Managing machine learning based channel state information reporting at user equipment

Abstract

This disclosure provides systems, apparatuses, apparatuses, and methods, including computer programs encoded on a storage medium, for managing ML-based CSI reporting at a UE. The UE (102) receives (406) an ML model performance report configuration from the network entity (104) and sends (408, 508c) an ML model performance report prepared according to the ML model performance report configuration to the network entity (104). The ML model performance report conveys the performance of the current ML model used to compress the CSI. The performance is based on a comparison of the CSI to the decompressed output of the current ML model.

Description

Cross - Reference to Related Applications

[0001] This application claims the benefit and priority of U.S. Provisional Application No. 63 / 422,863, titled "MANAGING MACHINE LEARNING BASED CHANNEL STATE INFORMATION REPORTING AT A USER EQUIPMENT", filed on Nov. 4, 2022, and U.S. Provisional Application No. 63 / 454,388, titled "MANAGING MACHINE LEARNING BASED CHANNEL STATE INFORMATION REPORTING AT A USER EQUIPMENT", filed on Mar. 24, 2023. The entire content of each of the above - mentioned applications is hereby expressly incorporated by reference herein. Technical Field

[0002] The present disclosure generally relates to wireless communication, and more particularly, to managing machine - learning (ML) - based channel state information (CSI) reporting at a user equipment (UE). Background Art

[0003] The 3rd Generation Partnership Project (3GPP) has specified a radio interface called the 5th Generation (5G) New Radio (5G NR). The architecture of a 5G NR wireless communication system includes a 5G Core (5GC) network, a 5G Radio Access Network (5G - RAN), user equipment (UE), etc. Compared with the previous generation of cellular communication systems, the 5G NR architecture seeks to provide increased data rates, reduced latency, and / or increased capacity.

[0004] Wireless communication systems can generally be configured to provide various telecommunication services (such as, for example, telephone, video, data, messaging, broadcasting, etc.) based on a multiple - access technology (such as orthogonal frequency - division multiple access (OFDMA) technology) that supports communication with multiple UEs. Improvements in mobile broadband continue the development of such wireless communication technologies. For example, machine - learning (ML) models may improve wireless performance, but ML models may also experience performance failures due to certain types of channel conditions or as a result of channel blockage. Further, UEs and / or networks may encounter difficulties in managing ML - based channel state information (CSI) reporting. Summary of the Invention

[0005] A simplified overview of one or more aspects is presented below to provide a basic understanding of such aspects. This overview is not an extensive review of all contemplated aspects. This overview neither identifies key or critical elements of all aspects nor describes the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.

[0006] A user equipment (UE) may utilize a machine learning (ML) model to compress channel state information (CSI), thereby generating an ML-based CSI report that is shorter than a non-ML-based CSI report. The CSI report is sent to a network entity (NE), such as a base station or an entity of a base station. The UE or the NE may assess the performance of using ML-based compression by comparing the result of decompressing the CSI with the uncompressed CSI. The performance of using ML compression may degrade over time or may be unsatisfactory for some types of channel conditions. For example, if an ML model is trained using offline field data associated with some channel conditions that do not include less common channel conditions (LCCC), the performance of compressing CSI using the trained model may be below a threshold when such LCCC conditions occur. Additionally, a change in the channel as a result of channel blocking may also cause a degradation in the performance of ML-based CSI compression.

[0007] Aspects presented herein propose that the UE optimizes the use of an ML model to compress CSI by monitoring the performance of the ML model, reporting the performance to the NE, and causing corrective actions if necessary. The NE may adjust the ML model based on detected performance failures. For example, the NE may update / switch the ML model or fallback to a non-ML communication technique with the UE. One or both of the UE and the NE may be able to detect an ML model failure. Whichever entity detects the ML model failure may then indicate the ML model failure to the other entity. Based on the ML model monitoring, the UE and the NE may adjust the CSI compression using the ML model.

[0008] According to some aspects, the UE receives an ML model performance report configuration from the NE and sends an ML model performance report prepared according to the ML model performance report configuration to the NE. The ML model performance report conveys the performance of the current ML model used for compressing CSI. This performance is based on a comparison of the CSI with the decompressed output of the current ML model.

[0009] According to some aspects, the NE sends an ML model performance report configuration to the UE and receives an ML model performance report prepared according to the ML model performance report configuration from the UE, as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1A diagram showing a wireless communication system including multiple user equipments (UEs) and network entities communicating via one or more cells.

[0011] Figures 2A to 2B A diagram showing an example process for machine learning (ML)-based channel state information (CSI) compression at a UE.

[0012] Figures 2C to 2D A diagram showing an example process for ML model performance evaluation at a UE.

[0013] Figures 2E to 2F A diagram showing an example process for ML model performance evaluation at a network entity.

[0014] Figures 3A to 3E A signaling diagram showing an example of ML model performance monitoring.

[0015] Figures 4A to 4B A flowchart of a wireless communication method associated with ML model performance reporting.

[0016] Figures 5A to 5B A flowchart of a wireless communication method associated with ML model performance reporting.

[0017] Figures 6A to 6C A flowchart of a wireless communication method based on conditions for performing ML model performance reporting.

[0018] Figure 7 A flowchart of a wireless communication method for requesting an ML-based CSI reporting configuration from a network.

[0019] Figure 8 A flowchart of a wireless communication method for receiving a reporting configuration based on a request.

[0020] Figure 9 A flowchart of a wireless communication method for reconfiguring a UE's reporting configuration based on a request.

[0021] Figure 10 A diagram showing a hardware implementation for an example UE device.

[0022] Figure 11 A diagram showing a hardware implementation for one or more example network entities. Detailed Description

[0023] Figure 1FIG. 100 shows a wireless communication system associated with multiple cells 190. The wireless communication system includes a user equipment (UE) 102 and a base station / network entity 104. Some base stations may include an aggregated base station architecture, while other base stations may include a disaggregated base station architecture. The aggregated base station architecture includes a radio unit (RU) 106, a distributed unit (DU) 108, and a central unit (CU) 110, which are configured to utilize a radio protocol stack physically or logically integrated within a single radio access network (RAN) node. The disaggregated base station architecture utilizes a protocol stack physically or logically distributed among two or more units (e.g., RU 106, DU 108, CU 110). For example, the CU 110 is implemented within a RAN node, and one or more DUs 108 may be co-located with the CU 110, or alternatively, may be geographically or virtually distributed among one or more other RAN nodes. The DU 108 may be implemented to communicate with one or more RUs 106. Each of the RU 106, DU 108, and CU 110 may be implemented as a virtual unit, such as a virtual radio unit (VRU), a virtual distributed unit (VDU), or a virtual central unit (VCU). The base station / network entity 104 (e.g., an aggregated base station or a disaggregated unit of a base station, such as RU 106, DU 108, or CU 110) may be referred to as a transmission and reception point (TRP).

[0024] The operation and / or network design of the base station 104 may be based on the aggregation characteristics of the base station functions. For example, a disaggregated base station architecture is utilized in an integrated access backhaul (IAB) network, an open radio access network (O-RAN) network, or a virtualized radio access network (vRAN) (which may also be referred to as a cloud radio access network (C-RAN)). The disaggregation may include distributing functions across two or more units at various physical locations, as well as virtually distributing the functions of at least one unit, which can achieve flexibility in network design. The various units of the disaggregated base station architecture or the disaggregated RAN architecture may be configured to communicate with at least one other unit either wired or wirelessly. For example, the base stations 104a / 104e and / or the RUs 106a to 106d may communicate with the UEs 102a to 102d and 102s via one or more radio frequency (RF) access links based on the Uu interface. In an example, multiple RUs 106 and / or base stations 104 may serve the UE 102 simultaneously, such as via intra-cell and / or inter-cell access links between the UE 102 and the RU 106 / base station 104.

[0025] RU 106, DU 108, and CU 110 may include (or may be coupled to) one or more interfaces configured to send or receive information / signals via a wired or wireless transmission medium. Base station 104 or any one of the one or more distributed base station units may be configured to communicate with one or more other base stations 104 or one or more other distributed base station units via a wired or wireless transmission medium. In an example, a processor, memory, and / or controller associated with the executable instructions of the interface may be configured to provide communication between base station 104 and / or one or more distributed base station units via a wired or wireless transmission medium. For example, a wired interface may be configured to send or receive information / signals via a wired transmission medium, such as via a fronthaul link 160 between RU 106d and the baseband unit (BBU) 112 of base station 104d associated with cell 190d. BBU 112 includes DU 108 and CU 110, which may also have a wired interface (e.g., a midhaul link) configured between DU 108 and CU 110 to send or receive information / signals between DU 108d and CU 110d. In a further example, a wireless interface that may include a receiver, transmitter, or transceiver (such as an RF transceiver) is configured to send and / or receive information / signals via a wireless transmission medium, such as information transmitted between RU 106a of cell 190a and base station 104e of cell 190e via an inter-cell communication beam 136 - 138 of RU 106a and base station 104e.

[0026] RU 106 may be configured to implement lower layer functions. For example, RU 106 is controlled by DU 108 and may correspond to a logical node hosting RF processing functions or lower layer PHY functions, such as performing fast Fourier transform (FFT), inverse FFT (iFFT), digital beamforming, physical random access channel (PRACH) extraction and filtering, etc. The functions of RU106 may be based on a functional division, such as a lower layer functional division.

[0027] RU 106 can send or receive over-the-air (OTA) communications with one or more UEs 102. For example, RU 106b of cell 190b communicates with UE 102b of cell 190b via the first communication beam set 132 of RU 106b and the second communication beam set 134b of UE 102b, and these two communication beam sets may correspond to inter-cell communication beams or in some examples correspond to cross-cell communication beams. For example, UE 102b of cell 190b can communicate with RU 106a of cell 190a via the third communication beam set 134a of UE 102b and the fourth communication beam set 136 of RU 106a. The real-time and non-real-time characteristics of the control plane and user plane communications of RU 106 can be controlled by the associated DU 108.

[0028] Any combination of RU 106, DU 108, and CU 110 or a reference to them individually can correspond to base station 104. Thus, base station 104 can include at least one of RU 106, DU 108, or CU 110. Base station 104 provides access to the core network for UE 102. Base station 104 may relay communications between UE 102 and the core network. Base station 104 can be associated with a macro cell of a high-power cellular base station and / or a small cell of a low-power cellular base station. For example, cell 190e may correspond to a macro cell, while cells 190a to 190d may correspond to small cells. Small cells include femtocells, picocells, microcells, etc. A cell structure including at least one macro cell and at least one small cell can be referred to as a "heterogeneous network".

[0029] Transmissions from UE 102 to base station 104 / RU 106 are referred to as uplink (UL) transmissions, while transmissions from base station 104 / RU 106 to UE 102 are referred to as downlink (DL) transmissions. Uplink transmissions can also be referred to as reverse link transmissions, while downlink transmissions can also be referred to as forward link transmissions. For example, RU 106d uses the antennas of base station 104d of cell 190d to send downlink / forward link communications to UE 102d or receive uplink / reverse link communications from UE 102d based on the Uu interface associated with the access link between UE 102d and base station 104d / RU 106d.

[0030] The communication link between the UE 102 and the base station 104 / RU 106 can be based on multiple-input multiple-output (MIMO) antenna technology, including spatial multiplexing, beamforming, and / or transmit diversity. The communication link can be associated with one or more carriers. The UE 102 and the base station 104 / RU 106 can utilize a spectral bandwidth of Y MHz per carrier (e.g., 5 MHz, 10 MHz, 15 MHz, 20 MHz, 100 MHz, 400 MHz, 800 MHz, 1600 MHz, 2000 MHz, etc.) allocated in carrier aggregation of up to a total of Yx MHz, where x component carriers (CCs) are used for communication in each of the uplink direction and the downlink direction. The carriers can be adjacent to each other along the spectrum or may not be adjacent to each other. In an example, the uplink carriers and the downlink carriers can be allocated in an asymmetric manner, and more or fewer carriers can be allocated for the uplink or the downlink. The component carriers can include a primary component carrier and one or more secondary component carriers. The primary component carrier can be associated with a primary cell (PCell), and the secondary component carriers can be associated with secondary cells (SCells).

[0031] Some UEs 102 (such as UEs 102a and 102s) can perform device-to-device (D2D) communication via a sidelink. For example, the sidelink communication / D2D link utilizes the spectrum of the wireless wide area network (WWAN) associated with the uplink communication and the downlink communication. The sidelink communication / D2D link can also use one or more sidelink channels, such as the physical sidelink broadcast channel (PSBCH), the physical sidelink discovery channel (PSDCH), the physical sidelink shared channel (PSSCH), and / or the physical sidelink control channel (PSCCH), to transmit information between the UEs 102a and 102s. Such sidelink / D2D communication can be performed via various wireless communication systems, such as a wireless fidelity (Wi-Fi) system, a Bluetooth system, a long term evolution (LTE) system, a new radio (NR) system, etc.

[0032] The electromagnetic spectrum is typically subdivided into different categories, bands, channels, etc. based on the different frequencies / wavelengths associated with the electromagnetic spectrum. Fifth-generation (5G) NR is typically associated with two operating frequency ranges (FRs) called Frequency Range 1 (FR1) and Frequency Range 2 (FR2). The FR1 ranges from 410 MHz to 7.125 GHz, and the FR2 ranges from 24.25 GHz to 71.0 GHz, where the FR2 includes FR2-1 (24.25 GHz to 52.6 GHz) and FR2-2 (52.6 GHz to 71.0 GHz). Although a portion of FR1 is actually greater than 6 GHz, FR1 is typically referred to as the "sub-6 GHz" band. In contrast, FR2 is typically referred to as the "millimeter wave" (mmW) band. FR2 is different from the "extremely high frequency" (EHF) band, but is an approximate subset of this band, where the EHF band ranges from 30 GHz to 300 GHz and is sometimes also referred to as the "millimeter wave" band. The frequencies between FR1 and FR2 are typically referred to as "mid-band" frequencies. The operating frequency range for mid-band frequencies can be referred to as Frequency Range 3 (FR3), which ranges from 7.125 GHz to 24.25 GHz. The bands within FR3 can include characteristics of FR1 and / or FR2. Thus, the characteristics of FR1 and / or FR2 can extend to mid-band frequencies. Higher operating frequency ranges have been identified to extend 5G NR communication above 52.6 GHz associated with the upper limit of FR2. Three of these higher operating frequency ranges include FR2-2 (ranging from 52.6 GHz - 71.0 GHz), FR4 (ranging from 71.0 GHz to 114.25 GHz), and FR5 (ranging from 114.25 GHz to 300 GHz). The upper limit of FR5 corresponds to the upper limit of the EHF band. Thus, unless explicitly stated otherwise herein, the term "sub-6 GHz" can refer to frequencies less than 6 GHz, frequencies within FR1, or frequencies that can include mid-band frequencies. Further, unless explicitly stated otherwise herein, the term "millimeter wave" or mmW refers to frequencies that can include mid-band frequencies, frequencies that can be within FR2-1, FR4, FR2-2, and / or FR5, or frequencies that can be within the EHF band.

[0033] UE 102 and base station 104 / RU 106 may each include multiple antennas. The multiple antennas may correspond to antenna elements, antenna panels, and / or antenna arrays that may facilitate beamforming operations. For example, RU 106b transmits a downlink beamformed signal to UE 102b based on a first communication beam set 132 in one or more transmission directions of RU 106b. UE 102b may receive the downlink beamformed signal from RU 106b based on a second communication beam set 134b in one or more reception directions of UE 102b. In a further example, UE 102b may also transmit an uplink beamformed signal to RU 106b based on a second communication beam set 134b in one or more transmission directions of UE 102b. RU 106b may receive the uplink beamformed signal from UE 102b in one or more reception directions of RU 106b.

[0034] UE 102b may perform beam training to determine the optimal reception and transmission directions for beamformed signals. The transmission and reception directions of UE 102 and base station 104 / RU 106 may be the same or may be different. In a further example, the beamformed signal may be transmitted between a first base station / RU 106a and a second base station 104e. For example, base station 104e of cell 190e may transmit a beamformed signal to RU 106a based on a communication beam 138 in one or more transmission directions of base station 104e. RU 106a may receive the beamformed signal from base station 104e of cell 190e based on an RU communication beam 136 in one or more reception directions of RU 106a. In a further example, base station 104e transmits a downlink beamformed signal to UE 102e based on a communication beam 138 in one or more transmission directions of base station 104e. UE 102e receives the downlink beamformed signal from base station 104e based on a UE communication beam 130 in one or more reception directions of UE 102e. UE 102e may also transmit an uplink beamformed signal to base station 104e based on a UE communication beam 130 in one or more transmission directions of UE 102e, such that base station 104e may receive the uplink beamformed signal from UE 102e in one or more reception directions of base station 104e.

[0035] Base station 104 may include and / or be referred to as a network entity. That is, a "network entity" may refer to base station 104 or at least one unit of base station 104, such as RU 106, DU 108, and / or CU 110. Base station 104 may also include and / or be referred to as a next-generation evolved Node B (ng-eNB), a generation Node B (gNB), an evolved Node B (eNB), an access point, a base station transceiver, a radio base station, a radio transceiver, a transceiver function, a basic service set (BSS), an extended service set (ESS), a TRP, a network node, a network device, or other related terms. The entity at base station 104 or base station 104 may be implemented as an IAB node, a relay node, a sidelink node, an aggregated (monolithic) base station having RU 106 and a BBU 112 including DU 108 and CU 110, or may be implemented as a disaggregated base station including one or more of RU 106, DU 108, and / or CU 110. The set of aggregated or disaggregated base stations may be referred to as a next-generation radio access network (NG-RAN). In some examples, UE 102a operates in dual connectivity (DC) with base station 104e and base station / RU 106a. In such a case, base station 104e may be the master node, and base station / RU 160a may be the secondary node.

[0036] Uplink / downlink signaling may also communicate via a satellite positioning system (SPS) 114. In an example, the SPS 114 of cell 190c may communicate with one or more UEs 102, such as UE 102c, and one or more base stations 104 / RUs 106, such as RU 106c. The SPS 114 may correspond to one or more of a global navigation satellite system (GNSS), a global positioning system (GPS), a non-terrestrial network (NTN), or other satellite positioning / positioning systems. The SPS 114 may be associated with an LTE signal, an NR signal (e.g., based on round-trip time (RTT) and / or multi-RTT), a wireless local area network (WLAN) signal, a terrestrial beacon system (TBS), sensor-based information, NR enhanced cell ID (NR E-CID) technology, downlink angle of departure (DL-AoD), downlink time difference of arrival (DL-TDOA), uplink time difference of arrival (UL-TDOA), uplink angle of arrival (UL-AoA), and / or other systems, signals, or sensors.

[0037] Still referring to Figure 1, in some aspects, any of the UEs 102 may include a model performance reporting component 140 configured to receive a machine learning (ML) model performance reporting configuration from a network entity; and send to the network entity an ML model performance report prepared according to the ML model performance reporting configuration, the ML model performance report conveying the performance of the current ML model for compressing channel state information (CSI), the performance being based on a comparison of the CSI with the decompressed output of the current ML model.

[0038] In some aspects, any of the base stations 104 or a network entity of the base stations 104 may include a model performance configuration component 150 configured to send an ML model performance reporting configuration to a UE; and receive from the UE an ML model performance report prepared according to the ML model performance reporting configuration, the ML model performance report conveying the performance of the current ML model for compressing CSI, the performance being based on a comparison of the CSI with the decompressed output of the current ML model.

[0039] Thus, Figure 1 a wireless communication system is described that may be implemented in conjunction with aspects of one or more other figures described herein (such as Figures 2A to 3E the aspects shown in). Further, although the following description may focus on 5G NR, the concepts described herein may be applicable to other similar fields, such as 5G-Advanced and future releases, LTE, LTE-advanced (LTE-A), and other wireless technologies such as 6G.

[0040] Similar to Figure 2D , Figure 2AFIG. 205 shows an example process for an ML-based CSI compressor and / or encoder at UE 102 and an ML-based CSI decompressor and / or decoder at network entity 104. UE 102 and network entity 104 (such as a base station or an entity of a base station) may perform multi-input multi-output (MIMO) communication, where network entity 104 may use CSI to select a digital precoder (e.g., a precoding matrix) for UE 102. Network entity 104 may configure CSI reporting from UE 102 via RRC signaling (e.g., CSI-ReportConfig), where UE 102 may use the first CSI-RS 240 as a channel measurement resource (CMR) for UE 102 to measure the downlink channel. Network entity 104 may also configure the second CSI-RS as an interference measurement resource (IMR) for UE 102 (e.g., via CSI-ReportConfig) to measure interference to the downlink channel. Thus, UE 102 may estimate 250a the channel between UE 102 and network entity 104 based on CSI-RS 240 and obtain (e.g., determine and / or generate) (raw) CSI.

[0041] Then, UE 102 performs CSI compression 270a (e.g., an AI / ML-based CSI generator) on the raw CSI to obtain compressed CSI. UE 102 includes the compressed CSI in a CSI report 285 and transmits 280a the CSI report 285 to network entity 104. In some implementations, network entity 104 includes a rank indicator (RI), a precoding matrix indicator (PMI), a channel quality indicator (CQI), a layer indicator (LI), and / or a layer 1 reference signal received power (L1-RSRP) in the CSI report 285, as described for Figure 2D . In other implementations, UE 102 avoids including RI, PMI, CQI, LI, L1-RSRP, layer 1 reference signal received quality (L1-RSRQ), and / or layer 1 signal-to-noise and interference ratio (L1-SINR) in the CSI report 285.

[0042] Similar to Figure 2A , Figure 2BFIG. 215 shows an example process for CSI-RS based AI / ML model performance monitoring and evaluation, except that UE 102 includes a neural network 270b for CSI decompression and a neural network performance evaluation 290 for evaluating or determining the performance of neural network 270a for CSI compression. When UE 102 performs CSI compression 270a to obtain compressed CSI or after that, UE 102 performs CSI decompression 270b on the compressed CSI to obtain decompressed CSI. Then, UE 102 performs neural network performance evaluation 290 based on the decompressed CSI (inferred CSI) and the original CSI (ground truth CSI) to evaluate the AI / ML model inference accuracy, as described for Figure 2E as described.

[0043] Similar to Figure 2F , Figure 2C FIG. 225 shows an example process for SRS based AI / ML model performance monitoring, except that network entity 104 includes a neural network 270b for CSI decompression as described for Figure 2A as described. Network entity 104 directly performs CSI compression 270a on the original CSI to obtain compressed CSI, and performs CSI decompression 270b on the compressed CSI to obtain decompressed CSI. Then, network entity 104 performs neural network performance evaluation 290 based on the decompressed CSI (inferred CSI) and the original CSI (ground truth CSI), as described for Figure 2F as described.

[0044] Figure 2D , Figure 2E and Figure 2F The pairs of CSI compression 270a and CSI decompression 270b in Figure 2A , Figure 2B and Figure 2C differ from the pairs of CSI compression 270a and CSI decompression 270b in Figure 2D , Figure 2E and Figure 2F in that the inputs and outputs of pairs 270a and 270b in Figure 2A , Figure 2B and Figure 2C are precoding matrices, while the inputs and outputs of pairs 270a and 270b in Figure 2D , Figure 2E and Figure 2F are channel matrices. Due to the different training input data types (channel matrix or precoding matrix) at the AI / ML model training stage, thus Figure 2D , Figure 2E and Figure 2F the pairs 270a and 270b in Figure 2A , Figure 2B and Figure 2CThe AI / ML model weighting parameters between 270a and 270b in

[0045] Figure 2D FIG. 235 shows an example process for an ML-based CSI compression and / or encoder at UE 102 and an ML-based CSI decompression and / or decoder on network entity 104. UE 102 and network entity 104 (such as a base station or an entity of a base station) may perform multi-input multi-output (MIMO) communication, where network entity 104 may use CSI to select a digital precoder (e.g., a precoding matrix) for UE 102. Network entity 104 may configure CSI reporting from UE 102 via RRC signaling (e.g., CSI-ReportConfig), where UE 102 may use the first CSI-RS 240 as a CMR for UE 102 to measure the downlink channel. Network entity 104 may also configure the second CSI-RS as an IMR for UE 102 to measure interference to the downlink channel (e.g., via CSI-ReportConfig). The first CSI-RS and the second CSI-RS may be the same CSI-RS or different CSI-RSs. Thus, UE 102 may estimate the channel between UE 102 and network entity 104 based on CSI-RS 240 and obtain (e.g., determine and / or generate) (raw) CSI.

[0046] Then, UE 102 performs eigenvector calculation 260a on each subband and performs CSI compression 270a (e.g., an AI / ML-based CSI generator) on the (raw) CSI to obtain compressed CSI. UE 102 includes the compressed CSI in CSI report 285 and transmits 280a CSI report 285 to network entity 104. In some implementations, UE 102 includes RI, PMI, CQI, LI, and / or L1-RSRP in CSI report 285. CQI may indicate the signal-to-interference-plus-noise ratio (SINR) used to determine the modulation and coding scheme (MCS). LI may indicate the strongest layer, such as for multi-user (MU)-MIMO pairing for low-rank transmission and precoder selection 260b, as for phase-tracking reference signal (PT-RS). In other implementations, UE 102 avoids including RI, PMI, CQI, LI, L1-RSRP, L1-RSRQ, and / or L1-SINR in CSI report 285.

[0047] The network entity 104 may configure the time-domain behavior for the transmission 280a of the CSI report 285 to the network entity 104 (e.g., based on CSI-ReportConfig), such as periodic, semi-persistent, or aperiodic reporting. In an example, the network entity 104 may activate / deactivate semi-persistent CSI reporting from the UE 102 using a MAC control element (MAC-CE). The network entity 104 may trigger semi-persistent CSI reporting or aperiodic CSI reporting from the UE 102 based on the transmission of downlink control information (DCI) to the UE 102. The network entity 104 may receive periodic CSI from the UE 102 on a physical uplink control channel (PUCCH) resource (e.g., configured via CSI-ReportConfig). CSI-ReportConfig may also be used to configure the PUCCH resource for the transmission 280a of semi-persistent CSI reports to the network entity 104. In other examples, the transmission 280a of semi-persistent CSI reports to the network entity 104 may be on a physical uplink shared channel (PUSCH) resource triggered by DCI. In still other examples, the transmission 280a of semi-persistent CSI reports to the network entity 104 may be on a PUCCH resource activated by a MAC-CE. The UE 102 may also transmit 280a an aperiodic CSI report on a PUSCH resource triggered by DCI.

[0048] For a first resource element (RE) k associated with the CSI-RS 240, the received signal at the UE 102 may be determined based on the following: where indicates the effective channel including analog beamforming weights, where the dimension is N Rx by N Tx , corresponding to the CSI-RS 240 at RE k, corresponding to the interference plus noise, N Rx corresponding to the first number of receive ports, and N Tx corresponding to the second number of transmit ports.

[0049] For a second RE k associated with the physical downlink shared channel (PDSCH), the received signal at the UE 102 may be determined based on the following: where indicates the precoder. The network entity 104 may select 260b the same precoder for subcarriers within a subband (e.g., bundled in a physical resource block (PRB)).

[0050] UE 102 may use a type 2 CSI codebook for CSI measurement and reporting, where the precoder can be based on: where corresponding to a broadband precoder of dimension N Tx multiplied by 2L, corresponding to a subband precoder of dimension 2L multiplied by v, where L indicates the number of beams and v indicates the number of layers, which may correspond to RI + 1. can be based on the codebook, while can be based on the power and angle associated with each transmission. Since is subband-based and there may be multiple subbands for CSI report 285, UE 102 may experience large overhead to send CSI report 285 to network entity 104.

[0051] CSI report 285 can be based on the bandwidth used for CSI-RS 240. In an example, the codebook that the network entity can use to select 260b W1 can be based on: where corresponds to the Kronecker product; L indicates the number of beams that can be configured via RRC signaling; N 1 and N 2 correspond to the number of ports, O 1 and O 2 correspond to the oversampling factors in the horizontal and vertical domains that can be configured via RRC signaling. The candidate values of the oversampling factor can be based on the CSI-RS port number indicated via N 1 and N 2 The codebook can include precoders with different values of m and n. In some examples, the candidate values can be predefined based on a standardization protocol.

[0052] An ML model can be implemented to compress the CSI associated with 270a channel estimation 250a. The first v columns of the eigenvector of 260a calculated for each subband of the average channel can be used as the input to the ML model. In an example, the eigenvector can be input to a neural network at UE 102 for compression 270a of the CSI encoder. UE 102 sends CSI report 285 including the compressed CSI to network entity 104.

[0053] The network entity 104 detects the CSI report 285 sent 280a from the UE 102 and decodes the CSI report 285 including the compressed CSI. The decoded CSI report 285 including the compressed CSI can be input to a neural network at the network entity 104 for CSI decompression 270a. That is, the neural network at the network entity 104 can decompress 270b the compressed CSI to determine the decompressed CSI. The network entity 104 can determine a feature vector that serves as an input to the compression 270a of the CSI encoder at the UE 102 based on the decompressed CSI. The network entity 104 can select 260b a precoder for each subband based on the determined / reported feature vector. In some implementations, in cases where the CSI report 285 includes RI, PMI, CQI, LI, and / or L1-RSRP, the network entity 104 can jointly determine a digital precoder (e.g., a precoding matrix) or perform precoder selection 260b using the decompressed CSI, RI, PMI, CQI, LI, and / or L1-RSRP.

[0054] Similar to Figure 2D , Figure 2EFIG. 245 shows an example process for CSI-RS based AI / ML model performance monitoring and / or evaluation, except that UE 102 includes a neural network 270b for CSI decompression and a neural network (e.g., ML model) performance evaluation 290 for evaluating or determining the performance of neural network 270a for CSI compression. When UE 102 performs CSI compression 270a to obtain compressed CSI or thereafter, UE 102 performs CSI decompression 270b on the compressed CSI to obtain decompressed CSI. Then, UE 102 performs neural network performance evaluation 290 based on the decompressed CSI (inferred CSI) and the original CSI (ground truth CSI) to evaluate the AI / ML model inference accuracy. In performance evaluation 290, UE 102 determines the AI / ML model performance metrics based on the original CSI (ground truth CSI) and the decompressed CSI (inferred CSI), and evaluates the performance metrics against a performance metric threshold. For example, if the performance metric is higher than or equal to the performance metric threshold, UE 102 determines that the performance of neural network 270a for CSI compression is good. Otherwise, if the performance metric is lower than the performance metric threshold, UE 102 determines that the performance of neural network 270a for CSI compression is poor. In some implementations, UE 102 receives the performance metric threshold from network entity 104. In other implementations, UE 102 pre-determines or pre-stores the performance metric threshold. In yet other implementations, the performance metric threshold is defined or pre-defined in 3GPP specifications. In some implementations, the performance metric is the cosine similarity (value) between the original CSI (ground truth CSI) and the decompressed CSI (inferred CSI), and the performance metric threshold is the cosine similarity threshold.

[0055] If the UE 102 unconditionally or continuously evaluates the performance of the neural network 270a for CSI compression as described above, the UE 102 consumes a large amount of battery power. To save battery power, the UE 102 may determine whether to evaluate the performance of the neural network 270a for CSI compression based on one or more system performance metrics such as system throughput, block error rate (BLER), maximum number of HARQ retransmissions, reference signal received power (RSRP), reference signal received quality (RSRQ), and / or signal-to-noise-and-interference ratio (SINR). For example, if the UE 102 determines that one or more system performance metrics meet the corresponding one or more criteria, the UE 102 determines to evaluate the performance of the neural network 270a for CSI compression. Otherwise, if the UE 102 determines that one or more system performance metrics do not meet the corresponding one or more criteria, the UE 102 determines not to evaluate or stop evaluating the performance of the neural network 270a for CSI compression. In some implementations, if the UE 102 determines to evaluate the performance of the neural network 270a for CSI compression, the UE 102 activates the neural network 270b for CSI decompression. Otherwise, if the UE 102 determines not to evaluate or stop evaluating the performance of the neural network 270a for CSI compression, the UE 102 avoids activating or deactivates the neural network 270b for CSI decompression. The UE 102 may receive one or more RRC messages from the network entity 104 that include the configuration of one or more criteria. For example, the one or more RRC messages include an RRCReconfiguration message and / or an RRCResume message.

[0056] For example, if the UE 102 detects or determines that the BLER of the DL transport block received by the UE 102, for example, for a first time period or immediately above or equal to a first BLER threshold, then the UE 102 determines to evaluate the performance of the neural network 270a for CSI compression. In response to determining to evaluate the performance of the neural network 270a for CSI compression, the UE 102 activates the neural network 270b for CSI decompression. Otherwise, the UE 102 determines not to evaluate or stops evaluating the performance of the neural network 270a for CSI compression. In response to determining not to evaluate the performance of the neural network 270a for CSI compression, the UE 102 avoids activating or deactivating the neural network 270b for CSI decompression. In some implementations, after activating the neural network 270b for CSI decompression, if the UE 102 detects or determines that the BLER of the DL transport block received by the UE 102, for example, for a second time period or immediately below a second BLER threshold, then the UE 102 determines not to evaluate or stops evaluating the performance of the neural network 270a for CSI compression. In some implementations, the first BLER threshold and the second BLER threshold are the same. In other implementations, the first BLER threshold and the second BLER threshold are different. In some implementations, the first time period and the second time period are the same. In other implementations, the first time period and the second time period are different. In some implementations, the UE 102 receives the configuration of the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period from the network entity 104. For example, the UE 102 receives an RRC message (e.g., an RRCReconfiguration message or an RRCResume message) including the configuration from the network entity 104. In other implementations, the UE 102 applies the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period predefined in the 3GPP specifications. In still other implementations, the UE 102 pre-determines and pre-stores the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period.

[0057] In another example, if the UE 102 detects or determines that the maximum number of HARQ retransmissions for one or more transport blocks received by the UE 102, e.g., for a first time period or immediately above or equal to a first HARQ retransmission threshold, then the UE 102 determines to evaluate the performance of the neural network 270a for CSI compression. In response to determining to evaluate the performance of the neural network 270a for CSI compression, the UE 102 activates the neural network 270b for CSI decompression. Otherwise, the UE 102 determines not to evaluate or stops evaluating the performance of the neural network 270a for CSI compression. In response to determining not to evaluate the performance of the neural network 270a for CSI compression, the UE 102 avoids activating or deactivating the neural network 270b for CSI decompression. In some implementations, after activating the neural network 270b for CSI decompression, if the UE 102 detects or determines that the maximum number of HARQ retransmissions for one or more transport blocks received by the UE 102, e.g., for a second time period or immediately below a second HARQ retransmission threshold, then the UE 102 determines not to evaluate or stops evaluating the performance of the neural network 270a for CSI compression. In some implementations, the first HARQ retransmission threshold and the second HARQ retransmission threshold are the same. In other implementations, the first HARQ retransmission threshold and the second HARQ retransmission threshold are different. In some implementations, the first time period and the second time period are the same. In other implementations, the first time period and the second time period are different. In some implementations, the UE 102 receives the configuration of the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period from the network entity 104. For example, the UE 102 receives an RRC message (e.g., an RRCReconfiguration message or an RRCResume message) including the configuration from the network entity 104. In other implementations, the UE 102 applies the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period predefined in the 3GPP specifications. In still other implementations, the UE 102 pre-determines and pre-stores the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period.

[0058] To save battery power, the UE 102 may evaluate the performance of the neural network 270a for CSI compression on a discontinuous basis instead of on a continuous basis. For example, the UE 102 receives multiple CSI-RSs from the network entity 104 at different time instances. The UE 102 uses some of the multiple CSI-RSs to evaluate the performance of the neural network 270a for CSI compression and does not use the remaining multiple CSI-RSs to evaluate the performance of the neural network 270a for CSI compression. For example, the UE 102 evaluates the performance of the neural network 270a for CSI compression only based on the x-th CSI-RS among every y CSI-RSs and does not use the remaining multiple CSI-RSs among every y CSI-RSs to evaluate the performance of the neural network 270a for CSI compression. x and y are integers, and 0 < x ≤ y and 1 < y.

[0059] Similar to Figure 2A , Figure 2F FIG. 255 shows an example process for SRS-based AI / ML model performance monitoring. The network entity 104 may send an RRC message including an SRS configuration (e.g., SRS-Config) to the UE 102 to configure the UE 102 to perform SRS transmission. The SRS transmission 220 at the UE 102 sends one or more SRSs 265 to the network entity 104 according to the SRS configuration, for example, and the network entity 104 receives the SRS 265 from the UE 102 according to the SRS configuration. In some implementations, the network entity 104 may send an activation command (e.g., MAC CE or DCI) to the UE 102 to activate the SRS configuration after sending the SRS configuration to the UE 102, and the UE 102 sends the SRS in response to the activation command. Then, the network entity 104 performs channel estimation 250b based on the SRS to obtain the raw CSI. After obtaining the raw CSI from the channel estimation 250b, the network entity 104 performs eigenvector calculation 260a for each subband and derives the raw precoding matrix (true precoding matrix), i.e., a plurality of eigenvectors, from the eigenvector calculation 260a. The network entity 104 performs CSI compression 270a (e.g., an AI / ML-based CSI generator) on the raw precoding matrix to obtain the compressed CSI. The network entity 104 derives the decompressed precoding matrix (inferred precoding matrix) for each subband from the compressed CSI. Finally, the network entity 104 performs neural network performance evaluation 290 based on the decompressed precoding matrix (inferred precoding matrix) and the raw precoding matrix (true precoding matrix) to evaluate the AI / ML model inference accuracy.

[0060] In performance evaluation 290, network entity 104 determines or generates AI / ML model performance metrics based on the original precoding matrix (true precoding matrix) and the decompressed CSI (inferred precoding matrix), and evaluates the performance metrics by comparing them with a performance metric threshold. For example, if the performance metric is higher than or equal to the performance metric threshold, network entity 104 determines that the neural network 270a for CSI compression has good performance. Otherwise, if the performance metric is lower than the performance metric threshold, network entity 104 determines that the neural network 270a for CSI compression has poor performance. In a case where network entity 104 determines that the neural network 270a for CSI compression has poor performance, network entity 104 may apply at least one of the following: update the ML model, switch the ML model, or fallback to non-ML CSI reporting. In some implementations, network entity 104 receives the performance metric threshold from an operation, administration, and maintenance (OAM) node or an AI / ML function node. In other implementations, network entity 104 pre-stores the performance metric threshold. In still other implementations, the performance metric threshold is defined or pre-defined in the 3GPP specification. In some implementations, the performance metric is the cosine similarity between the original CSI (true CSI) and the decompressed CSI (inferred CSI), and the performance metric threshold is a cosine similarity threshold.

[0061] If the network entity 104 unconditionally or continuously evaluates the performance of the neural network 270a for CSI compression as described above, the network entity 104 consumes a large amount of battery power. To save battery power, the network entity 104 may determine whether to evaluate the performance of the neural network 270a for CSI compression based on one or more system performance metrics such as system throughput, block error rate (BLER), maximum number of HARQ retransmissions, reference signal received power (RSRP), reference signal received quality (RSRQ), and / or signal-to-noise-and-interference ratio (SINR). For example, if the network entity 104 determines that one or more system performance metrics meet the corresponding one or more criteria, the network entity 104 determines to evaluate the performance of the neural network 270a for CSI compression. Otherwise, if the network entity 104 determines that one or more system performance metrics do not meet the corresponding one or more criteria, the network entity 104 determines not to evaluate or stop evaluating the performance of the neural network 270a for CSI compression. In some implementations, if the network entity 104 determines to evaluate the performance of the neural network 270a for CSI compression, the network entity 104 activates the neural network 270b for CSI decompression. Otherwise, if the network entity 104 determines not to evaluate or stop evaluating the performance of the neural network 270a for CSI compression, the network entity 104 avoids activating or deactivates the neural network 270b for CSI decompression. In some implementations, if the network entity 104 determines to evaluate the performance of the neural network 270a for CSI compression, the network entity 104 may send an SRS configuration and / or activation command to the UE 102. Otherwise, if the network entity 104 determines not to evaluate the performance of the neural network 270a for CSI compression, the network entity 104 may avoid sending an SRS configuration and / or activation command to the UE 102. Otherwise, if the network entity 104 determines not to evaluate or stop evaluating the performance of the neural network 270a for CSI compression, the network entity 104 may send an RRC message to the UE 102 to release the SRS configuration, or send a deactivation command (e.g., MAC CE or DCI) to the UE 102 to deactivate the SRS configuration.

[0062] For example, if network entity 104 detects or determines that the BLER of the DL transport block received by UE 102, for example, for a first time period or immediately above or equal to a first BLER threshold, then network entity 104 determines to evaluate the performance of neural network 270a for CSI compression. In response to determining to evaluate the performance of neural network 270a for CSI compression, network entity 104 activates neural network 270b for CSI decompression. Otherwise, network entity 104 determines not to evaluate the performance of neural network 270a for CSI compression. In response to determining not to evaluate the performance of neural network 270a for CSI compression, network entity 104 avoids activating or deactivating neural network 270b for CSI decompression. In some implementations, after activating neural network 270b for CSI decompression, if network entity 104 detects or determines that the BLER of the DL transport block received by UE 102, for example, for a second time period or immediately below a second BLER threshold, then network entity 104 determines not to evaluate the performance of neural network 270a for CSI compression. In some implementations, the first BLER threshold and the second BLER threshold are the same. In other implementations, the first BLER threshold and the second BLER threshold are different. In some implementations, the first time period and the second time period are the same. In other implementations, the first time period and the second time period are different. In some implementations, network entity 104 receives the configuration of the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period from an OAM node. In other implementations, network entity 104 applies the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period predefined in the 3GPP specifications. In still other implementations, network entity 104 pre-determines and pre-stores the first BLER threshold, the second BLER threshold, the first time period, and / or the second time period.

[0063] In another example, if network entity 104 detects or determines that the maximum number of HARQ retransmissions for one or more transport blocks to be sent to UE 102, e.g., for a first time period or immediately above or equal to a first HARQ retransmission threshold, then network entity 104 determines to evaluate the performance of neural network 270a for CSI compression. In response to determining to evaluate the performance of neural network 270a for CSI compression, network entity 104 activates neural network 270b for CSI decompression. Otherwise, network entity 104 determines not to evaluate or stops evaluating the performance of neural network 270a for CSI compression. In response to determining not to evaluate or stop evaluating the performance of neural network 270a for CSI compression, network entity 104 avoids activating or deactivating neural network 270b for CSI decompression. In some implementations, after activating neural network 270b for CSI decompression, if network entity 104 detects or determines that the maximum number of HARQ retransmissions for one or more transport blocks to be sent to UE 102, e.g., for a second time period or immediately below a second HARQ retransmission threshold, then network entity 104 determines not to evaluate or stops evaluating the performance of neural network 270a for CSI compression. In some implementations, the first HARQ retransmission threshold and the second HARQ retransmission threshold are the same. In other implementations, the first HARQ retransmission threshold and the second HARQ retransmission threshold are different. In some implementations, the first time period and the second time period are the same. In other implementations, the first time period and the second time period are different. In some implementations, network entity 104 receives the configuration of the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period from an OAM node. In other implementations, network entity 104 applies the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period predefined in the 3GPP specifications. In still other implementations, network entity 104 pre-determines and pre-stores the first HARQ retransmission threshold, the second HARQ retransmission threshold, the first time period, and / or the second time period.

[0064] To save battery power, network entity 104 may evaluate the performance of neural network 270a for CSI compression on a discontinuous basis instead of on a continuous basis. For example, network entity 104 receives multiple SRSs from UE 102 at different time instances. Network entity 104 uses some of the multiple SRSs to evaluate the performance of neural network 270a for CSI compression and does not use the remainder of the multiple SRSs to evaluate the performance of neural network 270a for CSI compression. For example, network entity 104 evaluates the performance of neural network 270a for CSI compression only based on the xth SRS out of every y SRSs and does not use the remainder of the multiple SRSs out of every y SRSs to evaluate the performance of neural network 270a for CSI compression. x and y are integers, and 0 < x ≤ y and 1 < y.

[0065] Figure 3A FIG. 305 is a signaling diagram showing an example of an AI / ML-based CSI report. Initially, UE 102 communicates 302 with network entity 104. For example, UE 102 transmits 302 UL data and / or DL data to network 104. For example, the UL data and / or DL data may include control plane messages such as radio resource control (RRC) messages. UE 102 may send 304 UE capability information (e.g., UECapabilityInformation message) including one or more CSI reporting capabilities to network entity 104. For simplicity of the following description, "capability" is used to represent "one or more capabilities". In some implementations, UE 102 includes other capabilities in the UE capability information. In some implementations, UE 102 receives a UE capability query message (e.g., UECapabilityEnquiry message) from network 104. In response, UE 102 sends 304 UE capability information including CSI reporting capabilities to network entity 104. In some implementations, UE 102 generates a container information element (IE) including CSI reporting capabilities and other capabilities (i.e., capabilities other than CSI reporting capabilities) and includes the container in the UE capability information. In an example, the container IE is a UE-NR-Capability IE or a UE-6G-Capability IE. Alternatively, network entity 104 receives 306 CSI reporting capabilities or container IEs from a network node different from UE 102, such as another base station (e.g., similar to base station 104) or a core network entity (e.g., access and mobility management function (AMF)).

[0066] In some implementations, the CSI reporting capabilities 304, 306 include non-ML-based CSI reporting capabilities. That is, the UE 102 indicates the capabilities of non-ML-based reporting in the non-ML-based reporting capabilities. Based on the non-ML-based CSI reporting capabilities, the network entity 104 sends 308 a configuration for non-ML-based CSI reporting to the UE 102 to configure the UE 102 to send non-ML-based CSI reports. For example, the configuration 308 includes a CSI reporting configuration (e.g., CSI-ReportConfig IE). After sending the configuration 308, the network entity 104 may send 312a CSI-RS to the UE 102 according to the configuration 308. After receiving the configuration 308, the UE 102 may receive the CSI-RS 312a and perform channel estimation and / or measurements based on the CSI-RS 312a according to the configuration 308. The UE 102 generates a non-ML-based CSI report based on the channel estimation and / or measurements and sends 314 the non-ML-based CSI report to the network entity 104. In some implementations, the UE 102 includes non-ML-based CSI in the non-ML-based CSI report. In some implementations, the non-ML-based CSI includes RI, PMI, CQI, LI, L1-RSRP, L1-RSRQ, and / or L1-SINR.

[0067] In some implementations, the network entity 104 may send 308 an RRC message to the UE 102, and the RRC message includes a configuration for non-ML-based CSI reporting. In an example, the RRC message may include an RRCReconfiguration message. In response to each of the RRC messages, the UE 102 may send an RRC response message (e.g., an RRCReconfigurationComplete message) to the network entity 104. In some cases, the UE 102 is in dual connectivity with the network entity 104 (e.g., operating as an SN) and another network entity similar to the network entity 104 (e.g., operating as an MN not shown in FIG. 3). In an example, the SN 104 sends an RRC message to the UE 102 as described above. In other examples, the SN 104 sends an RRC message to the UE 102 via the MN.

[0068] In some implementations, configuration 308 includes CSI resource configuration for configuring CSI-RS 312a. In some implementations, CSI-RS 312a includes periodic CSI-RS, semi-persistent CSI-RS, and / or aperiodic CS-RS. The CSI resource configuration may include CSI resource configuration for configuring periodic CSI-RS, CSI resource configuration for configuring semi-persistent CSI-RS, and / or CSI resource configuration for configuring aperiodic CS-RS. The network entity 104 may transmit 312a periodic CSI-RS on a periodic basis according to the CSI resource configuration for configuring periodic CSI-RS. The network entity 104 may transmit 312a semi-persistent CSI-RS on a semi-persistent basis according to the CSI resource configuration for configuring semi-persistent CSI-RS. The network entity 104 may transmit 312a aperiodic CSI-RS on a one-time basis so that the UE 102 transmits an aperiodic non-ML-based CSI report according to the aperiodic CSI resource configuration, as described below.

[0069] In some implementations, the network entity 104 may transmit CSI-RS 312a from antenna ports, where corresponds to the maximum number of the downlink layer configured in the configuration 308 or the CSI resource configuration. In some implementations, the network entity 104 may use a precoder to transmit CSI-RS 312a or some of CSI-RS 312a. In other implementations, the network entity 104 may transmit CSI-RS 312a or some of CSI-RS 312a without a precoder.

[0070] In some implementations, configuration 308 includes a semi-persistent non-ML-based CSI report configuration for configuring a semi-persistent non-ML-based CSI report, and UE 102 avoids sending the semi-persistent non-ML-based CSI report until it receives a trigger command from network entity 104 that triggers UE 102 to send the semi-persistent non-ML-based CSI report according to the semi-persistent non-ML-based CSI report configuration. After sending configuration 308, network entity 104 may send 310 a trigger command to UE 102 to trigger the semi-persistent non-ML-based CSI report. After receiving the trigger command 310 or in response to receiving the trigger command 310, UE 102 performs channel estimation and / or measurement on CSI-RS 312a, generates a semi-persistent non-ML-based CSI report, and sends 314 the semi-persistent non-ML-based CSI report to network entity 104. In some implementations, UE 102 (periodically) sends the semi-persistent non-ML-based CSI report on a PUCCH that can be configured in configuration 308. In other implementations, UE 102 (periodically) sends the semi-persistent non-ML-based CSI report on a PUSCH that can be configured in configuration 308 and / or the trigger command. In some implementations, the trigger command is a MAC CE. In other implementations, the trigger command is DCI. In some implementations, CSI-RS includes periodic CSI-RS and / or semi-persistent CSI-RS. In the case of semi-persistent CSI-RS, network entity 104 may send an activation command to UE 102 to indicate that the semi-persistent CSI-RS is activated. After receiving the activation command (e.g., in response to receiving the activation command), UE 102 determines that the transmission of the semi-persistent CSI-RS is activated. In some implementations, network entity 104 sends the activation command before or after sending the trigger command. Alternatively, network entity 104 may send 310 a MAC PDU to UE 102 that includes the activation command and the trigger command. In some implementations, the activation command is a MAC CE. In some implementations, UE 102 activates channel estimation and / or measurement based on CSI-RS or a part of CSI-RS (e.g., one or some) in response to receiving the trigger command. In other implementations, UE 102 activates channel estimation and / or measurement based on CSI-RS or a part of CSI-RS (e.g., one or some) in response to the semi-persistent non-ML-based CSI report configuration and before receiving the trigger command.

[0071] In other implementations, configuration 308 includes a periodic non-ML-based CSI report configuration for configuring a periodic non-ML-based CSI report, and UE 102 performs channel estimation and / or measurement based on CSI-RS, generates a non-ML-based CSI report based on the channel estimation and / or measurement, and transmits a periodic ML-based CSI report 314 based on or in response to the periodic non-ML-based CSI report configuration. In such cases, UE 102 activates performing channel estimation and / or measurement based on CSI-RS or a part of CSI-RS (e.g., one or some), generates a periodic non-ML-based CSI report based on the channel estimation and / or measurement, and transmits 314 the periodic non-ML-based CSI report to network entity 104 when receiving the periodic non-ML-based CSI report configuration. Accordingly, network entity 104 does not send a trigger command to UE 102 to trigger the transmission of a periodic non-ML-based CSI report. In some implementations, UE 102 transmits the periodic non-ML-based CSI report on a PUCCH that can be configured in configuration 308. In other implementations, UE 102 transmits the periodic non-ML-based CSI report on a PUSCH that can be configured in configuration 308 and / or DCI received by UE 102 from network entity 104. The DCI includes a UL grant for UE 102 to transmit user data, rather than a trigger command.

[0072] In yet other implementations, configuration 308 includes an aperiodic non-ML-based CSI report configuration for configuring an aperiodic non-ML-based CSI report. For each of the aperiodic non-ML-based CSI report configurations, UE 102 avoids transmitting an aperiodic non-ML-based CSI report until receiving a trigger command from network entity 104 that triggers UE 102 to transmit an aperiodic non-ML-based CSI report according to the aperiodic non-ML-based CSI report configuration. After transmitting the aperiodic non-ML-based CSI report configuration, network entity 104 may send 310 a trigger command to UE 102 that triggers UE 102 to transmit an aperiodic non-ML-based CSI report according to the aperiodic non-ML-based CSI report configuration. In response to the trigger command, UE 102 performs channel estimation and / or measurement on CSI-RS, generates a single aperiodic non-ML-based CSI report, and transmits the aperiodic non-ML-based CSI report according to the aperiodic non-ML-based CSI report configuration. In some implementations, the trigger command is a MAC CE. In other implementations, the trigger command is DCI. In some implementations, CSI-RS includes periodic CSI-RS, semi-persistent CSI-RS, or aperiodic CSI-RS.

[0073] Events 308, 310, 312a, and 314 are collectively referred to as the non-ML-based CSI reporting procedure 390 in Figure 3A .

[0074] In some implementations, the CSI reporting capabilities 304, 306 include ML-based CSI reporting capabilities. That is, UE 102 indicates the ability to perform ML-based CSI reporting in its ML-based CSI reporting capabilities. Based on this ML-based CSI reporting capability, network entity 104 sends 316 a configuration for ML-based CSI reporting to UE 102 to configure UE 102 to send an ML-based CSI report using a first ML model (e.g., neural network 270a or 270a for CSI compression). For example, UE 102 may indicate support for the first ML model or include a first identifier (ID) of the first ML model in the UE capability information, such that network entity 104 can determine to configure the first ML model based on this indication or the first ID. For example, the configuration 316 includes a CSI reporting configuration (e.g., the CSI-ReportConfig IE or a new RRC IE defined in 3GPP specification v18.0.0 and / or later versions). After sending the configuration 316, network entity 104 may send 312b CSI-RS to UE 102 at multiple time instances according to the configuration 316. After receiving the configuration 316, UE 102 receives the CSI-RS 312b and performs channel estimation and / or measurements based on the CSI-RS 312b. UE 102 generates an ML-based CSI report based on the channel estimation and / or measurements and the first ML model, and sends 324 the ML-based CSI report to network entity 104. In one implementation, network entity 104 may indicate the first ML model in the configuration 316. For example, network entity 104 includes the first ID in the configuration 316. In another implementation, network entity 104 does not configure an ML model in the configuration 316. In this case, UE 102 determines the first ML model based on a predetermined configuration stored in UE 102. In some implementations, network entity 104 enables or configures ML-based CSI compression for UE 102 in the configuration 316, and UE 102 generates compressed CSI based on the first ML model and sends the compressed CSI in the ML-based CSI report 324, as described for Figure 2D or Figure 2A .

[0075] In some implementations, network entity 104 may send a 316 RRC message to UE 102, and the RRC message includes a configuration for ML-based CSI reporting. In some implementations, configuration 316 includes a new CSI reporting configuration (e.g., CSI-ReportConfig IE). In other implementations, configuration 316 includes configuration parameters to reconfigure at least one CSI reporting configuration in configuration 308 for application to ML-based CSI reporting. In such cases, configuration 316 includes the at least one CSI reporting configuration. After applying the configuration parameters (e.g., in response to applying the configuration parameters), UE 102 stops applying the at least one CSI reporting configuration to non-ML-based reporting. After applying the configuration parameters (e.g., in response to applying the configuration parameters), UE 102 stops sending non-ML-based CSI reports according to the at least one CSI reporting configuration. In an example, the RRC message may include an RRCReconfiguration message. In response to each of the RRC messages, UE 102 may send an RRC response message (e.g., an RRCReconfigurationComplete message) to network entity 104. In some cases, UE 102 is in dual connectivity with network entity 104 (e.g., as an SN operation) and another network entity similar to network entity 104 (e.g., as an MN operation not shown in FIG. 3). In an example, SN 104 sends an RRC message to UE 102 as described above. In other examples, SN 104 sends an RRC message to UE 102 via MN.

[0076] In some implementations, configuration 316 includes a CSI resource configuration for configuring CSI-RS 312b. In some implementations, CSI-RS 312b includes periodic CSI-RS, semi-persistent CSI-RS, and / or aperiodic CS-RS. The CSI resource configuration may include a CSI resource configuration for configuring periodic CSI-RS, a CSI resource configuration for configuring semi-persistent CSI-RS, and / or a CSI resource configuration for configuring aperiodic CS-RS. Network entity 104 may send 312b periodic CSI-RS on a periodic basis according to the CSI resource configuration for configuring periodic CSI-RS. Network entity 104 may send 312b semi-persistent CSI-RS on a semi-persistent basis according to the CSI resource configuration for configuring semi-persistent CSI-RS. Network entity 104 may send 312b aperiodic CSI-RS on a one-time basis so that UE 102 can send an aperiodic ML-based CSI report according to the aperiodic CSI resource configuration, as described below.

[0077] In some implementations, network entity 104 may receive from One antenna port transmits CSI-RS 312b, where Corresponds to the maximum number of downlink layers configured in Configuration 316 or the CSI resource configuration. In some implementations, network entity 104 may use a precoder to transmit CSI-RS or some of the CSI-RS. In other implementations, network entity 104 may transmit CSI-RS or some of the CSI-RS without a precoder.

[0078] In some implementations, configuration 316 includes a semi-persistent ML-based CSI report configuration for configuring semi-persistent ML-based CSI reporting, and UE 102 avoids sending the semi-persistent ML-based CSI report until it receives a trigger command from network entity 104, which triggers UE 102 to send the semi-persistent ML-based CSI report according to the semi-persistent CSI report configuration. After sending configuration 316, network entity 104 may send 320 a trigger command to UE 102 to trigger the semi-persistent ML-based CSI report. After receiving the trigger command 320 or in response to receiving the trigger command 320, UE 102 performs channel estimation and / or measurement on CSI-RS 312b, generates a semi-persistent ML-based CSI report, and sends 324 the semi-persistent ML-based CSI report to network entity 104. In some implementations, UE 102 (periodically) sends the semi-persistent ML-based CSI report on a PUCCH that can be configured in configuration 316. In other implementations, UE 102 (periodically) sends the semi-persistent ML-based CSI report on a PUSCH that can be configured in configuration 316, the semi-persistent ML-based CSI report configuration, and / or the trigger command. In some implementations, the trigger command is a MAC CE. In other implementations, the trigger command is DCI. In some implementations, CSI-RS includes periodic CSI-RS and / or semi-persistent CSI-RS. In the case of semi-persistent CSI-RS, network entity 104 may send an activation command to UE 102 to indicate that the semi-persistent CSI-RS is activated. After receiving the activation command (e.g., in response to receiving the activation command), UE 102 determines that the transmission of the semi-persistent CSI-RS is activated. In some implementations, network entity 104 sends the activation command before or after sending the trigger command. Alternatively, network entity 104 may send 320 a MAC PDU including the activation command and the trigger command to UE 102. In some implementations, the activation command is a MAC CE. In some implementations, UE 102 activates performing channel estimation and / or measurement based on CSI-RS or a part of CSI-RS (e.g., one or some) in response to receiving the trigger command. In other implementations, UE 102 activates performing channel estimation and / or measurement based on CSI-RS or a part of CSI-RS (e.g., one or some) in response to the semi-persistent ML-based CSI report configuration and before receiving the trigger command.

[0079] In other implementations, configuration 316 includes a periodic ML-based CSI report configuration for configuring periodic ML-based CSI reporting, and UE 102 performs channel estimation and / or measurements based on CSI-RS, generates an ML-based CSI report based on the channel estimation and / or measurements, and transmits the periodic ML-based CSI report 324 based on or in response to the periodic ML-based CSI report configuration. In such cases, UE 102 activates channel estimation and / or measurements based on CSI-RS or a part of CSI-RS (e.g., one or some), generates a periodic ML-based CSI report based on the channel estimation and / or measurements, and transmits 324 the periodic ML-based CSI report to network entity 104 when receiving the periodic ML-based CSI report configuration. Thus, network entity 104 does not send a trigger command to UE 102 to trigger the transmission of the periodic ML-based CSI report. In some implementations, UE 102 transmits the periodic ML-based CSI report on a PUCCH that can be configured in configuration 316. In other implementations, UE 102 transmits the periodic ML-based CSI report on a PUSCH that can be configured in configuration 316 and / or in DCI received by UE 102 from network entity 104. The DCI includes a UL grant for UE 102 to transmit user data, rather than a trigger command.

[0080] In yet other implementations, configuration 316 includes an aperiodic ML-based CSI report configuration for configuring aperiodic ML-based CSI reporting. For each of the aperiodic ML-based CSI report configurations, UE 102 avoids transmitting an aperiodic ML-based CSI report until receiving a trigger command from network entity 104 that triggers UE 102 to transmit an aperiodic ML-based CSI report according to the aperiodic ML-based CSI report configuration. After transmitting the aperiodic ML-based CSI report configuration, network entity 104 may send 320 a trigger command to UE 102 that triggers UE 102 to transmit an aperiodic ML-based CSI report according to the aperiodic ML-based CSI report configuration. In response to the trigger command, UE 102 performs channel estimation and / or measurements on CSI-RS, generates a single aperiodic ML-based CSI report, and transmits the aperiodic ML-based CSI report according to the aperiodic ML-based CSI report configuration. In some implementations, the trigger command is a MAC CE. In other implementations, the trigger command is DCI. In some implementations, CSI-RS includes periodic CSI-RS, semi-persistent CSI-RS, or aperiodic CSI-RS.

[0081] In some implementations, if network entity 104 determines to configure UE 102 to send an ML-based CSI report, network entity 104 may send an RRC message 318 (e.g., an RRCReconfiguration message) to UE 102 to release at least one CSI report configuration in configuration 308. In some implementations, if configuration 316 and configuration 308 exceed the CSI reporting capabilities of UE 102, network entity 104 may send RRC message 318. If configuration 316 and configuration 308 do not exceed the CSI reporting capabilities of UE 102, network entity 104 may not send the RRC message. In some implementations, network entity 104 may still send a release indication because the ML-based CSI report configured in configuration 316 may replace the non-ML-based CSI report configured in the at least one CSI report configuration.

[0082] In other implementations, if network entity 104 determines to configure UE 102 to send an ML-based CSI report, network entity 104 may send an RRC message 318 (e.g., an RRCReconfiguration message) to UE 102 to reconfigure at least one CSI report configuration in configuration 308. In some implementations, RRC message 318 reconfigures the at least one CSI report configuration to prevent UE 102 from sending the non-ML-based CSI report configured in the at least one CSI report configuration. For example, the at least one CSI report configuration is configured for periodic CSI reporting, and network entity 104 may reconfigure the at least one CSI report configuration for semi-persistent CSI reporting or aperiodic CSI reporting. In some implementations, if configuration 316 and configuration 308 exceed the CSI reporting capabilities of UE 102, network entity 104 may send an RRC message. If configuration 316 and configuration 308 do not exceed the CSI reporting capabilities of UE 102, network entity 104 may not send the RRC message. In some implementations, network entity 104 may still send the RRC message because the ML-based CSI report configured in configuration 316 may replace the non-ML-based CSI report configured in the at least one CSI report configuration.

[0083] In yet other implementations, if network entity 104 determines to configure UE 102 to send ML-based CSI reports, network entity 104 may send an RRC message 318 to UE 102 to modify configuration 308. In some implementations, the RRC message modifies configuration 308 such that UE 102 sends non-ML-based CSI reports less frequently. Network entity 104 may do so because network entity 104 may use the ML-based CSI reports configured in configuration 316 in place of most of the non-ML-based CSI reports configured in configuration 308.

[0084] Events 312b, 316, 318, 320, and 324 are collectively referred to as the ML-based CSI reporting procedure 392 in Figure 3A this context.

[0085] In some implementations, procedure 390 may overlap with procedure 392, either fully or partially. In other implementations, procedure 390 does not overlap with procedure 392. In some implementations, configuration 308 and configuration 316 include at least one identical configuration. For example, CSI-RS 312a and CSI-RS 312b may include the same CSI-RS and / or different CSI-RS. To configure UE 102 to generate ML-based CSI reports and non-ML-based CSI reports based on the same CSI-RS, network entity 104 may send a CSI resource configuration (i.e., a single instance), each CSI resource configuration including a CSI resource configuration ID and configuring CSI-RS, and network entity 104 may include the CSI resource configuration ID in configuration 308 and configuration 316. UE 102 identifies the CSI resource configuration based on the (identical) CSI resource configuration ID. Thus, UE 102 receives the CSI-RS configured in the CSI resource configuration, performs channel estimation and / or measurements on the CSI-RS, and sends ML-based CSI reports and non-ML-based CSI reports based on the channel estimation and / or measurements. For each ML-based CSI report among the ML-based CSI reports, network entity 104 may obtain ML-based CSI (e.g., compressed CSI) from the ML-based CSI report and obtain reconstructed CSI (e.g., decompressed CSI) from the ML-based CSI and a first ML model (e.g., a neural network for decompressing 270b). For each non-ML-based CSI report among the non-ML-based CSI reports, network entity 104 also retrieves non-ML-based CSI from the non-ML-based CSI report.

[0086] In some implementations, the network entity 104 may determine 326 to perform ML model performance monitoring and / or evaluation on the first ML model after sending the configuration 316 to the UE 102 or in response to sending the configuration 316 to the UE 102. In other implementations, the network entity 104 may determine whether to perform ML model performance monitoring and / or evaluation based on one or more system performance metrics such as system throughput, BLER, maximum number of HARQ retransmissions, RSRP, RSRQ, and / or SINR), as described for Figure 2F the same. In some implementations, the network entity 104 performs a non-ML-based CSI reporting procedure 390 with the UE 102 in response to the determination. In response to determining to perform ML model performance monitoring and / or evaluation, the network entity 104 evaluates or determines 340a the performance of the first ML model based on the reconstructed CSI and the non-ML-based CSI for the same instance of the CSI-RS. In the performance monitoring and / or evaluation 340a, the network entity 104 determines AI / ML model performance metrics based on the non-ML-based CSI and the reconstructed CSI, and evaluates the performance metrics by comparing them with a performance metric threshold.

[0087] For example, if the performance metric is higher than or equal to the performance metric threshold, the network entity 104 determines that the performance of the first ML model (e.g., the neural network 270a for CSI compression) is good. Otherwise, if the performance metric is lower than the performance metric threshold, the network entity 104 determines that the performance of the first ML model is bad. In response to determining that the performance of the first ML model is bad, the network entity 104 sends a 342 command to the UE 102 to release or deactivate the configuration 316 (e.g., configure the UE 102 to stop using the first ML model or deactivate the first ML model) or replace the first ML model with a second ML model. In response to the command 342, the UE 102 releases or deactivates the configuration, or replaces the first ML model with a second ML model. The command can be a message (e.g., RRCReconfiguration message), MAC CE, or DCI. For example, the network entity 104 can include the configuration ID of the configuration 316 in the release information element (IE) in the message to configure the UE 102 to release the configuration 316. In another example, the network entity 104 can include the configuration ID of the configuration 316 in the MAC CE or DCI to configure the UE 102 to deactivate the configuration 316. In the case of replacing the first ML model with a second ML model, the network entity 104 can include the second ID of the second ML model in the command. In some implementations, the CSI reporting capability, container IE, or UE capability information indicates support for the second ML model or includes the second ID. The network entity 104 can send the command 342 to replace the first ML model with the second ML model because the network entity 104 determines that the UE 102 supports the second ML model based on the CSI reporting capability, container IE, or UE capability information. If the UE 102 does not support the second ML model, the network entity 104 can send the command 342 to release or deactivate the configuration 316.

[0088] In the case of replacing the first ML model with the second ML model, similar to event 324, the UE 102 sends an ML-based CSI report to the network entity 104.

[0089] Figure 3B is a signaling diagram 315 showing an example of ML model performance monitoring and reporting. Elements 302, 304, 306, 342, 390, and 392 have been described with respect to Figure 3A above.

[0090] After or during the ML-based CSI reporting procedure 392, the network entity 104 determines 325 to configure the UE 102 to perform ML model performance reporting. In some implementations, the UE capability information, the container IE, or the CSI reporting capability includes an indication of the UE 102's ability to support ML model performance reporting, monitoring, and / or evaluation. The network entity 104 may make the determination 325 based on the capability indicating support for ML model performance reporting, which includes support for ML model performance monitoring and / or evaluation.

[0091] Based on the determination 325, the network entity 104 sends 328 the configuration of the ML model performance reporting to the UE 102. In response to the configuration 328, the UE 102 activates 330 ML model performance monitoring and / or evaluation. In some implementations, the network entity 104 may include at least one ID, each ID identifying an ML model in the configuration 328, and the UE 102 activates 330 ML model performance monitoring and / or evaluation for / of the ML model identified by the at least one ID. For example, the ML model includes a first ML model, and the at least one ID includes a first ID. Thus, the UE 102 activates ML model performance monitoring and / or evaluation for / of the first ML model in event 330. In another example, the ML model includes a second ML model, and the at least one ID includes a second ID. Thus, the UE 102 activates ML model performance monitoring and / or evaluation for / of the second ML model in event 330. In other implementations, the network entity 104 does not include the ID of the ML model in the configuration 328, and the UE 102 activates ML model performance monitoring and / or evaluation for / of the ML model (i.e., the first ML model) that the UE 102 is using for ML-based CSI reporting 334 or used in the ML-based CSI reporting procedure 392.

[0092] After receiving the configuration 328, similar to events 310, 320, and / or 312 respectively, the UE 102 may receive 331 a trigger command from the network entity 104 and / or receive 332 CSI-RS. Thereafter, the UE 102 generates a non-ML-based CSI report and / or an ML-based CSI report based on the channel estimation and / or measurement of the CSI-RS 332, and sends 334 the non-ML-based CSI report and / or the ML-based CSI report to the network entity 104 similar to events 314 and / or 324. The UE 102 uses the first ML model to generate the ML-based CSI report 334, similar to event 324.

[0093] After activating ML model performance monitoring and / or evaluation for an ML model (e.g., in response to activating ML model performance monitoring and / or evaluation for an ML model), UE 102 performs ML model performance monitoring and / or evaluation based on CSI-RS 332. UE 102 generates an ML model performance report based on the results from the ML model performance monitoring and / or evaluation and sends 336 the ML model performance report to network entity 104. In some implementations, UE 102 includes at least one ID in the ML model performance report. Accordingly, network entity 104 determines the ML model performance report for the ML model (e.g., associated with the ML model) based on the at least one ID. In other implementations, UE 102 does not include the ID of the ML model in the ML model performance report, and network entity 104 determines the ML model performance report for the ML model (e.g., associated with the ML model) that UE 102 is using for ML-based CSI reporting 334 or in the ML-based CSI reporting process 392. Network entity 104 determines 340b the ML model performance based on the ML model performance report 336. Based on the determined ML model performance, network entity 104 may send 342 a command to UE 102 to release or deactivate configuration 316 or replace the first ML model with a second ML model, as described for Figure 3A above. For example, if network entity 104 determines based on the ML model performance report that the performance of the first ML model is poor and / or the performance of the second ML model is good or better than the first ML model, network entity 104 sends 342 a command to UE 102 to release or deactivate configuration 316 or replace the first ML model with a second ML model, as described for Figure 3A above.

[0094] In some implementations, based on each of the CSI-RS 332, UE 102 generates performance metrics from the ML model performance monitoring and / or evaluation. In some implementations, configuration 328 configures UE 102 to periodically send the ML model performance report, and UE 102 periodically sends, in event 336, the ML model performance report including the performance metrics to network entity 104.

[0095] In other implementations, configuration 328 configures event-triggered ML model performance reporting. For example, configuration 328 includes a performance metric threshold for enabling UE 102 to determine whether a reporting event has occurred. In one implementation, if the performance metric is below the performance metric threshold (e.g., the reporting event has occurred), then UE 102 sends an ML model performance report to network entity 104 in event 336. UE 102 may include the performance metric and / or an indication that the performance metric is below the performance metric threshold in the ML model performance report. Otherwise, if the performance metric is above or equal to the performance metric threshold, then UE 102 refrains from sending an ML model performance report to network entity 104. In another implementation, if the performance metric is above or equal to the performance metric threshold (e.g., the reporting event has occurred), then UE 102 sends an ML model performance report to network entity 104 in event 336. UE 102 may include the performance metric and / or an indication that the performance metric is above or equal to the performance metric threshold in the ML model performance report. Otherwise, if the performance metric is below the performance metric threshold, then UE 102 refrains from sending an ML model performance report to network entity 104. In some scenarios or implementations, configuration 328 does not include a performance metric threshold, and UE 102 pre-determines or pre-stores a performance metric threshold predefined in 3GPP specifications. In some implementations, UE 102 periodically sends an ML model performance report to network entity 104 in event 336 after detecting the occurrence of the event. Network entity 104 may configure UE 102 to do so in configuration 328. In other implementations, UE 102 sends N ML model performance reports to network entity 104 in event 336 after detecting the occurrence of the event. N is an integer greater than zero. Network entity 104 may configure N in configuration 328.

[0096] In some implementations, network entity 104 may include CSI resource configuration for configuring CSI-RS 332 in configuration 328. UE 102 uses the CSI resource configuration to receive CSI-RS 332. In other implementations, network entity 104 does not include CSI resource configuration in configuration 328. In such cases, CSI-RS 332 is configured in configuration 308 and / or configuration 316, and UE 102 receives CSI-RS 332 as described for event 312.

[0097] Figure 3C is signaling diagram 335 showing an example of ML model performance monitoring and reporting. Elements 302, 304, 306, 342, 390, and 392 have been described with respect to Figure 3A Elements 325, 328, 330, 332, 336, and 340b have been described with respect to Figure 3B has been described.

[0098] In signaling diagram 335, network entity 104 determines 337c to configure UE 102 to perform ML-based CSI reporting based on the ML model performance report. Based on this determination 337c, network entity 104 performs an ML-based CSI reporting procedure 392 with UE 102. For example, if network entity 104 determines based on the ML model performance report that the performance of the first ML model is good (i.e., the first ML model is suitable for communication between UE 102 and network entity 104), then network entity 104 performs an ML-based CSI reporting procedure 392 with UE 102. Otherwise, if network entity 104 determines that the performance of the first ML model is poor (i.e., the first ML model is not suitable for communication between UE 102 and network entity 104), then network entity 104 avoids configuring UE 102 to perform ML-based CSI reporting.

[0099] Figure 3D Signaling diagram 345 is an example showing SRS-based ML model performance monitoring. Elements 302, 304, 306, 326, 342, 390, and 392 have been described with respect to Figure 3A this.

[0100] Before, after, or in response to determination 326, network entity 104 sends 344 an SRS configuration (e.g., SRS-Config) to UE 102 to configure UE 102 to transmit 346 an SRS. In some implementations, the network entity sends a message (e.g., an RRCReconfiguration message) including the SRS configuration to UE 102. In response, UE 102 sends a response message (e.g., an RRCReconfigurationComplete message) to network entity 104. During or after procedure 392, UE 102 transmits 346 an SRS to network entity 104 according to the SRS configuration. Network entity 104 determines 340d the ML model performance of at least one ML model based on the SRS. In some implementations, the at least one ML model includes a first ML model and / or a second ML model. Based on the determined ML model performance, network entity 104 may send 342 a command to UE 102 to release or deactivate configuration 316 or replace the first ML model with the second ML model, as described for Figure 3A this.

[0101] In some implementations, network entity 104 performs ML model performance monitoring and / or evaluation based on SRS. Network entity 104 determines or generates performance metrics for each of at least one ML model. Network entity 104 determines the performance of each of at least one ML model based on the corresponding performance metrics and (the same) performance metric thresholds in event 340c. For example, if the performance metric of the first ML model is below the performance metric threshold, network entity 104 sends command 342 to UE 102 to release or deactivate configuration 316. In another example, if the performance metric of the first ML model is below the performance metric threshold while the performance metric of the second ML model is above the performance metric threshold, network entity 104 sends command 342 to replace the first ML model with the second ML model.

[0102] Figure 3E is signaling diagram 355 showing an example of SRS-based ML model performance monitoring. Elements 302, 304, 306, 326, 342, 390, and 392 have been described with respect to Figure 3A Elements 340d, 344, and 346 have been described with respect to Figure 3D

[0103] In signaling diagram 355, network entity 104 sends SRS configuration 344 to UE 102 before performing the ML-based CSI reporting procedure 392 with UE 102. Network entity 104 determines 337e to configure UE 102 to perform ML-based CSI reporting based on SRS 346. Based on this determination 337e, network entity 104 performs the ML-based CSI reporting procedure 392 with UE 102. For example, if network entity 104 determines based on SRS that the performance of the first ML model is good (i.e., suitable for communication between UE 102 and network entity 104), network entity 104 performs the ML-based CSI reporting procedure 392 with UE 102. Otherwise, if network entity 104 determines that the performance of the first ML model is poor (i.e., the first ML model is not suitable for communication between UE 102 and network entity 104), network entity 104 avoids configuring UE 102 to perform ML-based CSI reporting. Figures 3A to 3E shows an example process for ML model performance monitoring. Figures 4A to 9 shows Figures 3A to 3E one or more aspects of the method for implementing

[0104] Figures 4A to 4B shows flowcharts 400, 450 of the wireless communication method at the UE. Refer to Figures 3A to 3E and Figure 10 ​, This method can be executed by the UE 102, the UE equipment 1002, etc. The UE, the UE device, etc. can include memories 1026', 1006', 1016 and can correspond to the entire UE 102 or the entire UE equipment 1002, or components of the UE 102 or the UE equipment 1002 such as the radio baseband processor 1026 and / or the application processor 1006. The UE 102 can implement the flowcharts 400, 450 for managing ML-based CSI reporting with the RAN (e.g., the network entity 104).

[0105] Referring to flowchart 400, the UE 102 communicates 402 with the RAN. For example, in Figures 3A to 3E , the UE 102 communicates with the network entity 104. The UE 102 sends 404 the ML-based CSI reporting capability and the ML model performance monitoring capability to the RAN. For example, in Figures 3A to 3E , the UE 102 sends 304 UE capability information (e.g., CSI reporting capability) to the network entity 104. The UE102 receives 406 the configuration of the ML model performance report from the RAN. For example, in Figures 3B to 3C , the UE 102 receives 328 the configuration of the ML model performance report from the network entity 104. The UE 102 performs 408 ML model performance monitoring and reporting according to the configuration of the ML model performance report. For example, in Figures 3B to 3C , the UE 102 activates 330 ML model performance monitoring and / or evaluation and sends 336 the ML model performance report to the network entity 104. The UE 102 receives 410 the configuration for configuring the ML-based CSI reporting based on the ML model from the RAN. For example, in Figure 3A , the UE 102 receives 316 the configuration of the ML-based CSI reporting for the ML-based CSI reporting process 392. The UE 102 sends 412 the ML-based CSI report to the RAN according to the configuration for configuring the ML-based CSI report. For example, in Figure 3A , the UE 102 sends 324 the ML-based CSI report to the network entity 104 based on the configuration of the ML-based CSI reporting received 316 from the network entity 104.

[0106] Referring to flowchart 450, elements 402, 410, and 412 have been described with respect to Figure 4A . The UE 102 sends 405 the ML-based CSI reporting capability to the RAN. For example, in Figures 3A to 3E , the UE 102 sends 304 UE capability information (e.g., CSI reporting capability) to the network entity 104. The UE 102 performs 411 ML model performance monitoring and reporting according to the configuration of the ML-based CSI report from the RAN. For example, in Figures 3B to 3CIn this case, UE 102 activates 330 ML model performance monitoring and / or evaluation and sends a 336 ML model performance report to network entity 104.

[0107] Figures 5A to 5B Flowcharts 500 and 550 of a wireless communication method at a UE are shown. Refer to Figures 3A to 3E and Figure 10 This method can be executed by UE 102, UE equipment 1002, etc. This UE, this UE device, etc. can include memories 1026', 1006', 1016 and can correspond to the entire UE 102 or the entire UE equipment 1002, or components of UE 102 or UE equipment 1002 such as wireless baseband processor 1026 and / or application processor 1006. UE 102 can implement flowcharts 500 and 550 for sending an ML model performance report to a RAN (e.g., network entity 104).

[0108] Refer to flowchart 500. Elements 402 and 406 have been described with respect to Figure 4A UE 102 performs 508a ML model performance monitoring and / or evaluation on the ML model. For example, in Figures 3B to 3C In this case, UE 102 activates 330 ML model performance monitoring and / or evaluation and sends a 336 ML model performance report to network entity 104. UE 102 determines 508b whether the performance of the ML model meets the reporting criteria. If UE 102 determines 508b that the performance of the ML model meets the reporting criteria, then UE 102 sends an ML model performance report to the RAN. For example, in Figures 3B to 3C In this case, UE 102 sends a 336 ML model performance report to network entity 104. Otherwise, if UE 102 determines 508b that the performance of the ML model does not meet the reporting criteria, then UE 102 avoids 508d sending an ML model performance report to the RAN.

[0109] Refer to flowchart 550. Elements 402 and 406 have been described with respect to Figure 4A Element 508a has been described with respect to Figure 5A UE 102 sends 508e an ML model performance report to the RAN based on the ML model performance monitoring and / or evaluation. For example, in Figures 3B to 3C In this case, UE 102 sends a 336 ML model performance report to network entity 104.

[0110] Figures 6A to 6C Flowcharts 600, 630, and 660 of a wireless communication method at a UE are shown. Refer to Figures 3A to 3E and Figure 10, this method can be executed by the UE 102, UE equipment 1002, etc. The UE, the UE equipment, etc. may include memories 1026', 1006', 1016 and may correspond to the entire UE 102 or the entire UE equipment 1002, or components of the UE 102 or UE equipment 1002 such as the radio baseband processor 1026 and / or the application processor 1006. The UE 102 can implement flowcharts 600, 630, 660 for performing ML model performance monitoring and / or evaluation.

[0111] Referring to flowchart 600, elements 402 and 406 have been described with respect to Figure 4A Element 508a has been described with respect to Figure 5A Element 508e has been described with respect to Figure 5B The UE 102 determines 607a whether the conditions for performing ML model performance monitoring and / or evaluation are met. In some implementations, the UE 102 determines 607a whether the conditions for performing ML model performance monitoring and / or evaluation are met only when the UE 102 receives 406 a configuration for the ML model performance report from the RAN. If the UE 102 determines 607a that the conditions for performing ML model performance monitoring and / or evaluation are met, the UE 102 performs 508a ML model performance monitoring and / or evaluation on the ML model and may send 508e an ML model performance report to the RAN based on the ML model performance monitoring. Otherwise, if the UE 102 determines 607a that the conditions for performing ML model performance monitoring and / or evaluation are not met, the UE 102 avoids or stops 609 performing ML model performance monitoring.

[0112] Referring to flowchart 630, elements 402 and 410 have been described with respect to Figure 4A Element 508a has been described with respect to Figure 5A Element 508e has been described with respect to Figure 5B Element 609 has been described with respect to Figure 6A The UE102 determines 607b whether the conditions for performing ML-based CSI reporting and / or evaluation are met. In some implementations, the UE 102 determines 607b whether the conditions for performing ML-based CSI reporting and / or evaluation are met only when the UE 102 receives 410 a configuration for configuring the ML-based CSI report based on the ML model from the RAN.

[0113] If the UE 102 determines that 607b meets the conditions for performing ML-based CSI reporting and / or evaluation, the UE 102 performs 508a ML model performance monitoring and / or evaluation on the ML model, and may send a 508e ML model performance report to the RAN based on the ML model performance monitoring. Otherwise, if the UE 102 determines that 607b does not meet the conditions for performing ML-based CSI reporting and / or evaluation, the UE 102 avoids or stops 609 performing ML model performance monitoring.

[0114] Referring to flowchart 660, elements 402, 406, and 410 have been described with respect to Figure 4A The UE 102 determines whether 607c meets the conditions for performing ML model performance monitoring or ML-based CSI reporting. If the UE 102 determines that 607c meets the conditions, the UE 102 performs 608a ML model performance monitoring and / or ML-based CSI reporting, and may send a 613 ML model performance report or an ML-based CSI report to the RAN. Otherwise, if the UE 102 determines that 607c does not meet the conditions, the UE 102 ends 614 the method of flowchart 660.

[0115] Figure 7 A flowchart 700 of a wireless communication method at the UE is shown. Referring to Figures 3A to 3E and Figure 10 , the method may be performed by the UE 102, the UE device 1002, etc., which may include memories 1026', 1006', 1016 and may correspond to the entire UE 102 or the entire UE device 1002, or components of the UE 102 or the UE device 1002 such as the radio baseband processor 1026 and / or the application processor 1006. The UE 102 may implement flowchart 700 for ML-based CSI reporting to the RAN (e.g., network entity 104).

[0116] Referring to flowchart 700, elements 402, 410, and 412 have been described with respect to Figure 4Ais described. The UE 102 sends a preference indication 703 to the RAN (e.g., with respect to a non-ML-based CSI report) indicating that the UE 102 prefers an ML-based CSI report. In some implementations, the preference indication includes the ID of a first ML model in the preference indication. The RAN determines to configure the first ML model or configures the first ML model based on the ID. In some implementations, the preference indication is an RRC message, a MAC-CE, or UL control information (UCI) sent on the PUCCH. The RRC message can be a UEAssistanceInformation message or a different RRC message, such as a message defined in the 3GPP specifications. In some implementations, the UE 102 receives an RRC message (e.g., an RRCReconfiguration message or an RRCResume message) from the RAN, and the RRC message includes a preference indication configuration that permits or configures the UE 102 to send the preference indication. If the UE 102 does not receive the preference indication configuration, the UE 102 avoids sending the preference indication to the RAN. In some implementations, the UE 102 activates and / or performs ML performance monitoring and / or evaluation in response to receiving the preference indication configuration. Thus, the UE 102 can determine whether the UE 102 prefers an ML-based CSI report (e.g., using the first ML model) based on the ML performance monitoring and / or evaluation. If the UE 102 does not receive the preference indication configuration, the UE 102 can avoid activating and / or performing ML performance monitoring and / or evaluation. In other implementations, the UE 102 activates and / or performs ML performance monitoring and / or evaluation (e.g., using the first ML model) in response to receiving a configuration of a non-ML-based CSI report (e.g., event 308). If the UE 102 does not receive the configuration of the non-ML-based CSI report, the UE 102 can avoid activating and / or performing ML performance monitoring and / or evaluation.

[0117] Figure 8 shows a flowchart 800 of a wireless communication method at a UE. Referring to Figures 3A to 3E and Figure 10 , the method can be performed by the UE 102, the UE equipment 1002, etc., and the UE, the UE device, etc. can include memories 1026', 1006', 1016 and can correspond to the entire UE 102 or the entire UE equipment 1002, or components of the UE 102 or the UE equipment 1002 such as a wireless baseband processor 1026 and / or an application processor 1006. The UE 102 can implement the flowchart 800 for stopping the ML-based CSI report to the RAN (e.g., the network entity 104).

[0118] Referring to the flowchart 800, elements 402 and 410 have been described with respect to Figure 4Ais described. The UE 102 sends 803 to the RAN a preference indication indicating that the UE 102 does not prefer the ML-based CSI report. The UE 102 receives 810 from the RAN a configuration that configures the UE 102 to stop (e.g., release or deactivate) the ML-based CSI report (e.g., event 342) based on the ML model. In some implementations, the configuration 410 configures the UE 102 to perform the ML-based CSI report using the first ML model. In some implementations, the preference indication includes the ID of the first ML model in the preference indication. The configuration 810 may include the ID to indicate that the UE 102 stops using the first ML model for the ML-based CSI report. Regarding Figure 7 the examples and implementations described may also apply to Figure 8 .

[0119] Figure 9 A flowchart 900 of a wireless communication method at the UE is shown. Referring to Figures 3A to 3E and Figure 10 , the method may be performed by the UE 102, the UE equipment 1002, etc., and the UE, the UE device, etc. may include memories 1026', 1006', 1016 and may correspond to the entire UE 102 or the entire UE equipment 1002, or components of the UE 102 or the UE equipment 1002 such as the radio baseband processor 1026 and / or the application processor 1006. The UE 102 may implement the flowchart 900 for reconfiguring the ML model for the ML-based CSI report to the RAN (e.g., network entity 104).

[0120] Referring to the flowchart 900, the element 402 has been described with respect to Figure 4A . The UE 102 receives 910a from the RAN a first configuration for configuring the ML-based CSI report based on the first ML model. For example, in Figure 3A , the UE 102 receives 316 the ML-based CSI report configuration of the ML-based CSI report process 392. The UE 102 sends 912a to the RAN the ML-based CSI report according to the first configuration for configuring the ML-based CSI report. For example, in Figure 3A , the UE 102 sends 324 the ML-based CSI report for the ML-based CSI report process 392. The UE 102 may send 903 to the RAN a preference indication indicating that the UE 102 prefers to use the ML-based CSI report of the second ML model. The UE 102 receives 910b from the RAN a second configuration for configuring the ML-based CSI report based on the second ML model. The UE 102 sends 912b to the RAN the ML-based CSI report according to the second configuration.

[0121] In some implementations, the preference indication includes the ID of the second ML model in the preference indication. The RAN determines or configures the second ML model based on this ID. For Figure 7 the examples and implementations described Figure 9 are also applicable to

[0122]

[0123] Figure 10 In some implementations, UE 102 activates and / or performs ML performance monitoring and / or evaluation in response to receiving a preference indication configuration. Thus, UE 102 can determine whether UE 102 prefers ML-based CSI reporting (e.g., using the first ML model and / or the second ML model) based on the ML performance monitoring and / or evaluation. If UE 102 does not receive a preference indication configuration, UE 102 can avoid activating and / or performing ML performance monitoring and / or evaluation. In other implementations, UE 102 activates and / or performs ML performance monitoring and / or evaluation (e.g., using the first ML model and / or the second ML model) in response to receiving a configuration for non-ML-based CSI reporting (e.g., event 308) and / or a configuration for ML-based CSI reporting (e.g., event 316). If UE 102 does not receive a configuration for non-ML-based CSI reporting, UE 102 can avoid activating and / or performing ML performance monitoring and / or evaluation. In some implementations, UE 102 indicates the ID in a preference indication (the first preference indication). For example, UE 102 includes a first ID (e.g., an ML model ID) that identifies the first ML model in the first preference indication. When UE 102 performs ML-based CSI reporting according to a configuration (e.g., the first configuration), UE 102 may send a second preference indication to the RAN indicating that UE 102 prefers to use the second ML model for ML-based CSI reporting. For example, UE 102 includes a second ID (e.g., an ML model ID) that identifies the second ML model in the second preference indication. After sending the second preference indication, UE 102 receives a configuration from the RAN that configures UE 102 to perform ML-based CSI reporting using the second ML model instead of the first ML model.FIG. 1000 is an example showing a hardware implementation of UE device 1002. UE device 1002 may be UE 102, a component of UE 102, or may implement UE functions. UE device 1002 may include an application processor 1006, which may have on-chip memory 1006'. In an example, application processor 1006 may be coupled to a Secure Digital (SD) card 1008 and / or a display 1010. Application processor 1006 may also be coupled to a sensor module 1012, a power supply 1014, an additional memory module 1016, a camera 1018, and / or other related components. For example, sensor module 1012 may control a barometric pressure sensor / altimeter, motion sensors (such as an Inertial Management Unit (IMU)), a gyroscope, an accelerometer, a Light Detection and Ranging (LIDAR) device, a Radio Assisted Detection and Ranging (RADAR) device, a Sound Navigation and Ranging (SONAR) device, a magnetometer, an audio device, and / or other technologies for positioning.

[0124] UE device 1002 may further include a wireless baseband processor 1026, which may be referred to as a modem. Wireless baseband processor 1026 may have on-chip memory 1026'. Together with and similar to application processor 1006, wireless baseband processor 1026 may also be coupled to a sensor module 1012, a power supply 1014, an additional memory module 1016, a camera 1018, and / or other related components. Wireless baseband processor 1026 may additionally be coupled to one or more Subscriber Identity Module (SIM) cards 1020 and / or one or more transceivers 1030 (e.g., wireless RF transceivers).

[0125] Within one or more transceivers 1030, UE device 1002 may include a Bluetooth module 1032, a WLAN module 1034, an SPS module 1036 (e.g., GNSS module), and / or a cellular module 1038. Bluetooth module 1032, WLAN module 1034, SPS module 1036, and cellular module 1038 may each include an on-chip transceiver (TRX), or in some cases, include only a transmitter (TX) or only a receiver (RX). Bluetooth module 1032, WLAN module 1034, SPS module 1036, and cellular module 1038 may each include a dedicated antenna and / or utilize antenna 1040 to communicate with one or more other nodes. For example, UE device 1002 may communicate with another UE 102 (e.g., sidelink communication) and / or with a network entity 104 (e.g., uplink / downlink communication) via transceiver 1030 via antenna 1040, where network entity 104 may correspond to a base station or a unit of a base station, such as RU 106, DU 108, or CU 110.

[0126] The wireless baseband processor 1026 and the application processor 1006 may each separately include a computer-readable medium / memory 1026', 1006'. The additional memory module 1016 may also be regarded as a computer-readable medium / memory. Each computer-readable medium / memory 1026', 1006', 1016 may be non-transitory. The wireless baseband processor 1026 and the application processor 1006 may each be responsible for general processing, including executing software stored in the computer-readable medium / memory 1026', 1006', 1016. When executed by the wireless baseband processor 1026 / application processor 1006, the software causes the wireless baseband processor 1026 / application processor 1006 to perform the various functions described herein. The computer-readable medium / memory may also be used to store data manipulated by the wireless baseband processor 1026 / application processor 1006 when executing the software. The wireless baseband processor 1026 / application processor 1006 may be components of the UE 102. The UE device 1002 may be a processor chip (e.g., a modem and / or an application) and include only the wireless baseband processor 1026 and / or the application processor 1006. In other examples, the UE device 1002 may be the entire UE 102 and include additional modules of the device 1002.

[0127] As Figure 1 discussed and with respect to Figures 4A to 10 implemented, the model performance reporting component 140 is configured to receive an ML model performance reporting configuration from a network entity; and send an ML model performance report prepared according to the ML model performance reporting configuration to the network entity, the ML model performance report conveying the performance of the current ML model for compressing CSI, the performance being based on a comparison of the CSI with the decompressed output of the current ML model. The model performance reporting component 140 may be located within the application processor 1006 (e.g., at 140a), within the wireless baseband processor 1026 (e.g., at 140b), or within both the application processor 1006 and the wireless baseband processor 1026. The model performance reporting components 140a to 140b may be one or more hardware components specifically configured to execute the stated processes / algorithms, implemented by one or more processors configured to execute the stated processes / algorithms, stored in a computer-readable medium for implementation by one or more processors, or a combination thereof.

[0128] Figure 11FIG. 1100 is an example showing a hardware implementation of one or more network entities 104. The one or more network entities 104 may be a base station, a component of a base station, or may implement base station functions. The one or more network entities 104 may include or may correspond to at least one of RU 106, DU 108, or CU 110. The CU 110 may include a CU processor 1146, which may have on-chip memory 1146'. In some aspects, the CU 110 may further include an additional memory module 1156 and / or a communication interface 1148, both of which may be coupled to the CU processor 1146. The CU 110 may communicate with the DU 108 via an intermediate link 162 (such as an F1 interface between the communication interface 1148 of the CU 110 and the communication interface 1128 of the DU 108).

[0129] The DU 108 may include a DU processor 1126, which may have on-chip memory 1126'. In some aspects, the DU 108 may further include an additional memory module 1136 and / or a communication interface 1128, both of which may be coupled to the DU processor 1126. The DU 108 may communicate with the RU 106 via a fronthaul link 160 between the communication interface 1128 of the DU 108 and the communication interface 1108 of the RU 106.

[0130] The RU 106 may include an RU processor 1106, which may have on-chip memory 1106'. In some aspects, the RU 106 may further include an additional memory module 1116, a communication interface 1108, and one or more transceivers 1130, all of which may be coupled to the RU processor 1106. The RU 106 may further include an antenna 1140, which may be coupled to one or more transceivers 1130 such that the RU 106 may communicate with the UE 102 via the antenna 1140 through one or more transceivers 1130.

[0131] The on-chip memories 1106', 1126', 1146' and the additional memory modules 1116, 1136, 1156 can each be regarded as computer-readable media / memories. Each computer-readable media / memory can be non-transitory. Each of the processors 1106, 1126, 1146 is responsible for general processing, including executing software stored on the computer-readable media / memories. The software, when executed by the corresponding processors 1106, 1126, 1146, causes the processors 1106, 1126, 1146 to perform the various functions described herein. The computer-readable media / memories can also be used to store data manipulated by the processors 1106, 1126, 1146 when executing the software. In an example, the model performance configuration component 150 can be located at any one of one or more network entities 104, such as at the CU 110; at both the CU 110 and the DU 108; at each of the CU 110, the DU 108, and the RU 106; at the DU 108; at both the DU 108 and the RU 106; or at the RU 106.

[0132] As Figure 1 discussed, the model performance configuration component 150 is configured to send an ML model performance report configuration to the UE; and receive from the UE an ML model performance report prepared according to the ML model performance report configuration, where the ML model performance report conveys the performance of the current ML model for compressing CSI, and the performance is based on a comparison of the CSI with the decompressed output of the current ML model. The model performance configuration component 150 can be within one or more processors of one or more network entities 104, such as within the RU processor 1106 (e.g., at 150a), the DU processor 1126 (e.g., at 150b), and / or the CU processor 1146 (e.g., at 150c). The model performance configuration components 150a to 150c can be one or more hardware components specifically configured to execute the stated processes / algorithms, implemented by one or more processors 1106, 1126, 1146 configured to execute the stated processes / algorithms, stored in a computer-readable medium for implementation by one or more processors 1106, 1126, 1146, or a combination thereof.

[0133] The specific order or hierarchy of the boxes in the processes and flowcharts disclosed herein is illustrative of example methods. Thus, the specific order or hierarchy of the boxes in the processes and flowcharts can be rearranged. Some boxes can also be combined or deleted. The dashed lines can represent optional elements of the figure. The appended method claims present the elements of the various boxes in an example order and are not limited to the specific order or hierarchy presented in the claims, processes, and flowcharts.

[0134] The detailed description set forth herein describes various configurations in connection with the accompanying drawings, but is not intended to represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for providing a thorough explanation of the various concepts. However, the concepts may be practiced without these specific details. In some instances, well-known structures and components are shown in block diagram form in order to avoid obscuring such concepts.

[0135] Aspects of a wireless communication system, such as a telecommunications system, are presented with reference to various apparatuses and methods. These apparatuses and methods are described in the following detailed description and are illustrated in the accompanying drawings by various blocks, components, circuits, processes, call flows, systems, algorithms, etc. (collectively referred to as "elements"). These elements may be implemented using electronic hardware, computer software, or a combination thereof. Whether such elements are implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system.

[0136] An element, or any portion of an element, or any combination of elements may be implemented as a "processing system" that includes one or more processors. Examples of processors include a microprocessor, a microcontroller, a graphics processing unit (GPU), a central processing unit (CPU), an application processor, a digital signal processor (DSP), a reduced instruction set computing (RISC) processor, a system on a chip (SoC), a baseband processor, a field programmable gate array (FPGA), a programmable logic device (PLD), a state machine, gated logic, discrete hardware circuits, and other similar hardware configured to perform the various functions described throughout this disclosure. One or more processors in the processing system may execute software, which may be referred to as software, firmware, middleware, microcode, hardware description language, or other. Software should be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software components, applications, software applications, software packages, routines, subroutines, objects, executables, execution threads, processes, functions, or any combination thereof.

[0137] If the functions described herein are implemented in software, the functions may be stored on a computer-readable medium, such as a non-transitory computer-readable storage medium, or encoded as one or more instructions or code on the computer-readable medium. The computer-readable medium includes computer storage media and may include random access memory (RAM), read-only memory (ROM), electrically erasable programmable ROM (EEPROM), optical disk storage, magnetic disk storage, other magnetic storage devices, combinations of these types of computer-readable media, or any other medium that can be used to store computer-executable code in the form of computer-accessible instructions or data structures. The storage media may be any available media that is computer-accessible.

[0138] The aspects, implementations, and / or use cases described herein can be implemented across many different platform types, devices, systems, shapes, sizes, and packaging arrangements. For example, the aspects, implementations, and / or use cases can be realized via integrated chip implementations and other non-module component-based devices such as end-user devices, vehicles, communication devices, computing devices, industrial equipment, retail / purchasing devices, medical devices, artificial intelligence (AI)-enabled devices, machine learning (ML)-enabled devices, etc. The scope of the aspects, implementations, and / or use cases can range from chip-level or modular components to non-modular or non-chip-level implementations and further to aggregated, distributed, or original equipment manufacturer (OEM) devices or systems incorporating one or more of the technologies described herein.

[0139] Devices incorporating the aspects and features described herein may also include additional components and features for implementing and practicing the claimed and described aspects and features. For example, the transmission and reception of wireless signals necessarily involve many components for analog and digital purposes, such as hardware components, antennas, RF chains, power amplifiers, modulators, buffers, processors, interleavers, adders / summers, etc. The techniques described herein can be practiced in a wide variety of devices, chip-level components, systems, distributed arrangements, aggregated or disaggregated components, end-user devices, etc., in various configurations.

[0140] The description herein is provided to enable those skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects. Thus, the claims are not limited to the aspects described herein, but should be construed in view of the full scope of the present disclosure consistent with the language of the claims.

[0141] References to singular elements do not, unless expressly stated otherwise, mean "one and only one" but rather "one or more." Terms such as "if," "when," and "while" do not imply an immediate temporal relationship or reaction. That is, these phrases (e.g., "when") do not mean an immediate action in response to or during the occurrence of an action, but rather mean that if a certain condition is met, a certain action will occur, without requiring a specific or immediate temporal constraint on the occurrence of the action. As used in this disclosure, the terms "may," "might," and "can" generally carry certain connotations. For example, "may" refers to a permissive feature that may or may not occur, "might" refers to a feature that is likely to occur, and "can" refers to an ability (e.g., be able to). The phrase "for example" generally carries a similar connotation to "may," and thus, "may" is sometimes excluded from sentences that include "for example" or other similar phrases.

[0142] Unless otherwise expressly stated, the term "some" means one or more. Combinations such as "at least one of A, B, or C" or "one or more of A, B, or C" include any combination of A, B, and / or C, such as A and B, A and C, B and C, or A and B and C, and may include multiple A's, multiple B's, and / or multiple C's, or may include only A, only B, or only C. A set shall be interpreted as a set of elements having a number of one or more.

[0143] Unless otherwise expressly indicated, ordinal terms such as "first" and "second" do not necessarily imply an order in time, sequence, numerical value, etc., but are used to distinguish different instances of the terms or phrases following each ordinal term. Reference numerals used in the specification and drawings are sometimes cross-referenced between the drawings to indicate the same or similar features. Features that are identical in multiple drawings may be labeled with the same reference numeral in multiple drawings. Features that are similar but not identical between multiple drawings may be labeled with reference numerals having different leading digits but having one or more of the same trailing digits (e.g., 206, 306, 406, etc. may refer to similar features in the drawings). Sometimes, "X" is used to generally denote multiple variations of a feature. For example, "X06" may generally refer to all reference numbers ending with "06" (e.g., 206, 306, 406, etc.).

[0144] Structural equivalents and functional equivalents of elements of the various aspects described throughout this disclosure that are known or later become known to those of ordinary skill in the art are hereby expressly incorporated by reference herein and are covered by the claims. The words "module", "mechanism", "element", "device", etc. may not be substitutes for the word "component". Accordingly, no claim element shall be construed as means-plus-function unless the claim element is expressly recited using the phrase "means for...". As used herein, the phrase "based on" shall not be construed as a reference to a closed set of information, one or more conditions, one or more factors, etc. In other words, unless expressly stated otherwise, the phrase "based on A" (where "A" may be information, a condition, a factor, etc.) shall be construed as "at least based on A".

[0145] The following examples are illustrative only and may be combined with other examples or teachings described herein without limitation.

[0146] Example 1 is a method of wireless communication at a UE, including: A method of wireless communication performed by a UE, the method including: receiving an ML model performance report configuration from a network entity; and sending to the network entity an ML model performance report prepared according to the ML model performance report configuration, the ML model performance report conveying the performance of the current ML model for compressing CSI, the performance being based on a comparison of the CSI with the decompressed output of the current ML model.

[0147] Example 2 can be combined with Example 1 and includes triggering the sending of the ML model performance report by meeting a reporting criterion.

[0148] Example 3 can be combined with Example 1 and further includes performing an ML model performance monitoring process, wherein the sending of the ML model performance report is based on performing the ML model performance monitoring process.

[0149] Example 4 can be combined with Example 3 and further includes initiating the ML model performance monitoring process when a monitoring condition is met, the monitoring condition being indicated by the ML model performance report configuration.

[0150] Example 5 can be combined with Example 3 and further includes initiating the ML model performance monitoring process when an ML-based CSI reporting condition is met, the ML-based CSI reporting condition being indicated by an ML-based CSI report configuration.

[0151] Example 6 can be combined with Example 3 and includes the receipt of the ML model performance report configuration triggering the execution of the ML model performance monitoring process.

[0152] Example 7 can be combined with any one of Examples 1 to 6 and further includes sending to the network entity an ML-based CSI report conveying compressed CSI.

[0153] Example 8 can be combined with any one of Examples 1 to 7 and further includes receiving from the network entity a configuration of an ML-based CSI report.

[0154] Example 9 can be combined with Example 8 and further includes sending to the network entity an ML model performance report prepared according to the configuration of the ML-based CSI report.

[0155] Example 10 can be combined with any one of Examples 1 to 9 and further includes sending to the network entity UE capability information indicating at least one of: ML model performance monitoring capability, or ML-based CSI reporting capability.

[0156] Example 11 can be combined with any one of Examples 1 to 10 and further includes sending to the network entity a first indication that the UE prefers ML-based reporting over non-ML-based reporting.

[0157] Example 12 can be combined with any one of Examples 1 to 10 and further includes sending a second indication to the network entity that the UE prefers a non-ML-based report over an ML-based report.

[0158] Example 13 can be combined with any one of Examples 1 to 10 and further includes sending a third indication to the network entity that the UE prefers to replace the current ML model with a different ML model.

[0159] Example 14 is a method of wireless communication performed by a network entity, the method including: sending an ML model performance report configuration to a UE; and receiving, from the UE, an ML model performance report prepared according to the ML model performance report configuration, the ML model performance report conveying the performance of a current ML model for compressing CSI, the performance being based on a comparison of the CSI with a decompressed output of the current ML model.

[0160] Example 15 can be combined with Example 14 and includes that the reception of the ML model performance report is associated with meeting a reporting criterion.

[0161] Example 16 can be combined with Example 14 and includes that the reception of the ML model performance report is associated with an ML model performance monitoring process.

[0162] Example 17 can be combined with Example 16 and includes that the sending of the ML model performance report configuration triggers the ML model performance monitoring process.

[0163] Example 18 can be combined with any one of Examples 14 to 17 and further includes receiving, from the UE, an ML-based CSI report conveying compressed CSI.

[0164] Example 19 can be combined with any one of Examples 14 to 18 and further includes sending a configuration of the ML-based CSI report to the UE.

[0165] Example 20 can be combined with Example 19 and further includes receiving, from the UE, an ML model performance report prepared according to the configuration of the ML-based CSI report.

[0166] Example 21 can be combined with any one of Examples 14 to 20 and further includes receiving, from the UE, UE capability information indicating at least one of: ML model performance monitoring capability, or ML-based CSI reporting capability.

[0167] Example 22 can be combined with any one of Examples 14 to 21 and further includes receiving, from the UE, a first indication that the UE prefers an ML-based report over a non-ML-based report.

[0168] Example 23 can be combined with any of Examples 14 to 21 and further includes receiving, from the UE, a second indication that the UE prefers a non-ML-based report over an ML-based report.

[0169] Example 24 can be combined with any of Examples 14 to 21 and further includes receiving, from the UE, a third indication that the UE prefers to replace the current ML model with a different ML model.

[0170] Example 25 is an apparatus for wireless communication, configured to implement the method according to any of Examples 1 to 24.

[0171] Example 26 is an apparatus for wireless communication, comprising components for implementing the method according to any of Examples 1 to 24.

[0172] Example 27 is a non-transitory computer-readable medium storing computer-executable code that, when executed by a processor, causes the processor to implement the method according to any of Examples 1 to 24.

Claims

1. A method of wireless communication performed by a user equipment UE (102), the method comprising: receiving (406) a machine learning ML model performance report configuration from a network entity (104); and sending (408, 508c) to the network entity (104) an ML model performance report prepared according to the ML model performance report configuration, the ML model performance report conveying the performance of a current ML model for compressing channel state information CSI, the performance being based on a comparison of the CSI with the decompressed output of the current ML model.

2. The method according to claim 1, wherein the sending (408, 508c) of the ML model performance report is triggered by satisfaction of a reporting criterion (508b).

3. The method according to claim 1, further comprising: performing (508a) an ML model performance monitoring process, wherein the sending (408, 508c) of the ML model performance report is based on the performing (508a) of the ML model performance monitoring process.

4. The method according to claim 3, further comprising: initiating (508a) the ML model performance monitoring process when a monitoring condition (607a) is satisfied, the monitoring condition being indicated by the ML model performance report configuration.

5. The method according to claim 3, further comprising: initiating (508a) the ML model performance monitoring process when an ML-based CSI reporting condition (607b) is satisfied, the ML-based CSI reporting condition being indicated by an ML-based CSI reporting configuration.

6. The method according to claim 3, wherein the receiving (406) of the ML model performance report configuration triggers the performing (508a) of the ML model performance monitoring process.

7. The method according to any one of claims 1 to 6, further comprising: sending (412, 613) to the network entity (104) an ML-based CSI report conveying compressed CSI.

8. The method according to any one of claims 1 to 7, further comprising: receiving (410) a configuration of an ML-based CSI report from the network entity (104).

9. The method according to claim 8, further comprising: sending (411) to the network entity (104) the ML model performance report prepared according to the configuration of the ML-based CSI report.

10. The method according to any one of claims 1 to 9, further comprising: sending (404, 405) to the network entity (104) UE capability information indicating at least one of the following: ML model performance monitoring capability, or ML-based CSI reporting capability.

11. The method according to any one of claims 1 to 10, further comprising: sending (703) to the network entity (104) a first indication that the UE (102) prefers ML-based reporting over non-ML-based reporting.

12. The method according to any one of claims 1 to 10, further comprising: Send (803) a second indication to the network entity (104) that the UE (102) prefers a non-ML-based report over an ML-based report.

13. The method according to any one of claims 1 to 10, further comprising: Send (903) a third indication to the network entity (104) that the UE (102) prefers to replace the current ML model with a different ML model.

14. A method of wireless communication performed by a network entity (104), the method comprising: Send (406) a machine learning ML model performance report configuration to a user equipment UE (102); and Receive (408, 508c) from the UE (102) an ML model performance report prepared according to the ML model performance report configuration, the ML model performance report conveying the performance of the current ML model for compressing channel state information CSI, the performance being based on a comparison of the CSI with the decompressed output of the current ML model.

15. An apparatus for wireless communication, comprising a memory, a transceiver, and a processor, the processor coupled to the memory and the transceiver, the apparatus configured to implement the method according to any one of claims 1 to 14.

Images