement and computation based on AI / ML (including model activation time and inference time) See Figure 7An example illustrates the timing of CSI reporting, AI / ML model activation, and AI / ML model inference. For AI / ML-based CSI measurement and computation, a model activation process is required before performing subsequent model inference operations to load the AI / ML model into main memory (e.g., main memory in memory 11 or memory / storage device 740). Typically, AI / ML models are stored in read-only memory (ROM) (e.g., ROM in memory / storage device 740) or a UE-side database in file form (e.g., .pkl, .pmml, .mlmodel, .caffemodel format files) and loaded into memory at runtime. Therefore, using AI / ML models for CSI measurement and computation requires additional model activation time compared to traditional non-AI / ML models. During the execution of AI / ML-based CSI measurement and computation on the UE side, the CSI measurement and computation time can be determined by at least one of the following parameters: First time length T based on AI / ML-based CSI measurement and calculation proc,CSI,AI / ML Z Ref,AI / ML This indicates the next uplink symbol after the last symbol of the PDCCH (e.g., a CSI request) that triggers an AI / ML-based CSI report, and may be related to the value μ. T proc,CSI,AI / ML The time difference between the last symbol of the PDCCH that triggers the CSI report and the next uplink symbol after its completion. This first time length may include the time of at least one of the following operations: DCI decoding containing the CSI request, AI / ML model activation, AI / ML model switching, AI / ML model inference, and beamforming switching. proc,CSI,AI / ML and symbol Z Ref,AI / ML It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0158] The second time length T' based on AI / ML CSI measurement and calculation proc,CSI,AI / ML :Z' Ref,AI / ML This indicates the next uplink symbol after the last symbol of the reference signal (e.g., CSI-RS, SRS, DMRS, SSB, etc.) used for AI / ML-based CSI measurements and calculations, and may be related to the numerical value μ. T' proc,CSI,AI / ML The time difference between the last symbol of the reference signal used for CSI measurement and reporting and the next uplink symbol after its termination. This second time length may include the time of at least one of the following operations: partial or full activation of the AI / ML model, AI / ML model switching, and AI / ML model inference. proc,CSI,AI / ML and symbol Z' Ref,AI / MLIt can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0159] Time required to activate an AI / ML model X load,AI / ML : Indicates the time from the start of a CSI request or CSI-RS transfer to the successful loading of a specific AI / ML model into RAM. This time may be related to the value μ, where X load,AI / ML It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0160] Time required for AI / ML model inference X inf,AI / ML : Represents the time from when the AI / ML model starts inference to when it finishes inference. This time may be related to the value μ, where X inf,AI / ML It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0161] For example, when aperiodic CSI-RS is used for channel measurements in the nth triggered CSI report, the AI / ML model inference can start after the last symbol of the CSI-RS ends and after the AI / ML model activation is complete. The UE (e.g., UE 10) can receive aperiodic CSI-RS resources for channel measurements, aperiodic CSI-IM resources for interference measurements, and aperiodic NZP CSI-RS resources for interference measurements. Parameter Z Ref,AI / ML 'Used to indicate the next uplink symbol after the last symbol of the reference signal (e.g., CSI-RS, SRS, DMRS, SSB, etc.) used for AI / ML-based CSI measurements and calculations.' Ref,AI / ML It is Z' Ref In the AI / ML-based version of CSI measurement and calculation, Z Ref,AI / ML It is Z Ref The version in AI / ML-based CSI measurement and calculation.

[0162] In addition, parameter X load,AI / ML and X inf,AI / ML It can be determined based on one or more factors, such as the content of CSI measurements and calculations, AI / ML models, AI / ML functions, etc. The specific method depends on the implementation. load,AI / ML and X inf,AI / ML The value of parameter X can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20. If parameter X... load,AI / ML and X inf,AI / MLAs predefined by the wireless communication system, the UE (e.g., UE 10) can select from multiple X values ​​based on the numerical value μ and one or more factors (e.g., CSI measurement content, AI / ML model, AI / ML function, etc.). load,AI / ML Select one parameter from multiple X parameters inf,AI / ML Choose one parameter. The table below shows the parameter X, which is related to the numerical value μ and determined based on the factors mentioned above. load,AI / ML and X inf,AI / ML An example. Where X inf,AI / ML It can be represented as X infij X load,AI / ML It can be represented as X loadij Where i is an integer variable representing the row index and j is an integer variable representing the column index. In the table, Class A / B / C represents different types of CSI-ReportConfig, Type A / B / C represents different AI / ML functions, and Model A / B / C represents different AI / ML models, with the same meaning as explained above.

[0163] Table 5

[0164] The inference time of the AI / ML model must be started under two conditions: (1) the UE has received the CSI reference signal for CSI measurement sent by the gNB, such as CSI-RS, IM-RS or other signals; (2) the activation of the AI / ML model has been completed.

[0165] Once the AI / ML model is activated, considering the specific time length X required for AI / ML model activation... load,AI / ML Reasoning time X inf,AI / ML The timing of the gNB sending the CSI reference signal to the UE, and the relationship between these three time lengths can lead to two scenarios, denoted as Case 1 and Case 2 respectively: Scenario 1: The AI / ML model activation process is completed before the UE (e.g., UE 10) receives the CSI reference signal.

[0166] See Figure 7 In this case, when the UE (e.g., UE 10) receives the last symbol of the CSI reference signal (including CSI-RS, IM-RS, or other types of signals), the AI / ML model activation process has been completed and successfully loaded into RAM. Time Length X inf,AI / MLThe CSI reporting begins when the UE receives the last symbol of the CSI reference signal transmitted by the gNB and ends when the CSI measurement and calculation are completed using the AI / ML model and the inference result is obtained. The inference result includes the CSI measurement and calculation results obtained through AI / ML model inference. That is, for the nth triggered CSI report, X... inf,AI / ML Forming T' proc,CSI,AI / ML Therefore, T' proc,CSI,AI / ML The length of time is equal to X inf,AI / ML .

[0167] When X inf,AI / ML If the end time is earlier than the transmission time of the first symbol of the message carrying the CSI report using the uplink channel (e.g., PUSCH), the UE should provide a valid CSI report inferred using the AI / ML model for the nth triggered CSI report. Otherwise, if there is no multiplexing of HARQ-ACK or transport block on the PUSCH, the UE can ignore the CSI report inferred using the AI / ML model.

[0168] Scenario 2: The AI / ML model activation process is completed only after the UE (e.g., UE 10) receives the CSI reference signal.

[0169] See Figure 8 In this scenario, when the UE receives the last symbol of the CSI reference signal (including CSI-RS, IM-RS, or other types of signals), the AI / ML model activation process is still in progress and has not yet been successfully loaded into RAM. (Time length X) inf,AI / ML The process begins when the AI / ML model is successfully loaded into RAM and ends when CSI measurements and calculations are completed using the AI / ML model and the inference results are obtained. The inference results include the CSI measurements and calculations obtained through AI / ML model inference. That is, for the nth triggered CSI report, X... load,AI / ML and X inf,AI / ML Composition of T proc,CSI,AI / ML Therefore, X load,AI / ML With X inf,AI / ML The time length is continuous, with no time interval between them. T is calculated from the last symbol of the message carrying the CSI request. proc,CSI,AI / ML The length of time equals .

[0170] when If the end time is earlier than the transmission time of the first symbol of the message carrying the CSI report using the uplink channel (e.g., PUSCH), the UE (e.g., UE 10) should provide a valid CSI report inferred using the AI / ML model for the nth triggered CSI report. Otherwise, if there is no multiplexing of HARQ-ACK or transport block on the PUSCH, the UE may ignore the CSI report inferred using the AI / ML model.

[0171] Example 5: Comparison of time consumption (including model activation and inference time) for AI / ML-based CSI measurement and computation with that for non-AI / ML-based CSI measurement and computation.

[0172] Example 5 can be understood in conjunction with the foregoing content, except for the detailed description below. In this example, the difference between the time consumption of non-AI / ML-based CSI measurement and computation and the time consumption of AI / ML-based CSI measurement and computation is defined. This difference can be quantified by the number of OFDM symbols. This example addresses a scenario where AI / ML model activation is completed before the UE (e.g., UE 10) receives all CSI reference signals (e.g., CSI-RS, CSI-IM, etc.) sent by the gNB (e.g., gNB 20). Therefore, this difference represents the difference between the time consumed by non-AI / ML model execution and the time consumed by AI / ML model inference on the UE (e.g., UE 10).

[0173] During the CSI measurement and calculation process in a UE (e.g., UE 10) using an AI / ML model, the CSI measurement and calculation time should be determined by the following parameters: The difference between AI / ML model inference time and non-AI / ML execution time, expressed in terms of the number of OFDM symbols. Z: Parameter Z represents the difference between the number of OFDM symbols required for AI / ML model inference and the number of OFDM symbols required for corresponding non-AI / ML execution. Parameter Z may be related to the value μ, where Z can be obtained from predefined system parameters, or indicated by UE 10 through a reporting message sent to gNB 20.

[0174] For example, parameters Z can be calculated using the following formula:

[0175] Wherein, parameter T' proc,CSI,AI / ML In Example 4, the parameter T' is defined. proc,CSIThis represents the number of predefined OFDM symbols used to perform non-AI / ML-based CSI measurements and calculations, and can be considered a constant known to both the UE and gNB (e.g., UE 10 and gNB 20).

[0176] In addition, parameters Z can be determined based on one or more factors, such as CSI measurement content, AI / ML model, AI / ML functionality, etc. The specific method depends on the implementation. The value of Z can be obtained from predefined system parameters, or indicated by UE 10 through a reporting message sent to gNB 20.

[0177] UE (e.g., UE 10) can be based on a numerical value μ and one or more factors (e.g., CSI measurement content, AI / ML model, AI / ML function, etc.) from multiple parameters. One is determined from Z. The table below shows the parameters related to the numerical value μ and determined based on one or more factors (e.g., CSI measurement content, AI / ML model, AI / ML function, etc.). Example of Z. Parameters Z can be represented as Zij, where i is an integer variable representing the row index and j is an integer variable representing the column index. In the table, Class A / B / C represents different types of CSI-ReportConfig, Type A / B / C represents different AI / ML functions, and Model A / B / C represents different AI / ML models, with the same meaning as explained above.

[0178] Table 6

[0179] Scenario 1: When the AI / ML function is Class A and the value μ is 0, the AI / ML model is relatively simple, the time required for model activation and inference is less, and the time difference can be zero. Z 04 =0.

[0180] Scenario 2: When the AI / ML function is Class B and the value μ is 0, the CSI measurement / calculation metrics are relatively complex, and the model activation and inference processes take a long time. Therefore, the parameters... The value of Z is set to a large value, that is Z 05 =20.

[0181] Example 6: Time consumption for CSI measurement and calculation based on AI / ML (using a total time length) Example 6 is similar to Example 4 and can be understood in conjunction with Example 4. The differences are as follows: In this example, the AI / ML model activation process begins when the UE receives the CSI reference signal (including CSI-RS, IM-RS, or other types of signals), and the AI / ML model begins inference immediately after successful activation (e.g., loading into RAM). The time length X defined in Example 4... load,AI / ML and X inf,AI / ML It is continuous, with no time intervals in between. Therefore, there is no need to distinguish X. load,AI / ML End time and X inf,AI / ML The start time and the two time lengths can be combined into a single time length, which begins when the UE receives the CSI reference signal and ends when the AI / ML model completes the inference for CSI measurement and obtains the inference result for CSI reporting. This combined time length includes the AI / ML model activation and inference time.

[0182] During the process of UE performing CSI measurements and calculations using AI / ML models, the CSI measurement and calculation time should be determined by at least one of the following parameters: The time consumed by running the AI / ML model on the UE (User Experience) AI This parameter represents the total duration of AI / ML model operation on the UE (e.g., UE10), including AI / ML model activation, handover, and inference time, as well as DCI decoding and beamforming handover time including CSI requests. This time begins when the UE receives the CSI reference signal, where X AI It can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20.

[0183] Parameter X AI It can be determined based on one or more factors, such as CSI measurement content, AI / ML model, AI / ML functionality, etc. The specific method depends on the implementation. AI The value of parameter X can be obtained from predefined system parameters, or indicated by the UE 10 through a reporting message sent to gNB 20. If parameter X... AI As predefined by the wireless communication system, the UE (e.g., UE 10) can select from multiple X values ​​based on the numerical value μ and one or more factors (e.g., CSI measurement content, AI / ML model, AI / ML function, etc.). AI Choose one parameter. The table below shows the parameter X, which is related to the numerical value μ and is determined based on one or more factors (e.g., CSI measurement content, AI / ML model, AI / ML function, etc.). AI Example. Parameter X AI It can be represented as X AIijWhere i is an integer variable representing the row index and j is an integer variable representing the column index. In the table, Class A / B / C represents different types of CSI-ReportConfig, Type A / B / C represents different AI / ML functions, and Model A / B / C represents different AI / ML models, with the same meaning as explained above.

[0184] Table 7

[0185] In this embodiment, the AI / ML model activation process begins when the UE (e.g., UE 10) receives the last symbol of the CSI reference signal (including CSI-RS, IM-RS, and other types of signals). Time length X AI The CSI reporting begins when the UE (e.g., UE 10) receives the last symbol of the CSI reference signal sent by the gNB (e.g., gNB 20) and ends when the AI / ML model completes the inference for CSI measurement and calculation and obtains the inference result for CSI reporting. That is, for the nth triggered CSI reporting, the time length is X. AI Equal to T' proc,CSI,AI / ML .

[0186] When T' proc,CSI,AI / ML If the end time is earlier than the transmission time of the first symbol of the message carrying the CSI report using the uplink channel (e.g., PUSCH), the UE should provide a valid CSI report inferred using the AI / ML model for the nth triggered CSI report. Otherwise, if there is no multiplexing of HARQ-ACK or transport block on the PUSCH, the UE can ignore the CSI report inferred using the AI / ML model.

[0187] Figure 9 This is a block diagram of a wireless communication example system 700 according to an embodiment of this disclosure. The embodiments described herein can be implemented in this system by any suitably configured hardware and / or software. Figure 9 The system 700 shown includes RF circuitry 710, baseband circuitry 720, processing unit 730, memory / storage device 740, display 750, camera 760, sensor 770, and I / O interface 780, and these components are coupled to each other as illustrated.

[0188] The processing unit 730 may include circuitry, such as, but not limited to, one or more single-core or multi-core processors. The processor may include any combination of a general-purpose processor and a dedicated processor (e.g., a graphics processor and an application processor). The processor may be coupled to a memory / storage device and configured to execute instructions stored in the memory / storage device to enable various applications and / or operating systems running in the system to function.

[0189] Radio control functions may include, but are not limited to, signal modulation, encoding, decoding, and radio frequency shifting. In some embodiments, the baseband circuitry can provide communication compatible with one or more wireless technologies. For example, in some embodiments, the baseband circuitry can support communication with 5G NR, LTE, Evolved Universal Terrestrial Radio Access Network (EUTRAN) and / or other Wireless Metropolitan Area Networks (WMAN), Wireless Local Area Networks (WLAN), and Wireless Personal Area Networks (WPAN). Embodiments in which the baseband circuitry is configured to support multiple wireless protocol communications may be referred to as multimode baseband circuitry. In various embodiments, baseband circuitry 720 may include circuitry for processing signals that are not strictly baseband frequency signals. For example, in some embodiments, the baseband circuitry may include circuitry for processing intermediate frequency (IF) signals located between the baseband frequency and the radio frequency.

[0190] In various embodiments, system 700 may be a mobile computing device, such as, but not limited to, a laptop, tablet, netbook, ultrabook, smartphone, etc. In various embodiments, the system may have more or fewer components and / or a different architecture. Where appropriate, the methods described herein may be implemented as a computer program. This computer program may be stored on a storage medium, such as a non-transitory storage medium.

[0191] This disclosure describes a combination of technologies / processes that can be employed in 3GPP specifications to form a final product.

[0192] If the software functional units are implemented and used and sold as products, they can be stored in a computer-readable storage medium. Based on the above understanding, the technical solutions proposed in this disclosure can be implemented in whole or in part in the form of a software product, or a portion of it that is beneficial to the prior art can also be implemented in the form of a software product. This software product is stored in a computer's storage medium and includes multiple instructions for causing a computing device (e.g., a personal computer, server, or network device) to execute all or part of the steps disclosed in the embodiments of this disclosure. The storage medium includes a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a floppy disk, or other media capable of storing program code.

[0193] This disclosure proposes a novel framework for Channel State Information (CSI) measurement using an AI / ML model. The framework encompasses reference signal configuration, CSI measurement, and CSI reporting. This approach offers several advantages, such as reduced signaling overhead, good scalability, efficient resource allocation for different model requirements, and unification of single-sided and double-sided model processes into a single workflow via AI model IDs.

[0194] Although this disclosure has been described in conjunction with embodiments that are considered to be most practical and preferred, it should be understood that this disclosure is not limited to the disclosed embodiments, but is intended to cover various arrangements made without departing from the widest interpretation of the appended claims.

Claims

1. A channel state information (CSI) processing method based on artificial intelligence (AI) / machine learning (ML), executed by a base station, characterized in that, include: Receives CSI processing capabilities based on AI / ML reported by user equipment (UE); as well as Based on the AI / ML-based CSI processing capability, an AI / ML-based CSI measurement configuration is determined, wherein the AI / ML-based CSI measurement configuration includes one or more of the following: CSI resource scheduling configuration, CSI reporting configuration, and CSI measurement triggering configuration.

2. The method according to claim 1, characterized in that, Determining the AI / ML-based CSI measurement configuration based on the AI / ML-based CSI processing capabilities includes: The AI / ML-based CSI measurement configuration occupies less than or equal to the AI / ML-based CSI processing capacity. The weighted value of the AI / ML-based CSI processing resources occupied by the AI / ML-based CSI measurement configuration is less than or equal to the AI / ML-based CSI processing capacity; or The sum of the CSI processing resources occupied by the AI / ML-based CSI measurement configuration and the CSI processing resources occupied by the non-AI / ML-based CSI measurement is less than or equal to the processing capacity of the UE.

3. The method according to claim 1, characterized in that, The UE's AI / ML-based CSI processing capabilities include one or more of the following: The number of CPUs in the AI / ML-based CSI processing unit of the UE. The number of AI / ML-based CSI processing processes for the UE. The number of AI / ML-based CSI storage units in the UE. The number of AI / ML-based CSI computing units in the UE, and The number of AI / ML-based CSI power consumption units in the UE.

4. The method according to claim 3, characterized in that, The AI / ML-based CSI processing capability of the UE is determined based on one or more of the following: Subcarrier spacing (SCS) CSI parameter information based on AI / ML, and Features of CSI models based on AI / ML.

5. The method according to claim 2, characterized in that, The AI / ML-based CSI processing resources include one or more of the following: The number of CPUs of the AI / ML-based CSI processing unit of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI processing processes of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI storage units occupied by the AI / ML-based CSI measurement configuration of the UE; The number of AI / ML-based CSI calculation units of the UE occupied by the AI / ML-based CSI measurement configuration; and The number of AI / ML-based CSI power consumption units of the UE occupied by the AI / ML-based CSI measurement configuration.

6. The method according to claim 2, characterized in that, The AI / ML-based CSI processing resources are determined based on one or more of the following: Subcarrier spacing (SCS); CSI parameter information based on AI / ML; and CSIAI / ML model features.

7. The method according to claim 2, characterized in that, The AI / ML-based CSI processing resources of the UE are obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the corresponding AI / ML-based CSI processing resources of the UE are pre-configured or pre-defined by the base station; and The base station receives a message from the UE, wherein the message includes one or more of the following: The UE's AI / ML-based CSI processing resources; Instructions indicating that AI / ML-based CSI processing resources and non-AI / ML-based CSI processing resources are separated from each other.

8. The method according to claim 2, characterized in that, The AI / ML-based CSI processing resources of the UE are obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the corresponding AI / ML-based CSI processing resources of the UE are pre-configured or pre-defined by the base station; and The base station receives a message from the UE, wherein the message includes one or more of the following: The UE's AI / ML-based CSI processing resources; Indication information indicating that the AI / ML-based CSI processing resources and the non-AI / ML-based CSI processing resources belong to the same common CSI processing unit CPU resource pool.

9. A base station, characterized in that, include: A processor is configured to invoke and run a computer program stored in memory to cause the device in which the processor resides to perform the method of any one of claims 1 to 8.

10. A chip, characterized in that, include: The processor is configured to invoke and run a computer program stored in memory to cause the device in which the chip resides to perform the method of any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, It contains a computer program that causes a computer to perform the method according to any one of claims 1 to 8.

12. A computer program product, characterized in that, Includes a computer program, wherein the computer program causes a computer to perform the method of any one of claims 1 to 8.

13. A channel state information (CSI) processing method based on AI / ML, executed by the UE, characterized in that, include: The UE reports its CSI processing capabilities based on AI / ML; Receives AI / ML-based CSI measurement configuration from the base station, wherein the AI / ML-based CSI measurement configuration is based on the AI / ML-based CSI processing capability; and The first operation is performed according to the AI / ML-based CSI measurement configuration; The AI / ML-based CSI measurement configuration includes one or more of the following: CSI resource scheduling configuration, CSI reporting configuration, and CSI measurement triggering configuration.

14. The method according to claim 13, characterized in that, The AI / ML-based CSI measurement configuration is based on the AI / ML-based CSI processing capabilities and includes: The AI / ML-based CSI measurement configuration occupies AI / ML-based CSI processing resources that are less than or equal to the AI / ML-based CSI processing capacity; or The weighted value of the AI / ML-based CSI processing resources occupied by the AI / ML-based CSI measurement configuration is less than or equal to the AI / ML-based CSI processing capacity; or The sum of the CSI processing resources occupied by the AI / ML-based CSI measurement configuration and the CSI processing resources occupied by the non-AI / ML-based CSI measurement is less than or equal to the CSI processing capacity of the UE.

15. The method according to claim 13, characterized in that, The UE's AI / ML-based CSI processing capabilities include one or more of the following: The number of AI / ML-based CSI processing units (CPUs) in the UE; The number of AI / ML-based CSI processing processes for the UE; The number of AI / ML-based CSI storage units in the UE; The number of AI / ML-based CSI computing units in the UE; and The number of AI / ML-based CSI power consumption units in the UE.

16. The method according to claim 15, characterized in that, The AI / ML-based CSI processing capability of the UE is determined based on one or more of the following: Subcarrier spacing (SCS); CSI parameter information based on AI / ML; and CSIAI / ML model features.

17. The method according to claim 14, characterized in that, The AI / ML-based CSI processing resources include one or more of the following: The number of CPUs of the AI / ML-based CSI processing unit of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI processing processes of the UE occupied by the AI / ML-based CSI measurement configuration; The number of AI / ML-based CSI storage units occupied by the AI / ML-based CSI measurement configuration of the UE; The number of AI / ML-based CSI calculation units of the UE occupied by the AI / ML-based CSI measurement configuration; and The number of AI / ML-based CSI power consumption units of the UE occupied by the AI / ML-based CSI measurement configuration.

18. The method according to claim 17, characterized in that, The AI / ML-based CSI processing resources are determined based on one or more of the following: Subcarrier spacing (SCS); CSI parameter information based on AI / ML; and CSIAI / ML model features.

19. The method according to claim 17, characterized in that, The AI / ML-based CSI processing resources of the UE are obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the corresponding AI / ML-based CSI processing resources of the UE are pre-configured or pre-defined; and The base station receives a message from the UE, wherein the message includes one or more of the following: The UE's AI / ML-based CSI processing resources; Instructions indicating that AI / ML-based CSI processing resources and non-AI / ML-based CSI processing resources are separated from each other.

20. The method according to claim 17, characterized in that, The AI / ML-based CSI processing resources of the UE are obtained through one or more of the following: The AI / ML-based CSI measurement configuration and the corresponding AI / ML-based CSI processing resources of the UE are pre-configured or pre-defined; and The base station receives a message from the UE, wherein the message includes one or more of the following: The UE's AI / ML-based CSI processing resources; Indication information indicating that the AI / ML-based CSI processing resources and the non-AI / ML-based CSI processing resources belong to the same common CSI processing unit CPU resource pool.

21. The method according to claim 13, characterized in that, When the AI / ML-based CSI measurement configuration exceeds the UE's AI / ML-based CSI processing capability, the UE discards the AI / ML-based CSI measurement configuration that exceeds the AI / ML-based CSI processing capability; and When the AI / ML-based CSI measurement configuration is less than or equal to the UE's AI / ML-based CSI processing capability, the UE performs the first operation according to the AI / ML-based CSI measurement configuration.

22. The method according to claim 13, characterized in that, When the weighted sum calculated based on the AI / ML-based CSI measurement configuration and its weights exceeds the UE's AI / ML-based CSI processing capability, the UE discards the AI / ML-based CSI measurement configuration that exceeds the AI / ML-based CSI processing capability; and When the weighted sum is less than or equal to the UE's AI / ML-based CSI processing capability, the UE performs the first operation according to the AI / ML-based CSI measurement configuration.

23. A UE, characterized in that, include: A processor is configured to invoke and run a computer program stored in memory to cause the device in which the processor resides to perform the method of any one of claims 13 to 22.

24. A chip, characterized in that, include: The processor is configured to invoke and run a computer program stored in memory to cause the device in which the chip resides to perform the method of any one of claims 13 to 22.

25. A computer-readable storage medium, characterized in that, It contains a computer program that causes a computer to perform the method of any one of claims 13 to 22.

26. A computer program product, characterized in that, Includes a computer program, wherein the computer program causes a computer to perform the method of any one of claims 13 to 22.

27. A computer program, characterized in that, The computer program causes the computer to perform the method according to any one of claims 13 to 22.

28. A channel state information (CSI) processing method based on AI / ML, executed by the UE, characterized in that, include: Receive AI / ML-based CSI measurement configuration and AI / ML-based CSI reference signal; Wherein, the time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time; and The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability.

29. The method according to claim 28, characterized in that, The first processing time includes one or more of the following: The time required to decode the AI / ML-based CSI measurement configuration; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; as well as The duration required for UE antenna switching.

30. The method according to claim 28, characterized in that, The second processing time includes one or more of the following: Used for part or all of the time required to activate AI / ML models; The time required for AI / ML model switching; and The time required for AI / ML model inference.

31. The method according to claim 28, characterized in that, The first processing time and / or the second processing time are determined by one or more of the following: CSI measurement data based on AI / ML; CSI measurement features or functions based on AI / ML; CSI measurement time based on AI / ML; AI / ML model parameters used for CSI measurements; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; and Subcarrier spacing (SCS).

32. The method according to claim 28, characterized in that, The first processing time and / or the second processing time are obtained by one or more of the following: Predefined or preconfigured; Empirical values ​​obtained based on statistics of the time consumed in executing the AI / ML-based CSI measurement configuration; and The values ​​of the first processing time and / or the second processing time carried in the uplink message, wherein the uplink message is transmitted in UE capability reporting, scheduling request, physical uplink shared channel PUSCH or physical uplink control channel PUCCH.

33. The method according to claim 30, characterized in that, The second processing time includes a portion or all of the time required for AI / ML model activation; If the AI / ML model activation begins after receiving the AI / ML-based CSI reference signal, the second processing time includes the total duration required for AI / ML model activation. If the AI / ML model activation begins before receiving the AI / ML-based CSI reference signal and completes after receiving the AI / ML-based CSI reference signal, then the second processing time includes a portion of the time required for AI / ML model activation; and If the AI / ML model activation is completed before the AI / ML-based CSI reference signal is received, the second processing time does not include the time required for AI / ML model activation.

34. A UE, characterized in that, include: A processor is configured to invoke and run a computer program stored in memory to cause the device in which the processor resides to perform the method of any one of claims 28 to 33.

35. A chip, characterized in that, include: The processor is configured to invoke and run a computer program stored in memory to cause the device in which the chip resides to perform the method of any one of claims 28 to 33.

36. A computer-readable storage medium, characterized in that, It contains a computer program that causes a computer to perform the method of any one of claims 28 to 33.

37. A computer program product, characterized in that, Includes a computer program, wherein the computer program causes a computer to perform the method of any one of claims 28 to 33.

38. A computer program, characterized in that, The computer program causes the computer to perform the method according to any one of claims 28 to 33.

39. A channel state information (CSI) processing method based on AI / ML, executed by a base station, characterized in that, include: The base station sends an AI / ML-based CSI measurement configuration and an AI / ML-based CSI reference signal to enable the UE to perform CSI measurements based on the AI / ML-based CSI measurement configuration and the AI / ML-based CSI reference signal. The time interval from receiving the AI / ML-based CSI measurement configuration to CSI reporting is less than or equal to a first processing time, or the time interval from receiving the AI / ML-based CSI reference signal to CSI reporting is less than or equal to a second processing time. The first processing time and the second processing time are determined by the UE's AI / ML-based CSI processing capability. as well as Receive the results reported by the CSI.

40. The method according to claim 39, characterized in that, The first processing time includes one or more of the following: The time required to decode the AI / ML-based CSI measurement configuration; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; as well as The duration required for UE antenna switching.

41. The method according to claim 39, characterized in that, The second processing time includes one or more of the following: Used for part or all of the time required to activate AI / ML models; The time required for AI / ML model switching; and The time required for AI / ML model inference.

42. The method according to claim 39, characterized in that, The first processing time and / or the second processing time are determined by one or more of the following: CSI measurement data based on AI / ML; CSI measurement features or functions based on AI / ML; CSI measurement time based on AI / ML; AI / ML model parameters used for CSI measurements; The time required for AI / ML model activation; Time required for AI / ML model switching; The time required for AI / ML model inference; and Subcarrier spacing (SCS).

43. The method according to claim 39, characterized in that, The first processing time and / or the second processing time are obtained by one or more of the following: Predefined or preconfigured; Empirical values ​​obtained based on statistics of the time consumed in executing the AI / ML-based CSI measurement configuration; and The values ​​of the first processing time and / or the second processing time carried in the uplink message, wherein the uplink message is transmitted in UE capability reporting, scheduling request, physical uplink shared channel PUSCH or physical uplink control channel PUCCH.

44. The method according to claim 41, characterized in that, The second processing time includes a portion or all of the time required for AI / ML model activation; If the AI / ML model activation begins after receiving the AI / ML-based CSI reference signal, the second processing time includes the total duration required for AI / ML model activation. If the AI / ML model activation begins before receiving the AI / ML-based CSI reference signal and completes after receiving the AI / ML-based CSI reference signal, then the second processing time includes a portion of the time required for AI / ML model activation; and If the AI / ML model activation is completed before the AI / ML-based CSI reference signal is received, the second processing time does not include the time required for AI / ML model activation.

45. A base station, characterized in that, include: A processor is configured to invoke and run a computer program stored in memory to cause the device in which the processor resides to perform the method of any one of claims 39 to 44.

46. ​​A chip, characterized in that, include: The processor is configured to invoke and run a computer program stored in memory to cause the device in which the chip resides to perform the method of any one of claims 39 to 44.

47. A computer-readable storage medium, characterized in that, It contains a computer program that causes a computer to perform the method of any one of claims 39 to 44.

48. A computer program product, characterized in that, Includes a computer program, wherein the computer program causes a computer to perform the method of any one of claims 39 to 44.

49. A computer program, characterized in that, The computer program causes the computer to perform the method according to any one of claims 39 to 44.

Images