Model monitoring for ML-based CSI compression

Abstract

The UE (102) receives (310a-310b) at least one downlink signal from the network entity (104) to monitor (316a-316b) performance of an ML model for CSI compression. Monitoring (316a-316b) the performance is based on a measurement of the at least one downlink signal. The UE transmits (318, 414a-414b) information associated with the performance of the ML model for CSI compression to the network entity (104) based on the measurement of the at least one downlink signal. When the information associated with the performance of the ML model indicates a performance failure, the UE communicates (322) with the network entity (104), where the communication (322) applies at least one of an update to the ML model, a handover of the ML model to a different ML model, or a non-ML CSI report.

Description

Technical Field

[0001] The present disclosure relates generally to wireless communications and, more particularly, to performing fault monitoring of machine learning (ML) models. Background Art

[0002] The 3rd Generation Partnership Project (3GPP) specifies a radio interface called fifth generation (5G) New Radio (NR) (5G NR). The architecture of a 5G NR wireless communication system may include a 5G core (5GC) network, a 5G radio access network (5G-RAN), user equipment (UE), etc. Compared to other types of wireless communication systems, the 5G NR architecture may provide increased data rates, reduced latency, and / or increased capacity.

[0003] Wireless communication systems may generally be configured to provide various telecommunication services (e.g., telephony, video, data, messaging, broadcast, etc.) based on multiple access technologies (such as orthogonal frequency division multiple access (OFDMA) technologies) that support communication with multiple UEs. Improvements in mobile broadband have helped to continue the advancement of such wireless communication technologies. For example, a machine learning (ML) model may improve wireless performance, but the ML model may also experience performance failures for certain types of channel conditions or due to channel congestion. Summary of the invention

[0004] The following presents a simplified overview of one or more aspects in order to provide a basic understanding of such aspects. This overview is not an extensive review of all contemplated aspects. This overview neither identifies the key or important 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 a more detailed description presented later.

[0005] A user equipment (UE) may utilize a machine learning (ML) model to perform channel state information (CSI) compression for transmitting compressed CSI reports to a network entity, such as a base station or an entity of a base station. However, some ML models may experience performance failures for certain types of channel conditions. For example, an ML model may be trained using offline field data associated with some channel conditions, but for another less common channel condition, offline field data may be more difficult to obtain, which may result in performance failures of the ML model during the inference phase. In addition, the channel may experience changes due to channel obstruction, which may also cause the ML model to experience performance failures.

[0006] Various aspects of the present disclosure address the above and other deficiencies by configuring a UE to monitor the performance of an ML model and indicate to a network entity when a performance failure of the ML model occurs so that the UE and / or the network entity can adjust the ML model. For example, the UE and the network entity can update / switch the ML model or fall back to a non-ML communication technology. Either the UE or the network entity can be able to detect an ML model failure. Whichever entity detects the ML model failure can then indicate the ML model failure to the other entity. Based on the ML model failure detection and reporting, the UE and the network entity can adjust communications managed by the ML model.

[0007] According to some aspects, a UE receives at least one downlink signal from a network entity to monitor performance of an ML model for CSI compression. While monitoring the performance, the UE measures at least one downlink signal. The UE transmits information associated with the performance of the ML model for CSI compression to the network entity based on the measured value of the at least one downlink signal. When the information associated with the performance of the ML model indicates a performance failure, the UE communicates with the network entity. The communication applies at least one of: updating the ML model, switching the ML model, or using non-ML CSI reporting.

[0008] According to some aspects, a network entity transmits at least one downlink signal to a UE to monitor performance of an ML model for CSI compression. The network entity receives information associated with performance of the ML model for CSI compression from the UE based on a measurement of the at least one downlink signal. When the information associated with the performance of the ML model indicates a performance failure, the network entity modifies communications with the UE. The communications modification applies at least one of: an update to the ML model, switching the ML model to a different ML model, or using non-ML CSI reporting.

[0009] To accomplish the foregoing and related purposes, one or more aspects correspond to the features described below and particularly pointed out in the claims. One or more aspects can be implemented by any of an apparatus, a method, a component for executing the method, and / or a non-transitory computer-readable medium. The following description and the accompanying drawings set forth in detail certain illustrative features of one or more aspects. However, these features indicate only a few of the various ways in which the principles of the various aspects can be employed. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0011] Figure 2A diagram illustrating an example process for machine learning (ML) based channel state information (CSI) encoder compression at a UE and ML based CSI decoder decompression at a network entity.

[0012] Figure 3 is a signaling diagram illustrating an example of UE-based ML model monitoring.

[0013] Figure 4 is a signaling diagram illustrating an example of network-based ML model monitoring.

[0014] Figure 5 is a flow chart of a method for performance failure monitoring for ML performed by a UE.

[0015] Figure 6 is a flow chart of a method for performance fault monitoring for ML performed by a network entity.

[0016] Figure 7 is a diagram illustrating an example of a hardware implementation of an example UE equipment.

[0017] Figure 8 is a diagram illustrating an example of a hardware implementation of one or more example network entities. DETAILED DESCRIPTION

[0018] Figure 1 A diagram 100 of a wireless communication system associated with a plurality of cells 190 is shown. The wireless communication system includes a user equipment (UE) 102 and base stations 104, some of which 104a include an aggregated base station architecture and other base stations 104b include a decomposed base station architecture. The aggregated base station architecture includes a radio unit (RU) 106, a distributed unit (DU) 108, and a centralized 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 decomposed base station architecture utilizes a protocol stack physically or logically distributed between two or more units (e.g., RU 106, DU 108, CU 110). For example, CU 110 is implemented within a RAN node, and one or more DUs 108 may be co-located with CU 110, or alternatively, may be geographically or virtually distributed in one or more other RAN nodes. 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).

[0019] The operation and / or network design of the base station 104 can be based on the aggregated nature of the base station functions. For example, a decomposed 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). Decomposition may include distributing functions between two or more units located at various physical locations, and virtually distributing the functions of at least one unit, which enables flexibility in network design. Various units of a decomposed base station architecture or a decomposed RAN architecture may be configured to communicate with at least one other unit in wired or wireless communication. For example, CU 110a communicates with DU 108a-108b via a corresponding midhaul link 162 based on an F1 interface. DU 108a-108b may communicate with RU 106a and RU106b-106c via corresponding fronthaul links 160, respectively. RU 106a-106c can communicate with corresponding UE 102a-102c and 102s via one or more radio frequency (RF) access links based on Uu interface. In an example, multiple RU 106 and / or base station 104 can provide services for UE 102 at the same time, such as the access link of RU 106a of cell 190a and the UE 102a of cell 190e served by base station 104a of cell 190e at the same time.

[0020] One or more CUs 110, such as CU 110a or CU 110d, may communicate directly with the core network 120 via a backhaul link 164. For example, CU 110d communicates with the core network 120 via a backhaul link 164 based on a next generation (NG) interface. One or more CUs 110 may also communicate indirectly with the core network 120 through one or more decomposed base station units, such as a near real-time RAN intelligent controller (RIC) 128 via an E2 link and a service management and orchestration (SMO) framework 116 that may be associated with a non-real-time RIC 118. The near real-time RIC 128 may communicate with the SMO framework 116 and / or the non-real-time RIC 118 via an A1 link. The SMO framework 116 and / or the non-real-time RIC 118 may also communicate with an open cloud (O-cloud) 130 via an O2 link. One or more CUs 110 may further communicate with each other via a backhaul link 164 based on an Xn interface. For example, the CU 110d of the base station 104a communicates with the CU 110a of the base station 104b via the backhaul link 164 based on the Xn interface. Similarly, the base station 104a of the cell 190e can communicate with the CU 110a of the base station 104b via the backhaul link 164 based on the Xn interface.

[0021] RU 106, DU 108 and CU 110 and near real-time RIC 128, non-real-time RIC 118 and / or SMO framework 116 may include (or may be coupled to) one or more interfaces configured to transmit or receive information / signals via a wired or wireless transmission medium. Base station 104 or any one of the one or more decomposed base station units may be configured to communicate with one or more other base stations 104 or one or more other decomposed base station units via a wired or wireless transmission medium. In an example, a processor, memory and / or controller associated with executable instructions of the interface may be configured to provide communication between base station 104 and / or one or more decomposed base station units via a wired or wireless transmission medium. For example, a wired interface may be configured to transmit or receive information / signals via a wired transmission medium, such as a fronthaul link 160 between a RU 106d and a baseband unit (BBU) 112 for a cell 190d, or more specifically, a fronthaul link 160 between a RU 106d and a DU 108d. The BBU 112 includes the DU 108d and the CU 110d, which may also have a wired interface configured between the DU 108d and the CU 110d to transmit or receive information / signals between the DU 108d and the CU 110d based on the midhaul link 162. In a further example, a wireless interface, which may include a receiver, a transmitter, or a transceiver (such as an RF transceiver), may be configured to transmit or receive information / signals via a wireless transmission medium, such as for information transmitted between the RU 106a of the cell 190a and the base station 104a of the cell 190e via cross-cell communication beams of the RU 106a and the base station 104a.

[0022] One or more high-level control functions (such as functions related to radio resource control (RRC), packet data convergence protocol (PDCP), service data adaptation protocol (SDAP), etc.) can be hosted at CU 110. Each control function can be associated with an interface for transmitting signals based on one or more other control functions hosted at CU 110. User plane functions (such as central unit-user plane (CU-UP) functions), control plane functions (such as central unit-control plane (CU-CP) functions), or a combination thereof can be implemented based on CU 110. For example, CU 110 may include one or more CU-UP processes and / or one or more CU-CP processes. When implemented in an O-RAN configuration, the CU-UP function can be based on bidirectional communication with the CU-CP function via an interface (such as an E1 interface (not shown)).

[0023] The CU 110 may communicate with the DU 108 for network control and signal transmission. The DU 108 is a logical unit of the base station 104 that is configured to perform one or more base station functions. For example, the DU 108 may control the operation of one or more RUs 106. One or more of the following may be hosted at the DU 108: a radio link control (RLC) layer, a media access control (MAC) layer, or one or more higher physical (PHY) layers, such as forward error correction (FEC) modules for encoding / decoding, scrambling, modulation / demodulation, etc. The DU 108 may host such functions based on the functional division of the DU 108. The DU 108 may similarly host one or more lower PHY layers, where each lower layer or module may be implemented based on an interface for communicating with other layers and modules hosted at the DU 108, or based on a control function hosted at the CU 110.

[0024] The RU 106 may be configured to implement lower layer functions. For example, the RU 106 is controlled by the 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 the RU 106 may be based on functional partitioning, such as lower layer functional partitioning.

[0025] The RU 106 may transmit or receive over-the-air (OTA) communications with one or more UEs 102. For example, the RU 106b of the cell 190b communicates with the UE 102b of the cell 190b via the first communication beam set 132 of the RU 106b and the second communication beam set 134 of the UE 102b, which may correspond to inter-cell communication beams or inter-cell communication beams. Both real-time and non-real-time features of the control plane and user plane communications of the RU 106 may be controlled by the associated DU 108. Thus, the DU 108 and the CU 110 may be used in a cloud-based RAN architecture (such as a vRAN architecture), and the SMO framework 116 may be used to support non-virtualized and virtualized RAN network elements. For non-virtualized network elements, the SMO framework 116 may support the deployment of dedicated physical resources for RAN coverage, where the dedicated physical resources may be managed through an operation and maintenance interface (such as an O1 interface). For virtualized network elements, the SMO framework 116 can interact with a cloud computing platform (such as O-cloud 130) via an O2 link (e.g., a cloud computing platform interface) to manage the network elements. The virtualized network elements may include, but are not limited to, RU 106, DU 108, CU 110, near real-time RIC 128, etc.

[0026] The SMO framework 116 may be configured to communicate directly with one or more RUs 106 using an O1 link. The non-real-time RIC 118 of the SMO framework 116 may also be configured to support the functionality of the SMO framework 116. For example, the non-real-time RIC 118 implements logic functions that are capable of controlling non-real-time RAN features and resources, features / applications of the near real-time RIC 128, and / or artificial intelligence / machine learning (AI / ML) processes. The non-real-time RIC 118 may communicate (or couple) with the near real-time RIC 128, such as through an A1 interface. The near real-time RIC 128 may implement logic functions that are capable of controlling near real-time RAN features and resources based on data collection and interaction through an E2 interface (such as an E2 interface between the near real-time RIC 128 and the CU 110a and the DU 108b).

[0027] The non-real-time RIC 118 may receive parameters or other information from an external server to generate an AI / ML model for deployment in the near-real-time RIC 128. For example, the non-real-time RIC 118 receives parameters or other information from the O-cloud 130 via the O2 link to deploy the AI / ML model to the real-time RIC 128 via the A1 link. The near-real-time RIC 128 may utilize the parameters and / or other information received from the non-real-time RIC 118 or the SMO framework 116 via the A1 link to perform near-real-time functions. The near-real-time RIC 128 and the non-real-time RIC 115 may be configured to adjust the performance of the RAN. For example, the non-real-time RIC 116 monitors patterns and long-term trends to improve the performance of the RAN. The non-real-time RIC 116 may also deploy the AI / ML model through the SMO framework 116 for implementing corrective actions, such as initiating reconfiguration of the O1 link or instructing the management process of the A1 link.

[0028] Any combination of RU 106, DU 108, and CU 110 or any reference thereto individually may correspond to base station 104. Therefore, base station 104 may include at least one of RU 106, DU 108, or CU 110. Base station 104 provides UE 102 with access to core network 120. That is, base station 104 may relay communications between UE 102 and core network 120. Base station 104 may 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 corresponds to a macro cell, and cells 190a-190d may correspond to a small cell. Small cells include femto cells, pico cells, micro cells, etc. A cell structure including at least one macro cell and at least one small cell may be referred to as a "heterogeneous network."

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

[0030] The communication link between UE 102 and 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. UE 102 and base station 104 / RU 106 can utilize Y MHz (e.g., 5 MHz, 10 MHz, 15 MHz, 20 MHz, 100 MHz, 400 MHz, etc.) spectrum bandwidth allocated per carrier in up to a total of Yx MHz carrier aggregation, where x component carriers (CCs) are used for communication in each direction of the uplink direction and the downlink direction. The carriers may be adjacent to each other along the spectrum, or may not be adjacent to each other. In an example, uplink carriers and downlink carriers may be allocated in an asymmetric manner, and more or fewer carriers may be allocated for uplink or downlink. A component carrier may include a primary component carrier and one or more secondary component carriers. The primary component carrier may be associated with a primary cell (PCell), and the secondary component carrier may be associated with a secondary cell (SCell).

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

[0032] The electromagnetic spectrum is often subdivided into different categories, bands, channels, etc. based on different frequencies / wavelengths associated with the electromagnetic spectrum. Fifth Generation (5G) NR is often associated with two operating bands, referred to as Frequency Range 1 (FR1) and Frequency Range 2 (FR2). FR1 ranges from 410 MHz – 7.125 GHz, and FR2 ranges from 24.25 GHz – 52.6 GHz. FR1 is often referred to as the “sub-6 GHz” band, although a portion of FR1 is actually greater than 6 GHz. In contrast, FR2 is often referred to as the “millimeter wave” (mmW) band. FR2 is not the same as the “extremely high frequency” (EHF) band, but is an approximate subset of that band, the EHF band ranges from 30 GHz – 300 GHz, and is sometimes also referred to as the “millimeter wave” band. Frequencies between FR1 and FR2 are often referred to as “mid-band” frequencies. The operating band of mid-band frequencies may be referred to as frequency range 3 (FR3), which ranges from 7.125 GHz – 24.25 GHz. The frequency bands within FR3 may include characteristics of FR1 and / or FR2. Thus, the features of FR1 and / or FR2 may be extended to mid-band frequencies. Higher operating bands have been identified to extend 5G NR communications above 52.6 GHz associated with the upper limit of FR2. Three of these higher operating bands include FR2-2 (ranging from 52.6 GHz – 71 GHz), FR4 (ranging from 71 GHz – 114.25 GHz), and FR5 (ranging from 114.25 GHz – 300 GHz). The upper limit of FR5 corresponds to the upper limit of the EHF band. Therefore, unless otherwise expressly stated herein, the term “below 6 GHz” may refer to frequencies less than 6 GHz, frequencies within FR1, or frequencies that may include mid-band frequencies. Further, unless otherwise expressly stated herein, the term "millimeter wave" or mmW refers to frequencies that may include mid-band frequencies, frequencies that may be within FR2, FR4, FR2-2 and / or FR5, or frequencies that may be within the EHF band.

[0033] UE 102 and base station 104 / RU 106 may each include multiple antennas. 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 beam set 132 in one or more transmission directions of RU 106b. UE 102b may receive a downlink beamformed signal from RU 106b based on a second beam set 134 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 beam set 134 in one or more transmission directions of UE 102b. RU 106b may receive an uplink beamformed signal from UE 102b in one or more reception directions of RU 106b. UE 102b may perform beam training to determine the optimal reception and transmission direction of the beamformed signal. The transmission and reception directions of the UE 102 and the base station 104 / RU 106 may be the same, or may be different. In a further example, a beamformed signal may be transmitted between the first base station 104a and the second base station 104b. For example, the RU 106a of the cell 190a may transmit a beamformed signal to the base station 104a of the cell 190e based on the RU beam set 136 in one or more transmission directions of the RU 106a. The base station 104a of the cell 190e may receive a beamformed signal from the RU 106a based on the base station beam set 138 in one or more reception directions of the base station 104a. Similarly, the base station 104a of the cell 190e may transmit a beamformed signal to the RU 106a based on the base station beam set 138 in one or more transmission directions of the base station 104a. The RU 106a may receive a beamformed signal from the base station 104a of the cell 190e based on the RU beam set 136 in one or more reception directions of the RU 106a.

[0034] The base station 104 may include and / or be referred to as a next generation evolved Node B (ng-eNB), a generation NB (gNB), an evolved NB (eNB), an access point, a base transceiver station, a radio base station, a radio transceiver, a transceiver function, a basic service set (BSS), an extended service set (ESS), a transmission reception point (TRP), a network node, a network entity, a network device, or other related terms. The base station 104 or an entity at the base station 104 may be implemented as an IAB node, a relay node, a side link node, a converged (monolithic) base station having a RU 106 and a BBU including a DU 108 and a CU 110, or as a decomposed base station 104b including one or more of the RU 106, the DU 108, and / or the CU 110. The set of converged or decomposed base stations 104a-104b may be referred to as a next generation radio access network (NG-RAN).

[0035] The core network 120 may include an access and mobility management function (AMF) 121, a session management function (SMF) 122, a user plane function (UPF) 123, a unified data management (UDM) 124, a gateway mobile location center (GMLC) 125, and / or a location management function (LMF) 126. The core network 120 may also include one or more location servers (which may include the GMLC 125 and the LMF 126), as well as other functional entities. For example, the one or more location servers include one or more location / positioning servers, and in addition to one or more of a positioning determination entity (PDE), a serving mobile location center (SMLC), a mobile positioning center (MPC), etc., the one or more location / positioning servers may also include the GMLC 125 and the LMF 126.

[0036] AMF 121 is a control node that handles signal transmission between UE 102 and core network 120. AMF 121 supports registration management, connection management, mobility management, and other functions. SMF 122 supports session management and other functions. UPF 123 supports packet routing, packet forwarding, and other functions. UDM 124 supports the generation of authentication and key agreement (AKA) credentials, user identity handling, access authorization, and subscription management. GMLC 125 provides an interface for clients / applications (e.g., emergency services) to access UE positioning information. LMF 126 receives measurement and assistance information from NG-RAN and UE 102 via AMF 121 to calculate the positioning of UE 102. NG-RAN can use one or more positioning methods to determine the location of UE 102. Positioning UE 102 can involve signal measurement, positioning estimation, and optional speed calculation based on measurement. Signal measurement can be performed by UE 102 and / or serving base station 104 / RU 106.

[0037] The transmitted signal may also be based on one or more of a satellite positioning system (SPS) 114, such as a signal measured for positioning. In an example, the SPS 114 of the cell 190c may communicate with one or more UEs 102 (such as UE 102c) and one or more base stations 104 / RU 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 / position systems. The SPS 114 may be associated with LTE signals, NR signals (e.g., based on round trip time (RTT) and / or multiple RTT), wireless local area network (WLAN) signals, ground beacon systems (TBS), sensor-based information, NR enhanced cell ID (NR E-CID) technology, downlink departure angle (DL-AoD), downlink arrival time difference (DL-TDOA), uplink arrival time difference (UL-TDOA), uplink arrival angle (UL-AoA), and / or other systems, signals, or sensors.

[0038] UE 102 may be configured as a cellular phone, a smart phone, a Session Initiation Protocol (SIP) phone, a laptop computer, a personal digital assistant (PDA), a satellite radio, a GPS, a multimedia device, a video device, a digital audio player (e.g., a Moving Picture Experts Group (MPEG) Audio Layer 3 (MP3) player), a camera, a game console, a tablet computer, a smart device, a wearable device, a vehicle, a utility meter, a gas pump, a home appliance, a healthcare device, a sensor / actuator, a display, or any other device with similar functionality. Some of UE 102 may be referred to as Internet of Things (IoT) devices, such as parking meters, gas pumps, home appliances, vehicles, healthcare equipment, etc. UE 102 may also be referred to as a station (STA), a mobile station, a subscriber station, a mobile unit, a subscriber unit, a wireless unit, a remote unit, a mobile device, a wireless device, a wireless communication device, a remote device, a mobile subscriber station, an access terminal, a mobile terminal, a wireless terminal, a remote terminal, a handheld device, a mobile client, a client, or other similar terms. The term UE may also apply to a roadside unit (RSU), which may communicate with other RSU UEs, non-RSU UEs, the base station 104 , and / or entities at the base station 104 , such as the RU 106 .

[0039] Still refer to Figure 1In certain aspects, the UE 102 may include a machine learning (ML) performance failure component 140 configured to receive at least one downlink signal from a network entity to monitor performance of an ML model for channel state information (CSI) compression, wherein the monitored performance is based on a measurement of the at least one downlink signal; transmit information associated with the performance of the ML model for CSI compression to the network entity based on the measurement of the at least one downlink signal; and communicate with the network entity when the information associated with the performance of the ML model indicates a performance failure, wherein the communication applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-MLCSI report.

[0040] In certain aspects, the base station 104 or a network entity of the base station 104 may include an ML model adjustment component 150 configured to transmit at least one downlink signal to a UE to monitor performance of an ML model for CSI compression; receive information associated with the performance of the ML model for CSI compression from the UE based on a measurement of the at least one downlink signal; and modify communications with the UE when the information associated with the performance of the ML model indicates a performance failure, wherein the communications modification applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or non-ML CSI reporting.

[0041] therefore, Figure 1 A wireless communication system is described that can incorporate aspects of one or more of the other figures described herein such as Figure 2-Figure 4 In addition, 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 versions, LTE, LTE-Advanced (LTE-A), and other wireless technologies.

[0042] Figure 2Diagram 200 shows an example process for ML-based CSI encoder compression at UE 102 and ML-based CSI decoder decompression at network entity 104. UE 102 and network entity 104, such as a base station or an entity of a base station, may perform multiple-input multiple-output (MIMO) communication, where network entity 104 may use CSI to select a digital precoder for UE 102. Network entity 104 may configure CSI reports from UE 102 via RRC signaling (e.g., CSI-reportConfig), where UE 102 may use channel state information reference signal (CSI-RS) 245 as channel measurement resources (CMR) for UE 102 to measure downlink channels. Network entity 104 may also configure (e.g., via CSI-reportConfig) interference measurement resources (IMR) for UE 102 to measure interference to the downlink channel. Thus, UE 102 may estimate 250 a channel between UE 102 and network entity 104 based on CSI-RS 245.

[0043] The UE 102 may determine the CSI using the CMR and / or IMR configured by the network entity 104 and include the CSI in a CSI report 285 that is transmitted 280a to the network entity 104 after calculating 260a the eigenvectors for each subband and compressing 270a the CSI encoder. The CSI may include a rank indicator (RI), a precoder matrix indicator (PMI), a channel quality indicator (CQI), and / or a layer indicator (LI). The network entity 104 may use the RI and PMI to determine the digital precoder. The CQI may indicate a signal to interference plus noise ratio (SINR) for determining a modulation and coding scheme (MCS). The LI may indicate the strongest layer, such as a multi-user (MU)-MIMO pairing for low rank transmission with precoder selection 260b for a phase tracking reference signal (PT-RS).

[0044] The network entity 104 may configure (e.g., based on CSI-reportConfig) a time domain behavior, such as periodic, semi-persistent, or aperiodic reporting, for transmission 280a of a CSI report 285 to the network entity 104. 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 aperiodic CSI reporting from the UE 102 based on transmission of downlink control information (DCI) to the UE 102. The network entity 104 may receive periodic CSI reports 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 PUCCH resources for transmission 280a of a semi-persistent CSI report to the network entity 104. In other examples, transmission 280a of the semi-persistent CSI report to the network entity 104 may be on a physical uplink shared channel (PUSCH) resource triggered by the DCI. The UE 102 may likewise transmit 280a an aperiodic CSI report on a PUSCH resource triggered by the DCI.

[0045] For a first resource element (RE) k associated with CSI-RS 245, the received signal at UE 102 may be determined based on: Among them, H k Indicates the effective channel including the simulated beamforming weights, dimension N Rx × N Tx , X k Corresponding to CSI-RS 245 at RE k, Corresponding to interference plus noise, N Rx corresponds to the first number of receiving ports, and N Tx Corresponding to the second number of transmission ports.

[0046] For a second RE k associated with a physical downlink shared channel (PDSCH), the signal received at the UE 102 may be determined based on: Where W k Indicating a precoder. The network entity 104 may select 260b the same precoder for subcarriers within a subband (eg, bundled in a physical resource block (PRB)).

[0047] The UE 102 may use a type 2 CSI codebook for CSI measurement and reporting, where the precoder may be based on: Where W1 and dimension are N Tx 2L, and W2 corresponds to a subband precoder of dimension 2L xv, L indicating the number of beams, and v indicating the number of layers that may correspond to RI+1. W1 may be codebook based, while W2 may be based on the power and angle associated with each transmission. Since W2 is subband based, and there may be multiple subbands for CSI report 285, UE 102 may experience large overhead to transmit 280a the CSI report 285 to the network entity 104.

[0048] The CSI report 285 may be based on the bandwidth used for the CSI-RS 245. In an example, the codebook that the network entity may use for selection 260b of W1 may be based on: in Corresponding to the Kronecker product, L indicates the number of beams configurable via RRC signaling, N1 and N2 correspond to the number of ports, O1 and O2 correspond to the oversampling factors in the horizontal and vertical domains, which can be configured via RRC signaling. Candidate values ​​for the oversampling factor can be based on the number of CSI-RS ports indicated via N1 and N2. The codebook may include precoders with different values ​​of m and n. In some examples, the candidate values ​​can be predefined based on a standardized protocol.

[0049] The ML model may be implemented to compress 270a a CSI encoder associated with the channel estimate 250. The first v columns of the feature vector calculated 260a for each subband of the average channel may be used as input to the ML model. In an example, the feature vector may be input to a neural network at the UE 102 for compression 270a of the CSI encoder. The UE 102 transmits 280a a CSI report 285 including the compressed CSI encoder to the network entity 104.

[0050] The network entity 104 detects 280b a CSI report 285 transmitted 280a from the UE 102 and decodes the CSI report 285 including the compressed CSI encoder. The decoded CSI report 285 including the compressed CSI encoder may be input to a neural network at the network entity 104 for decompression 270a. That is, the neural network at the network entity 104 may decompress 270b the compressed CSI encoder to determine a decompressed CSI decoder. The network entity 104 may determine a feature vector from the decompressed CSI decoder, which is used as an input for compression 270a of the CSI encoder at the UE 102. The network entity 104 may select 260b a precoder for each subband based on the determined / reported feature vectors.

[0051] Some ML models may experience performance failures for certain types of channel conditions. For example, an ML model may be trained using offline field data associated with certain channel conditions, but for other less common channel conditions, offline field data may be more difficult to obtain, which may cause performance failures of the ML model during the inference phase. In addition, a channel may experience changes in channel conditions due to channel obstruction, which may also cause the ML model to experience performance failures.

[0052] Therefore, the network entity 104 may configure the UE 102 to monitor the performance of the ML model and indicate to the network entity 104 when a performance failure of the ML model occurs so that the UE 102 and / or the network entity 104 can adjust the ML model. For example, the UE 102 and the network entity 104 can update / switch the ML model or fall back to a non-ML based communication technology. Both the UE 102 and the network entity 104 may be able to detect the ML model failure. Whichever entity detects the ML model failure, the ML model failure can then be indicated to the other entity. Based on the ML model failure detection and reporting, the UE 102 and the network entity 104 can adjust the communication associated with the ML model. That is, the UE 102 and / or the network entity 104 can perform ML model performance failure detection, failure event reporting, and ML model update / switch. Figure 2 CSI compression / decompression using ML models is described. Figure 3-Figure 4 Describes monitoring the performance of ML models.

[0053] Figure 3300 is a signaling diagram illustrating an example of UE-based ML model monitoring. The network entity 104 may transmit 306 control signaling to the UE 102 to enable ML-based CSI compression and configure downlink signals / parameters for ML model monitoring. The downlink signals for ML model monitoring may correspond to CSI-RS, PDSCH transmission, etc. The parameters for ML model monitoring may correspond to ML model performance failure detection counters, ML model performance failure detection thresholds, uplink resources for transmitting ML model performance failure reports, etc.

[0054] The network entity 104 may transmit 306 control signaling to the UE 102 via RRC signaling, MAC-CE, or DCI, and based on the information about Figure 2 The described ML-based CSI reporting continues to be communicated 308 with the UE 102. In an example, the RRC signaling may include an RRCReconfiguration message. In some cases, the UE 102 communicates with the network entity 104 (e.g., operating as a secondary node (SN)) and similar to the network entity 104 (e.g., operating as a secondary node (SN)). Figure 3 102 is in dual connectivity with another network entity (operated by a master node (MN) not shown in FIG. 104 ). In an example, the SN transmits control signaling to the UE 102, as described above. In other examples, the SN transmits control signaling to the UE 102 via the MN. Communicating 308 with the UE 102 based on the ML-based CSI report can be independent of transmitting 310a-310b one or more downlink signals for ML model monitoring to the UE 102, or simultaneously with transmitting 310a-310b one or more downlink signals for ML model monitoring to the UE 102. For example, the communication 308 based on the ML-based CSI report and the transmission 310a-310b of the one or more downlink signals can be associated with the perforated resources. The UE 102 measures the one or more downlink signals transmitted 310a-310b from the network entity 104 for ML model performance fault monitoring 316a-316b and / or detection 316b. For example, if the UE 102 identifies N consecutive ML model performance failure instances based on the ML model performance failure monitoring 316 a - 316 b , the UE 102 may determine that an ML model performance failure event is detected 316 b and transmit 318 an ML model performance failure event report to the network entity 104 .

[0055] The downlink signal transmitted 310a-310b for ML model monitoring may be a precoded CSI-RS that may be transmitted on a periodic basis. In other examples, the CSI-RS may be transmitted to the UE 102 on a semi-persistent basis. The network entity 104 may receive a CSI-RS signal from the N R Antenna ports transmit CSI-RS, where NR Corresponds to the maximum number of downlink layers configured by RRC signaling. The network entity 104 may transmit the CSI-RS using the precoder for the most recent ML-based CSI report.

[0056] UE 102 may calculate cosine similarity or squared cosine similarity based on the normalized eigenvector and the estimated channel determined by UE 102 from the CSI-RS for ML-based CSI reporting and the normalized estimated channel. This calculation may correspond to: in The jth column indicating the estimated channel at the i-th subband from the CSI-RS for model performance failure monitoring, N s Corresponding to the number of subbands, corresponds to the estimated channel at the i-th subband from the CSI-RS for ML-based CSI feedback, and Corresponding to the jth column of the eigenvector of the estimated channel at the i-th subband of the CSI-RS for ML-based CSI feedback. If the cosine similarity or the squared cosine similarity is below a threshold, the UE 102 may determine that an ML model performance failure instance has occurred. The threshold may be predefined or may be configured via RRC signaling from the network entity 104.

[0057] The UE 102 may calculate a block error rate (BLER) (e.g., an assumed BLER) based on the CSI-RS for ML model monitoring and at least one of the CQI and RI for the most recent ML-based CSI report, a target spectral efficiency per layer, or a target spectral efficiency per number of layers. The target spectral efficiency may be predefined or configured by the network entity 104 based on RRC signaling. The UE 102 may calculate a block error rate (BLER) based on the CSI-RS for ML model monitoring and at least one of the CQI and RI for the most recent ML-based CSI report. The assumed BLER is determined for each port, where Corresponding to the number of layers indicated by the RI in the most recent ML-based CSI report. If the assumed BLER is above the threshold (or otherwise satisfies the first threshold criterion), the UE 102 may determine that an ML model performance failure instance has occurred. The threshold for BLER for CQI selection or for CQI selection plus an offset may be predefined or configured via RRC signaling from the network entity 104.

[0058] One or more downlink signals transmitted 310a-310b for ML model monitoring may correspond to a non-precoded CSI-RS. The non-precoded CSI-RS may be the same as the CSI-RS used for ML-based CSI reporting. In other examples, the CSI-RS may be a dedicated CSI-RS for ML model monitoring, which may be based on the same number of ports as the CSI-RS used for ML-based CSI reporting. The network entity 104 may transmit an indication of an ML model for decompression to the UE 102. The UE 102 may calculate cosine similarity or squared cosine similarity based on a first eigenvector of a channel estimated from the CSI-RS and a second eigenvector for ML-based CSI compression and decompression associated with an ML model for compressing and decompressing CSI. If the cosine similarity or squared cosine similarity is below a threshold (or otherwise satisfies a second threshold criterion), the UE 102 may determine that an ML model performance failure instance has occurred. The threshold may be predefined or may be configured via RRC signaling from the network entity 104 .

[0059] One or more downlink signals transmitted 310a-310b for ML model performance fault monitoring may correspond to a PDSCH transmission, which may include a demodulation reference signal (DMRS) in the PDSCH transmission. For different MIMO transmission schemes, such as MU-MIMO or single user (SU)-MIMO, the PDSCH of the precoder reported in the most recent ML-based CSI report is used for ML model monitoring. Therefore, the network entity 104 may indicate in the DCI that schedules the PDSCH whether the PDSCH (e.g., the DMRS in the PDSCH) is used for ML model monitoring. A 1-bit field may be included in the DCI (e.g., DCI format 1_1 or DCI format 1_2) to provide an indication of whether the PDSCH is used for ML model monitoring. The UE 102 may calculate the cosine similarity or squared cosine similarity of the eigenvector and estimated channel of the DMRS for the PDSCH and the CSI-RS for ML-based CSI reporting. If the cosine similarity or squared cosine similarity is below a threshold (or otherwise satisfies a second threshold criterion), the UE 102 may determine that an ML model performance failure instance has occurred. The threshold may be predefined or may be configured via RRC signaling from the network entity 104.

[0060] UE 102 may calculate an assumed BLER based on the DMRS in the PDSCH and the CQI and RI in the most recent ML-based CSI report. In other examples, UE 102 may calculate an assumed BLER based on the DMRS in the PDSCH and the scheduled MCS. If the assumed BLER is above the threshold (or otherwise satisfies a third threshold criterion), UE 102 may determine that an ML model performance failure instance has occurred. The threshold for CQI selection or for CQI selection plus an offset BLER may be predefined or configured via RRC signaling from network entity 104. The most recent ML-based CSI report may correspond to a most recent ML-based CSI report using an ML model Y time slots implemented prior to a time slot including one or more downlink signals transmitted 310a-310b for ML model monitoring, where Y may be predefined (e.g., Y=0) or may be configured via RRC signaling from network entity 104.

[0061] After determining that N consecutive ML model performance failure instances have occurred, the UE 102 may detect 316b an ML model performance failure. In an example, N may correspond to one of 1, ..., 32. In another example, N may be based on a predefined protocol. In yet another example, N may be configured via RRC signaling from the network entity 104. The UE 102 may increment an ML model performance failure detection counter to count the number of consecutive ML model performance failure instances. For example, when the UE 102 receives 306 control signaling from the network entity 104, the UE 102 may initialize the ML model performance failure detection counter to an initial value (e.g., zero). After each ML model performance failure instance, the UE 102 increments the ML model performance failure detection counter by one. If the ML model performance failure detection counter reaches N, the UE 102 declares an ML model performance failure event. The network entity 104 may configure the time interval for ML model performance failure monitoring 316a-316b based on the RRC signaling. In another example, the time interval for ML model performance fault monitoring 316a-316b can be determined by the UE 102 based on the time interval for one or more downlink signals.

[0062] After detecting 316b the ML model performance failure event, the UE 102 transmits 318 an ML model performance failure event report to the network entity 104, and the network entity 104 transmits 320 a response to the ML model performance failure event report to the UE 102. The response transmitted 320 to the UE 102 may indicate that communication between the network entity 104 and the UE 102 will fall back to a non-ML based CSI reporting technique, or that the ML model will be updated or switched. Thus, the UE 102 and the network entity 104 may continue to communicate 322 based on the non-ML based CSI reporting technique or the updated / switched ML model, as indicated in the response transmitted 320 to the UE 102.

[0063] If the UE 102 determines that there is no ML model performance failure after receiving 310a-310b the downlink signal from the network entity 104, the UE 102 may reset the ML model performance failure detection counter to zero. In another example, if the UE 102 determines that there is no ML model performance failure and the ML model performance failure detection counter is greater than one, the UE 102 may decrement the ML model performance failure detection counter. In still another example, if the UE 102 determines that there is no ML model performance failure for M consecutive instances (e.g., opportunities, time slots, etc.), the UE 102 may reset the ML model performance failure detection counter to zero. In yet another example, if the UE 102 determines that there is no ML model performance failure for M consecutive instances (e.g., opportunities or time slots) and the ML model performance failure detection counter is greater than one, the UE 102 may decrement the ML model performance failure detection counter. The value of M may correspond to one of 1, ..., 32, may be based on a predefined protocol, or may be configured by the network entity 104 through control signaling. The network entity 104 may configure M to be less than or equal to N in some cases, or to be greater than N in other cases. Figure 3 Model performance failure monitoring from the UE side of the communication environment is shown. Figure 4 Model performance fault monitoring from the network entity side of a communication environment is shown.

[0064] Figure 4 4 is a signaling diagram 400 illustrating an example of network-based ML model monitoring. Elements 308, 310a-310b, and 322 in the signaling diagram 400 have been described with respect to Figure 3Similar to the control signaling transmitted 306 from the network entity 104 to the UE 102 in the signaling diagram 300, the control signaling transmitted 406 from the network entity 104 to the UE 102 also enables ML-based CSI compression and configures downlink signals / parameters. However, the control signaling transmitted 406 in the signaling diagram 400 also configures UE feedback for ML model monitoring.

[0065] The network entity 104 may configure the UE 102 to report uncompressed CSI based on at least one of the CSI-RS resources configured for ML-based CSI compression. The network entity 104 may configure at least one reporting parameter using RRC signaling. The configured parameters may correspond to a subband size for uncompressed CSI, a first number of bits for amplitude quantization of each coefficient, a second number of bits for phase quantization of each coefficient, a third number of the strongest coefficients reported, and the like. In other examples, the parameters may be based on a predefined protocol, such as, the reported CSI corresponds to wideband CSI, where the number of subband sizes is 1, the number of bits for phase and amplitude quantization is equal to 3, the number of the strongest coefficients reported is the maximum number of downlink layers multiplied by the number of antenna ports for CSI-RS, and the like.

[0066] The UE 102 may determine 412a-412b feedback information for ML model monitoring based on one or more downlink signals received 310a-310b from the network entity 104 for ML model monitoring. After determining 412a-412b feedback information for ML model monitoring, the UE 102 transmits 414a-414b the feedback information (e.g., BLER, uncompressed CSI, etc.) to the network entity 104 for ML model monitoring. For example, the UE 102 may report uncompressed CSI on PUCCH / PUSCH resources. The network entity 104 may configure or indicate to the UE 102 the PUCCH / PUSCH resources for reporting uncompressed CSI reports. The uncompressed CSI reports transmitted 414a-414b to the network entity 104 as feedback information may correspond to an eigenvector of an estimated channel of the CSI-RS, where corresponds to the coefficient at row i and column j in the eigenvector matrix, such that = UE 102 may include the strongest set of coefficients or all coefficients in the datagrams with quantized amplitudes that are transmitted 414a-414b to network entity 104. and Phase in the feedback information.

[0067] The network entity 104 may calculate the cosine similarity or squared cosine similarity of the uncompressed CSI and the decompressed ML-based CSI reported in the feedback information 414a-414b received from the UE 102. If the cosine similarity or squared cosine similarity is below a threshold (or otherwise satisfies a fourth threshold criterion), the network entity 104 may determine that an ML model performance failure instance has occurred. That is, the network entity 104 interprets the feedback information received 414a-414b from the UE 102 for ML model performance failure monitoring 416a-416b and / or detection 416b.

[0068] In an example, after determining that N consecutive ML model performance failure instances have occurred, the network entity 104 may determine that an ML model performance failure event has been detected 416b. Similar to the multiple instances of downlink signal transmission 310a-310b and multiple occasions of ML model performance monitoring 316a-316b that may occur in the signaling diagram 300 before the ML model performance failure event is detected 316b at the UE 102, the signaling diagram 400 may similarly include multiple instances of downlink signal transmission 310a-310b, multiple occasions of determining 412a-412b and transmitting 414a-414b feedback information to the network entity 104, and multiple occasions of ML model performance monitoring 416a-416b before the ML model performance failure event is detected 416b at the network entity 104. Upon detecting 416b the ML model performance failure event, the network entity may transmit 420 control signaling to the UE 102 to fall back to a non-ML based CSI reporting technique or to update / switch the ML model. The control signaling transmitted 420 to the UE 102 in the signaling diagram 400 may correspond to the same or similar control signaling as transmitted 320 to the UE in the signaling diagram 300. The control signaling may correspond to RRC signaling, MAC-CE, or DCI. The UE 102 and the network entity 104 may continue to communicate 322 based on the non-ML based CSI reporting technique or the updated / switched ML model.

[0069] The network entity 104 may configure the UE 102 to include BLER information (e.g., an assumed BLER) in the feedback information transmitted 414a-414b to the network entity 104 based on one or more measurements of the downlink signal. In an example, the downlink signal may correspond to a CSI-RS based on a CSI-RS resource configured by the network entity 104 via RRC signaling, MAC-CE, or DCI. In other examples, the downlink signal corresponds to a PDSCH transmission or a DCI that schedules the PDSCH transmission. Therefore, the network entity 104 may then indicate whether to report the assumed BLER via the PDSCH. The UE 102 may measure the assumed BLER based on at least one of a target spectral efficiency per layer or a target spectral efficiency per number of layers. The UE 102 may determine the target spectral efficiency per layer from the most recently reported CQI, and determine the number of layers from the most recently reported RI for ML-based CSI reporting. The network entity 104 may configure the target spectral efficiency per layer and the target spectral efficiency per number of layers via RRC signaling, MAC-CE, or DCI.

[0070] The UE 102 may transmit 414a-414b the assumed BLER to the network entity 104 on the PUCCH / PUSCH resources. The network entity may configure or indicate the PUCCH / PUSCH resources via RRC signaling, MAC-CE, or DCI. The number of reported bits for the assumed BLER may be based on a predefined protocol or may be configured by the network entity 104 via RRC signaling. The UE 102 may explicitly report the assumed BLER or indicate whether the assumed BLER exceeds a threshold, where the threshold may be predefined (e.g., a target BLER for CQI reporting for ML-based CSI reporting) or may be configured by the network entity via RRC signaling. The RRC signaling may indicate an RRC reconfiguration message from the network entity 104 to the UE 102, a system information block (SIB), where the SIB may be a predefined SIB (e.g., SIB1) or a different SIB transmitted by the network entity 104. The network entity 104 may also determine the UE capabilities via UE capability report signaling or from another network entity or the core network (e.g., AMF). Figure 3-Figure 4 ML model performance failure monitoring is shown. Figure 5-Figure 6 Shown is the method for implementing Figure 3-Figure 4 Specifically, Figure 5 The UE 102 is shown Figure 3-Figure 4 Implementation of one or more aspects. Figure 6 The network entity 104 is shown Figure 3-Figure 4 Implementation of one or more aspects.

[0071] Figure 5A flowchart 500 of a method for performance failure monitoring of ML performed by UE 102 is shown. Figure 1-Figure 4 The method may be performed by UE 102, UE equipment 702, etc., which may include a memory 724' and may correspond to the entire UE 102 or UE equipment 702, or a component of the UE 102 or UE equipment 702 such as a wireless baseband processor 724.

[0072] The UE 102 receives 506 control signaling that at least one of: activates an ML model for CSI compression, indicates DL signal information for at least one downlink (DL) signal, configures one or more parameters for monitoring performance of the ML model, configures feedback information to be included in information associated with performance of the ML model, or indicates a CSI decoder for CSI compression. For example, referring to Figure 3 , the UE 102 receives 306 control signaling from the network entity 104 to enable ML-based CSI compression and configure downlink signals / parameters for ML model monitoring. Figure 4 As another example, the UE 102 receives 406 control signaling from the network entity 104 that also configures UE feedback for ML model monitoring.

[0073] UE 102 receives 510 at least one DL signal from a network entity to monitor the performance of the ML model for CSI compression—the monitoring performance is based on a measurement value of the at least one DL signal. Figure 3-Figure 4 , UE 102 receives 310a - 310b downlink signals from network entity 104 for ML model monitoring.

[0074] UE 102 determines 516 whether an ML model performance failure has occurred. Figure 3 , UE 102 monitors 316a-316b for ML model performance failures. If UE 102 determines 516 that no ML model performance failure has occurred (or if the UE does not perform ML model performance failure monitoring, such as Figure 4 ), the UE 102 returns to receiving 510 at least one downlink signal to monitor the performance of the ML model.

[0075] If the UE 102 determines 516 that an ML model performance failure has occurred, the UE 102 transmits 518a information associated with the performance of the ML model for CSI compression to a network entity based on the measurement value of at least one DL signal. Figure 3 , the UE 102 transmits 318 an ML model performance failure event report to the network entity 104 based on the detection 316b of the ML model performance failure. Figure 4 As another example, based on the UE 102 determining 412a - 412b feedback information for ML model monitoring, the UE 102 transmits 414a - 414b feedback information (eg, BLER, uncompressed CSI, etc.) to the network entity 104 for ML model monitoring.

[0076] If the UE does not receive a response to the information associated with the performance of the ML model within a configured or predefined duration and if the number of retransmissions of the information associated with the performance of the ML model is lower than the maximum number of retransmissions, the UE 102 retransmits 518b the information associated with the performance of the ML model. Figure 3 , transmission 318 may be a retransmission of the ML model performance failure event report. Figure 4 As another example, transmissions 414a-414b may be retransmissions of feedback information for ML model monitoring.

[0077] The UE 102 receives 520 a response to the information associated with the performance of the ML model within a configured or predefined duration after transmitting the information associated with the performance of the ML model—the response indicating a communication that applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. Figure 3 , the UE 102 receives 320 a response to the ML model performance failure event report. Figure 4 , the UE 102 receives 420 a fallback indication to non-ML based CSI reporting or to an updated / switched ML model.

[0078] When the information associated with the performance of the ML model indicates a performance failure, the UE 102 communicates 522 with a network entity—the communication applying at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. Figure 3-Figure 4 , the UE 102 communicates 322 with the network entity 104 based on the non-ML based CSI report or the updated / switched ML model. Figure 5 A method from the UE side of a wireless communication link is described, while Figure 6 A method from the network side of a wireless communication link is described.

[0079] Figure 6 is a flow chart 600 of a method for performance fault monitoring of ML performed by the network entity 104. Figure 1-Figure 4The method may be performed by a base station or one or more network entities 104 at a base station, which may correspond to a RU 106, a DU 108, a CU 110, a RU processor 842, a DU processor 832, a CU processor 812, etc. A base station or one or more network entities 104 at a base station may include a memory 812' / 832' / 842', which may correspond to the entirety of one or more network entities 104 or base stations, or a component of one or more network entities 104 or base stations, such as a RU processor 842, a DU processor 832, or a CU processor 812.

[0080] The base station or one or more network entities 104 of the base station transmits 606 control signaling that performs at least one of: activating an ML model for CSI compression, indicating DL signal information for at least one DL signal, configuring one or more parameters for monitoring performance of the ML model, configuring feedback information to be included in information associated with performance of the ML model, or instructing a CSI decoder for CSI compression. For example, referring to Figure 3 , the network entity 104 transmits 306 control signaling to the UE 102 to enable ML-based CSI compression and configure downlink signals / parameters for ML model monitoring. Figure 4 As another example, the network entity 104 transmits 406 control signaling to the UE 102 that also configures UE feedback for ML model monitoring.

[0081] The base station or one or more network entities 104 of the base station transmits 610 at least one downlink signal to the UE to monitor the performance of the ML model for CSI compression. Figure 3-Figure 4 , the network entity 104 transmits 310a - 310b downlink signals to the UE 102 for ML model monitoring.

[0082] The base station or one or more network entities 104 of the base station determines 616 whether an ML model performance failure has occurred. Figure 4 , the network entity 104 monitors 416a-416b the ML model performance failure. If the base station or one or more network entities 104 of the base station determine 516 that no ML model performance failure has occurred, the base station or one or more network entities 104 of the base station returns to transmitting 610 at least one downlink signal to monitor the performance of the ML model.

[0083] If the base station or one or more network entities 104 of the base station determines 616 that an ML model performance failure has occurred (or if the base station does not perform ML model performance failure monitoring, such as Figure 3), the base station or one or more network entities 104 of the base station receives 618 information associated with the performance of the ML model for CSI compression from the UE based on the measurement value of at least one downlink signal. Figure 3 , the network entity 104 receives 318 an ML model performance failure event report from the UE 102 based on the detection 316b of the ML model performance failure. Figure 4 As another example, the network entity 104 receives 414a-414b feedback information (eg, BLER, uncompressed CSI, etc.) from the UE 102 for ML model monitoring for the network entity 104 to determine 416a-416b and / or detect 416b ML model performance failures.

[0084] The base station or one or more network entities 104 of the base station transmits 620 a response to the information associated with the performance of the ML model—the response indicating a communication modification to apply at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. Figure 3 , the network entity 104 transmits 320 a response to the ML model performance failure event report. Figure 4 As another example, the network entity 104 transmits 420 a fallback indication to a non-ML based CSI report or to an updated / switched ML model.

[0085] When the information associated with the performance of the ML model indicates a performance failure, the base station or one or more network entities 104 of the base station modify 622 communications with the UE—the communication modification applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or non-ML CSI reporting. Figure 3-Figure 4 , modifying network entity communications 322 with the UE 102 based on the non-ML based CSI reports or the updated / switched ML model. Figure 7 The UE equipment 702 described in the flowchart 500 may execute the method of the flowchart 500. The base station or one or more network entities 104 of the base station, such as Figure 8 As described in the flowchart 600, the method of the flowchart 600 can be performed.

[0086] Figure 7700 is a diagram illustrating an example of a hardware implementation of a UE equipment 702. The equipment 702 may be a UE 102, a component of a UE, or may implement UE functionality. In some aspects, the equipment 702 may include a wireless baseband processor 724 (also referred to as a modem) coupled to one or more transceivers 722 (e.g., a wireless RF transceiver). The wireless baseband processor 724 may include an on-chip memory 724'. In some aspects, the equipment 702 may further include one or more subscriber identity modules (SIM) cards 720 and an application processor 706 coupled to a secure digital (SD) card 708 and a screen 710. The application processor 706 may include an on-chip memory 706'.

[0087] Equipment 702 may further include a Bluetooth module 712, a WLAN module 714, an SPS module 716 (e.g., a GNSS module), and a cellular module 717 located within one or more transceivers 722. The Bluetooth module 712, the WLAN module 714, the SPS module 716, and the cellular module 717 may include an on-chip transceiver (TRX) (or in some cases, only a receiver (RX)). The Bluetooth module 712, the WLAN module 714, the SPS module 716, and the cellular module 717 may include their own dedicated antennas and / or communicate using antenna 780. Equipment 702 may further include one or more sensor modules 718 (e.g., an atmospheric pressure sensor / altimeter; a motion sensor such as an inertial management unit (IMU), a gyroscope, and / or an accelerometer; light detection and ranging (LIDAR), radio-aided detection and ranging (RADAR), sound navigation and ranging (SONAR), a magnetometer, audio, and / or other technologies for positioning), an additional module for memory 726, a power supply 730, and / or a camera 732.

[0088] The wireless baseband processor 724 communicates with another UE 102 and / or with a RU associated with the network entity 104 via one or more antennas 780 through the transceiver 722. The wireless baseband processor 724 and the application processor 706 can each include a computer-readable medium / memory 724', 706', respectively. The additional modules of the memory 726 can also be regarded as computer-readable media / memory. Each computer-readable medium / memory 724', 706', 726 can be non-temporary. The wireless baseband processor 724 and the application processor 706 are each responsible for general processing, including executing software stored on the computer-readable medium / memory. When the software is executed by the wireless baseband processor 724 / application processor 706, it causes the wireless baseband processor 724 / application processor 706 to perform the various functions described. The computer-readable medium / memory can also be used to store data manipulated by the wireless baseband processor 724 / application processor 706 when executing the software. The wireless baseband processor 724 / application processor 706 can be a component of the UE 102. The equipment 702 may be a processor chip (modem and / or applications) and include only the wireless baseband processor 724 and / or the application processor 706 , and in another configuration, the equipment 702 may be the entire UE 102 and include additional modules of the equipment 702 .

[0089] As discussed, the ML model performance failure component 140 is configured to receive at least one downlink signal from a network entity to monitor the performance of an ML model for channel state information (CSI) compression, wherein the monitored performance is based on a measurement of the at least one downlink signal; transmit information associated with the performance of the ML model for CSI compression to the network entity based on the measurement of the at least one downlink signal; and communicate with the network entity when the information associated with the performance of the ML model indicates a performance failure, wherein the communication applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The ML model performance failure component 140 may be located within the wireless baseband processor 724, the application processor 706, or within both the wireless baseband processor 724 and the application processor 706. The ML model performance failure component 140 may be one or more hardware components explicitly configured to perform the process / algorithm, implemented by one or more processors configured to perform the process / algorithm, stored in a computer-readable medium for implementation by one or more processors, or some combination thereof.

[0090] As shown, the equipment 702 may include various components configured for various functions. In one configuration, the equipment 702, and specifically the wireless baseband processor 724 and / or the application processor 706, includes: a component for receiving at least one downlink signal from a network entity to monitor the performance of an ML model for CSI compression, wherein the monitoring performance is based on a measurement value of the at least one downlink signal; a component for transmitting information associated with the performance of the ML model for CSI compression to the network entity based on the measurement value of the at least one downlink signal; and a component for communicating with the network entity when the information associated with the performance of the ML model indicates a performance failure, wherein the communication applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The equipment 702 further includes a component for receiving control signaling, which performs at least one of the following: activating the ML model for CSI compression, indicating downlink signal information for at least one downlink signal, configuring one or more parameters for monitoring the performance of the ML model, configuring feedback information to be included in the information associated with the performance of the ML model, or instructing a CSI decoder for CSI compression. The means for communicating with the network entity is further configured to: transmit the performance fault report to the network entity via at least one of MAC-CE, PUCCH or PRACH. The apparatus 702 further includes: a means for transmitting a dedicated SR for uplink resources for transmitting the performance fault report to the network entity; a means for receiving a configuration of uplink resources for transmitting the performance fault report to the network entity; and a means for transmitting the performance fault report to the network entity via MAC-CE on the uplink resources. The apparatus 702 further includes a means for receiving a response to the information associated with the performance of the ML model within a configured or predefined duration after transmitting the information associated with the performance of the ML model, wherein the response indicates a communication applying at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The apparatus 702 further includes a means for retransmitting the information associated with the performance of the ML model if the UE does not receive a response to the information associated with the performance of the ML model within a configured or predefined duration and if the number of retransmissions of the information associated with the performance of the ML model is less than a maximum number of retransmissions. The component may be the ML model performance fault component 140 of the equipment 702 configured to perform the function recited by the component.

[0091] Figure 88 is a diagram 800 showing an example of a hardware implementation of one or more network entities 104. One or more network entities 104 may be a BS, a component of a BS, or may implement BS functionality. One or more network entities 104 may include at least one of a CU 810, a DU 830, or a RU 840. For example, component 199 may be located at one or more network entities 104, such as a CU 810; both a CU 810 and a DU 830; each of a CU 810, a DU 830, and a RU 840; a DU 830; both a DU 830 and a RU 840; or a RU 840.

[0092] The CU 810 may include a CU processor 812. The CU processor 812 may include an on-chip memory 812'. In some aspects, the CU 810 may further include an additional memory module 814 and a communication interface 818. The CU 810 communicates with the DU 830 via a midhaul link 162 such as an F1 interface. The DU 830 may include a DU processor 832. The DU processor 832 may include an on-chip memory 832'. In some aspects, the DU 830 may further include an additional memory module 834 and a communication interface 838. The DU 830 communicates with the RU 840 via a fronthaul link 160. The RU 840 may include a RU processor 842. The RU processor 842 may include an on-chip memory 842'. In some aspects, the RU 840 may further include an additional memory module 844, one or more transceivers 846, an antenna 880, and a communication interface 848. The RU 840 communicates wirelessly with the UE 102.

[0093] On-chip memory 812', 832', 842' and additional memory modules 814, 834, 844 can each be considered as a computer-readable medium / memory. Each computer-readable medium / memory can be non-temporary. Each of the processors 812, 832, 842 is responsible for general processing, including executing software stored on the computer-readable medium / memory. The software, when executed by the corresponding processor, enables the processor to perform the various functions described above. The computer-readable medium / memory can also be used to store data manipulated by the processor when executing the software.

[0094] As discussed, the ML model adjustment component 150 is configured to transmit at least one downlink signal to the UE to monitor the performance of the ML model for CSI compression; receive information associated with the performance of the ML model for CSI compression from the UE based on the measurement value of the at least one downlink signal; and modify the communication with the UE when the information associated with the performance of the ML model indicates a performance failure, wherein the communication modification applies at least one of the following: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The ML model adjustment component 150 can be located within one or more processors of one or more of the CU 810, the DU 830, and the RU 840. The ML model adjustment component 150 can be one or more hardware components explicitly configured to perform the process / algorithm, implemented by one or more processors configured to perform the process / algorithm, stored in a computer-readable medium for implementation by one or more processors, or some combination thereof.

[0095] The one or more network entities 104 may include various components configured for various functions. In one configuration, the one or more network entities 104 include means for: transmitting at least one downlink signal to a UE to monitor the performance of an ML model for CSI compression; receiving information associated with the performance of the ML model for CSI compression from the UE based on a measurement of the at least one downlink signal; and modifying communications with the UE when the information associated with the performance of the ML model indicates a performance failure, wherein the communication modification applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The one or more network entities 104 further include means for transmitting control signaling, the control signaling performing at least one of: activating the ML model for CSI compression, indicating downlink signal information for at least one downlink signal, configuring one or more parameters for monitoring the performance of the ML model, configuring feedback information to be included in the information associated with the performance of the ML model, or indicating a CSI decoder for CSI compression. The means for modifying communications with the UE is further configured to: receive a performance failure report from the UE via at least one of MAC-CE, PUCCH, or PRACH. The one or more network entities 104 further include: a means for receiving a dedicated scheduling request for uplink resources for receiving performance failure reports from the UE; a means for transmitting a configuration of uplink resources for receiving performance failure reports from the UE; and a means for receiving the performance failure reports from the UE via a MAC-CE on the uplink resources. The one or more network entities 104 further include a means for transmitting a response to information associated with the performance of the ML model, wherein the response indicates a communication modification to apply at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report. The means may be an ML model adjustment component 150 of the one or more network entities 104 configured to perform the functions recited by the means.

[0096] The specific order or hierarchy of the boxes in the process and flow chart disclosed herein is an illustration of an example method. Therefore, the specific order or hierarchy of the boxes in the process and flow chart can be rearranged. Some boxes can also be merged or deleted. Dashed lines can indicate optional elements of the figure. The attached method claims present elements of each box in an example order and are not limited to the specific order or hierarchy presented in the claims, process and flow chart.

[0097] The detailed description set forth herein describes various configurations in conjunction with the accompanying drawings, but does not represent the only configuration in which the concepts described herein can be practiced. The detailed description includes specific details for providing a comprehensive explanation of the various concepts. However, these concepts can be practiced without using these specific details. In some cases, well-known structures and components are shown in block diagram form in order to avoid blurring such concepts.

[0098] Various aspects of wireless communication systems (such as telecommunication systems) are presented with reference to various equipment and methods. These equipment and methods are described in the detailed description that follows 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 on the specific application and design constraints imposed on the overall system.

[0099] Elements, or any part of elements or any combination of elements can be implemented as a "processing system" including one or more processors. Examples of processors include microprocessors, microcontrollers, graphics processing units (GPUs), central processing units (CPUs), application processors, digital signal processors (DSPs), reduced instruction set computing (RISC) processors, systems on chip (SoCs), baseband processors, field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gating logic, discrete hardware circuits, and other similar hardware configured to perform various functions described throughout this disclosure. One or more processors in a processing system can execute software, which can be referred to as software, firmware, middleware, microcode, hardware description language, or other. Software should be broadly interpreted as meaning instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software components, applications, software applications, software packages, routines, subroutines, objects, executables, execution threads, processes, functions, or any combination thereof.

[0100] 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 codes on the computer-readable medium. Computer-readable media include 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 medium can be any available medium that is accessible to a computer.

[0101] The aspects, implementations, and / or use cases described herein may be implemented across many different platform types, devices, systems, shapes, sizes, and packaging arrangements. For example, the aspects, implementations, and / or use cases may be generated via integrated chip implementations and other non-module component-based devices such as end-user devices, vehicles, communication devices, computing devices, industrial equipment, retail / procurement devices, medical devices, artificial intelligence (AI)-enabled devices, machine learning (ML)-enabled devices, and the like. The aspects, implementations, and / or use cases may 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 techniques described herein.

[0102] Devices incorporating the various aspects and features described herein may also include additional components and features for implementing and practicing the various aspects and features claimed and described. For example, the transmission and reception of wireless signals necessarily include 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 may 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.

[0103] 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 may be applied to other aspects. Therefore, the claims are not limited to the various aspects described herein, but should be interpreted in view of the full scope of the disclosure consistent with the language of the claims.

[0104] Unless explicitly stated, references to singular elements do not mean "one and only one", but rather "one or more". Terms such as "if", "when ..." and "at ..." do not imply an immediate temporal relationship or reaction. That is, these phrases (such as "when ...") do not imply an immediate action in response to the occurrence of an action or during the occurrence of an action, but simply mean that if a certain condition is met, a certain action will occur, but no specific or immediate time constraints are required for the occurrence of the action. Unless explicitly stated otherwise, the term "some" refers to 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, multiple B, and / or multiple C, or may include only A, only B, or only C. A set should be interpreted as a set of elements in which the number of elements is one or more.

[0105] Unless expressly indicated otherwise, ordinal terms such as "first" and "second" do not necessarily imply an order in time, sequence, value, etc., but are used to distinguish different instances of the term or phrase following each ordinal term.

[0106] The structural equivalents and functional equivalents of the elements of various aspects described in the entire disclosure known or later learned by those of ordinary skill in the art are expressly incorporated herein by reference and are covered by the claims. The words "module", "mechanism", "element", "device", etc. may not be substitutes for the word "component". Therefore, unless the phrase "component for ..." is used to clearly state the claim element, any claim element shall not be interpreted as a means plus function. As used herein, the phrase "based on" should not be interpreted as a reference to a closed information set, one or more conditions, one or more factors, etc. In other words, unless clearly stated differently, the phrase "based on A" (where "A" can be information, conditions, factors, etc.) should be interpreted as "at least based on A".

[0107] The following examples are merely illustrative and may be combined with other examples or teachings described herein without limitation.

[0108] Example 1 is a method of wireless communication at a UE, comprising: receiving at least one downlink signal from a network entity to monitor performance of an ML model for CSI compression, and including monitoring the performance based on a measurement value of the at least one downlink signal; transmitting information associated with the performance of the ML model for CSI compression to the network entity based on the measurement value of the at least one downlink signal; and communicating with the network entity when the information associated with the performance of the ML model indicates a performance failure, and including the communication applying at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report.

[0109] Example 2 may be combined with Example 1 and further include receiving control signaling that performs at least one of: activating an ML model for CSI compression, indicating downlink signal information for at least one downlink signal, configuring one or more parameters for monitoring performance of the ML model, configuring feedback information to be included in information associated with performance of the ML model, or indicating a CSI decoder for CSI compression.

[0110] Example 3 may be combined with any of Examples 1-2 and include that the information associated with the performance of the ML model corresponds to at least one of: feedback information or a performance failure report associated with detection of a performance failure, the feedback information indicating whether the performance of the ML model contributed to the performance failure.

[0111] Example 4 may be combined with Example 3 and include communicating with the network entity further comprising: transmitting the performance failure report to the network entity via at least one of MAC-CE, PUCCH, or PRACH.

[0112] Example 5 can be combined with any of Examples 3-4 and further includes transmitting a dedicated scheduling request for uplink resources for transmitting a performance fault report to a network entity; receiving a configuration of uplink resources for transmitting a performance fault report to the network entity; and transmitting the performance fault report to the network entity on the uplink resources.

[0113] Example 6 may be combined with any of Examples 1-5 and further include receiving a response to the information associated with the performance of the ML model within a configured or predefined duration after transmitting the information associated with the performance of the ML model, and including a communication in which the response indicates application of at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report.

[0114] Example 7 can be combined with any of Examples 1-5 and further include retransmitting the information associated with the performance of the ML model if the UE does not receive a response to the information associated with the performance of the ML model within a configured or predefined duration and if the number of retransmissions of the information associated with the performance of the ML model is lower than a maximum number of retransmissions.

[0115] Example 8 can be combined with any of Examples 1-7 and includes at least one downlink signal for monitoring the performance of the ML model corresponding to at least one of: DMRS, PDSCH transmission, or CSI-RS transmitted from the same number of antenna ports as the maximum number of downlink layers configured based on RRC signaling.

[0116] Example 9 can be combined with any of Examples 1-8 and includes applying at least one of the cosine similarity or squared cosine similarity for monitoring the performance of the ML model to at least one of: a normalized feature vector, an estimated channel of at least one downlink signal, or a normalized channel of at least one downlink signal, and including that a performance failure of the ML model is based on at least one of the cosine similarity or squared cosine similarity satisfying a threshold criterion.

[0117] Example 10 may be combined with any of Examples 1-9 and include that a performance failure of the ML model is based on a number of consecutive failure detection instances associated with a measurement value of the at least one downlink signal.

[0118] Example 11 can be combined with any of Examples 1-10 and includes decompression of the CSI based on at least one downlink signal and at least one of: a subband size for the decompressed CSI, a first bit number for amplitude quantization, a second bit number for phase quantization, or a third number of coefficients associated with at least one downlink signal.

[0119] Example 12 may be combined with any of Examples 1-11 and include monitoring the performance of the ML model being based on a BLER associated with a measurement of at least one downlink signal, and including a performance failure of the ML model corresponding to the BLER satisfying a BLER threshold criterion.

[0120] Example 13 may be combined with Example 12 and include that the BLER is based on at least one of: a first target average spectral efficiency per layer or a second target average spectral efficiency per number of layers.

[0121] Example 14 may be combined with any of Examples 12-13 and include that the information associated with the performance of the ML model includes the BLER.

[0122] Example 15 is a method of wireless communication at a network entity, comprising: transmitting at least one downlink signal to a UE to monitor performance of an ML model for CSI compression; receiving information associated with the performance of the ML model for CSI compression from the UE based on a measurement value of the at least one downlink signal; and modifying communications with the UE when the information associated with the performance of the ML model indicates a performance failure, and including that the communication modification applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or non-ML CSI reporting.

[0123] Example 16 may be combined with Example 15 and further include transmitting control signaling that performs at least one of: activating an ML model for CSI compression, indicating downlink signal information for at least one downlink signal, configuring one or more parameters for monitoring performance of the ML model, configuring feedback information to be included in information associated with performance of the ML model, or indicating a CSI decoder for CSI compression.

[0124] Example 17 may be combined with any of Examples 15-16 and include the information associated with the performance of the ML model corresponding to at least one of: feedback information or a performance failure report associated with detection of a performance failure, the feedback information indicating whether the performance of the ML model contributed to the performance failure.

[0125] Example 18 may be combined with Example 17 and include modifying the communication with the UE further comprising: receiving a performance failure report from the UE via at least one of MAC-CE, PUCCH, or PRACH.

[0126] Example 19 may be combined with any of Examples 15-18 and further include receiving a dedicated scheduling request for uplink resources for receiving a performance failure report from a UE; transmitting a configuration of uplink resources for receiving a performance failure report from the UE; and receiving a performance failure report from the UE on the uplink resources.

[0127] Example 20 may be combined with any of Examples 15-19 and further include transmitting a response to information associated with performance of the ML model, and including the response indicating a communication modification that applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report.

[0128] Example 21 can be combined with any of Examples 15-20 and include at least one downlink signal for monitoring the performance of the ML model corresponding to at least one of: DMRS, PDSCH transmission, or CSI-RS transmitted from the same number of antenna ports as the maximum number of downlink layers configured based on RRC signaling.

[0129] Example 22 can be combined with any of Examples 15-21 and include applying at least one of the cosine similarity or the squared cosine similarity for monitoring the performance of the ML model to at least one of: a normalized feature vector, an estimated channel of at least one downlink signal, or a normalized channel of at least one downlink signal, and including that a performance failure of the ML model is based on at least one of the cosine similarity or the squared cosine similarity satisfying a threshold criterion.

[0130] Example 23 may be combined with any of Examples 15-22 and include that a performance failure of the ML model is based on a number of consecutive failure detection instances associated with receiving feedback information from the UE.

[0131] Example 24 can be combined with any of Examples 15-23 and include decompression of the CSI based on at least one downlink signal and at least one of: a subband size for the decompressed CSI, a first bit number for amplitude quantization, a second bit number for phase quantization, or a third number of coefficients associated with at least one downlink signal.

[0132] Example 25 may be combined with any of Examples 15-24 and include monitoring the performance of the ML model being based on a BLER associated with a measurement of at least one downlink signal, and including a performance failure of the ML model corresponding to the BLER satisfying a BLER threshold criterion.

[0133] Example 26 may be combined with Example 25 and include that the BLER is based on at least one of: a first target average spectral efficiency per layer or a second target average spectral efficiency per number of layers.

[0134] Example 27 may be combined with any of Examples 25-26 and include that the information associated with the performance of the ML model comprises the BLER.

[0135] Example 28 is an apparatus for wireless communication for implementing the method as described in any of Examples 1-27.

[0136] Example 29 is an apparatus for wireless communications including means for implementing the method of any of Examples 1-27.

[0137] Example 30 is a non-transitory computer readable medium storing computer executable code, which, when executed by at least one processor, causes the at least one processor to implement the method as described in any one of Examples 1-27.

Claims

1. A method of wireless communication at a user equipment (UE), comprising: receiving at least one downlink signal from a network entity to monitor performance of a machine learning (ML) model for channel state information (CSI) compression, wherein monitoring the performance is based on measurements of the at least one downlink signal; transmitting, to the network entity, information associated with the performance of the ML model for the CSI compression based on the measured value of the at least one downlink signal; as well as When the information associated with the performance of the ML model indicates a performance failure, communicating with the network entity, wherein the communication applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or a non-ML CSI report.

2. The method of claim 1, further comprising: receiving control signaling that at least one of: activates the ML model for the CSI compression, indicates downlink signal information for the at least one downlink signal, configures one or more parameters for monitoring the performance of the ML model, configures feedback information to be included in the information associated with the performance of the ML model, or indicates a CSI decoder for the CSI compression.

3. The method of any one of claims 1-2, wherein the information associated with the performance of the ML model corresponds to at least one of: the feedback information or a performance failure report associated with the detection of the performance failure, the feedback information indicating whether the performance of the ML model contributed to the performance failure.

4. The method of claim 3, wherein communicating with the network entity further comprises: The performance failure report is transmitted to the network entity via at least one of a Medium Access Control-Control Element (MAC-CE), a Physical Uplink Control Channel (PUCCH), or a Physical Random Access Channel (PRACH).

5. The method according to any one of claims 3 to 4, further comprising: transmitting a dedicated scheduling request (SR) for uplink resources for transmitting the performance fault report to the network entity; receiving a configuration of the uplink resources for transmitting the performance fault report to the network entity; as well as The performance fault report is transmitted to the network entity on the uplink resources.

6. The method of any one of claims 1 to 5, further comprising: Receiving a response to the information associated with the performance of the ML model within a configured or predefined duration after transmitting the information associated with the performance of the ML model, wherein the response indicates the communication applying at least one of: the update to the ML model, the switching of the ML model to the different ML model, or the non-ML CSI report.

7. The method of any one of claims 1 to 5, further comprising: If the UE does not receive a response to the information associated with the performance of the ML model within a configured or predefined duration and if the number of retransmissions of the information associated with the performance of the ML model is lower than a maximum number of retransmissions, retransmitting the information associated with the performance of the ML model.

8. The method of any one of claims 1-7, wherein the at least one downlink signal used to monitor the performance of the ML model corresponds to at least one of: a demodulation reference signal (DMRS) transmitted from the same number of antenna ports as the maximum number of downlink layers configured based on radio resource control (RRC) signaling, a physical downlink shared channel (PDSCH) transmission, or a channel state information reference signal (CSI-RS).

9. The method of any one of claims 1-8, wherein at least one of cosine similarity or squared cosine similarity used to monitor the performance of the ML model applies at least one of: a normalized feature vector, an estimated channel of the at least one downlink signal, or a normalized channel of the at least one downlink signal, and wherein the performance failure of the ML model is based on the at least one of the cosine similarity or the squared cosine similarity satisfying a threshold criterion.

10. The method of any one of claims 1-9, wherein the performance failure of the ML model is based on a number of consecutive failure detection instances associated with the measurements of the at least one downlink signal.

11. The method of any one of claims 1-10, wherein the decompression of the CSI is based on the at least one downlink signal and at least one of: a subband size for the decompressed CSI, a first number of bits for amplitude quantization, a second number of bits for phase quantization, or a third number of coefficients associated with the at least one downlink signal.

12. The method of any one of claims 1-11, wherein monitoring the performance of the ML model is based on a block error rate (BLER) associated with the measured values ​​of the at least one downlink signal, and wherein the performance failure of the ML model corresponds to the BLER satisfying a BLER threshold criterion.

13. The method of claim 12, wherein the BLER is based on at least one of: a first target average spectral efficiency per layer or a second target average spectral efficiency per number of layers.

14. The method of any one of claims 12-13, wherein the information associated with the performance of the ML model comprises the BLER.

15. A method of wireless communication at a network entity, comprising: transmitting at least one downlink signal to a user equipment (UE) to monitor performance of a machine learning (ML) model for channel state information (CSI) compression; receiving, from the UE, information associated with the performance of the ML model for the CSI compression based on a measurement value of the at least one downlink signal; as well as When the information associated with the performance of the ML model indicates a performance failure, modifying communications with the UE, wherein the communications modification applies at least one of: an update to the ML model, a switch of the ML model to a different ML model, or non-ML CSI reporting.

16. The method of claim 15, further comprising: Transmitting control signaling, the control signaling performs at least one of the following: activating the ML model for the CSI compression, indicating downlink signal information for the at least one downlink signal, configuring one or more parameters for monitoring the performance of the ML model, configuring feedback information to be included in the information associated with the performance of the ML model, or indicating a CSI decoder for the CSI compression.

17. The method of any one of claims 15-16, wherein the information associated with the performance of the ML model corresponds to at least one of: the feedback information or a performance failure report associated with the detection of the performance failure, the feedback information indicating whether the performance of the ML model contributed to the performance failure.

18. The method of any one of claims 15-17, further comprising: transmitting a response to the information associated with the performance of the ML model, wherein the response indicates the communication modification to apply at least one of: the update to the ML model, the switching of the ML model to the different ML model, or the non-ML CSI report.

19. An apparatus for wireless communication, comprising a memory and at least one processor, the at least one processor being coupled to the memory and configured to implement the method according to any one of claims 1-18.

Images