AI / ML model monitoring operations for NR air interfaces

Abstract

The present disclosure relates to a 5G or 6G communication system for supporting a higher data transmission rate. The AI / ML monitoring operation is performed based on a received monitoring configuration that forms part of a configuration for using an AI / ML model for communication system operation. Based on the monitoring configuration, AI / ML model assistance information is reported that includes AI / ML model monitoring results from the AI / ML monitoring operations. And receiving AI / ML model management and adaptation information based on the AI / ML model monitoring results, and executing AI / ML model management and adaptation operation. The AI / ML model management and adaptation information may include a parameter characterizing an AI / ML model management and adaptation action, or an indication of an AI / ML model management and adaptation action. The AI / ML model management and adaptation actions may include one of model switching, model optimization or update, or model migration.

Description

Technical Field

[0001] The present disclosure relates generally to artificial intelligence / machine learning (AI / ML) monitoring operations, and more particularly to AI / ML model management and adaptation operations for one or more user devices in a wireless communication system. Background Art

[0002] 5G mobile communication technology defines a wide frequency band, making high transmission rates and new services possible, and can be implemented not only in "below 6 GHz" frequency bands such as 3.5 GHz, but also in "above 6 GHz" frequency bands (including 28 GHz and 39 GHz) called millimeter waves. In addition, 6G mobile communication technology (called super 5G system) is also considered to be implemented in terahertz (THz) frequency bands (for example, 95 GHz to 3 THz frequency bands) to achieve 50 times the transmission rate of 5G mobile communication technology and one-tenth the ultra-low latency of 5G mobile communication technology.

[0003] At the beginning of the development of 5G mobile communication technology, in order to support services and to meet performance requirements related to enhanced mobile broadband (eMBB), ultra-reliable low-latency communication (URLLC) and massive machine type communication (mMTC), standardization has been carried out on the following items: beamforming and massive MIMO for alleviating radio wave path loss and increasing radio wave transmission distance in millimeter waves, dynamic operation for efficiently utilizing millimeter wave resources and time slot formats to support parameter sets (for example, operating multiple subcarrier spacings), initial access technology for supporting multi-beam transmission and broadband, definition and operation of BWP (bandwidth part), new channel coding methods such as LDPC (low-density parity check) codes for large-scale data transmission and polarization codes for highly reliable transmission of control information, L2 preprocessing, and network slicing for providing dedicated networks dedicated to specific services.

[0004] Currently, discussions on updates and performance enhancements of initial 5G mobile communication technologies are being conducted in consideration of services that will be supported by 5G mobile communication technologies, and there is already physical layer standardization on technologies such as: V2X (Vehicle to Everything) technology that assists driving determination and enhances user convenience by autonomous vehicles based on information sent by the vehicle about the vehicle's location and status, NR-U (Unlicensed New Radio) technology that aims to operate a system that complies with various regulatory requirements in non-enabled frequency bands, NR UE power saving technology, and Non-Terrestrial Network (NTN) technology for UE-satellite direct communication that provides coverage and positioning in areas where communication with terrestrial networks is unavailable.

[0005] Moreover, standardization in air interface architecture / protocol has been conducted with respect to the following technologies, such as Industrial Internet of Things (IIoT) technology for supporting new services through interaction and integration with other industries, IAB (Integrated Access and Backhaul) technology for providing nodes for network service area expansion by supporting wireless backhaul links and access links in an integrated manner, mobility enhancement technology including conditional switching and DAPS (Dual Active Protocol Stack) switching, and two-step random access (2-step RACH for NR) technology for simplifying the random access process. Standardization of system architecture / services for a 5G reference architecture (e.g., a service-based architecture or a service-based interface) for combining network function virtualization (NFV) and software-defined network (SDN) technologies, and mobile edge computing (MEC) for receiving services based on UE location has also been conducted.

[0006] With the commercialization of 5G mobile communication systems, the number of connected devices that have been exponentially increasing will be connected to the communication network, so it can be expected that enhanced functions and performance of 5G mobile communication systems and integrated operations of connected devices will be necessary. To this end, new research is arranged on extended reality (XR) for efficiently supporting AR (augmented reality), VR (virtual reality), MR (mixed reality), etc., 5G performance improvement and complexity reduction by using artificial intelligence (AI) and machine learning (ML), AI service support, metaverse service support, and drone communication.

[0007] In addition, such development of 5G mobile communication systems will serve as a basis for developing not only new waveforms for providing coverage in the terahertz band of 6G mobile communication technology, multi-antenna transmission technology such as full-dimensional MIMO (FD-MIMO), array antennas and large antennas, metamaterial-based lenses and antennas for improving coverage of terahertz band signals, high-dimensional spatial multiplexing technology using OAM (orbital angular momentum) and RIS (reconfigurable smart surface), but also full-duplex technology for improving the frequency efficiency of 6G mobile communication technology and updating system networks, AI-based communication technology for achieving system optimization by utilizing satellites and AI (artificial intelligence) from the design stage and internalizing end-to-end AI support functions, and next-generation distributed computing technology for implementing services at a complexity level that exceeds the limitations of UE operating functions by utilizing ultra-high performance communication and computing resources. Summary of the invention

[0008] Technical issues

[0009] The present disclosure relates to wireless communication systems, and more particularly, to AI / ML model monitoring.

[0010] Solutions to the problem

[0011] The AI / ML monitoring operation is based on a received monitoring configuration, which forms part of a configuration for using an AI / ML model for communication system operation. Based on the monitoring configuration, AI / ML model auxiliary information is reported, which AI / ML model auxiliary information includes AI / ML model monitoring results from the AI / ML monitoring operation. AI / ML model management and adaptation information based on these AI / ML model monitoring results is received, and AI / ML model management and adaptation operations are performed. The AI / ML model management and adaptation information may include parameters characterizing AI / ML model management and adaptation actions, or indications of AI / ML model management and adaptation actions. The AI / ML model management and adaptation actions may include one of model switching, model optimization or updating, or model migration.

[0012] In a first embodiment, a method includes: at a user equipment (UE), performing an AI / ML monitoring operation based on a received monitoring configuration, the monitoring configuration forming part of a usage configuration for an AI / ML model of the operation. The method also includes reporting AI / ML model auxiliary information based on the monitoring configuration, the AI / ML model auxiliary information including AI / ML model monitoring results from the AI / ML monitoring operation. The method also includes receiving AI / ML model management and adaptation information based on the AI / ML model monitoring results. The method further includes performing AI / ML model management and adaptation operations based on the AI / ML model management and adaptation information.

[0013] In a second embodiment, the UE includes a transceiver and a processor. The processor is configured to perform an AI / ML monitoring operation based on a received monitoring configuration, the monitoring configuration forming part of a usage configuration of an AI / ML model for the operation. The transceiver is configured to report AI / ML model auxiliary information based on the monitoring configuration, the AI / ML model auxiliary information including AI / ML model monitoring results from the AI / ML monitoring operation. The transceiver is also configured to receive AI / ML model management and adaptation information based on the AI / ML model monitoring results. The processor is also configured to perform AI / ML model management and adaptation operations based on the AI / ML model management and adaptation information.

[0014] In a third embodiment, the base station includes a transceiver configured to send a monitoring configuration for monitoring an AI / ML monitoring operation to a UE. The monitoring configuration forms part of a usage configuration for an AI / ML model for the operation. The transceiver is also configured to receive AI / ML usage assistance information from the UE, the AI / ML usage assistance information including AI / ML model monitoring results from the AI / ML monitoring operation. The base station also includes a processor configured to evaluate UE-specific usage performance of the AI / ML model for the operation based on the AI / ML model monitoring results. The processor is also configured to determine AI / ML model management and adaptation information corresponding to the AI / ML model monitoring results.

[0015] In any of the foregoing embodiments, the AI / ML model management and adaptation information may include parameters characterizing AI / ML model management and adaptation actions.

[0016] In any of the foregoing embodiments, the AI / ML model management and adaptation information may include indications of AI / ML model management and adaptation actions.

[0017] In the above embodiments, when the indication of the AI / ML model management and adaptation action includes an indication of model switching, the UE may select an AI / ML model to be applied at the UE from the trained models; when the indication of the AI / ML model management and adaptation action includes an indication of model optimization or update, the UE may optimize the AI / ML model by one or both of retraining using new training data or revalidating using new validation data; when the indication of the AI / ML model management and adaptation action includes an indication of model updating, the UE may reconstruct or prepare a new AI / ML model to be applied on the UE; and when the indication of the AI / ML model management and adaptation action includes an indication of model migration, the UE may apply the received AI / ML model parameters.

[0018] In any of the foregoing embodiments, the monitoring configuration may include monitoring resources for monitoring in the time domain or the frequency domain or the spatial domain, the number of reports for the operation, and the type of reports.

[0019] In the above embodiment, the monitoring configuration may include a condition that triggers the UE to perform one of the following operations: reporting AI / ML monitoring results, or autonomously performing AI / ML model management and adaptation.

[0020] For those skilled in the art, other technical features may be apparent from the following drawings, descriptions and claims.

[0021] Before proceeding to the following detailed description, it may be beneficial to set forth the definitions of certain words and phrases used in this patent document. The term "coupled" and its derivatives refer to any direct or indirect communication between two or more elements, whether or not these elements are in physical contact with each other. The terms "send", "receive" and "communication" and their derivatives include direct and indirect communication. The terms "include" and "comprise" and their derivatives mean including but not limited to. The term "or" has compatibility, meaning and / or. The term "association" and its derivatives refer to including, being included therein, interconnected with it, containing, contained within, connected to or connected with it, coupled to or coupled with it, communicable, collaborative, interwoven, juxtaposed, close to, bound or bound with it, having, having characteristics, having a relationship with it or having an association with it, etc. The term "controller" refers to any device, system or part thereof that controls at least one operation. Such a controller can be implemented with hardware or a combination of hardware and software and / or firmware. The functions associated with any particular controller can be centralized or distributed, whether local or remote. When a list of items is used, the phrase "at least one" means that different combinations of one or more of the listed items can be used, and only one item in the list may be required. For example, "at least one of A, B, and C" includes any of the following combinations: A, B, C, A and B, A and C, B and C, and A and B and C. Likewise, the term "set" means one or more. Thus, a set of items can be a single item or a set of two or more items.

[0022] Moreover, various functions as described below can be implemented or supported by one or more computer programs, each of which is formed by a computer-readable program code and implemented in a computer-readable medium. The terms "application" and "program" refer to one or more computer programs, software components, instruction sets, processes, functions, objects, categories, instances, related data, or parts thereof adapted to be implemented in a suitable computer-readable program code. The phrase "computer-readable program code" includes any type of computer code, including source code, object code, and executable code. The phrase "computer-readable medium" includes any type of medium that can be accessed by a computer, such as a read-only memory (ROM), a random access memory (RAM), a hard drive, a compact disc (CD), a digital video disc (DVD), or any other type of memory. "Non-transient" computer-readable media excludes wired, wireless, optical, or other communication links that convey instantaneous electrical signals or other signals. Non-transient computer-readable media include media in which data can be permanently stored and media in which data can be stored and rewritten later, such as rewritable optical discs or erasable memory devices.

[0023] Definitions for certain words and phrases are provided throughout this patent document, those of skill in the art should understand that in many, if not most instances, such definitions apply to prior, as well as future uses of such defined words and phrases.

[0024] Advantageous Effects of the Invention

[0025] According to the embodiments of the present disclosure, wireless communication can be performed efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] For a more complete understanding of the present disclosure and its advantageous effects, reference is now made to the following description in conjunction with the accompanying drawings, in which:

[0027] Figure 1 An exemplary networking system utilizing AI / ML-enabled UE capabilities in a cellular system according to various embodiments of the present disclosure is illustrated;

[0028] Figure 2 An exemplary base station (BS) utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure is illustrated;

[0029] Figure 3 An exemplary electronic device for communicating in a networked computing system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure is shown;

[0030] Figure 4 A high-level diagram showing an overall setup for AI / ML-related UE capability migration according to various embodiments of the present disclosure;

[0031] Figure 5 is a high-level flow chart for NW-centric AI / ML model monitoring of UE behavior in a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure;

[0032] Figure 6 is a high-level flow chart for NW behavior in NW-centric AI / ML model monitoring in a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure;

[0033] Figure 7 is a high-level flow chart for UE behavior in UE-centric AI / ML model monitoring in a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure; and

[0034] Figure 8is a high-level flow chart of NW behaviors in UE-centric AI / ML model monitoring in a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure. DETAILED DESCRIPTION

[0035] In order to meet the growing demand for wireless data traffic since the deployment of 4G communication systems and to enable various vertical applications, 6G / 5G / NR communication systems have been developed and are currently being deployed. 6G / 5G / NR communication systems are believed to be implemented in higher frequency (millimeter wave) bands (such as 28 gigahertz (GHz) or 60 GHz bands) to achieve higher data rates, or in lower frequency bands (such as 6 GHz bands) to achieve robust coverage and mobility support. In order to reduce the propagation loss of radio waves and increase the transmission distance, in 5G / NR communication systems, beamforming, massive multiple-input multiple-output (MIMO), full-dimensional multiple-input multiple-output (FD-MIMO), array antennas, analog beamforming, and massive antenna technology are being discussed.

[0036] In addition, in 6G / 5G / NR communication systems, development is being advanced to optimize system networks based on the following advanced technologies: miniaturized base stations (cells), cloud radio access networks (RANs), ultra-dense networks, device-to-device (D2D) communications, wireless backhaul, mobile networks, collaborative communications, coordinated multi-point (CoMP), and receiver interference cancellation.

[0037] The discussion of 6G and 5G systems and frequency bands associated therewith is for reference only, as certain embodiments of the present disclosure may be implemented in 6G / 5G systems. However, the present disclosure is not limited to 6G / 5G systems or frequency bands associated therewith, and embodiments of the present disclosure may be used in conjunction with any frequency band. For example, aspects of the present disclosure may also be applied to the deployment of 6G / 5G communication systems, 6G, or even later versions that may use terahertz (THz) frequency bands.

[0038] The drawings and various embodiments used to describe the principles of the present disclosure herein are only examples and should not be interpreted as limiting the scope of the present disclosure in any way. In addition, those skilled in the art will appreciate that the principles of the present disclosure can be implemented in any appropriately arranged wireless communication system.

[0039] refer to:

[0040] [1]3GPP RP-213599, New SI: Study on Artificial Intelligence(AI) / MachineLearning(ML) for NR Air Interface.

[0041] [2]3GPP, TS 38.300v17.1.0, 5G; NR; NR and NG-RAN Overall description; Stage-2.

[0042] [3]3GPP, TS 38.331v17.1.0, 5G; NR; Radio Resource Control (RRC); Protocol specification.

[0043] [4]3GPP, TS 38.306v17.1.0, 5G; NR; User Equipment (UE) radio accesscapabilities.

[0044] [5]3GPP, TS 38.214v17.2.0, 5G; NR; Physical layer procedures for data.

[0045] [6]3GPP, TS 37.355v17.1.0, LTE; 5G; LTE Positioning Protocol (LPP).

[0046] The above reference(s) are incorporated herein by reference.

[0047] abbreviation:

[0048] Artificial Intelligence

[0049] ML Machine Learning

[0050] BS Base Station

[0051] NW Network

[0052] UE User Equipment

[0053] NR New Radio

[0054] 3GPP 3rd Generation Partnership Project

[0055] CSI Channel State Information

[0056] RTT Round Trip Time

[0057] TDOA Time Difference of Arrival

[0058] TDD Time Division Duplex

[0059] FDD Frequency Division Duplex

[0060] FR Frequency Range

[0061] NLOS Non-Line-of-Sight

[0062] FLOPS (Floating Point Operations Per Second)

[0063] NN Neural Network

[0064] CNN Convolution Neural Network

[0065] RNN Recurrent Neural Network

[0066] LSTM (Long Short-Term Memory)

[0067] BiLSTM Bidirectional Long Short-Term Memory

[0068] SGCS Squared Generalized Cosine Similarity

[0069] GCS Generalized Cosine Similarity

[0070] UMa Urban Macrocell

[0071] UMi Urban Microcell

[0072] InH Indoor Hotspot

[0073] RMaRural Macrocell

[0074] ISD Inter-Site Distance

[0075] DCI Downlink Control Information

[0076] RS Reference Signal

[0077] QCL Quasi Co-located

[0078] SSB Synchronization Signal Block

[0079] SRS Sounding Reference Signal

[0080] CQI Channel Quality Indicator

[0081] RI Rank Indicator

[0082] RSRP Reference Signal Received Power

[0083] RSRQ Reference Signal Received Quality

[0084] SINR Signal to Inference and Noise Ratio

[0085] CRI CSI-RS Resource Indicator

[0086] SSBRI SSB Resource Indicator

[0087] SLI Strongest Layer Indicator

[0088] LI layer indicator (Layer Indicator)

[0089] PUCCH Physical Uplink Control Channel

[0090] PUSCH Physical Uplink Shared Channel

[0091] KPI Key Performance Indicator

[0092] PRS Positioning Reference Signal

[0093] 3GPP (3rd Generation Partnership Project) develops technical specifications and standards to define the new 5th generation (5G) wireless access technology, referred to as 5G NR (New Radio). As recent advances in AI / ML bring new opportunities for improved performance in wireless communications, a new study item [1] has been agreed in Release 18 to study selected use cases where AI / ML algorithms can be applied to achieve performance improvements and / or reduce complexity and overhead. The initial set of use cases includes CSI feedback enhancements, e.g., reducing overhead, improving accuracy, and making predictions; beam management, e.g., beam prediction in the time and / or spatial domain for reducing overhead and latency, improving beam selection accuracy; and positioning accuracy enhancement for different scenarios, including, for example, those with severe NLOS conditions.

[0094] In order to apply AI / ML algorithms on the UE / NW side for selected use cases, relevant features to support AI / ML applications need to be developed and specified for each use case. The UE and NW should communicate about the UE capabilities of the relevant AI / ML features so that the UE can configure the AI / ML operations appropriately. Accordingly, it is expected that the UE capabilities to support the relevant AI / ML features will be specified.

[0095] AI / ML models are expected to improve the performance of various operations and use cases, including CSI feedback enhancement, beam management, and positioning. Compared with existing mechanisms, AI / ML-based CSI compression is expected to provide more accurate and less-overhead CSI reports than current Type I or Type II codebooks; AI / ML-based beam management is expected to identify better beams for DL / UL Rx / Tx than current beam management mechanisms based on beam measurement and reporting, or predict the next candidate beam before the current beam pairing fails; AI / ML-based positioning is expected to provide more accurate UE location than traditional non-AI / ML-based positioning methods (e.g., multi-RTT, DL-TDOA, UL-TDOA, DL-AoD, UL-AoA). However, as application scenarios, channel models, and system parameters change, AI / ML models may not always maintain good performance, resulting in the need for corresponding adjustments. In order to monitor the performance of the AI / ML model and adjust the behavior of the NW / UE accordingly, it is expected that a process for model monitoring will be designed and specified.

[0096] The present disclosure specifies UE capabilities in terms of AI / ML related features from different perspectives including use cases, AI / ML models, training / inference, and model management.

[0097] In this disclosure, processes for AI / ML model monitoring for various use cases are specified, including a NW-centric AI / ML model monitoring process and a UE-centric AI / ML model monitoring process.

[0098] Systems and methods consistent with embodiments of the present disclosure are described in detail below. Although multiple embodiments are described, it should be understood that the present disclosure is not limited to any one embodiment, but rather encompasses numerous alternatives, modifications, and equivalents. In addition, in order to thoroughly understand the embodiments disclosed herein, many specific details are set forth in the following description, but some embodiments may be implemented without some or all of these details. Moreover, for the sake of clarity, certain technical materials known in the relevant art are not described in detail so as not to unnecessarily obscure the present disclosure.

[0099] The following Figures 1 to 3 Various embodiments are described that may be implemented using Orthogonal Frequency Division Multiplexing (OFDM) or Orthogonal Frequency Division Multiple Access (OFDMA) communication techniques in a wireless communication system. Figures 1 to 3 The description is not meant to imply physical or architectural limitations to the manner in which different embodiments may be implemented. Different embodiments of the present disclosure may be implemented in any suitably configured communications system.

[0100] Figure 1An exemplary networking system utilizing AI / ML-enabled UE capabilities in a cellular system according to various embodiments of the present invention is illustrated. Figure 1 The embodiment of the wireless network shown is for example only. Other embodiments of the wireless network 100 may be used without departing from the scope of the present invention.

[0101] like Figure 1 As shown, the wireless network includes gNB 101 (e.g., base station, BS), gNB 102, and gNB 103. gNB 101 communicates with gNB 102 and gNB 103. gNB 101 also communicates with at least one network 130 (e.g., the Internet, a proprietary Internet Protocol (IP) network, or other data network).

[0102] gNB 102 provides wireless broadband access to network 130 for a first plurality of user equipment (UE) within coverage area 120 of gNB 102. The first plurality of UEs includes UE 111, which may be located in a small business; UE 112, which may be located in an enterprise; UE 113, which may be a WiFi hotspot; UE 114, which may be located in a first residence; UE 115, which may be located in a second residence; and UE 116, which may be a mobile device (e.g., a cell phone, a wireless laptop, a wireless PDA, etc.). gNB 103 provides wireless broadband access to network 130 for a second plurality of UEs within coverage area 125 of gNB 103. The second plurality of UEs includes UE 115 and UE 116. In some embodiments, one or more of gNBs 101 to 103 may communicate with each other and with UEs 111 to 116 using 5G / NR, Long Term Evolution (LTE), Long Term Evolution Advanced (LTE-A), WiMAX, WiFi, or other wireless communication technologies.

[0103] Depending on the network type, the term "base station" or "BS" can refer to any component (or collection of components) configured to provide wireless access to a network, such as a transmission point (TP), a transmission reception point (TRP), an enhanced base station (eNodeB or eNB), a 5G / NR base station (gNB), a macrocell, a femtocell, a WiFi access point (AP), or other wirelessly enabled devices. The base station can provide wireless access according to one or more wireless communication protocols such as: 5G / NR 3rd Generation Partnership Project (3GPP) NR, Long Term Evolution (LTE), LTE Advanced (LTE-A), High Speed ​​Packet Access (HSPA), Wi-Fi 802.11a / b / g / n / ac, etc. For convenience, the terms "BS" and "TRP" are used interchangeably in this patent document to refer to network infrastructure components that provide wireless access to remote terminals. In addition, depending on the network type, the term "user equipment" or "UE" can refer to any component, such as a "mobile station", "user station", "remote terminal", "wireless terminal", "reception point", or "user equipment". For convenience, the terms "user equipment" and "UE" are used in this patent document to refer to a remote wireless device that wirelessly accesses a BS, regardless of whether the UE is a mobile device (such as a mobile phone or smartphone) or is generally considered a fixed device (such as a desktop computer or vending machine).

[0104] The dashed lines show the approximate extents of coverage areas 120 and 125, which are shown as approximately circular for purposes of illustration and explanation only. It should be clearly understood that the coverage areas associated with a gNB, such as coverage areas 120 and 125, may have other shapes, including irregular shapes, depending on the configuration of the gNB and changes in the radio environment associated with natural and artificial obstacles.

[0105] As described in more detail below, one or more of the UEs 111 to 116 include circuits, programs, or a combination thereof for AI / ML model management and adaptation operations for one or more UEs in a wireless communication system. In certain embodiments, and one or more of the gNBs 101 to 103 include circuits, programs, or a combination thereof for supporting AI / ML model management and adaptation operations for one or more UEs in a wireless communication system.

[0106] although Figure 1 An example of a wireless network is shown, but Figure 1Various changes may be made. For example, the wireless network may include any number of gNBs and any number of UEs in any suitable arrangement. In addition, gNB 101 may communicate directly with any number of UEs and provide wireless broadband access to network 130 for these UEs. Similarly, each gNB 102-103 may communicate directly with network 130 and provide direct wireless broadband access to network 130 for the UEs. In addition, gNBs 101, 102, and / or 103 may provide access to other external networks or added external networks, such as an external telephone network or other type of data network.

[0107] Figure 2 An exemplary base station (BS) utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure is illustrated. Figure 2 The embodiment of the gNB 102 shown in FIG. 1 is for example only, and Figure 1 gNBs 101 and 103 may have the same or similar configurations. However, the configurations of gNBs vary, and Figure 2 The scope of the present invention is not limited to any particular implementation of the gNB.

[0108] like Figure 2 As shown, the gNB 102 includes multiple antennas 205a to 205n, multiple transceivers 210a to 210n, a controller / processor 225, a memory 230, and a backhaul or network interface 235.

[0109] The transceivers 210a to 210n receive incoming RF signals from the antennas 205a to 205n, such as signals sent by UEs in the network 100. The transceivers 210a to 210n downconvert the incoming RF signals to generate IF or baseband signals. The IF or baseband signals are processed by receive (RX) processing circuits in the transceivers 210a to 210n and / or the controller / processor 225, which generate processed baseband signals by filtering, decoding and / or digitizing the baseband or IF signals. The controller / processor 225 may further process the baseband signals.

[0110] Transmit (TX) processing circuitry in transceivers 210a to 210n and / or controller / processor 225 receives analog or digital data (e.g., voice data, web page data, email, or interactive video game data) from controller / processor 225. TX processing circuitry encodes, multiplexes, and / or digitizes the outgoing baseband data to generate processed baseband or IF signals. Transceivers 210a to 210n up-convert the baseband or IF signals to RF signals that are transmitted via antennas 205a to 205n.

[0111] The controller / processor 225 may include one or more processors or other processing devices that control the overall operation of the gNB 102. For example, the controller / processor 225 may control the reception of UL channel signals and the transmission of DL channel signals using the transceivers 210a to 210n in accordance with well-known principles. The controller / processor 225 may also support additional functionality, such as more advanced wireless communication functionality. For example, the controller / processor 225 may support beamforming or directional routing operations in which outgoing signals from / incoming signals to multiple antennas 205a to 205n are assigned different weights to effectively steer the outgoing signals in a desired direction. The controller / processor 225 may support any of a variety of other functions in the gNB 102.

[0112] The controller / processor 225 is also capable of running programs and other processes resident in the memory 230, such as processes for supporting AI / ML model management and adaptation operations for one or more UEs in the wireless communication system. The controller / processor 225 can move data into or out of the memory 230 as required to run the processes.

[0113] The controller / processor 225 is also coupled to a backhaul or network interface 235. The backhaul or network interface 235 allows the gNB 102 to communicate with other devices or systems via a backhaul connection or network. The interface 235 may support communication via any appropriate (one or more) wired or wireless connection. For example, when the gNB 102 is implemented as part of a cellular communication system (e.g., a system supporting 5G / NR, LTE, or LTE-A), the interface 235 may allow the gNB 102 to communicate with other gNBs via a wired or wireless backhaul connection. When the gNB 102 is implemented as an access point, the interface 235 may allow the gNB 102 to communicate via a wired or wireless local area network or via a wired or wireless connection to a larger network (e.g., the Internet). The interface 235 includes any appropriate structure that supports communication via a wired or wireless connection (e.g., Ethernet or a transceiver).

[0114] Memory 230 is coupled to controller / processor 225. A portion of memory 230 may include RAM, while another portion of memory 230 may contain flash memory or other ROM.

[0115] although Figure 2 An example of a gNB 102 is shown, but the Figure 2 For example, gNB 102 may include Figure 2 Any number of each component shown in Figure 2The various components in may be combined, further subdivided, or omitted, and additional components may be added according to specific needs.

[0116] Figure 3 An exemplary electronic device for communicating with AI / ML-enabled UE capabilities in a networked computing system according to various embodiments of the present disclosure is illustrated. Figure 3 The embodiment of UE 116 shown in FIG. 1 is for example only, and Figure 1 UEs 111 to 115 may have the same or similar configurations. However, the configurations of UEs vary, and Figure 3 The scope of the present disclosure is not limited to any particular implementation of the UE.

[0117] like Figure 3 As shown, UE 116 includes (one or more) antennas 305, (one or more) transceivers 310, and microphone 320. UE 116 also includes a speaker 330, a processor 340, an input / output (I / O) interface (IF) 345, an input 350, a display 355, and a memory 360. The memory 360 includes an operating system (OS) 361 and one or more applications 362.

[0118] Transceiver(s) 310 receive incoming RF signals from antenna 305 transmitted by a gNB of network 100. Transceiver(s) 310 down-convert the incoming RF signals to generate an intermediate frequency (IF) or baseband signal. The IF or baseband signal is processed by RX processing circuitry in transceiver(s) 310 and / or processor 340, which generates a processed baseband signal by filtering, decoding, and / or digitizing the baseband or IF signal. The RX processing circuitry transmits the processed baseband signal to speaker 330 (e.g., for voice data) or is processed by processor 340 (e.g., for web browsing data).

[0119] The TX processing circuit in the transceiver(s) 310 and / or processor 340 receives analog or digital voice data from the microphone 320, or receives other outgoing baseband data (e.g., web page data, email, or interactive video game data) from the processor 340. The TX processing circuit encodes, multiplexes, and / or digitizes the outgoing baseband data to generate a processed baseband or IF signal. The transceiver(s) 310 up-converts the baseband or IF signal to an RF signal that is transmitted via the antenna(s) 305.

[0120] The processor 340 may include one or more processors or other processing devices and runs an OS 361 stored in the memory 360 to control the overall operation of the UE 116. For example, the processor 340 may control the reception of DL channel signals and the transmission of UL channel signals using the transceiver(s) 310 according to well-known principles. In some embodiments, the processor 340 includes at least one microprocessor or microcontroller.

[0121] The processor 340 is also capable of running other processes and programs resident in the memory 360. The processor 340 can move data into or out of the memory 360 as required to run a process, such as a process for AI / ML model management and adaptation operations in a wireless communication system. In some embodiments, the processor 340 is configured to run an application 362 based on the OS 361 or in response to a signal received from a gNB or operator. The processor 340 is also coupled to an I / O interface 345, which provides the UE 116 with the ability to connect to other devices, such as laptops and handheld computers. The I / O interface 345 is a communication path between these accessories and the processor 340.

[0122] Processor 340 is also coupled to input 350 including, for example, a touch screen, keyboard, etc., and display 355. An operator of UE 116 may enter data into UE 116 using input 350. Display 355 may be a liquid crystal display, a light emitting diode display, or other display capable of presenting text and / or at least limited graphics, such as from a website.

[0123] Memory 360 is coupled to processor 340. A portion of memory 360 may include random access memory (RAM), and another portion of memory 360 may include flash memory or other read-only memory (ROM).

[0124] although Figure 3 An example of UE 116 is shown, but the Figure 3 Make various changes. For example, Figure 3 Various components in may be combined, further subdivided, or omitted, and additional components may be added as required. As a specific example, processor 340 may be divided into multiple processors, such as one or more central processing units (CPUs) and one or more graphics processing units (GPUs). In another example, transceiver(s) 310 may include any number of transceivers and signal processing chains, and may be connected to any number of antennas. In addition, although Figure 3 The UE 116 is shown configured as a mobile phone or smartphone, but the UE may be configured to operate as other types of mobile or stationary devices.

[0125] Figure 4 A high-level diagram of an overall setup for AI / ML-related UE capability migration according to various embodiments of the present disclosure is shown. Figure 4 The embodiments shown are for example only. Other embodiments may be used without departing from the scope of the present disclosure.

[0126] When the network 401 requires (additional) UE radio access capability information, the network 401 initiates the procedure 400 towards the UE 402 in the RRC_CONNECTED state. Figure 4 As shown, the network 401 may request the UE 402 to provide radio access capability information by sending a UECapabilityEnquiry message 403 after establishing access stratum (AS) security. The UE 402 replies with a UECapabilityInformation message 404. In one embodiment, in the information element (IE) UE-NR-Capability within the UECapabilityInformation message 404, the UE 402 may provide capability information of the UE regarding supported AI / ML related features, including AI / ML use cases and / or operations for specific use cases, types / structures of AI / ML models and / or types of training / inference, and / or related operations for model management, etc.

[0127] In one embodiment, the UE capabilities supporting AI / ML use cases may be defined for each UE and / or may be different in time division duplex (TDD) and frequency division duplex (FDD), and / or may be different in frequency range 1 (FR1) and frequency range 2 (FR2). In one example, for each use case, such as CSI compression, CSI prediction, beam management, and positioning, a one-bit indication is used to indicate whether the UE supports AI / ML-based operations for each use case:

[0128]

[0129] In another example, the UE may list the use cases supported by the UE:

[0130]

[0131] In one embodiment, the UE capabilities supporting AI / ML sub-use cases may be defined for each UE and / or different in TDD and FDD and / or different in FR1 and FR2. In one example of a use case for CSI compression, the UE may indicate the type of AI / ML supported by the UE (e.g., unilateral model and / or bilateral model) and / or the type of CSI compression supported by the UE (e.g., spatial-frequency domain CSI compression and / or time-space frequency domain CSI compression). In one example of a use case for CSI prediction, the UE may indicate the type of AI / ML-based CSI prediction supported by the UE, for example, spatial domain CSI prediction and / or time domain prediction. In one example of a beam management use case, the UE may indicate the type of AI / ML-based beam management supported by the UE, for example, spatial domain beam management and / or time domain beam management. In one example of a positioning use case, the UE may indicate the type of AI / ML-based positioning supported by the UE, for example, direct AI / ML-based positioning and / or AI / ML-assisted positioning. In another example, the UE may indicate whether the UE-side model and / or the NW-side model are supported for the supported (sub) use case. In one example, a UE supporting a specific (sub) use case mandatorily or optionally supports one or more associated operations for the use case, including data collection / delivery, inference pre-processing and post-processing, auxiliary information reporting, model generalization, model switching / activation / deactivation, model update, model transfer, model monitoring:

[0132]

[0133] In one embodiment, the UE capability of supporting AI / ML model switching / activation / deactivation may be defined for each UE and / or each (sub) use case, and / or may be different in TDD and FDD, and / or may be different in FR1 and FR2. In one example, the UE capability of supporting AI / ML model switching / activation / deactivation may be defined as a mandatory or conditionally mandatory capability or an optional capability. In the case of conditional mandatory, a UE supporting one or more AI / ML (sub) use cases mandatorily supports AI / ML model switching / activation / deactivation. A UE indicating support for switching / activation / deactivation of an AI / ML model is able to switch / activate / deactivate within a set of AI / ML models according to NW configuration or dynamic indication, where the AI / ML models may be predefined and well-trained for various deployment scenarios, channel models, carrier frequencies or system parameters.

[0134] In one embodiment, the UE capability of supporting configurable AI / ML models can be defined for each UE and / or each (sub) use case, and / or different in TDD and FDD, and / or different in FR1 and FR2. In one example, the UE capability of supporting configurable AI / ML models can be defined as a mandatory or conditionally mandatory capability or an optional capability. In the case of conditional mandatory, a UE that supports one or more AI / ML (sub) use cases mandatorily supports configurable AI / ML models. It means that a UE that supports configurable AI / ML models is able to build an AI / ML model according to the NW configuration on the NN structure (including NN type and number of layers, etc.).

[0135] In one embodiment, the UE capability indication of supported AI / ML models / structures may be defined for each UE and / or each (sub) use case, and / or may be different in TDD and FDD, and / or may be different in FR1 and FR2. In one example, the UE is able to indicate the maximum computational complexity (e.g., FLOPS) supported by the UE. In one example, the UE is able to indicate the NN structures supported by the UE, such as CNN, RNN, LSTM, BiLSTM, transformer, inception. In one example, the UE is able to indicate the maximum number of NN layers supported by the UE, and / or the maximum number of kernels per layer, and / or the maximum number of weights per layer, and / or the maximum number of branches, and / or the maximum number of real-valued / complex-valued model parameters.

[0136] In one embodiment, the UE capability indication for the supported learning type may be defined for each UE and / or each (sub) use case, and / or different in TDD and FDD, and / or different in FR1 and FR2. In one example, the UE is able to indicate whether the UE supports offline learning, and / or online learning, and / or supervised learning, and / or unsupervised learning, and / or semi-supervised learning, and / or joint learning, and / or reinforcement learning, and / or transfer learning.

[0137] In one embodiment, the UE capability to support AI / ML model updates may be defined for each UE and / or each (sub) use case, and / or may be different in TDD and FDD, and / or may be different in FR1 and FR2. In one example, the UE capability to support AI / ML model updates may be defined as a mandatory or conditionally mandatory capability or an optional capability. In the case of conditional mandatory, a UE supporting one or more AI / ML (sub) use cases mandatorily supports AI / ML model updates. Indicating that a UE supporting AI / ML model updates is able to update / adjust AI / ML model parameters by retraining with a new data set. In one example, the UE may indicate which part of the AI / ML model the UE supports updating, such as weights, and / or layers, and / or model structure.

[0138] In one embodiment, the UE capability to support AI / ML model migration may be defined for each UE and / or each (sub) use case, and / or may be different in TDD and FDD, and / or may be different in FR1 and FR2. In one example, the UE capability to support AI / ML model migration may be defined as a mandatory or conditionally mandatory capability or an optional capability. In the case of conditional mandatory, a UE that supports a bilateral AI / ML model for a certain (sub) use case mandatorily supports AI / ML model migration. Indicating that a UE that supports AI / ML model migration is capable of transmitting and / or receiving AI / ML model parameters over the air interface. In one example, the UE may indicate whether the UE supports full model and / or partial model migration.

[0139] In one embodiment, the UE capability to support AI / ML model monitoring may be defined for each UE and / or each (sub) use case, and / or may be different in TDD and FDD, and / or may be different in FR1 and FR2. In one example, the UE capability to support AI / ML model monitoring may be defined as a mandatory or conditionally mandatory capability or an optional capability. In the case of conditional mandatory, a UE supporting one or more AI / ML (sub) use cases mandatorily supports AI / ML model monitoring. The UE is able to indicate that the UE supports AI / ML model monitoring operations, for example, operations to monitor the inference performance based on AI / ML use cases:

[0140]

[0141] NW-centric AI / ML model monitoring

[0142] In one embodiment, AI / ML monitoring is NW-centric. The NW can monitor the performance of the AI / ML model on the UE / NW side based on the auxiliary information reported by the UE and / or based on the NW-side statistics that can reflect the performance of the AI / ML model for specific use cases (e.g., CSI compression, CSI prediction, spatial / temporal beam prediction, and positioning). Based on these indicators, the NW can decide on the operations related to the AI / ML model on the NW / UE side and send the necessary control signaling to the UE.

[0143] Figure 5 is a high-level flow chart of UE behavior in NW-centric AI / ML model monitoring within a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure. Figure 5 The embodiments are for example only, and one or more other (processes) may be used. Figure 5 The scope of the present disclosure is not limited to any particular process.

[0144] exist Figure 5In the UE behavior shown, the UE applies the AI / ML model for a specific use case and performs related operations according to the NW configuration including the AI / ML model monitoring configuration (operation 501). The configuration can be sent in an RRC message (e.g., RRCSetup, RRCReconfiguration). In operation 502, if configured, the UE reports auxiliary information for AI / ML model monitoring. In operation 503, the UE receives signaling / configuration from the NW to trigger fallback to the traditional operation of the use case or trigger AI / ML model adaptation including model switching / update / optimization / migration, and the UE performs the following operations accordingly:

[0145] -Model switching: UE selects a qualified AI / ML model to be applied on the UE side from the trained models;

[0146] -Model optimization: The UE optimizes the current AI / ML model on the UE side by retraining and / or revalidating it with new training / validation data;

[0147] -Model update: UE rebuilds / prepares new AI / ML models to be applied on the UE side;

[0148] -Model migration: UE directly applies AI / ML model parameters migrated from NW;

[0149] - Fallback to legacy operation: The UE disables the AI / ML model for the use case and enables legacy operation (e.g., CSI measurement and reporting using Type I and / or Type II codebooks, beam management and reporting, DL-PRS measurement and reporting).

[0150] Figure 6 is a high-level flow chart of NW behaviors in NW-centric AI / ML model monitoring within a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure. Figure 6 The embodiments are for example only and other processes may be used. Figure 6 The scope of the present disclosure is not limited to any particular process. For example, one or more base stations within the NW may perform Figure 6 processing.

[0151] exist Figure 6In the NW behavior shown, in operation 601, the NW configures the UE to apply the AI / ML model for a specific use case and perform related operations including AI / ML model monitoring. The configuration can be sent in an RRC message (e.g., RRCSetup, RRCReconfiguration). In operation 602, if configured, the NW receives auxiliary information reported by the UE for AI / ML model monitoring. In operation 603, the NW monitors / evaluates the performance of the AI / ML model for a specific UE of the use case based on the auxiliary information and / or statistics of performance indicators reported by the UE. In operation 604, the NW decides on the operation of the AI / ML model on the NW / UE side and sends signaling / configuration to the UE to trigger fallback to traditional operations of the use case or trigger AI / ML model adaptation including model switching / update / optimization / migration.

[0152] In one embodiment of NW-centric AI / ML model monitoring, at operation 506, the NW evaluates the performance of the NW / UE-side AI / ML model based on statistics of NW-side performance indicators (e.g., average user throughput, average UPT, user scheduling delay, PDSCH decoding bit error rate from HARQ ACK feedback, PUSCH decoding frame error rate, etc.). The NW can also compare the performance of the AI / ML model for a specific UE with other UEs and / or traditional operations (e.g., CSI reporting based on Class I / II codebooks).

[0153] In one embodiment of NW-centric AI / ML model monitoring for CSI compression, the NW can configure reciprocal DL RS and UL RS for model monitoring. The NW can configure the UE to report auxiliary information and evaluate the performance of the AI / ML model based on the auxiliary information reported by the UE.

[0154] For CSI compression, in one example of operation 501 / 601, the NW may configure the UE to perform model monitoring and / or report model monitoring assistance information, including AI / ML generated CSI feedback. The model monitoring configuration may include monitoring RS configuration and / or assistance information reporting configuration.

[0155] - In the monitoring RS configuration, the NW may configure aperiodic / periodic / semi-persistent CSI-RS or SSB as a DL monitoring RS, and / or configure aperiodic / periodic / semi-persistent SRS as a UL monitoring RS. The RS configuration may include: one or more fields in IE NZP-CSI-RS-Resource[3], and / or one or more fields in SRS-Resource[3], and / or the ID of the DL-UL monitoring RS pair, the ID of each pair of DL monitoring RS and the ID of the UL monitoring RS (e.g., nzp-CSI-RS-ResourceID, srs-ResourceID), the periodicity of the DL-UL pair, the QCL information in each DL monitoring RS, the spatial relationship information in each UL monitoring RS, the resource element mapping of RS resources in the time domain and the frequency domain, the scrambling / sequence ID used to generate the RS, and one or more of the number of ports. In one example, the DL-UL monitoring RS pair may be configured in a manner in which the ID of the DL monitoring RS resource is associated with the ID of the UL monitoring RS resource. In another example, the DL-UL monitoring RS pair may be configured in such a way that the ID of the DL monitoring RS resource set is associated with the ID of the UL monitoring RS resource set. In one example, the use of the DL / UL RS resource (set) may be explicitly indicated as AI-ML-monitoring. In another example, the periodic / aperiodic / semi-persistent monitoring window may be configured with a duration and / or periodicity and / or a start time offset in SFN / subframe / time slot / symbol, and the DL / UL RS resource (set) mapped to the monitoring window is implicitly indicated as monitoring RS.

[0156] -In one example, the auxiliary information may include a CSI feedback report, i.e., an auxiliary information reporting configuration including an AI / ML-based CSI reporting configuration (e.g., CSI ReportConfig[3]). In another example, the auxiliary information may include an average RS measurement amount. In this case, the auxiliary information reporting configuration may include one or more of a report type (e.g., periodic, semi-persistent, aperiodic), a report periodicity, a report amount (e.g., an average of any one of a feature vector, an amplitude coefficient, a phase coefficient, CQI, RI, SLI, LI, CRI, SSBRI, RSRP, SINR), a quantization level, an overhead limit (e.g., a maximum number of bits for a CSI measurement report), and group-based report enable / disable.

[0157] For CSI compression, in one example of operation 501, according to the monitoring RS configuration, the UE sends a UL monitoring RS for the NW to measure CSI, measures a DL monitoring RS that is reciprocal with the indicated UL monitoring RS, and generates CSI feedback through an AI / ML model. The UE may report the CSI feedback generated by AI / ML according to the AI / ML-based CSI reporting configuration.

[0158] For CSI compression, in one example of operation 502 / 602, if configured to report auxiliary information, the UE may report the auxiliary information in PUCCH, and / or PUSCH and / or RRC message (e.g., MeasurementReport and / or UEAssistanceInformation) according to the predefined process and / or the auxiliary information report configuration received in operation 402. In one example, if configured to report a CSI feedback report for DL ​​monitoring RS measurement, the UE reports CSI feedback via PUCCH and / or PUSCH. In another example, if configured to report other auxiliary information other than CSI feedback report, the UE reports the auxiliary information in a monitoring report in an RRC message (e.g., MeasurementReport and / or UEAssistanceInformation). In the monitoring report, the UE may provide the physical cell ID for which the report is being performed, and / or the ID of the RS being measured and / or the number of configured reports.

[0159] For CSI compression, in one example of operation 603, the NW evaluates the performance of the AI / ML model based on the measurements of the UL monitoring RS providing the true CSI and the reported CSI feedback of the DL monitoring RS providing the estimated CSI.

[0160] In another embodiment of NW-centric AI / ML model monitoring for use cases (e.g., CSI compression, CSI prediction, spatial beam prediction, temporal beam prediction, positioning), in operation 501 / 601, the NW may configure the UE to measure the DL monitoring RS, and configure the UE to report auxiliary information, including performance evaluation results and / or CSI feedback and / or UE GNSS position; in operation 603, the NW may evaluate the performance of the AI / ML model and determine operations regarding the AI / ML model based on the auxiliary information reported by the UE.

[0161] In one example of operation 501 / 601, the model monitoring configuration may include a target value of a KPI, parameters related to a monitoring window and / or a DL monitoring RS (e.g., CSI-RS, SSB, DL-PRS), and / or a configuration using a legacy operation (e.g., CSI measurement and reporting using a type I and / or type II codebook, beam management and reporting, DL-PRS measurement and reporting), and / or an auxiliary information reporting configuration:

[0162] -KPI may include one or more of SGCS, GCS, overhead size (e.g., number of bits), RSRP, RSRQ, SINR, L1-RSRP, L1-SINR, RI, CQI, SLI, offset of amplitude / phase coefficients, and horizontal positioning accuracy in meters.

[0163] - Parameters related to the monitoring window may include duration and / or periodicity and / or offset with respect to SFN (System Frame Number) / time slot / symbol.

[0164] -The monitoring RS configuration may include a CSI-RS (set) configuration (e.g., IE NZP-CSI-RS-Resource[3]) and / or a DL-PRS (set) configuration (e.g., IE NR-DL-PRS-Resource[6]), and / or one or more of the following: an ID of a DL monitoring RS (e.g., nzp-CSI-ResourceId), QCL information of each DL monitoring RS, resource element mapping of RS resources in the time and frequency domains, and a scrambling ID used to generate the RS.

[0165] - For DL ​​monitoring RS (e.g., CSI-RS, SSB), the NW can configure CSI measurement and reporting using Type I and / or Type II codebooks [5], and enable / disable codebook-based CSI feedback reporting for the configured monitoring RS.

[0166] - The auxiliary information report configuration may include one or more fields in the IE ReportConfigNR[3] and / or one or more of the following: report type (e.g., periodic, event triggered), report interval indicating the interval between reports, report amount indicating the number of reports, number of reports, triggering event, maximum number of RSs to report, number of RSs for combining / averaging / evaluation, and enabling / disabling CSI feedback reporting for monitoring RS measurements.

[0167] - If codebook-based CSI measurement and reporting is configured, the reported quantities may include AI / ML-based quantities and / or codebook-based quantities. Each set of quantities may include RS-specific values ​​and / or average values ​​of the KPI, and / or RS-specific offsets and / or average offsets to the KPI target value. The reported quantities may also include the UE GNSS position.

[0168] - For event triggered reports, trigger events may be defined in such a way that each trigger event consists of one or more conditions and associated parameters. In one example, the condition of a trigger event may be defined as the RS-specific value / average value of the measured KPI being less than or greater than a target value by an offset, or being outside a target value range, wherein the target value and the offset are indicated in the report configuration associated with the trigger event. In another example, the condition of a trigger event may be defined as the RS-specific value / average value of the measured KPI being less than or greater than a value evaluated on a conventional operation by an offset.

[0169] In one example of operation 501, if configured to evaluate the performance of the AI / ML model for a specific use case (e.g., CSI compression, CSI prediction, spatial / temporal beam prediction, positioning), the UE applies the monitoring configuration. If a monitoring window is configured, the UE may treat all DL RSs configured during the monitoring window as monitoring RSs and measure the DL RSs. If a specific DL monitoring RS is configured, the UE may measure the configured DL monitoring RS. Based on the measurement of the monitoring RS, the UE may generate measurement results, for example, via the AI / ML model and / or via CSI feedback of the type I / II codebook, and evaluate RS-specific performance and / or average performance.

[0170] In one example of operation 501, the UE evaluates the KPI for CSI compression. For the bilateral AI / ML model for CSI compression, the AI / ML CSI encoder is deployed at the UE end, and the AI / ML CSI decoder is deployed at the NW end. In a method for evaluating the performance of the AI / ML model, for a UE that also has a reference model of an AI / ML CSI decoder, the UE can generate an estimated feature vector by a reference CSI decoder, and use the feature vector estimated by the reference CSI decoder and the actual feature vector measured on the DL monitoring RS to calculate the overhead size of the AI / ML-based CSI report and / or SGCS / GCS. If CSI measurement and reporting using type I and / or type II codebooks is configured, the UE can perform AI / ML CSI feedback generation and CSI feedback generation based on type I and / or type II codebooks, and use the estimated feature vector generated by the codebook and the actual feature vector measured on the DL monitoring RS to calculate the overhead size of the codebook-based CSI report and / or SGCS / GCS.

[0171] In one example of operation 501, in order to evaluate the performance of the AI / ML model in CSI prediction, the UE may measure the DL monitoring RS of the future SFN / time slot / symbol for which the CSI is predicted by the UE, and generate CSI feedback if configured to report. Based on the measurement of the DL monitoring RS, the UE evaluates the KPI of the CSI prediction. The UE may calculate the SGCS / GCS using the eigenvectors measured by the monitoring RS and the corresponding predicted eigenvectors. The UE may compare the values ​​of the RSRP / RSRQ / SINR / RI / CQI / SLI / amplitude coefficient / phase / coefficient measured by the monitoring RS with the corresponding predicted values.

[0172] In one example of operation 501, in order to evaluate the performance of the AI / ML model in terms of spatial / temporal beam prediction, the UE may measure the L1-RSRP and / or L1-SINR for monitoring the RS. Based on the measurement, the UE evaluates the KPI by comparing the measured values ​​of L1-RSRP and / or L1-SINR with the predicted values. In one example, the UE may select the N best beams with the N highest L1-RSRP / L1-SINR from all measured beams, compare with the N predicted values, and evaluate the offset and / or prediction accuracy. In another example, the UE may also evaluate the performance in terms of overhead size, for example, comparing the estimated overhead size of the traditional beam measurement report for monitoring the RS and the estimated overhead size of the AI / ML-based beam prediction.

[0173] In one example of operation 502 / 602, if configured to report auxiliary information, the UE may report auxiliary information including CSI feedback and / or KPI evaluation results. The UE sends auxiliary information reports periodically, when requested by the network, or irregularly when any triggering event is met according to the auxiliary information reporting configuration received in operation 402. The UE sends auxiliary information reports in PUCCH, and / or PUSCH and / or RRC messages (e.g., MeasurementReport and / or UEAssistanceInformation) according to a predefined procedure. In one example, if reporting CSI feedback reports for monitoring RS measurements is enabled, the UE reports CSI feedback via PUCCH and / or PUSCH. In another example, the UE reports KPI evaluation results in an RRC message (e.g., MeasurementReport and / or UEAssistanceInformation). In the auxiliary information report, the UE may provide the physical cell ID for which the report is being performed, and / or the ID of the RS being measured and / or the number of configured reports. If configured to report an average number, the UE averages the KPIs of multiple RSs, where the number of RSs for result merging / averaging / evaluation is configured. In the example of CSI compression using a bilateral AI / ML model, if the UE uses a reference CSI decoder to evaluate the performance of the AI / ML model, the UE can report to the NW which reference CSI decoder was used in the evaluation. For example, the UE may indicate the ID of the reference CSI decoder and / or parameters and / or other model descriptions characterizing the reference CSI decoder in the report. If CSI measurement and reporting using Type I and / or Type II codebooks is configured, and if reporting of codebook-based CSI feedback for monitoring RSs is enabled, the UE may report codebook-based CSI feedback according to the CSI measurement and reporting configuration as in conventional operations.

[0174] In one example of operation 603, the NW further evaluates the performance of the AI / ML model based on the auxiliary information reported by the UE including the UE-side evaluation results and / or CSI feedback.

[0175] For operation 503 / 604, in the example where the use case falls back to legacy operation (e.g., CSI compression, CSI prediction, spatial / temporal beam prediction, positioning), the NW may disable the AI / ML model in an RRC message (e.g., RRCReconfiguration) and include a configuration for legacy operation. Alternatively, the NW may use a MAC CE with a subheader containing an LCID or eLCID or use a new bit or a repurposed bit in the DCI to indicate AI / ML model disabling.

[0176] For operation 503 / 604, as an example of AI / ML model switching, when the AI / ML model is already trained and the associated ID is predefined or preconfigured, the ID of the AI / ML model to be switched to can be indicated in an RRC message (e.g., RRCReconfiguration), and / or in a MAC CE with a subheader containing an LCID or eLCID, and / or using a new bit or a diverted bit in the DCI.

[0177] In another example of AI / ML model switching / updating in operation 503 / 604, the NW may send parameters characterizing the model to be switched / updated in an RRC message (e.g., RRCReconfiguration):

[0178] -Parameters characterizing the model may include one or more of the following: type of NN, type / number of NN layers, number of NN weights, type / number of kernels, complexity of FLOPs, type of learning, type of loss function, and KPIs for training / validation. The KPI for CSI feedback enhancement may be any of SGCS, GCS, overhead size, RSRP, RSRQ, SINR, RI, and CQI. The KPI for user positioning may be the horizontal position accuracy in meters. The KPI for beam management may be one or more of RSRP, RSRQ, SINR, L1-RSRP, L1-SINR of the selected beam and / or the average beam failure rate.

[0179] -Parameters used to characterize the model to be switched / updated may also refer to one or more deployment scenarios (e.g., UMa, UMi, InH, RMa), carrier frequency, ISD, antenna parameters, UE speed, and other system parameters to which the AI / ML model is applicable.

[0180] Based on the indication parameters characterizing the AI / ML model, the UE may perform the AI / ML model switching / updating according to a predefined procedure, or according to an explicit switching / updating indication sent together with the characterizing parameters, or until the UE implements it.

[0181] In another example of model migration of operation 503 / 604, the NW can migrate parameters through RRC messages (e.g., RRCReconfiguration) to directly apply to the AI / ML model on the UE side. The parameters may include the type / dimension of input / output data, pre-processing / post-processing of input / output data, type of NN, number of layers, structure and weight of each layer, activation function, number of kernels, structure and weight of each kernel, etc.

[0182] UE-centric AI / ML model monitoring

[0183] For UE-centric AI / ML model monitoring, the UE monitors and evaluates the performance of the AI / ML model for specific use cases (e.g., CSI compression, CSI prediction, spatial / temporal beam prediction, positioning) based on the parameters / configuration indicated by the NW, autonomously decides to fall back to traditional operations or perform model switching / update / migration / optimization, and notifies the NW of the UE behavior when necessary.

[0184] Figure 7 is a high-level flow chart of UE behavior in UE-centric AI / ML model monitoring within a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure. Figure 7 The embodiments are for example only and other processes may be used. Figure 7 The scope of the present disclosure is not limited to any particular process.

[0185] exist Figure 7 In the UE behavior shown, in operation 701, the UE applies the AI / ML model to a specific use case and performs related operations (e.g., AI / ML model monitoring) according to the NW configuration including the model monitoring configuration. In operation 702, the UE evaluates the performance of the AI / ML model based on the model monitoring configuration and decides the operation to be notified to the NW (e.g., model switching / update / optimization / migration or fallback to traditional operation). In operation 703, the UE sends a monitoring report including a model switching / update / optimization / migration indication and / or a performance evaluation result (e.g., a KPI gap), and / or a request to fall back to traditional operation. In operation 704, the UE may receive auxiliary information (e.g., recommended AI / ML model parameters) for model switching / update / optimization / migration or (re) configuration for traditional operation. In operation 705, the UE performs model switching / update / optimization / migration based on the performance evaluation results and / or auxiliary information during monitoring (if applicable), or the UE performs a fallback to traditional operation based on the (re) configuration (if configured).

[0186] Figure 8 is a high-level flow chart of NW behaviors in UE-centric AI / ML model monitoring within a communication system utilizing AI / ML-enabled UE capabilities according to various embodiments of the present disclosure. Figure 8 The embodiments are for example only and other processes may be used. Figure 8 The scope of the present invention is not limited to any particular process. Figure 8 The process may be performed, for example, at one or more base stations within the NW.

[0187] exist Figure 8In the NW behavior shown, in operation 801, the NW configures the UE to apply the AI / ML model for a specific use case and perform related operations including AI / ML model monitoring. In operation 802, the NW receives a monitoring report from the UE including a model switching / update / optimization / migration indication and / or a performance evaluation result (e.g., a KPI gap), and / or a request to fall back to traditional operation. In operation 803, if a model switching / update / optimization / migration indication is received in the monitoring report, the NW may send auxiliary information (e.g., recommended AI / ML model parameters) for model switching / update / optimization / migration, or if a request to fall back to traditional operation is received in the monitoring report, the NW sends a (re)configuration for traditional operation.

[0188] For UE-centric AI / ML model monitoring of use cases (e.g., CSI compression, CSI prediction, spatial beam prediction, temporal beam prediction), in one example of operation 701 / 801, the model monitoring configuration may include target values ​​for KPIs and / or parameters related to the monitoring window and / or DL ​​monitoring RS (e.g., CSI-RS, SSB), and / or configurations for using legacy operations (e.g., CSI measurement and reporting using Type I and / or Type II codebooks, beam management and reporting), and / or monitoring report configuration, and / or model management / adaptation configuration.

[0189] - The KPI may include one or more of SGCS, GCS, overhead size (e.g., number of bits), RSRP, RSRQ, SINR, L1-RSRP, L1-SINR, RI, CQI, SLI, offset of amplitude / phase coefficients.

[0190] - Parameters related to the monitoring window may include duration and / or period, and / or offset of SFN (System Frame Number) / time slot / symbol.

[0191] -The monitoring RS configuration may include a CSI-RS (set) configuration (e.g., IE NZP-CSI-RS-Resource[3]), and / or one or more of the following: an ID of a DL monitoring RS (e.g., nzp-CSI-ResourceId), QCL information for each DL monitoring RS, resource element mapping of RS resources in the time and frequency domains, and a scrambling ID for generating RS.

[0192] -For DL ​​monitoring RS (e.g., CSI-RS, SSB), the NW can configure CSI measurement and reporting using Type I and / or Type II codebooks [5], and enable / disable codebook-based CSI feedback reporting for the configured monitoring RS.

[0193] - The monitoring report configuration may include one or more fields in IE ReportConfigNR[3] and / or one or more of the following: report type (e.g., periodic, conditional event triggered), report interval indicating the interval between reports, report amount indicating the number of reports, number of reports, conditional event, maximum number of RSs to report, number of RSs for merging / averaging / evaluation, enable / disable CSI feedback reporting for monitoring RS metrics.

[0194] -If codebook-based CSI measurement and reporting is configured, the reported quantities may include AI / ML-based quantities and / or codebook-based quantities. Each set of quantities may include RS-specific values ​​and / or average values ​​of the KPI, and / or RS-specific offsets and / or average offsets to the KPI target value.

[0195] -For conditional event triggered reporting or conditional event triggered AI / ML model management / adaptation, conditional events can be defined in a way that each event consists of one or more of a condition, related parameters, and related operations that notify the NW when the event condition is met. In one example, the condition of the event can be defined as a RS-specific value / average value of the measurement of the KPI that is less than or greater than the target KPI value by an offset, or exceeds the KPI value range. In another example, the condition of the event can be defined as a RS-specific value / average value of the measurement of the KPI that is less than the value evaluated in the traditional operation by an offset. The target value and / or offset are indicated in the report configuration associated with the event.

[0196] - In one example, the relevant action to notify the NW immediately upon satisfying each periodic monitoring report or any conditional event is one of the following:

[0197] -Model switching: The UE selects a qualified AI / ML model to be applied on the UE side from the trained models;

[0198] -Model optimization: The UE optimizes the current AI / ML model on the UE side by retraining and / or revalidating with new training / validation data;

[0199] -Model update: UE rebuilds / prepares new AI / ML models to be applied on the UE side;

[0200] -Model migration: UE requests NW to migrate AI / ML model parameters to be directly applied to the UE side;

[0201] - Fallback to legacy operation: The UE requests to disable the AI / ML model for the use case and enable legacy operations (e.g., CSI measurement and reporting using Type I and / or Type II codebooks, beam management and reporting).

[0202] In one example of operation 701, if a monitoring window is configured, the UE may regard all DL RSs configured during the monitoring window as monitoring RSs and measure the DL RSs. If a specific DL monitoring RS is configured, the UE may measure the configured DL monitoring RS. Based on the measurement of the monitoring RS, the UE may generate a measurement result, for example, via an AI / ML model and / or CSI feedback via a type I / II codebook, and evaluate RS-specific performance and / or average performance.

[0203] In one example of operation 702, the UE evaluates the KPI of CSI compression. For the bilateral AI / ML model for CSI compression, the AI / ML CSI encoder is deployed at the UE end and the AI / ML CSI decoder is deployed at the NW end. In a method for evaluating the performance of the AI / ML model, for a UE that also has a reference model of an AI / ML CSI decoder, the UE can generate an estimated feature vector using a reference CSI decoder, and use the feature vector estimated by the reference CSI decoder and the actual feature vector measured on the DL monitoring RS to calculate the overhead size of the AI / ML-based CSI report and / or SGCS / GCS. If CSI measurement and reporting using type I and / or type II codebooks is configured, the UE can perform AI / ML CSI feedback generation and CSI feedback generation based on type I and / or type II codebooks, and use the estimated feature vector generated by the codebook and the actual feature vector measured on the DL monitoring RS to calculate the overhead size of the codebook-based CSI report and / or SGCS / GCS.

[0204] In one example of operation 702, to evaluate the performance of the AI / ML model in CSI prediction, the UE may measure the measured DL monitoring RS of the future SFN / time slot / symbol for which the CSI is predicted by the UE, and generate CSI feedback if configured to report. Based on the measurement of the DL monitoring RS, the UE evaluates the KPI for CSI prediction. The UE may calculate SGCS / GCS using the eigenvectors of the monitoring RS measurements and the corresponding predicted eigenvectors. The UE may compare the monitoring RS measurements of RSRP / RSRQ / SINR / RI / CQI / SLI / amplitude coefficient / phase / coefficient with the corresponding predicted values.

[0205] In one example of operation 702, in order to evaluate the performance of the AI / ML model in terms of spatial / temporal beam prediction, the UE may measure the L1-RSRP and / or L1-SINR for monitoring the RS. Based on the measurement, the UE evaluates the KPI by comparing the measured values ​​of L1-RSRP and / or L1-SINR with the predicted values. In one example, the UE may select the N best beams with the N highest L1-RSRP / L1-SINR from all measured beams, compare the N best beams with the N predicted values, and evaluate the offset and / or prediction accuracy. In another example, the UE may also evaluate the performance in terms of overhead size, for example, comparing the evaluated overhead size of the traditional beam measurement report for monitoring the RS and the estimated overhead size of the AI / ML-based beam prediction.

[0206] In one example of operation 702, the UE may compare the evaluation KPI value of the AI / ML model with the configured target KPI value, and / or compare the evaluation KPI value of the AI / ML model with the evaluation KPI value of the traditional operation (e.g., CSI measurement and reporting, beam management and reporting using type I and / or type II codebooks). In one example, the UE decides the operation to be notified to the NW at each periodic monitoring report based on the comparison. In another example, if a conditional event is configured, the UE notifies the associated operation defined for the conditional event when any conditional event is met.

[0207] In one example of operation 703 / 802, the UE sends a monitoring report periodically or when any conditional event is met according to the monitoring report configuration received in operation 701. The UE sends a monitoring report in a PUCCH, and / or PUSCH and / or an RRC message (e.g., MeasurementReport and / or UEAssistanceInformation) according to a predefined procedure. In one example, if the CSI feedback report for monitoring RS measurement is enabled, the UE reports the CSI feedback via PUCCH and / or PUSCH. In another example, the UE reports the KPI evaluation results in an RRC message (e.g., MeasurementReport and / or UEAssistanceInformation). In the monitoring report, the UE may provide the physical cell ID for which the report is being performed, and / or the operation to be notified to the NW (i.e., model switching / update / optimization / migration or fallback to legacy), and / or the ID of the measured RS and / or the number of configured reports. If CSI measurement and reporting using Type I and / or Type II codebooks is configured, and if reporting codebook-based CSI feedback for monitoring RSs is enabled, the UE may report codebook-based CSI feedback according to the CSI measurement and reporting configuration.

[0208] For operation 704 / 803, in an example of fallback to legacy operation for use cases (e.g., CSI compression, CSI prediction, spatial / temporal beam prediction, positioning), the NW may indicate AI / ML model disabling in an RRC message (e.g., RRCReconfiguration) and include CSI measurement and reporting configuration for legacy operation. Alternatively, the NW may use a MAC CE with a subheader containing an LCID or eLCID, or use a new bit or a diverted bit in the DCI to indicate AI / ML model disabling.

[0209] For operation 704 / 803, as an example of a NW providing auxiliary information for AI / ML model switching, when the AI / ML mode is trained and the associated ID is predefined or preconfigured, the ID of the AI / ML model to be switched to can be indicated in an RRC message (e.g., RRCReconfiguration), and / or in a MAC CE with a subheader containing an LCID or eLCID, and / or using a new bit or a diverted bit in the DCI.

[0210] In another example of a NW providing auxiliary information for AI / ML model switching / updating at operation 704 / 803, the NW may provide parameters characterizing the switched / updated model in an RRC message (e.g., RRCReconfiguration):

[0211] -Parameters characterizing the model may include one or more of the following: type of NN, type / number of NN layers, number of NN weights, type / number of kernels, complexity of FLOPs, type of learning, type of loss function, and performance targets in training / validation. The performance target for CSI feedback enhancement may be one or more of the values ​​of SGCS, GCS, overhead size, RSRP, RSRQ, SINR, RI, CQI. The performance target for positioning may be the horizontal position accuracy in meters. The performance target for beam management may be one or more of the values ​​of RSRP, RSRQ, SINR of the selected beam and / or the average beam failure rate.

[0212] -Parameters characterizing the model to be switched / updated may also refer to one or more of deployment scenarios (e.g., UMa, UMi, InH, RMa), carrier frequency, ISD, antenna parameters, UE speed, and other system parameters to which the AI / ML model is applicable.

[0213] In another example of model migration of operation 704 / 803, the NW may provide auxiliary information through an RRC message (e.g., RRCReconfiguration), which includes parameters of the AI / ML model to be directly applied to the UE side. The parameters may include the type / dimension of input / output data, pre-processing / post-processing of input / output data, type of NN, number of layers, structure and weight of each layer, activation function, number of kernels, structure and weight of each kernel, etc.

[0214] In one example of operation 705, if the UE does not receive auxiliary information for model switching / update / optimization / migration from the NW at operation 704, the UE may, for example, perform model switching / update / optimization based on the AI / ML performance evaluation results during monitoring and / or until the UE implements it alone. In another example of operation 705, if the UE receives auxiliary information for model switching / update / optimization / migration from the NW at operation 704, the UE may choose to use the auxiliary information (if applicable) to perform model switching / update / optimization / migration, and / or perform model switching / update / optimization / migration based on the AI / ML performance evaluation results during monitoring, and / or fall back to traditional operation based on (re) configuration (if configured). In another example of operation 705, if conditional event-triggered AI / ML model management and adaptation is configured, and if the conditional event is met according to the AI / ML performance evaluation based on the model monitoring results, the UE performs the AI / ML model management / adaptation action associated with the met conditional event.

[0215] For illustrative purposes, the steps of the above algorithm are described sequentially. However, some of these steps can be performed in parallel with each other. The operation diagrams illustrate example methods that can be implemented according to the principles of the present invention, and various changes can be made to the methods shown in the flow charts. For example, although shown as a series of steps, the individual steps in each figure can overlap, occur in parallel, occur in different orders, or occur multiple times. In another example, steps can be omitted or replaced by other steps.

[0216] Although the present disclosure has been described with exemplary embodiments, various changes and modifications may be suggested to one skilled in the art. The present disclosure is intended to include such changes and modifications as fall within the scope of the appended claims.

Claims

1. A method performed by a user equipment (UE) in a wireless communication system, the method comprising: performing, at a user equipment (UE), an artificial intelligence / machine learning AI / ML monitoring operation based on a received monitoring configuration, the monitoring configuration forming part of a usage configuration for an AI / ML model for the operation; Based on the monitoring configuration, reporting AI / ML model auxiliary information, the AI / ML model auxiliary information including AI / ML model monitoring results from the AI / ML monitoring operation; Based on the AI / ML model monitoring results, receiving AI / ML model management and adaptation information; and Based on the AI / ML model management and adaptation information, AI / ML model management and adaptation operations are performed.

2. The method according to claim 1, wherein: The AI / ML model management and adaptation information includes an indication of the AI / ML model management and adaptation action and a parameter characterizing the AI / ML model management and adaptation action, Wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model switching, the method further includes selecting, at the UE, an AI / ML model to be applied at the UE from the trained models, Wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model optimization or update, the method further includes: optimizing the AI / ML model at the UE by one or both of retraining using new training data or revalidating using new validation data, Wherein, when the indication of the AI / ML model management and adaptation action includes an indication of a model update, the method further includes one of rebuilding or preparing, at the UE, a new AI / ML model to be applied at the UE, Wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model migration, the method further includes applying the received AI / ML model parameters at the UE, and The monitoring configuration includes resources for monitoring in the time domain, frequency domain or spatial domain, the number of reports and the type of reports for the operation, or conditions for triggering the UE to perform one of the following operations: reporting AI / ML monitoring results, or autonomously performing AI / ML model management and adaptation.

3. The method according to claim 1, further comprising: When the first condition is met, report the AI / ML monitoring results; Requesting, by the UE, AI / ML model management and adaptation information based on the AI / ML model monitoring result, wherein the request for the AI / ML model management and adaptation information specifies an AI / ML model management and adaptation action for the operation; When the second condition is met, autonomously performing AI / ML model management and adaptation actions for the operation at the UE; receiving, at the UE, a request for UE capabilities for AI / ML functions; as well as The UE sends information about the UE capability of the AI / ML function.

4. The method according to claim 3, wherein: The UE capabilities of the AI / ML functions include: Supported AI / ML-based operations, or The type or structure of AI / ML models supported, or the types of training or inference supported, or Supported operations for model management and adaptation.

5. A user equipment (UE) in a wireless communication system, the UE comprising: Transceiver; as well as A processor, the processor being configured to: performing, at the UE, an artificial intelligence / machine learning (AI / ML) monitoring operation based on the received monitoring configuration, the monitoring configuration forming part of a usage configuration for an AI / ML model for the operation, and reporting AI / ML model auxiliary information based on the monitoring configuration, the AI / ML model auxiliary information including AI / ML model monitoring results from the AI / ML monitoring operation, wherein the transceiver is configured to receive AI / ML model management and adaptation information based on the AI / ML model monitoring results, and The processor is further configured to perform AI / ML model management and adaptation operations based on the AI / ML model management and adaptation information.

6. The UE according to claim 5, wherein: The AI / ML model management and adaptation information includes an indication of the AI / ML model management and adaptation action and a parameter characterizing the AI / ML model management and adaptation action, The UE is configured to: when the indication of the AI / ML model management and adaptation action includes an indication of model switching, select an AI / ML model to be applied at the UE from the trained models, The UE is configured to: when the indication of the AI / ML model management and adaptation action includes an indication of model optimization or update, optimize the AI / ML model by retraining using new training data or revalidating using new validation data, or both; The UE is configured to: when the indication of the AI / ML model management and adaptation action includes an indication of a model update, one of rebuilding or preparing a new AI / ML model to be applied at the UE, The UE is configured to: when the indication of the AI / ML model management and adaptation action includes an indication of model migration, apply the received AI / ML model parameters, The monitoring configuration includes resources for monitoring in the time domain, frequency domain, or spatial domain, the number of reports and the type of reports for the operation, or a condition for triggering the UE to perform one of the following operations: reporting AI / ML monitoring results, or autonomously performing AI / ML model management and adaptation, and 7. The UE according to claim 5, wherein: The processor is further configured to perform one of the following operations: When the first condition is met, report the AI / ML monitoring results, requesting AI / ML model management and adaptation information based on the AI / ML model monitoring result, wherein the request for the AI / ML model management and adaptation information specifies an AI / ML model management and adaptation action for the operation, or When the second condition is met, an AI / ML model management and adaptation action for the operation is autonomously performed at the UE.

8. The UE according to claim 5, wherein: The transceiver is further configured to: receiving a request for UE capabilities for AI / ML functions; and Send information about the UE capabilities of the AI / ML function.

9. A method performed by a base station in a wireless communication system, the method comprising: sending, from a base station to a user equipment (UE), a monitoring configuration for monitoring an artificial intelligence / machine learning (AI / ML) monitoring operation, the monitoring configuration forming part of a usage configuration for an AI / ML model for the operation; receiving AI / ML usage assistance information from the UE, the AI / ML usage assistance information comprising an AI / ML model monitoring result from an AI / ML monitoring operation; evaluating a UE-specific usage performance of the AI / ML model for the operation based on the AI / ML model monitoring results; and Determine AI / ML model management and adaptation information corresponding to the AI / ML model monitoring results.

10. The method according to claim 9, further comprising: Sending an indication of an AI / ML model management and adaptation action and parameters characterizing the AI / ML model management and adaptation action to the UE; Sending a request for UE capabilities for AI / ML functions to the UE; as well as Information on UE capability of the AI / ML function is received from the UE at the base station.

11. The base station according to claim 9, wherein: The UE capabilities of the AI / ML functions include: Supported AI / ML-based operations, The type or structure of AI / ML models supported, the types of training or inference supported, or Supported operations for model management and adaptation, wherein the AI / ML model management and adaptation information includes an indication of AI / ML model management and adaptation actions, and wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model switching, the UE selects an AI / ML model to be applied at the UE from the trained models, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model optimization or update, the UE optimizes the AI / ML model by one or both of retraining using new training data or revalidating using new validation data, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of a model update, the UE one of rebuilding or preparing a new AI / ML model to be applied at the UE, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model migration, the UE applies the received AI / ML model parameters, and The monitoring configuration includes resources for monitoring in the time domain, frequency domain or spatial domain, the number of reports and the type of reports for the operation, or conditions for triggering the UE to perform one of the following operations: reporting AI / ML monitoring results, or autonomously performing AI / ML model management and adaptation.

12. A base station in a wireless communication system, the base station comprising: A transceiver, the transceiver being configured to: sending, from a base station to a user equipment (UE), a monitoring configuration for monitoring an artificial intelligence / machine learning (AI / ML) monitoring operation, the monitoring configuration forming part of a usage configuration for an AI / ML model for the operation, and receiving AI / ML usage assistance information from the UE, the AI / ML usage assistance information comprising an AI / ML model monitoring result from the AI / ML monitoring operation; and A processor, the processor being configured to: evaluating the UE-specific usage performance of the AI / ML model for the operation based on the AI / ML model monitoring results, and Determine AI / ML model management and adaptation information corresponding to the AI / ML model monitoring result.

13. The base station according to claim 12, wherein: The transceiver is further configured to: sending an indication of an AI / ML model management and adaptation action and a parameter characterizing the AI / ML model management and adaptation action to the UE, sending a request for UE capabilities for AI / ML functions to the UE; and The information of the UE capability of the AI / ML function is received from the UE at the base station.

14. The base station according to claim 12, wherein: The UE capabilities of the AI / ML functions include: Supported AI / ML-based operations, The type or structure of AI / ML models supported, the types of training or inference supported, or Supported operations for model management and adaptation.

15. The base station according to claim 12, wherein: The AI / ML model management and adaptation information includes indications of AI / ML model management and adaptation actions, and wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model switching, the UE selects an AI / ML model to be applied at the UE from the trained models, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model optimization or update, the UE optimizes the AI / ML model by one or both of retraining using new training data or revalidating using new validation data, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of a model update, the UE one of rebuilding or preparing a new AI / ML model to be applied at the UE, wherein, when the indication of the AI / ML model management and adaptation action includes an indication of model migration, the UE applies the received AI / ML model parameters, and The monitoring configuration includes resources for monitoring in the time domain, frequency domain or spatial domain, the number of reports and the type of reports for the operation, or conditions for triggering the UE to perform one of the following operations: reporting AI / ML monitoring results, or autonomously performing AI / ML model management and adaptation.

Images