Method and apparatus for enabling ai / ML functionalities in wireless networks

Abstract

A User Equipment (UE) for enabling Artificial Intelligence (AI) / Machine Learning (ML) functionalities is provided. The UE receives, from a Base Station (BS), a request for an AI / ML-based UE capability report; transmits, to the BS, the AI / ML-based capability report that indicates at least one AI / ML functionality supported by the UE; receives, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; selects, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE; transmits, to the BS, information of the at least one available functionality; receives, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality; and activates the at least one designated functionality. In addition, a method and a Base Station (BS) are also provided.

Description

METHOD AND APPARATUS FOR ENABLING AI / ML FUNCTIONALITIES IN WIRELESS NETWORKS

[0001] The present disclosure is related to wireless communication and, more specifically, to a User Equipment (UE), Base Station (BS), and method for enabling Artificial Intelligence (AI) / Machine Learning (ML) functionalities in the wireless communication networks.

[0002] Various efforts have been made to improve different aspects of wireless communication for the cellular wireless communication systems, such as the 5thGeneration (5G) New Radio (NR), by improving data rate, latency, reliability, and mobility. The 5G NR system is designed to provide flexibility and configurability to optimize network services and types, accommodating various use cases, such as enhanced Mobile Broadband (eMBB), massive Machine-Type Communication (mMTC), and Ultra-Reliable and Low-Latency Communication (URLLC). As the demand for radio access continues to grow, however, there exists a need for further improvements in the next-generation wireless communication systems, such as improvements in a beam management procedure.Summery of Invention

[0003] The present disclosure is directed to a User Equipment (UE), a Base Station (BS), and a method for enabling Artificial Intelligence (AI) / Machine Learning (ML) functionalities in the wireless communication networks.

[0004] In a first aspect of the present disclosure, a UE for enabling AI / ML functionalities is provided. The UE includes at least one processor and at least one non-transitory computer-readable medium coupled to the at least one processor and storing one or more computer-executable instructions that, when executed by the at least one processor, cause the UE to: receive, from a BS, a request for an AI / ML-based UE capability report; transmit, to the BS, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; receive, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; select, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE; transmit, to the BS, information of the at least one available functionality; receive, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality; and activate the at least one designated functionality.

[0005] In an implementation of the first aspect, the one or more computer-executable instructions, when executed by the at least one processor, further cause the UE to: transmit, to the BS, a deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE; receive, from the BS, a confirmation message corresponding to the deactivation request; and deactivate the at least one functionality to be deactivated.

[0006] In another implementation of the first aspect, the one or more computer-executable instructions, when executed by the at least one processor, further cause the UE to: transmit, to the BS, Hybrid Automatic Repeat reQuest-Acknowledgment (HARQ-ACK) information corresponding to the activation command. The at least one designated functionality is activated no later than a time duration after the last symbol of a Physical Uplink Control Channel (PUCCH) or a Physical Uplink Shared Channel (PUSCH) with the HARQ-ACK information.

[0007] In another implementation of the first aspect, the deactivation request includes UE assistance information (UAI) or an output of an AI / ML model executed by the UE.

[0008] In another implementation of the first aspect, the output indicates that a confidence level is below a preset threshold.

[0009] In another implementation of the first aspect, the UAI includes a deactivation reason of the at least one functionality to be deactivated.

[0010] In another implementation of the first aspect, the deactivation reason includes a remaining battery capacity being below a preset threshold.

[0011] In a second aspect of the present disclosure, a BS for supporting AI / ML functionalities for a UE is provided. The BS includes at least one processor; and at least one non-transitory computer-readable medium coupled to the at least one processor and storing one or more computer-executable instructions that, when executed by the at least one processor, cause the BS to: transmit, to the UE, a request for an AI / ML-based UE capability report; receive, from the UE, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; transmit, to the UE, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; receive, from the UE, information of at least one available functionality selected from the at least one AI / ML functionality supported by the UE; and transmit, to the UE, an activation command for activating at least one designated functionality of the at least one available functionality.

[0012] In an implementation of the second aspect, the one or more computer-executable instructions, when executed by the at least one processor, further cause the BS to: receive, from the UE, a deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE; and transmit, to the UE, a confirmation message corresponding to the deactivation request.

[0013] In another implementation of the second aspect, the one or more computer-executable instructions, when executed by the at least one processor, further cause the BS to: receive, from the UE, HARQ-ACK information corresponding to the activation command. The at least one designated functionality is activated by the UE no later than a time duration after the last symbol of a PUCCH or a PUSCH with the HARQ-ACK information.

[0014] In another implementation of the second aspect, the deactivation request includes UAI or an output of an AI / ML model executed by the UE.

[0015] In another implementation of the second aspect, the output indicates that a confidence level is below a preset threshold.

[0016] In another implementation of the second aspect, the UAI includes a deactivation reason of the at least one functionality to be deactivated.

[0017] In another implementation of the second aspect, the deactivation reason includes a remaining battery capacity being below a preset threshold.

[0018] In a third aspect of the present disclosure, a method performed by a UE for enabling AI / ML functionalities is provided. The method includes: receiving, from a Base Station (BS), a request for an AI / ML-based UE capability report; transmitting, to the BS, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; receiving, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; selecting, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE; transmitting, to the BS, information of the at least one available functionality; receiving, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality; and activating the at least one designated functionality.

[0019] Aspects of the present disclosure are best understood from the following detailed disclosure and the corresponding figures. Various features are not drawn to scale and dimensions of various features may be arbitrarily increased or reduced for clarity of discussion.

[0020] FIG. 1 is a diagram illustrating an Artificial Intelligence (AI) / Machine Learning (ML)-based positioning method, according to an example implementation of the present disclosure.

[0021] FIG. 2 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0022] FIG. 3 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0023] FIG. 4 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0024] FIG. 5 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0025] FIG. 6 is a flowchart illustrating a method / process performed by a user equipment (UE) for enabling AI / ML functionalities, according to an example implementation of the present disclosure.

[0026] FIG. 7 is a block diagram illustrating a node for wireless communication in accordance with various aspects of the present disclosure.

[0027] Some of the abbreviations used in the present disclosure include: Abbreviation　　　　Full name 3GPP　　　　3rd Generation Partnership Project 5G　　　　5th Generation A-CSI　　　　Aperiodic Channel State Information ACK　　　　Acknowledgment AI　　　　Artificial Intelligence ARFCN　　　　Absolute Radio-Frequency Channel Number BA　　　　Bandwidth Adaptation BFR　　　　Beam Failure Recovery BM　　　　Beam Management BS　　　　Base Station BWP　　　　Bandwidth Part CA　　　　Carrier Aggregation CB　　　　CodeBook CC　　　　Component Carrier CCE　　　　Control Channel Element CD-SSB　　　　Cell-Defining Synchronization Signal Block CE　　　　Control Element CIR　　　　Channel Impulse Response CN　　　　Core Network CORESET　　　　Control resource set COT　　　　Channel Occupancy Time CPE　　　　Customer Premises Equipment CRC　　　　Cyclic Redundancy Check C-RNTI　　　　Cell-Radio Network Temporary Identifier CS-RNTI　　　　Configured Scheduling Radio Network Temporary Identifier CSI　　　　Channel State Information CSI-RS　　　　Channel State Information-Reference Signal CSS　　　　Common Search Space DC　　　　Dual Connectivity DCI　　　　Downlink Control Information DCP　　　　DCI with CRC scrambled by PS-RNTI DL　　　　Downlink DL-AOD　　　　Downlink Angle-Of-Departure DL-TDOA　　　　Downlink Time-Difference-Of-Arrival DP　　　　Delay Profile DRX　　　　Discontinuous Reception eNB　　　　Evolved Node B E-UTRA　　　　Evolved Universal Terrestrial Radio Access FDM　　　　Frequency-Division Multiplexing FR　　　　Frequency Range FR1　　　　Frequency Range 1 FR1-2　　　　Frequency Range 1-2 FR2　　　　Frequency Range 2 FR2-2　　　　Frequency Range 2-2 FWA　　　　Fixed Wireless Access GC-PDCCH　　　　Group Common Physical Downlink Control Channel gNB　　　　Next Generation Node B GNSS　　　　Global Navigation Satellite System GSCN　　　　Global Synchronization Channel Number GSO　　　　GeoSynchronous Orbit GW　　　　GateWay HARQ　　　　Hybrid Automatic Repeat Request HARQ-ACK　　　　HARQ Acknowledgement ID　　　　Identifier IE　　　　Information Element IIoT　　　　Industrial Internet of Things LCM　　　　Life Cycle Management LMF　　　　Location Management Function LOS　　　　Line of Sight LPP　　　　Long Term Evolution Positioning Protocol LSB　　　　Least Significant Bit LTE　　　　Long Term Evolution L1 / L2 / L3　　　　Layer 1 / Layer 2 / Layer 3 MAC　　　　Medium Access Control MAC CE　　　　MAC Control Element MCG　　　　Master Cell Group MCS　　　　Modulation and Coding Scheme MCS-C-RNTI　　　　Modulation Coding Scheme Cell Radio Network Temporary Identifier MIB　　　　Master Information Block MIMO　　　　Multiple Input Multiple Output ML　　　　Machine Learning MPE　　　　Maximum Power Extrapolation MSB　　　　Most Significant Bit Msg1　　　　Message 1 MsgA　　　　Message A MsgB　　　　Message B Multi-RTT　　　　Multi-Round-Trip Time multi-TRP　　　　multiple Transmission and Reception Point NACK　　　　Negative Acknowledgment NAS　　　　Non-Access Stratum NCGI　　　　NR Cell Global Identifier NDI　　　　New Data Indicator NGSO　　　　Non-GeoSynchronous Orbit NES　　　　Network Energy Saving NLOS　　　　Non-Line of Sight NG-RAN　　　　Next Generation RAN non-CB　　　　non-CodeBook NR　　　　New Radio NRPPa　　　　New Radio Position Protocol annex NTN　　　　Non-Terrestrial Network NW　　　　Network OAM　　　　Operations, Administration and Maintenance OFDM　　　　Orthogonal Frequency Division Multiplexing OSI　　　　Other SI / On-demand SI OTT　　　　Over the Top PBCH　　　　Physical Broadcast Channel PC　　　　Power Control PCell　　　　Primary Cell PCI　　　　Physical Cell Identity PDCCH　　　　Physical Downlink Control Channel PDP　　　　Power Delay Profile PDSCH　　　　Physical Downlink Shared Channel PDU　　　　Protocol Data Unit PH　　　　Power Headroom PHR　　　　Power Headroom Report PHY　　　　Physical (layer) PL-RS　　　　Path-Loss Reference Signal P-MPR　　　　Power management Maximum Power Reduction PRACH　　　　Physical Random Access Channel PRS　　　　Positioning Reference Signal PRU　　　　Positioning Reference Unit PS　　　　Power Saving PSS　　　　Primary Synchronization Signal PSCell　　　　Primary Secondary Cell PTAG　　　　Primary Timing Advance Group PUCCH　　　　Physical Uplink Control Channel PUSCH　　　　Physical Uplink Shared Channel QCL　　　　Quasi Co-Location RA　　　　Random Access RACH　　　　Random Access Channel RAN　　　　Radio Access Network RAR　　　　Random Access Response Rel　　　　Release RLF　　　　Radio Link Failure RMSI　　　　Remaining Minimum System Information RNA　　　　RAN Notification Area RNAU　　　　RAN Notification Area Update RNTI　　　　Radio Network Temporary Identifier RRC　　　　Radio Resource Control RRM　　　　Radio Resource Management RS　　　　Reference Signal RSRP　　　　Reference Signal Received Power RTT　　　　Round-Trip Time RV　　　　Redundancy Version Rx　　　　Reception SCell　　　　Secondary Cell SCG　　　　Secondary Cell Group SCS　　　　Subcarrier Spacing SDM　　　　Spatial Division Multiplexing SFN　　　　System Frame Number SI　　　　System Information SIB　　　　System Information Block SIB1　　　　System Information Block 1 SINR　　　　Signal to Inference plus Noise Ratio SMTC　　　　SSB Measurement Timing Configuration SpCell　　　　Special Cell SP-CSI　　　　Semi-Persistent Channel State Information SR　　　　Scheduling Request SRS　　　　Sounding Reference Signal SRI　　　　SRS Resource Indicator SS　　　　Synchronization Signal SSB　　　　Synchronization Signal Block SSS　　　　Secondary Synchronization Signal STAG　　　　Secondary Timing Advance Group STxMP　　　　Simultaneous Transmission with Multi-Panels TA　　　　Timing Advance TAG　　　　Timing Advance Group TB　　　　Transport Block TBS　　　　Transport Block Size TCI　　　　Transmission Configuration Indicator TDM　　　　Time Division Multiplexing TNL　　　　Transport Network Layer TPC　　　　Transmission Power Control TPMI　　　　Transmit Precoder Matrix Indication TR　　　　Technical Report TRI　　　　Transmit Rank Indication TS　　　　Technical Specification Tx　　　　Transmission UAI　　　　User Assistance Information UCI　　　　Uplink Control Information UE　　　　User Equipment UL　　　　Uplink UL-AoA　　　　UL-Angle of Arrival UL-TDOA　　　　UL-Time Difference of Arrival URLLC　　　　Ultra-Reliable and Low-Latency Communication USS　　　　UE-Specific Search Space WCDMA　　　　Wideband Code Division Multiple Access WG 　　　　Working Group WI　　　　Working Item WUS　　　　Wake-Up Signal XR　　　　eXtended Reality ZP-CSI-RS　　　　Zero Power CSI-RS

[0028] The following contains specific information related to implementations of the present disclosure. The drawings and their accompanying detailed disclosure are merely directed to implementations. However, the present disclosure is not limited to these implementations. Other variations and implementations of the present disclosure will be obvious to those skilled in the art.

[0029] Unless noted otherwise, like or corresponding elements among the drawings may be indicated by like or corresponding reference numerals. Moreover, the drawings and illustrations in the present disclosure are generally not to scale and are not intended to correspond to actual relative dimensions.

[0030] For the purposes of consistency and ease of understanding, like features may be identified (although, in some examples, not illustrated) by the same numerals in the drawings. However, the features in different implementations may be different in other respects and may not be narrowly confined to what is illustrated in the drawings.

[0031] References to “one implementation,” “an implementation,” “example implementation,” “various implementations,” “some implementations,” “implementations of the present application,” etc., may indicate that the implementation(s) of the present application so described may include a particular feature, structure, or characteristic, but not every possible implementation of the present application necessarily includes the particular feature, structure, or characteristic. Further, repeated use of the phrase “In some implementations,” or “in an example implementation,” “an implementation,” do not necessarily refer to the same implementation, although they may. Moreover, any use of phrases like “implementations” in connection with “the present application” are never meant to characterize that all implementations of the present application must include the particular feature, structure, or characteristic, and should instead be understood to mean “at least some implementations of the present application” includes the stated particular feature, structure, or characteristic. The term “coupled” is defined as connected, whether directly or indirectly through intervening components, and is not necessarily limited to physical connections. The term “comprising,” when utilized, means “including, but not necessarily limited to”; it specifically indicates open-ended inclusion or membership in the so-described combination, group, series, and the equivalent.

[0032] The expression “at least one of A, B and C” or “at least one of the following: A, B and C” means “only A, or only B, or only C, or any combination of A, B and C.” The terms “system” and “network” may be used interchangeably. The term “and / or” is only an association relationship for describing associated objects and represents that three relationships may exist such that A and / or B may indicate that A exists alone, A and B exist at the same time, or B exists alone. The character “ / ” generally represents that the associated objects are in an “or” relationship.

[0033] For the purposes of explanation and non-limitation, specific details, such as functional entities, techniques, protocols, and standards, are set forth for providing an understanding of the disclosed technology. In other examples, detailed disclosure of well-known methods, technologies, systems, and architectures are omitted so as not to obscure the present disclosure with unnecessary details.

[0034] Persons skilled in the art will immediately recognize that any network function(s) or algorithm(s) disclosed may be implemented by hardware, software, or a combination of software and hardware. Disclosed functions may correspond to modules which may be software, hardware, firmware, or any combination thereof.

[0035] A software implementation may include computer executable instructions stored on a computer-readable medium, such as memory or other type of storage devices. One or more microprocessors or general-purpose computers with communication processing capability may be programmed with corresponding executable instructions and perform the disclosed network function(s) or algorithm(s).

[0036] The microprocessors or general-purpose computers may include Application-Specific Integrated Circuits (ASICs), programmable logic arrays, and / or one or more Digital Signal Processor (DSPs). Although some of the disclosed implementations are oriented to software installed and executing on computer hardware, alternative implementations implemented as firmware, as hardware, or as a combination of hardware and software are well within the scope of the present disclosure. The computer-readable medium includes but is not limited to Random Access Memory (RAM), Read Only Memory (ROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), flash memory, Compact Disc Read-Only Memory (CD-ROM), magnetic cassettes, magnetic tape, magnetic disk storage, or any other equivalent medium capable of storing computer-readable instructions.

[0037] A radio communication network architecture such as a Long-Term Evolution (LTE) system, an LTE-Advanced (LTE-A) system, an LTE-Advanced Pro system, or a 5G NR Radio Access Network (RAN) typically includes at least one base station (BS), at least one UE, and one or more optional network elements that provide connection within a network. The UE communicates with the network such as a Core Network (CN), an Evolved Packet Core (EPC) network, an Evolved Universal Terrestrial RAN (E-UTRAN), a 5G Core (5GC), or an internet via a RAN established by one or more BSs.

[0038] A UE may include, but is not limited to, a mobile station, a mobile terminal or device, or a user communication radio terminal. The UE may be a portable radio equipment that includes, but is not limited to, a mobile phone, a tablet, a wearable device, a sensor, a vehicle, or a Personal Digital Assistant (PDA) with wireless communication capability. The UE is configured to receive and transmit signals over an air interface to one or more cells in a RAN.

[0039] The BS may be configured to provide communication services according to at least a Radio Access Technology (RAT) such as Worldwide Interoperability for Microwave Access (WiMAX), Global System for Mobile communications (GSM) that is often referred to as 2G, GSM Enhanced Data rates for GSM Evolution (EDGE) RAN (GERAN), General Packet Radio Service (GPRS), Universal Mobile Telecommunication System (UMTS) that is often referred to as 3G based on basic wideband-code division multiple access (W-CDMA), high-speed packet access (HSPA), LTE, LTE-A, evolved LTE (eLTE) that is LTE connected to 5GC, NR (often referred to as 5G), and / or LTE-A Pro. However, the scope of the present disclosure is not limited to these protocols.

[0040] The BS may include, but is not limited to, a node B (NB) in the UMTS, an evolved node B (eNB) in LTE or LTE-A, a radio network controller (RNC) in UMTS, a BS controller (BSC) in the GSM / GERAN, an ng-eNB in an Evolved Universal Terrestrial Radio Access (E-UTRA) BS in connection with 5GC, a next generation Node B (gNB) in the 5G-RAN, or any other apparatus capable of controlling radio communication and managing radio resources within a cell. The BS may serve one or more UEs via a radio interface. Although the gNB is used as an example in some implementations within the present disclosure, it should be noted that the disclosed implementations may also be applied to other types of base stations.

[0041] The BS may be operable to provide radio coverage to a specific geographical area using multiple cells forming the RAN. The BS may support the operations of the cells. Each cell may be operable to provide services to at least one UE within its radio coverage.

[0042] Each cell (may often referred to as a serving cell) may provide services to one or more UEs within the cell’s radio coverage, such that each cell schedules the DL (and optionally UL resources) to at least one UE within its radio coverage for DL (and optionally UL packet transmissions from the UE). The BS may communicate with one or more UEs in the radio communication system via the cells.

[0043] A cell may allocate sidelink (SL) resources for supporting the Proximity Services (ProSe) or Vehicle to Everything (V2X) services. Each cell may have overlapped coverage areas with other cells.

[0044] In Multi-RAT Dual Connectivity (MR-DC) cases, the primary cell of a Master Cell Group (MCG) or a Secondary Cell Group (SCG) may be referred to as a Special Cell (SpCell). A Primary Cell (PCell) may include the SpCell of an MCG. A Primary SCG Cell (PSCell) may include the SpCell of an SCG. MCG may include a group of serving cells associated with the Master Node (MN), including the SpCell and optionally one or more Secondary Cells (SCells). An SCG may include a group of serving cells associated with the Secondary Node (SN), including the SpCell and optionally one or more SCells.

[0045] As discussed above, the frame structure for NR may support flexible configurations for accommodating various next generation (e.g., 5G) communication requirements, such as Enhanced Mobile Broadband (eMBB), Massive Machine Type Communication (mMTC), and Ultra-Reliable and Low-Latency Communication (URLLC), while fulfilling high reliability, high data rate, and low latency requirements. The Orthogonal Frequency-Division Multiplexing (OFDM) technology in the 3GPP may serve as a baseline for an NR waveform. The scalable OFDM numerology, such as adaptive sub-carrier spacing, channel bandwidth, and Cyclic Prefix (CP), may also be used.

[0046] Two coding schemes may be considered for NR, specifically, Low-Density Parity-Check (LDPC) code and Polar Code. The coding scheme adaption may be configured based on channel conditions and / or service applications.

[0047] At least the DL transmission data, a guard period, and UL transmission data should be included in a transmission time interval (TTI) of a single NR frame. The respective portions of the DL transmission data, the guard period, and the UL transmission data should also be configurable based on, for example, the network dynamics of NR. SL resources may also be provided in an NR frame to support ProSe services or V2X services.

[0048] Any two or more than two of the following paragraphs, (sub)-bullets, points, actions, behaviors, terms, or claims described in the present disclosure may be combined logically, reasonably, and properly to form a specific method.

[0049] Any sentence, paragraph, (sub)-bullet, point, action, behaviors, terms, or claims described in the present disclosure may be implemented independently and separately to form a specific method.

[0050] Dependency, e.g., “based on”, “more specifically”, “preferably”, “in one embodiment”, “in some implementations”, etc., in the present disclosure is just one possible example which would not restrict the specific method.

[0051] In some implementations, all the designs / embodiment / implementations introduced within this disclosure are not limited to be applied for dealing with the problems discussed within this disclosure. For example, the described embodiments may be applied to solve other problems that exist in the RAN of wireless communication systems. In some implementations, all of the numbers listed within the designs / embodiment / implementations introduced within this disclosure are just examples and for illustration, for example, of how the described methods are executed.

[0052] In some implementations, a UE, Positioning Reference Unit (PRU), or target device may receive a request from a gNB, NW, or Location Management Function (LMF) to transmit an Artificial Intelligence (AI) / Machine Learning (ML) capability report for performing AI / ML operations on specific use cases, such as beam management, positioning, Channel State Information (CSI) prediction, or CSI compression.

[0053] In some implementations, the UE, PRU, or target device may transmit the AI / ML capability report to the gNB, NW, or LMF. The AI / ML capability report may indicate which AI / ML-enabled functionalities the UE, PRU, or target device supports. Each AI / ML-enabled functionality may be associated with a legacy positioning method, such as NR Downlink Time-Difference-Of-Arrival (DL-TDOA), NR Downlink Angle of Departure (DL-AoD), NR Multi-Round-Trip Time (Multi-RTT), or NR UL positioning method.

[0054] In some implementations, after the UE, PRU, or target device transmits the AI / ML capability report to the gNB, NW, or LMF, the UE, PRU, or target device may receive a configuration for at least one AI / ML-enabled functionality to perform AI / ML operations.

[0055] In some implementations, if the at least one AI / ML-enabled functionality indicated by the gNB, NW, or LMF is available, the UE, PRU, or target device may inform the gNB, NW, or LMF that the indicated AI / ML-enabled functionality is available.

[0056] In some implementations, the gNB, NW, or LMF may transmit a message to activate the at least one AI / ML-enabled functionality for model inference. The message may include a configuration including the index of the at least one AI / ML-enabled functionality.

[0057] In some implementations, the UE, PRU, or target device may transmit a deactivation request to the gNB, NW, or LMF to terminate the at least one AI / ML-enabled functionality. The deactivation request may be a report of model output indicating that the confidence of the model output is zero, or the deactivation request may be UAI specifying a reason for terminating the at least one AI / ML-enabled functionality. The reason for terminating the at least one AI / ML-enabled functionality may include low battery capacity or overheating. In some implementations, if the UE, PRU, or target device receives a confirmation message from the gNB, NW, or LMF, the UE, PRU, or target device may terminate the at least one AI / ML-enabled functionality.

[0058] In some implementations, to enhance positioning accuracy, AI / ML feature(s) may be applied for positioning in the Cases A to E. Specifically, Case A involves UE-based positioning with a UE-side model, utilizing direct AI / ML positioning, Case B involves UE-assisted / LMF-based positioning with an LMF-side model, utilizing direct AI / ML positioning, Case C involves NG-RAN node-assisted positioning with an LMF-side model, utilizing direct AI / ML positioning, Case D involves UE-assisted / LMF-based positioning with a UE-side model, utilizing AI / ML-assisted positioning, and Case E involves NG-RAN node-assisted positioning with a gNB-side model, utilizing AI / ML-assisted positioning. Among the Cases A to E, two inference categories corresponding to AI / ML model inference may be defined, namely direct AI / ML positioning and AI / ML assisted positioning. Cases A to C may correspond to direct AI / ML positioning while Cases D and E may correspond to AI / ML assisted positioning.

[0059] In some implementations, in the inference category of direct AI / ML positioning, the UE, PRU, or LMF may directly generate location information for the UE based on AI / ML model(s) and / or AI / ML-enabled functionalities.

[0060] FIG. 1 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0061] As illustrated in FIG. 1, for Case A, the UE or PRU 10 may measure the Positioning Reference Signal(s) (PRS(s)) 13 transmitted from one or multiple Transmission Reception Points (TRPs) 11. Based on measurement results related to these PRS(s) 13, the AI / ML model inside the UE or PRU 10 may infer location information 14 for the UE or PRU 10. The measurement results may serve as the model input for the AI / ML model inside the UE or PRU 10 to generate location information 14 for the UE or PRU 10. After the UE or PRU 10 generates the location information 14 for the UE or PRU 10, the UE or PRU 10 may transmit the location information 14 to the LMF 12 via the Long-Term Evolution Positioning Protocol (LPP).

[0062] FIG. 2 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0063] As illustrated in FIG. 2, for Case B, the UE or PRU 20 may measure the PRS(s) 23 transmitted from one or multiple TRPs 21. The UE or PRU may transmit one or more measurement reports 24 to the LMF 22 via the LPP. Based on the measurement results carried in the measurement report(s) 24 transmitted by the UE or PRU 20, the AI / ML model inside the LMF 22 may infer location information for the UE or PRU 20. The measurement results carried in the measurement report(s) 24 may serve as the model input for the AI / ML model inside the LMF 22 to generate location information for the UE or PRU 20.

[0064] FIG. 3 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0065] As illustrated in FIG. 3, for Case C, the gNB or LMF 32 may indicate the UE or PRU 30 to transmit Sounding Reference Signal(s) (SRS(s)) 33 to one or multiple TRPs 31 or the gNB / NW. Each TRP 31 or the gNB / NW may measure the SRS(s) 33. After the TRP(s) 31 or gNB / NW measures the SRS(s) 33 transmitted by the UE or PRU 30, the TRP 31 or gNB may transmit one or more measurement reports 34 to the LMF 32 via the New Radio Position Protocol annex (NRPPa). Based on the measurement results carried in the measurement report(s) 34 transmitted by the TRP(s) 31 or gNB, the AI / ML model inside the LMF 32 may infer location information for the UE or PRU 30. The measurement results carried in the measurement report(s) 34 may serve as the model input for the AI / ML model inside the LMF 32 to generate location information for the UE or PRU 30.

[0066] In some implementations, in the inference category of AI / ML assisted positioning, the UE, PRU, gNB, NW, or TRP may generate one or more pieces of assistance information (e.g., Line of Sight (LOS) / Non-Line of Sight (NLOS) identification, timing, angle of measurement, or likelihood of measurement(s)) based on AI / ML model(s) to assist in deriving location information for the UE or PRU.

[0067] FIG. 4 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0068] As illustrated in FIG. 4, for Case D, the UE or PRU 40 may measure the PRS(s) 43 transmitted from one or multiple TRPs 41. Based on the measurement results related to these PRS(s) 43 (e.g., Channel Impulse Response (CIR), Display Profile (DP), Power Delay Profile (PDP)), the AI / ML model inside the UE or PRU 40 may infer assistance information 44 (e.g., LOS / NLOS identification, timing, angle of measurement(s), or likelihood of measurement(s)). The measurement results may serve as the model input for the AI / ML model inside the UE or PRU 40 to generate the assistance information 44. After the UE or PRU 40 generates the assistance information 44, the UE or PRU 40 may transmit the assistance information 44 to the LMF 42 via the LPP for the LMF 42 to calculate location information for the UE or PRU 40.

[0069] FIG. 5 is a diagram illustrating an AI / ML-based positioning method, according to an example implementation of the present disclosure.

[0070] As illustrated in FIG. 5, for Case E, the gNB or LMF 52 may indicate the UE or PRU 50 to transmit SRS(s) 53 to one or multiple TRPs 51 or the gNB / NW. Each TRP 51 or the gNB / NW may measure the SRS(s) 53. After the TRP(s) 51 or gNB / NW measures the SRS(s) 53 transmitted by the UE or PRU 50, the AI / ML model inside the TRP 51 or gNB may infer assistance information 54 (e.g., LOS / NLOS identification, timing, angle of measurement(s), or likelihood of measurement(s)). After the TRP 51 or gNB generates the assistance information 54, the TRP 51 or gNB may transmit the assistance information 54 to the LMF 52 via the NRPPa for the LMF 52 to calculate location information for the UE or PRU 50.

[0071] In some implementations, when applying AI / ML technologies for positioning to calculate location information for the UE or PRU, different entities (e.g., the UE, PRU, TRP, gNB / NW, or LMF) may be equipped with AI / ML model(s) depending on specific use cases (e.g., Case A, Case B, Case C, Case D, and Case E). The model output of AI / ML model(s) may include location information for the UE or PRU and / or assistance information (e.g., LOS / NLOS indicator) used for calculating location information for the UE or PRU. To support these diverse use cases, managing AI / ML model / functionality (e.g., activation, deactivation, switching, fallback) may present a challenge for AI / ML-based positioning. The purpose of AI / ML-based positioning is to enhance positioning accuracy. If the AI / ML model or functionality becomes unsuitable for positioning (e.g., the performance of the model output degrades), determining how the UE or PRU reacts or how the gNB / NW or LMF informs the UE or PRU of such conditions is a critical issue. The corresponding procedure may be designed based on different positioning use cases.

[0072] In some implementations, when applying AI / ML technologies for positioning, managing AI / ML models or functionalities, such as activation, deactivation, switching, or fallback, may be a critical issue. Immediately deactivating unsuitable AI / ML models or functionalities may ensure that AI / ML-based positioning provides better accuracy than legacy positioning methods, such as DL-TDOA or UL-TDOA. For model management, based on diverse positioning use cases, AI / ML models may be managed at different entities, such as the UE, PRU, TRP, gNB, or LMF.

[0073] In some implementations, an AI / ML-enabled functionality may be mapped to a legacy positioning method (e.g., NR DL-TDOA, NR DL-AoD, NR Multi-RTT, or NR UL positioning method). In some implementations, an AI / ML-enabled functionality may be mapped to multiple legacy positioning methods. Each Information Element (IE), such as ProvideCapabilities message, corresponding to a legacy positioning method, may include at least one of the following: 　　　　- whether the UE or PRU supports applying AI / ML operation for the legacy positioning method; and 　　　　- the positioning accuracy when applying AI / ML operation for the legacy positioning method.

[0074] In some implementations, an AI / ML-enabled functionality may not be mapped to any legacy positioning method. The UE or PRU may determine which legacy positioning method to select for the AI / ML-enabled functionality. Specifically, an IE may include at least the information of the positioning accuracy when applying AI / ML operation for the selected legacy positioning method.

[0075] In some implementations, an AI / ML-enabled functionality or feature may be enabled to be performed by one or more AI / ML models. One or more AI / ML-enabled functionalities or feature groups may be associated with one or more use cases, such as beam management, positioning, CSI prediction, or CSI compression. If an AI / ML-enabled functionality is supported by a UE or PRU, the UE or PRU is capable of supporting the AI / ML-enabled functionality. If an AI / ML-enabled functionality is activated, the AI / ML-enabled functionality is indicated to perform model inference. If an AI / ML-enabled functionality is deactivated, the AI / ML-enabled functionality is indicated not to perform model inference, and the deactivated AI / ML-enabled functionality may return to the model training phase or be terminated. For example, if an activated AI / ML-enabled functionality is deactivated due to poor performance, the deactivated AI / ML-enabled functionality may return to the model training phase or be terminated. For another example, if an activated AI / ML-enabled functionality is deactivated due to low battery capacity, the deactivated AI / ML-enabled functionality may be terminated. In some implementations, the gNB or LMF may configure whether the training phase continues, is re-performed, or is terminated when an AI / ML-enabled functionality is deactivated. If an AI / ML-enabled functionality is available, the model training or testing of the AI / ML-enabled functionality is successful, meaning the AI / ML model is ready for model inference.

[0076] In some implementations, if the gNB, NW, or LMF indicates the UE to perform AI / ML operation for specific use cases, such as beam management, positioning, CSI prediction, or CSI compression, the UE may receive a list of AI / ML-enabled functionalities from the gNB or NW via an RRC message or a MAC CE, or from the LMF via an LPP message. The list of AI / ML-enabled functionalities may be associated with a specific use case or with multiple use cases. After the UE receives the list of AI / ML-enabled functionalities, the UE may transmit a corresponding response to the gNB or NW via a MAC CE or PUCCH, or to the LMF via an LPP message. The corresponding response may indicate which AI / ML-enabled functionalities the UE supports. The UE may initiate an SR request or a RACH procedure to request UL resources to transmit the response after receiving the list of AI / ML-enabled functionalities.

[0077] In some implementations, the gNB, NW, or LMF may transmit an instruction to the UE to initiate UE-supported AI / ML-enabled functionalities. The instruction may specify which UE-supported AI / ML-enabled functionalities are initiated. For example, the UE may receive a MAC CE carrying the instruction from the gNB or NW, where the MAC CE may include at least one of the following: 　　　　- which UE-supported AI / ML-enabled functionalities are initiated, indicated by an identifier / index of the AI / ML model / functionality or a bitmap where each bit corresponds to an AI / ML-enabled functionality (e.g., with a value of ‘1’ indicating initiation); and 　　　　- a threshold used to determine whether the initiated UE-supported AI / ML-enabled functionalities are available after model training.

[0078] The UE may be configured with a timer via RRC signaling. When the UE receives the MAC CE carrying the AI / ML-enabled functionality initiation instruction, the UE may start the timer. If no initiated UE-supported AI / ML-enabled functionalities pass the test (e.g., the performance is below the threshold indicated in the MAC CE or RRC message) before the timer expires, the UE may determine that these functionalities are unavailable. The UE may then report which initiated UE-supported AI / ML-enabled functionalities are unavailable or available and terminate the unavailable functionalities. The UE may initiate an SR request or RACH procedure to request UL resources to transmit the report to the gNB or NW.

[0079] In some implementations, the UE may receive an LPP message carrying the instruction from the LMF, where the LPP message may include at least one of the following: 　　　　- which UE-supported AI / ML-enabled functionalities are initiated; and 　　　　- a threshold used to determine whether the initiated UE-supported AI / ML-enabled functionalities are available after model training.

[0080] The UE may be configured with a timer via LPP. When the UE receives the LPP message carrying the AI / ML-enabled functionality initiation instruction, the UE may start the timer. If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance is below the threshold indicated in the LPP message or RRC message) before the timer expires, the UE may determine that these functionalities are unavailable. The UE may then report which initiated UE-supported AI / ML-enabled functionalities are unavailable or available via LPP and terminate the unavailable functionalities.

[0081] In some implementations, the gNB, NW, or LMF may transmit one or more configurations to the UE to initiate UE-supported AI / ML-enabled functionalities. Each configuration may be associated with at least one UE-supported AI / ML-enabled functionality. For example, each configuration may include one or more indices related to the UE-supported AI / ML-enabled functionalities, where the index may be a functionality index or an association index. The association index may associate measurement configuration and report configuration for at least one AI / ML-enabled functionality and / or associate the functionality with a ground truth label dataset. The configuration may implicitly indicate which UE-supported AI / ML-enabled functionalities are initiated.

[0082] In some implementations, the UE may receive one or more configurations from the gNB or NW via RRC signaling. Each configuration may be associated with at least one AI / ML-enabled functionality and may include at least one of the following: 　　　　- one index associated with an AI / ML-enabled functionality, which may be a functionality index or an association index; 　　　　- a list of indices associated with multiple AI / ML-enabled functionalities, where each index may be a functionality index or an association index; 　　　　- PRS configuration(s) corresponding to at least one AI / ML-enabled functionality; 　　　　- the type of reporting for model output; 　　　　- ground truth label dataset index; 　　　　- a threshold used for model monitoring; 　　　　- a counter (maximum number of counter iterations) used for model monitoring; 　　　　- a TRP index; and 　　　　- a time duration for model training.

[0083] If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance is below the threshold indicated in the RRC message) during the configured time duration (from the slot, subframe, or frame when the UE receives the configuration), the UE may determine that these functionalities are unavailable. The UE may then report which initiated UE-supported AI / ML-enabled functionalities are unavailable or available and terminate the unavailable functionalities. The UE may initiate an SR request or RACH procedure to request UL resources to transmit the report to the gNB or NW.

[0084] In some implementations, the UE may receive one or more configurations from the LMF via LPP. Each configuration may be associated with at least one AI / ML-enabled functionality and may include at least one of the following: 　　　　- one index associated with an AI / ML-enabled functionality, which may be a functionality index or an association index; 　　　　- a list of indices associated with multiple AI / ML-enabled functionalities, where each index may be a functionality index or an association index; 　　　　- PRS configuration(s) corresponding to at least one AI / ML-enabled functionality; 　　　　- the type of reporting for model output; 　　　　- ground truth label dataset index; 　　　　- a threshold used for model monitoring; 　　　　- a counter (maximum number of counter iterations) used for model monitoring; 　　　　- a TRP index; and 　　　　- a time duration for model training.

[0085] If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance is below the threshold indicated in the LPP message) during the configured time duration (from the slot, subframe, or frame when the UE receives the configuration), the UE may determine that these functionalities are unavailable. The UE may then report which initiated UE-supported AI / ML-enabled functionalities are unavailable or available via LPP and terminate the unavailable functionalities.

[0086] In some implementations, a UE may report the AI / ML-enabled functionalities that the UE supports. The report may be a UE capability report or an AI / ML capability report which is different from the UE capability report. The UE may transmit the report, indicating the AI / ML-enabled functionalities that the UE supports, to the LMF via an LPP message, or to the gNB or NW via a MAC CE, PUCCH, or RRC message.

[0087] In some implementations, the gNB, NW, or LMF may transmit an instruction to the UE to initiate UE-supported AI / ML-enabled functionalities. The instruction may specify which UE-supported AI / ML-enabled functionalities are initiated.

[0088] In some implementations, the UE may receive a MAC CE carrying the instruction from the gNB or NW. The MAC CE may include at least one of the following: 　　　　- which UE-supported AI / ML-enabled functionalities are initiated; and 　　　　- a threshold used to determine whether the initiated UE-supported AI / ML-enabled functionalities are available after model training.

[0089] The UE may be configured with a timer via RRC signaling. When the UE receives the MAC CE carrying the AI / ML-enabled functionality initiation instruction, the UE may start the timer. If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance of an AI / ML-enabled functionality is below the threshold indicated in the MAC CE or RRC message) before the timer expires, the UE may determine that the failed functionalities are unavailable. The UE may then report the unavailable or available initiated UE-supported AI / ML-enabled functionalities. The UE may terminate the unavailable initiated UE-supported AI / ML-enabled functionalities. The UE may initiate an SR request or a RACH procedure to request UL resources to transmit the report to the gNB or NW.

[0090] In some implementations, the UE may receive an LPP message carrying the instruction from the LMF. The LPP message may include at least one of the following: 　　　　- which UE-supported AI / ML-enabled functionalities are initiated; and 　　　　- a threshold used to determine whether the initiated UE-supported AI / ML-enabled functionalities are available after model training.

[0091] The UE may be configured with a timer via LPP. When the UE receives the LPP message carrying the AI / ML-enabled functionality initiation instruction, the UE may start the timer. If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance of an AI / ML-enabled functionality is below the threshold indicated in the LPP message or RRC message) before the timer expires, the UE may determine that the failed functionalities are unavailable. The UE may then report the unavailable or available initiated UE-supported AI / ML-enabled functionalities via LPP. The UE may terminate the unavailable initiated UE-supported AI / ML-enabled functionalities.

[0092] In some implementations, the gNB, NW, or LMF may transmit one or more configurations to the UE to initiate UE-supported AI / ML-enabled functionalities. Each configuration may be associated with at least one UE-supported AI / ML-enabled functionality. For example, each configuration may include one or more indices related to the UE-supported AI / ML-enabled functionalities, where the index may be a functionality index or an association index. The association index may associate measurement configuration and report configuration for at least one AI / ML-enabled functionality or associate the functionality with a ground truth label dataset. The configuration may implicitly indicate which UE-supported AI / ML-enabled functionalities are initiated.

[0093] In some implementations, the UE may receive one or more configurations from the gNB or NW via RRC signaling. Each configuration may be associated with at least one AI / ML-enabled functionality and may include at least one of the following: 　　　　- one index associated with an AI / ML-enabled functionality, which may be a functionality index or an association index; 　　　　- a list of indices associated with multiple AI / ML-enabled functionalities, where each index may be a functionality index or an association index; 　　　　- PRS configuration(s) corresponding to at least one AI / ML-enabled functionality; 　　　　- the type of reporting for model output; 　　　　- ground truth label dataset index; 　　　　- a threshold used for model monitoring; 　　　　- a counter (maximum number of counter iterations) used for model monitoring; 　　　　- a TRP index; and 　　　　- a time duration for model training.

[0094] If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance of an AI / ML-enabled functionality is below the threshold indicated in the RRC message) during the configured time duration (from the slot, subframe, or frame when the UE receives the configuration), the UE may determine that the failed functionalities are unavailable. The UE may then report the unavailable or available initiated UE-supported AI / ML-enabled functionalities. The UE may terminate the unavailable initiated UE-supported AI / ML-enabled functionalities. The UE may initiate an SR request or a RACH procedure to request UL resources to transmit the report to the gNB or NW.

[0095] In some implementations, the UE may receive one or more configurations from the LMF via LPP. Each configuration may be associated with at least one AI / ML-enabled functionality and may include at least one of the following: 　　　　- one index associated with an AI / ML-enabled functionality, which may be a functionality index or an association index; 　　　　- a list of indices associated with multiple AI / ML-enabled functionalities, where each index may be a functionality index or an association index; 　　　　- PRS configuration(s) corresponding to at least one AI / ML-enabled functionality; 　　　　- the type of reporting for model output; 　　　　- ground truth label dataset index; 　　　　- a threshold used for model monitoring; 　　　　- a counter (maximum number of counter iterations) used for model monitoring; 　　　　- a TRP index; and 　　　　- a time duration for model training.

[0096] If any initiated UE-supported AI / ML-enabled functionalities fail the test (e.g., the performance of an AI / ML-enabled functionality is below the threshold indicated in the LPP message) during the configured time duration (from the slot, subframe, or frame when the UE receives the configuration), the UE may determine that the failed functionalities are unavailable. The UE may then report the unavailable or available initiated UE-supported AI / ML-enabled functionalities via LPP. The UE may terminate the unavailable initiated UE-supported AI / ML-enabled functionalities.

[0097] Two cases described above may apply to a UE-side model, namely Case A: “UE-based positioning with a UE-side model, direct AI / ML positioning,” and Case D: “UE-assisted / LMF-based positioning with a UE-side model, AI / ML-assisted positioning.”

[0098] For a UE-side model, the AI / ML model corresponding to different AI / ML-enabled functionalities may be trained at the UE. When the AI / ML model or AI / ML-enabled functionality becomes available, an entity (e.g., the UE, gNB, NW, or LMF) may activate the AI / ML model or AI / ML-enabled functionality to perform model inference. After the entity activates the AI / ML model or AI / ML-enabled functionality for model inference, the entity may monitor the performance of the activated AI / ML model or AI / ML-enabled functionality. If the performance of the activated AI / ML model or AI / ML-enabled functionality degrades (e.g., the model output performance falls below a configured threshold), the entity (e.g., the UE, PRU, gNB, NW, or LMF) may deactivate the AI / ML model or AI / ML-enabled functionality. After deactivation, the entity may terminate the AI / ML model or AI / ML-enabled functionality and revert to legacy positioning methods (e.g., DL-TDOA, DL-AoD, DL-Enhanced Cell Identity (DL-ECID), or UL-AoA) or return the deactivated AI / ML model or AI / ML-enabled functionality to the model training phase. In some implementations, a counter may be configured with AI / ML-enabled functionality configurations and initialized to zero upon receiving an activation command. The counter increments by one if a configured number of continuous monitored or predicted results (e.g., N samples, where N is configurable) falls within a low-confidence level. When the counter reaches a configured maximum value, the entity may transmit a deactivation request to relevant entities. The counter may reset upon receiving a reconfiguration of AI / ML-enabled functionalities or any existing trigger enforced MAC reset.

[0099] In some implementations, to apply one or more UE-side AI / ML-enabled functionalities or AI / ML models for positioning, at least one of the following steps (A) to (E) is necessary:

[0100] (A) UE-supported AI / ML functionalities or models reporting.

[0101] In some implementations, an entity (e.g., the gNB, NW, or LMF) may request the UE or PRU to transmit the AI / ML-enabled functionalities or models supported by the UE or PRU.

[0102] In some implementations, the UE or PRU may receive a Request message (e.g., RequestCapabilities message or RequestAIMLCapabilities message) from the LMF. The UE or PRU may generate a response message (e.g., ProvideCapabilities message or ProvideAIMLCapabilities message) as a response.

[0103] For example, the target device (e.g., UE / PRU) may: 1> for each (AI / ML-enabled) functionality for which a request for AI / ML capabilities is included in the received message: 　　　　2> if the target device (e.g., UE / PRU) supports this (AI / ML-enabled) functionality for AI / ML operation (or enables this (AI / ML-enabled) functionality for AI / ML operation): 　　　　　　　　3> include the AI / ML capabilities of the device for that AI / ML-enabled / supported functionality in the response message; 1> set the IE LPP-TransactionID in the response message to the same value as the IE LPP-TransactionID in the received message; 1> deliver the response message to lower layers for transmission.

[0104] In some implementations, the UE or PRU may receive a Request message (e.g., RequestCapabilities message) from the LMF. The UE or PRU may generate a response message (e.g., ProvideCapabilities message) as a response.

[0105] For example, the target device (e.g., UE / PRU) may: 1> for each positioning method for which a request for capabilities is included in the received message: 　　　　2> if the target device (e.g., UE / PRU) supports this positioning method: 　　　　　　　　3> include the capabilities of the device for that supported positioning method in the response message; 2> set the IE LPP-TransactionID in the response message to the same value as the IE LPP-TransactionID in the received message; 2> deliver the response message to lower layers for transmission.

[0106] The IE corresponding to each positioning method (e.g., NR-ECID-ProvideCapabilities, NR-DL-TDOA-ProvideCapabilities, NR-DL-AoD-ProvideCapabilities, or NR-UL-ProvideCapabilities) may include a field indicating whether the positioning method (e.g., NR ECID, NR DL-TDOA, NR DL-AoD, NR Multi-RTT, or NR UL positioning method) supports or enables AI / ML operation. For example, the IE corresponding to NR-ECID may include a field indicating whether NR-ECID supports or enables AI / ML operation (e.g., AI-ML-operation ENUMERATED {supported} or AI-ML-operation ENUMERATED {enabled}). For example, the IE corresponding to DL-TDOA / AoD may include a field indicating whether DL-TDOA / AoD supports or enables AI / ML operation (e.g., AI-ML-operation ENUMERATED {supported} or AI-ML-operation ENUMERATED {enabled}). For example, the IE corresponding to Multi-RTT may include a field indicating whether Multi-RTT supports or enables AI / ML operation (e.g., AI-ML-operation ENUMERATED {supported} or AI-ML-operation ENUMERATED {enabled}). For example, the IE corresponding to UL positioning method(s) may include a field indicating whether UL positioning method(s) support or enable AI / ML operation (e.g., AI-ML-operation ENUMERATED {supported} or AI-ML-operation ENUMERATED {enabled}). In some implementations, if the UE or PRU does not enable AI / ML operation for a positioning method (e.g., NR DL-TDOA, NR ECID, NR DL-AoD, NR Multi-RTT or NR UL positioning method), the UE or PRU may not set the field (e.g., AI-ML-operation ENUMERATED {enabled}) in the IE for that positioning method to enabled. Even if the UE or PRU supports AI / ML operation for a positioning method, the UE or PRU may not enable AI / ML operation for that method.

[0107] In some implementations, the gNB or NW may request the UE or PRU to report the AI / ML capabilities. The UE or PRU may transmit a report to the gNB or NW. The report may include an RRC IE (e.g., UE-NR-Capability) to convey NR UE Radio Access Capability Parameters. The RRC IE may include a field (e.g., AI-ML-Parameters or phy-Parameters) to convey information related to AI / ML operation, which may include information whether the UE or PRU supports specific AI / ML functionalities or models (e.g., functionality1 ENUMERATED {supported}, functionality2 ENUMERATED {supported}, functionality3 ENUMERATED {supported}, model1 ENUMERATED {supported}, model2 ENUMERATED {supported}).

[0108] In some implementations, the UE may report additional conditions for the AI / ML-enabled functionalities supported by the UE, including at least one of the following: 　　　　- the expected accuracy of the AI / ML-enabled functionalities; and 　　　　- the latency of the AI / ML-enabled functionalities, for example, defined as the time gap between the initiation of the AI / ML-enabled functionality and the generation of the model output by the UE.

[0109] In some implementations, the UE or PRU may transmit a report of supported AI / ML functionalities or models to an entity (e.g., the gNB, NW, or LMF) without a request from the entity.

[0110] In some implementations, when triggered to transmit a response message, the target device (e.g., UE / PRU may: 1> for each (AI / ML-enabled) functionality which is enabled to perform AI / ML operation or which supports AI / ML operation: 　　　　2> set the corresponding IE to include the device’s AI / ML capabilities; 1> deliver the response message to lower layers for transmission

[0111] In some implementations, when triggered to transmit a response message, the target device (e.g., UE / PRU may: 1> for each positioning method whose capabilities are to be indicated: 　　　　2> set the corresponding IE to include the device’s capabilities; 　　　　2> if OTDOA capabilities are to be indicated: 　　　　　　　　3> include the IE supportedBandListEUTRA; 1> deliver the response message to lower layers for transmission.

[0112] The corresponding IE may include a field indicating whether the positioning method (e.g., NR-ECID, NR DL-TDOA, NR DL-AoD, NR Multi-RTT, or NR UL positioning method) supports or enables AI / ML operation.

[0113] In some implementations, the UE or PRU may transmit a report to the gNB or NW without a request from the gNB or NW. The report may include an RRC IE (e.g., UE-NR-Capability) to convey NR UE Radio Access Capability Parameters. The RRC IE may include a field (e.g., AI-ML-Parameters or phy-Parameters) to convey information related to AI / ML operation, such as whether the UE or PRU supports specific AI / ML functionalities or models.

[0114] (B) AI / ML functionalities / models training.

[0115] In some implementations, the UE or PRU may train the AI / ML model or models corresponding to one or more AI / ML-enabled functionalities reported to the gNB, NW, or LMF, based on one or more configurations transmitted by the gNB, NW, or LMF.

[0116] In some implementations, the UE or PRU may receive one or more configurations, each associated with one or more AI / ML-enabled functionalities. For example, each configuration ID may correspond to an AI / ML-enabled functionality ID. For example, each configuration ID may include a field indicating an AI / ML-enabled functionality ID or a list of AI / ML-enabled functionality IDs. Each configuration may serve as the configuration for measurement or measurement reporting. Each configuration may be used for AI / ML model training corresponding to one or more AI / ML models associated with one or more AI / ML-enabled functionalities. The one or more configurations may be transmitted by the gNB, NW, or LMF.

[0117] The UE or PRU may notify the gNB, NW, or LMF that the training is completed to facilitate the activation of AI / ML-based positioning for inference output. The notification may be performed via at least one of the following options (a) to (c).

[0118] Option (a): An existing measurement report configuration. For instance, if the UE or PRU transmits a report with a specific value or a reserved value, the gNB, NW, or LMF may determine that the training is ready for inference.

[0119] Option (b) A new additional report configuration that the UE may be configured to indicate the training result and corresponding parameters (e.g., confidence level of training, resolution level of training).

[0120] Option (c) An L1 or L2 indication transmitted via UCI or UL MAC CE.

[0121] If the UE or PRU does not notify the gNB, NW, or LMF of the completion of training after a defined or configured period, the gNB, NW, or LMF may assume the AI / ML functionality is disabled, even if the UE previously indicated support or enablement.

[0122] If the UE or PRU does not notify the gNB, NW, or LMF of the completion of training after a defined or configured period, the gNB, NW, or LMF may transmit deactivation signaling to cancel the support of AI / ML-based functionality for positioning.

[0123] (C) AI / ML functionalities or models testing.

[0124] In some implementations, the UE or PRU may test whether the AI / ML model corresponding to one or more AI / ML-enabled functionalities reported to the gNB, NW, or LMF is available for model inference (e.g., for positioning).

[0125] In some implementations, if the AI / ML model corresponding to one or more AI / ML-enabled functionalities supported by the UE or PRU passes the testing (e.g., the performance of the model output of the AI / ML model corresponding to the one or more AI / ML-enabled functionalities exceeds a configured threshold), the UE or PRU may inform the gNB, NW, or LMF of the available AI / ML-enabled functionalities.

[0126] In some implementations, the UE or PRU may report the status of AI / ML models corresponding to one or more AI / ML-enabled functionalities supported by the UE or PRU to the gNB or NW.

[0127] For example, when the status of AI / ML models corresponding to one or more AI / ML-enabled functionalities supported by the UE or PRU is available, the UE or PRU may transmit an SR request to the gNB or NW or perform a RACH procedure for the UL transmission corresponding to a status report related to each AI / ML-enabled functionality supported by the UE or PRU. The status report may indicate the AI / ML-enabled functionalities that are ready to use or available. The status report may indicate the priority of each AI / ML-enabled functionality available for model inference or the confidence of each AI / ML-enabled functionality available for model inference. The status report may include information about the available AI / ML-enabled functionalities that can be used simultaneously. After the UE or PRU transmits the status report, the UE or PRU may expect to receive a response from the gNB, NW, or LMF. The response may indicate the available AI / ML-enabled functionality or functionalities to be used for model inference by the UE or activate the available AI / ML-enabled functionality or functionalities. The response may be a DCI or MAC CE transmitted by the gNB or NW. The DCI or MAC CE may include one or more fields indicating the available AI / ML-enabled functionality or functionalities to be activated. The response may be an LPP message transmitted by the LMF. The LPP message may include one or more fields indicating the available AI / ML-enabled functionality or functionalities to be activated. If the response indicates that no available AI / ML-enabled functionalities can be applied for model inference by the UE or PRU, the UE or PRU may fall back to non-AI / ML operation (e.g., legacy positioning methods), terminating all AI / ML-enabled functionalities. Alternatively, if the response indicates that no available AI / ML-enabled functionalities can be applied for model inference by the UE or PRU, the UE may request additional time for model training via the response.

[0128] For example, the response may include information indicating that the UE requires additional time for model training. For another example, the response may include information specifying the amount of time the UE requires for model training. The UE may then receive an instruction transmitted from the gNB, NW, or LMF. The instruction may direct the UE to continue performing model training for the AI / ML-enabled functionalities or to terminate all AI / ML-enabled functionalities. If the instruction directs the UE to perform model training for the AI / ML-enabled functionalities, the instruction may specify a timing window for performing model training. If the UE has no available AI / ML-enabled functionalities during the timing window, the UE may terminate all AI / ML-enabled functionalities. If the UE has any available AI / ML-enabled functionalities during the timing window, the UE may inform the gNB, NW, or LMF of the available AI / ML-enabled functionalities.

[0129] In some implementations, the UE may be configured with a timer via an RRC message or LPP message. When the timer expires, the UE may transmit a response to the gNB, NW, or LMF indicating the status of AI / ML-enabled functionalities (e.g., whether the AI / ML-enabled functionalities are available). If the response indicates that no available AI / ML-enabled functionalities can be applied for model inference by the UE or PRU, the response may indicate at least one legacy positioning method (e.g., DL ECID, DL-TDOA, or DL-AoD) to the UE or PRU. The UE or PRU may fall back to the indicated legacy positioning method or methods. For example, the UE or PRU may receive one or more configurations corresponding to different positioning methods, and the UE or PRU may perform AI / ML-enabled functionalities based on the one or more configurations. If the UE or PRU has no available AI / ML-enabled functionalities when the configured timer expires, the UE may transmit a response message to the gNB, NW, or LMF indicating the configuration to be used to apply a legacy positioning method. The UE or PRU may apply the legacy positioning method (e.g., NR ECID, NR DL TDOA, NR DL-AoD, NR Multi-RTT, or NR UL positioning method) associated with the indicated configuration for non-AI / ML-based positioning. The UE or PRU may indicate the required duration to complete the AI / ML training (e.g., to make the AI / ML-enabled functionalities available), and the gNB, NW, or LMF may configure the timer accordingly. The indication may be transmitted via UAI or together with the supporting report.

[0130] For example, when the status of at least one AI / ML model corresponding to one or more AI / ML-enabled functionalities supported by the UE is available, the UE or PRU may transmit an SR request to the gNB or NW or perform a RACH procedure for the UL transmission corresponding to an activation request. The activation request may inform the gNB, NW, or LMF of the available AI / ML-enabled functionality or functionalities that the UE intends to activate. After the UE or PRU transmits the activation request to the gNB, NW, or LMF, the UE may expect to receive a response from the gNB, NW, or LMF. The response may be a DCI or MAC CE transmitted by the gNB or NW. The DCI or MAC CE may include a field indicating whether to activate the one or more available AI / ML-enabled functionalities requested by the UE or PRU for model inference. The response may be an LPP message transmitted by the LMF. The LPP message may include a field indicating whether to activate the one or more available AI / ML-enabled functionalities requested by the UE or PRU for model inference.

[0131] In some implementations, if the UE or PRU receives a response indicating that the UE or PRU cannot activate the one or more available AI / ML-enabled functionalities requested, the UE or PRU may fall back to non-AI / ML operation (e.g., legacy positioning methods or the legacy positioning method indicated in the response). The response may indicate a legacy positioning method (e.g., DL ECID, DL-TDOA, or DL-AoD), implying that the gNB, NW, or LMF directs the UE or PRU to fall back to non-AI / ML operation. Alternatively, the response may include the configuration corresponding to the legacy positioning method indicated by the gNB, NW, or LMF. When the UE or PRU receives the response, the UE may release the configurations corresponding to the AI / ML-enabled functionalities.

[0132] In some implementations, if the UE or PRU receives a response indicating that the UE or PRU cannot activate the one or more available AI / ML-enabled functionalities requested, the UE may retransmit an SR request to the gNB or NW or perform a RACH procedure for the UL transmission corresponding to an activation request to request activation of another available AI / ML-enabled functionality or functionalities. If no other available AI / ML-enabled functionalities exist, the UE or PRU may fall back to non-AI / ML operation (e.g., legacy positioning methods) and inform the gNB, NW, or LMF of the fallback to non-AI / ML operation.

[0133] For example, when the status of at least one AI / ML model corresponding to one or more AI / ML-enabled functionalities supported by the UE is available, the UE or PRU may automatically activate the available AI / ML-enabled functionalities supported by the UE. The UE or PRU may transmit a termination request to the gNB, NW, or LMF to terminate unavailable AI / ML-enabled functionalities supported by the UE.

[0134] In some implementations, the number of available AI / ML-enabled functionalities that the UE can activate may exceed one. In some implementations, the UE or PRU may report the maximum number of AI / ML-enabled functionalities supported by the UE or PRU to the gNB, NW, or LMF in the AI / ML capability report. In some implementations, the UE or PRU may report the maximum number of AI / ML-enabled functionalities that can be activated simultaneously to the gNB, NW, or LMF in the AI / ML capability report. In some implementations, the gNB, NW, or LMF may activate AI / ML-enabled positioning without designating the functionality, leaving the UE or PRU to determine the one or more functionalities to activate.

[0135] (D) AI / ML functionalities or models inference.

[0136] In some implementations, the UE or PRU may receive a message from the gNB, NW, or LMF indicating the AI / ML-enabled functionality or functionalities to activate for model inference. The message transmitted by the gNB or NW may be a DCI or a MAC CE. The message transmitted by the LMF may be an LPP message.

[0137] If the available AI / ML-enabled functionality or functionalities are indicated by a MAC CE for activation for model inference, the MAC CE may include information related to the configuration used for model inference (e.g., the content of model output for reporting, the periodicity of PRS used for model input, the pattern of PRS used for model input, the type of assistance information corresponding to the model output such as confidence of model output, the threshold used for model monitoring). For example, a functionality ID may be appended in the MAC CE, and the UE may apply the corresponding configuration to perform inference generation and reporting.

[0138] If the available AI / ML-enabled functionality or functionalities are indicated by an LPP message for activation for model inference, the LPP message may include information related to the configuration used for model inference (e.g., the content of model output for reporting, the periodicity of PRS used for model input, the pattern of PRS used for model input, the type of assistance information corresponding to the model output such as confidence of model output, the threshold used for model monitoring).

[0139] If the available AI / ML-enabled functionality or functionalities are indicated by a MAC CE, a DCI, or an LPP message for activation for model inference, the model inference corresponding to the indicated available AI / ML-enabled functionality or functionalities may be performed based on the received configuration corresponding to the indicated available AI / ML-enabled functionality or functionalities. The configuration of each AI / ML-enabled functionality supported by the UE or PRU may be received before receiving the DCI, MAC CE, or LPP message activating the available AI / ML-enabled functionality or functionalities for model inference. The configuration may be a report configuration used for model output or a measurement configuration used for model input. For example, the activation of available AI / ML-enabled functionality or functionalities may be configured by RRC signaling, which may further include a configuration or reconfiguration of report configuration used for model inference output.

[0140] For the available AI / ML-enabled functionality or functionalities not indicated in the message for performing model inference, the UE or PRU may terminate the functionality or functionalities. The UE or PRU may further suspend the training and testing of the corresponding AI / ML-enabled functionality or functionalities.

[0141] In some implementations, after model training and testing, the UE or PRU may decide to activate one or more AI / ML-enabled functionalities for model inference (e.g., for positioning).

[0142] In some implementations, when the UE or PRU decides to activate one or more AI / ML-enabled functionalities, the UE or PRU may perform model inference corresponding to the one or more AI / ML-enabled functionalities based on the configurations related to the one or more AI / ML-enabled functionalities. The configurations may be transmitted to the UE or PRU by the gNB, NW, or LMF after the UE or PRU reports the AI / ML-enabled functionality or functionalities supported or decided to activate. The configurations may be transmitted to the UE via an RRC message, MAC CE, or LPP message. After the UE or PRU decides to activate one or more AI / ML-enabled functionalities, the UE or PRU may report the AI / ML-enabled functionality or functionalities intended for activation to the gNB or NW (e.g., via MAC CE or PUCCH) or LMF (e.g., via LPP message) to request the corresponding configurations related to at least one of the following: 　　　　- reporting of model output; 　　　　- measurement for the model input (e.g., PRS configuration for the model input of the activated AI / ML-enabled functionalities); 　　　　- additional condition of the NW side; and 　　　　- parameters of model monitoring (e.g., the periodicity of monitoring report, the threshold used to determine whether the AI / ML-enabled functionality is applicable for AI / ML operation).

[0143] Before the UE or PRU reports the AI / ML-enabled functionality or functionalities intended for activation to the gNB or NW, the UE or PRU may transmit an SR request to the gNB or NW or initiate a RACH procedure to request UL resources for transmitting the report.

[0144] The corresponding configurations may be associated with the AI / ML-enabled functionality or functionalities activated by the UE or PRU. For example, each corresponding configuration may be associated with an AI / ML-enabled functionality index.

[0145] (E) AI / ML functionalities or models monitoring.

[0146] In some implementations, the UE or PRU may monitor the performance of the model output of each activated AI / ML-enabled functionality.

[0147] In some implementations, the UE or PRU may be configured with a threshold to determine whether each activated AI / ML-enabled functionality remains applicable for model inference. The threshold may be an RSRP value, an error rate value, a confidence value, a quality of model output, a quality of measurement, or a probability value, but which is not limited thereto.

[0148] In some implementations, the UE or PRU may be configured with one or more configurations for deriving the ground-truth label corresponding to the model output of each activated AI / ML-enabled functionality used for model inference for positioning. When the value of the performance metric calculated by the model output and the ground-truth label is less than the configured threshold, the UE or PRU may decide to deactivate all activated AI / ML-enabled functionalities. When the UE or PRU decides to deactivate all activated AI / ML-enabled functionalities, the UE or PRU may terminate the deactivated AI / ML-enabled functionalities and fall back to the legacy positioning method(s). The UE or PRU may transmit a message to the gNB, NW, or LMF to inform the gNB, NW, or LMF of the decision to terminate all activated AI / ML-enabled functionalities and fall back to the legacy positioning method(s). The message may include at least one of the following: 　　　　- the functionality index or indices for the corresponding activated AI / ML-enabled functionality or functionalities that the UE or PRU decides to terminate; and 　　　　- the legacy positioning method (e.g., DL ECID, DL-TDOA) that the UE or PRU intends to apply for positioning after terminating all activated AI / ML-enabled functionalities.

[0149] In some implementations, when the UE or PRU decides to deactivate all activated AI / ML-enabled functionalities, the UE or PRU may direct the deactivated AI / ML-enabled functionalities to return to the model training phase. The UE or PRU may transmit a message to the gNB, NW, or LMF to inform the gNB, NW, or LMF of the decision to direct all deactivated AI / ML-enabled functionalities to return to the model training phase.

[0150] In some implementations, the UE or PRU may be configured with one or more configurations for deriving the ground-truth label corresponding to the model output of each activated AI / ML-enabled functionality used for model inference for positioning. The UE or PRU may be configured with a maximum counter number (e.g., N times) and one or more counters, where each counter may be associated with each activated AI / ML-enabled functionality used for model inference for positioning. When the value of the performance metric calculated by the model output and the ground-truth label is less than the configured threshold, the UE or PRU may increase the counter number by 1 for the counter of the corresponding activated AI / ML-enabled functionality. If the counter number of the corresponding AI / ML-enabled functionality equals the maximum counter number (e.g., N times), the UE or PRU may deactivate the corresponding activated AI / ML-enabled functionality or functionalities. The counter number may be set to zero when the corresponding AI / ML-enabled functionality is activated. The counter may be reset when the corresponding AI / ML-enabled functionality is deactivated. When the UE or PRU decides to deactivate the corresponding activated AI / ML-enabled functionality or functionalities, the UE or PRU may terminate the deactivated AI / ML-enabled functionalities and fall back to the legacy positioning method or methods. The UE or PRU may transmit a message to the gNB, NW, or LMF to inform the gNB, NW, or LMF of the decision to terminate the corresponding activated AI / ML-enabled functionality or functionalities and fall back to the legacy positioning method or methods. The message may include at least one of the following: 　　　　- the functionality index or indices for the corresponding activated AI / ML-enabled functionality or functionalities that the UE or PRU decides to terminate; and 　　　　- the legacy positioning method (e.g., DL ECID, DL-TDOA) that the UE or PRU intends to apply for positioning after terminating the activated AI / ML-enabled functionality or functionalities.

[0151] In some implementations, when the UE or PRU decides to deactivate all activated AI / ML-enabled functionalities, the UE or PRU may direct the deactivated AI / ML-enabled functionalities to return to the model training phase. The UE or PRU may transmit a message to the gNB, NW, or LMF to inform the gNB, NW, or LMF of the decision to direct all deactivated AI / ML-enabled functionalities to return to the model training phase.

[0152] In some implementations, the UE or PRU may transmit a report to the gNB, NW, or LMF. The report may include a list of available AI / ML-enabled functionalities. The gNB, NW, or LMF may indicate the AI / ML-enabled functionality or functionalities for the UE or PRU to activate for model inference. The gNB, NW, or LMF may transmit one or more configurations corresponding to the activated AI / ML-enabled functionality or functionalities to the UE or PRU. The configuration may include at least one of the following: 　　　　- the measurement configuration for deriving the ground-truth label corresponding to the activated AI / ML-enabled functionality or functionalities; and 　　　　- the configuration for calculating the AI / ML performance monitoring metric corresponding to the activated AI / ML-enabled functionality or functionalities.

[0153] In some implementations, the report may be transmitted via UAI, MAC CE, PUSCH, PUCCH, or LPP message.

[0154] In some implementations, the UE or PRU may transmit a request to the gNB, NW, or LMF for deactivating the activated AI / ML-enabled functionality or functionalities. The request may include at least one of the following: 　　　　- the functionality index or indices of the activated AI / ML-enabled functionality or functionalities that the UE or PRU intends to terminate or direct to return to the model training phase; 　　　　- the reason for the UE or PRU requesting to deactivate or terminate the AI / ML-enabled functionality or functionalities, such as low battery capacity or overheating, where the reason may be enumerated by several causes, and the UE may indicate the cause number to represent the reason; 　　　　- the functionality index or indices of the currently available AI / ML-enabled functionality or functionalities; 　　　　- the functionality index or indices of the AI / ML-enabled functionality or functionalities that the UE or PRU intends to switch to; 　　　　- whether to fall back to a legacy positioning method (e.g., DL ECID, DL TDOA, DL AoD, Multi-RTT, or UL positioning methods); and 　　　　- the legacy positioning method that the UE or PRU suggests for fallback.

[0155] In some implementations, the request may be transmitted to the gNB or NW via UAI, MAC CE, PUSCH, or PUCCH. The request may be transmitted to the LMF via an LPP message.

[0156] In some implementations, the UE or PRU may transmit a request to the gNB, NW, or LMF for deactivating the activated AI / ML-enabled functionality or functionalities. The UE or PRU may receive a response transmitted from the gNB, NW, or LMF. The UE or PRU may deactivate the AI / ML-enabled functionality or functionalities indicated in the request or response and activate the AI / ML-enabled functionality or functionalities indicated in the request or response no later than a time duration after the last symbol of a PUCCH or PUSCH with HARQ-ACK information of the response. The response may be transmitted by the gNB or NW via PDSCH or MAC CE. The response may be transmitted by the LMF via an LPP message. The request may indicate whether the activated AI / ML-enabled functionalities that the UE or PRU intends to deactivate should return to the model training phase or be terminated.

[0157] A case described above may apply to a NW-side model, namely Case E: “NG-RAN node assisted positioning with gNB-side model, AI / ML assisted positioning.”

[0158] In some implementations, the AI / ML-enabled functionality may be managed by the gNB or NW. For the NW-side model, the management of AI / ML-enabled functionalities may be implemented by the gNB or NW.

[0159] In some implementations, for the NW-side model, the AI / ML model or models corresponding to different AI / ML-enabled functionalities may be trained or tested at the gNB or NW. When the AI / ML-enabled functionalities are available, the gNB or NW may activate the available AI / ML-enabled functionalities. The gNB or NW may start performing model inference for the activated AI / ML-enabled functionality or functionalities. After the gNB or NW starts performing model inference for the activated AI / ML-enabled functionality or functionalities, the gNB or NW may begin to monitor the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities. If the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities degrades, the gNB or NW may deactivate or terminate the corresponding activated AI / ML-enabled functionality or functionalities. If the gNB or NW deactivates the corresponding activated AI / ML-enabled functionality or functionalities, the gNB or NW may direct the deactivated AI / ML-enabled functionality or functionalities to return to the model training phase or terminate the deactivated AI / ML-enabled functionality or functionalities.

[0160] In some implementations, to apply the NW-side model for AI / ML operation for use cases such as beam management, positioning, or CSI prediction, at least one of the following steps (A) to (D) is necessary.

[0161] (A) AI / ML functionalities or models training.

[0162] In some implementations, the gNB or NW may train the AI / ML model(s) corresponding to one or more AI / ML-enabled functionalities. To train the AI / ML model(s) corresponding to one or more AI / ML-enabled functionalities at the NW side, the gNB or NW may transmit a request to the LMF for providing ground truth label data via an NRPPa message. The request may include at least one of the AI / ML-enabled functionality indices. The ground truth label data transmitted by the LMF may be associated with at least one of the AI / ML-enabled functionality indices. The request may include information indicating whether the transmission of the ground truth data is aperiodic or periodic.

[0163] (B) AI / ML functionalities or models testing.

[0164] In some implementations, the gNB or NW may test the AI / ML model(s) corresponding to one or more AI / ML-enabled functionalities. When the AI / ML model(s) corresponding to one or more AI / ML-enabled functionalities pass the test, the AI / ML model(s) corresponding to one or more AI / ML-enabled functionalities may become available. The gNB or NW may activate one or more available AI / ML-enabled functionalities. The gNB or NW may transmit an NRPPa message to the LMF to stop the periodic ground truth label data transmission or to provide information indicating the AI / ML-enabled functionality or functionalities that are activated.

[0165] (C) AI / ML functionalities or models inference.

[0166] In some implementations, the gNB or NW may perform model inference for the activated AI / ML-enabled functionality or functionalities. Before the gNB or NW performs model inference, the gNB or NW may transmit the corresponding configuration to the UE or PRU for SRS transmission. The purpose of the SRS transmitted to the gNB or NW may be set to positioning or beam management. To support the operation, the gNB may configure an SRS for the UE or PRU with two separate reporting types (e.g., one for model training and the other for model inference monitoring). The gNB may control the activation or deactivation of the two reporting types via a DCI trigger.

[0167] (D) AI / ML functionalities or models monitoring.

[0168] In some implementations, the gNB or NW may monitor the performance of the model output of each activated AI / ML-enabled functionality used to perform model inference. If the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities degrades, the gNB or NW may deactivate the activated AI / ML-enabled functionality or functionalities. If any available AI / ML-enabled functionalities remain, the gNB or NW may activate the available AI / ML-enabled functionalities or terminate the AI / ML operation for positioning. If no available AI / ML-enabled functionalities remain, the gNB or NW may terminate the AI / ML operation for positioning. When the gNB or NW terminates the AI / ML operation for positioning, the gNB or NW may transmit a message to inform the LMF of the termination of the AI / ML operation for positioning via the NRPPa protocol. The gNB or NW may transmit an NRPPa message to inform the LMF of the activated AI / ML-enabled functionalities that are deactivated. The message may indicate whether the gNB or NW directs the deactivated AI / ML-enabled functionalities to return to the model training phase or terminates the deactivated AI / ML-enabled functionalities.

[0169] Two cases described above may apply to an LMF-side model, namely Case B: “UE-assisted / LMF-based positioning with LMF-side model, direct AI / ML positioning,” and Case C: “NG-RAN node assisted positioning with LMF-side model, direct AI / ML positioning.”

[0170] In some implementations, the AI / ML-enabled functionality or functionalities may be managed by the LMF. For the LMF-side model, the management of AI / ML-enabled functionality or functionalities may be implemented by the LMF.

[0171] In some implementations, for the LMF-side model, the AI / ML model(s) corresponding to different AI / ML-enabled functionality or functionalities may be trained or tested at the LMF. When the AI / ML-enabled functionality or functionalities are available, the LMF may activate the available AI / ML-enabled functionality or functionalities. The LMF may start performing model inference for the activated AI / ML-enabled functionality or functionalities. After the LMF starts performing model inference for the activated AI / ML-enabled functionality or functionalities, the LMF may begin to monitor the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities. If the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities degrades, the LMF may deactivate the corresponding AI / ML-enabled functionality or functionalities. If the LMF deactivates the corresponding AI / ML-enabled functionality or functionalities, the LMF may direct the deactivated AI / ML-enabled functionality or functionalities to return to the model training phase or terminate the deactivated AI / ML-enabled functionality or functionalities.

[0172] In some implementations, to apply the LMF-side model for AI / ML operation for use cases such as beam management, positioning, or CSI prediction, at least one of the following steps (A) to (D) is necessary.

[0173] (A) AI / ML functionalities or models training.

[0174] In some implementations, the LMF may train the AI / ML model(s) corresponding to one or more AI / ML-enabled functionality or functionalities. To train the AI / ML model(s) corresponding to one or more AI / ML-enabled functionality or functionalities at the LMF, the LMF may transmit a request to the UE or PRU for providing ground truth label data via an LPP message. The request may include at least one of the AI / ML-enabled functionality indices. The ground truth label data transmitted by the UE or PRU may be associated with at least one of the AI / ML-enabled functionality indices. The request may include information indicating whether the transmission of the ground truth label data is aperiodic or periodic.

[0175] (B) AI / ML functionalities or models testing.

[0176] In some implementations, the LMF may test the AI / ML model(s) corresponding to one or more AI / ML-enabled functionality or functionalities. When the AI / ML model(s) corresponding to one or more AI / ML-enabled functionality or functionalities pass the test, the AI / ML model(s) corresponding to one or more AI / ML-enabled functionality or functionalities become available. The LMF may activate one or more available AI / ML-enabled functionality or functionalities. The LMF may transmit an LPP message to the UE or PRU to stop the periodic ground truth label data transmission or to provide information indicating the AI / ML-enabled functionality or functionalities that are activated. The LMF may configure the corresponding measurement configuration or measurement report configuration to the UE or PRU for the activated AI / ML-enabled functionality or functionalities.

[0177] (C) AI / ML functionalities or models inference.

[0178] In some implementations, the LMF may perform model inference for the activated AI / ML-enabled functionality or functionalities.

[0179] Before the LMF performs model inference, the LMF may request the UE(s), PRU(s), gNB and / or NW to perform PRS or SRS measurement or to provide the corresponding measurement report from different entities such as the gNB, NW, UE(s), or PRU(s). The measurement result included in the corresponding measurement report may be used as the model input for the corresponding AI / ML-enabled functionality or functionalities. The measurement report from the UE(s), PRU(s), gNB, and / or NW may include at least one AI / ML-enabled functionality index. The measurement configuration configured to the UE, PRU, and / or LMF may be specific to the AI / ML-enabled functionality or functionalities, and the corresponding measurement may be activated by the LMF. For example, the LMF may transmit a measurement request to the UE(s), PRU(s), gNB, and / or NW.

[0180] (D) AI / ML functionalities or models monitoring.

[0181] In some implementations, the LMF may monitor the performance of the model output of each activated AI / ML-enabled functionality used to perform model inference. If the performance of the model output corresponding to the activated AI / ML-enabled functionality or functionalities degrades, the LMF may deactivate the activated AI / ML-enabled functionality or functionalities. If the LMF deactivates the corresponding AI / ML-enabled functionality or functionalities, the LMF may direct the deactivated AI / ML-enabled functionality or functionalities to return to the model training phase or terminate the deactivated AI / ML-enabled functionality or functionalities.

[0182] In some implementations, when the LMF deactivates all activated AI / ML-enabled functionality or functionalities, the LMF may terminate the activated AI / ML-enabled functionality or functionalities or direct the activated AI / ML-enabled functionality or functionalities to return to the model training phase. If any available AI / ML-enabled functionality or functionalities remain, the LMF may activate the available AI / ML-enabled functionality or functionalities or terminate the AI / ML operation (e.g., for positioning). If no available AI / ML-enabled functionality or functionalities remain, the LMF may terminate the AI / ML operation (e.g., for positioning).

[0183] In some implementations, when the LMF terminates the AI / ML operation (e.g., for positioning,) the LMF may transmit a message to inform the UE(s), PRU(s), gNB, and / or NW of the termination of the AI / ML operation (e.g., for positioning) via the LPP protocol or NRPPa protocol.

[0184] For example, the LMF may transmit an NRPPa message to inform the gNB or NW that AI / ML-based positioning is terminated. If the gNB or NW receives the termination message related to AI / ML-based positioning, the gNB or NW may terminate the corresponding measurement, measurement report, or ground truth label transmission.

[0185] For example, the LMF may transmit an LPP message to inform the UE(s) or PRU(s) that AI / ML-based positioning is terminated. If the UE(s) or PRU(s) receive the termination message related to AI / ML-based positioning, the UE(s) or PRU(s) may terminate the corresponding measurement, measurement report, or ground truth label transmission.

[0186] In some implementations, when the LMF deactivates the activated AI / ML-enabled functionality or functionalities or activates other available AI / ML-enabled functionality or functionalities, the LMF may transmit a message to inform the UE(s), PRU(s), gNB, and / or NW of the activated AI / ML-enabled functionality or functionalities that are deactivated or the available AI / ML-enabled functionality or functionalities that are activated. The LMF may inform the UE(s), PRU(s), gNB, and / or NW that the deactivated AI / ML-enabled functionality or functionalities return to the model training phase or are terminated.

[0187] For example, the LMF may transmit an NRPPa message to inform the gNB or NW of the activated AI / ML-enabled functionality or functionalities that are deactivated or the available AI / ML-enabled functionality or functionalities that are activated. If the gNB or NW receives the deactivation message related to the activated AI / ML-enabled functionality or functionalities, the gNB or NW may terminate the corresponding measurement, measurement report, or ground truth label transmission for model monitoring. If the gNB or NW receives the activation message related to the available AI / ML-enabled functionality or functionalities, the gNB or NW may start the corresponding measurement, measurement report, or ground truth label transmission. The deactivation command and activation command may be included in the same NRPPa message or in separate NRPPa messages.

[0188] For example, the LMF may transmit an LPP message to inform the UE(s) or PRU(s) of the activated AI / ML-enabled functionality or functionalities that are deactivated or the available AI / ML-enabled functionality or functionalities that are activated. If the UE(s) or PRU(s) receive the deactivation message related to the activated AI / ML-enabled functionality or functionalities, the UE(s) or PRU(s) may terminate the corresponding measurement, measurement report, or ground truth label transmission for model monitoring. If the UE(s) or PRU(s) receive the activation message related to the available AI / ML-enabled functionality or functionalities, the UE(s) or PRU(s) may start the corresponding measurement, measurement report, or ground truth label transmission. The deactivation command and activation command may be included in the same LPP message or in separate LPP messages.

[0189] FIG. 6 is a flowchart illustrating a method / process 600 performed by a UE for enabling AI / ML functionalities, according to an example implementation of the present disclosure.

[0190] In action 610, the process 600 may start by receiving, from a BS, a request for an AI / ML-based UE capability report. From an aspect of the BS, the BS may transmit, to the UE, the request for the AI / ML-based UE capability report.

[0191] In action 620, the process 600 may transmit, to the BS, the AI / ML-based capability report, where the AI / ML-based capability report indicates at least one AI / ML functionality supported by the UE. From the aspect of the BS, the BS may receive, from the UE, the AI / ML-based capability report.

[0192] In action 630, the process 600 may receive, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE. From the aspect of the BS, the BS may transmit, to the UE, the list of configurations corresponding to the at least one AI / ML functionality supported by the UE.

[0193] In action 640, the process 600 may select, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE.

[0194] In action 650, the process 600 may transmit, to the BS, information of the at least one available functionality. From the aspect of the BS, the BS may receive, from the UE, the information of at least one available functionality selected from the at least one AI / ML functionality supported by the UE.

[0195] In action 660, the process 600 may receive, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality. From the aspect of the BS, the BS may transmit, to the UE, the activation command for activating at least one designated functionality of the at least one available functionality. In some implementations, in response to the activation command, the UE may transmit Hybrid Automatic Repeat reQuest-Acknowledgment (HARQ-ACK) information corresponding to the activation command to the BS.

[0196] In action 670, the process 600 may activate the at least one designated functionality. In some implementations, the at least one designated functionality is activated no later than a (e.g., preset) time duration after the last symbol of the PUCCH or the PUSCH with the HARQ-ACK information transmitted in response to the activation command.

[0197] In some implementations, after the at least one designated functionality is activated, the UE may transmit, to the BS, a deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE. The UE may then receive, from the BS, a confirmation message corresponding to the deactivation request, and deactivate the at least one functionality to be deactivated.

[0198] From the aspect of the BS, the BS may receive, from the UE the deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE. In response to the deactivation request, the BS may transmit, to the UE, the confirmation message corresponding to the deactivation request.

[0199] In some implementations, the deactivation request may include UE assistance information (UAI) or an output of an AI / ML model executed by the UE. The UAI may include, for example, a deactivation reason (e.g., a remaining battery capacity of the UE being below a preset threshold) of the at least one functionality to be deactivated. The output of the AI / ML model executed by the UE may, for example, indicate that a confidence level is below a preset threshold.

[0200] FIG. 7 is a block diagram illustrating a node 700 for wireless communication in accordance with various aspects of the present disclosure. As illustrated in FIG. 7, a node 700 may include a transceiver 720, a processor 728, a memory 734, one or more presentation components 738, and at least one antenna 736. The node 700 may also include a radio frequency (RF) spectrum band module, a BS communications module, a network communications module, and a system communications management module, Input / Output (I / O) ports, I / O components, and a power supply (not illustrated in FIG. 7).

[0201] Each of the components may directly or indirectly communicate with each other over one or more buses 740. The node 700 may be a UE or a BS that performs various functions disclosed with reference to FIGS. 1 to 6.

[0202] The transceiver 720 has a transmitter 722 (e.g., transmitting / transmission circuitry) and a receiver 724 (e.g., receiving / reception circuitry) and may be configured to transmit and / or receive time and / or frequency resource partitioning information. The transceiver 720 may be configured to transmit in different types of subframes and slots including, but not limited to, usable, non-usable, and flexibly usable subframes and slot formats. The transceiver 720 may be configured to receive data and control channels.

[0203] The node 700 may include a variety of computer-readable media. Computer-readable media may be any available media that may be accessed by the node 700 and include volatile (and / or non-volatile) media and removable (and / or non-removable) media.

[0204] The computer-readable media may include computer-storage media and communication media. Computer-storage media may include both volatile (and / or non-volatile media), and removable (and / or non-removable) media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or data.

[0205] Computer-storage media may include RAM, ROM, EPROM, EEPROM, flash memory (or other memory technology), CD-ROM, Digital Versatile Disks (DVD) (or other optical disk storage), magnetic cassettes, magnetic tape, magnetic disk storage (or other magnetic storage devices), etc. Computer-storage media may not include a propagated data signal. Communication media may typically embody computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave, or other transport mechanisms and include any information delivery media.

[0206] The term “modulated data signal” may mean a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. Communication media may include wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above listed components should also be included within the scope of computer-readable media.

[0207] The memory 734 may include computer-storage media in the form of volatile and / or non-volatile memory. The memory 734 may be removable, non-removable, or a combination thereof. Example memory may include solid-state memory, hard drives, optical-disc drives, etc. As illustrated in FIG. 7, the memory 734 may store a computer-readable and / or computer-executable instructions 732 (e.g., software codes) that are configured to, when executed, cause the processor 728 to perform various functions disclosed herein, for example, with reference to FIGS. 1 to 6. Alternatively, the instructions 732 may not be directly executable by the processor 728 but may be configured to cause the node 700 (e.g., when compiled and executed) to perform various functions disclosed herein.

[0208] The processor 728 (e.g., having processing circuitry) may include an intelligent hardware device, e.g., a Central Processing Unit (CPU), a microcontroller, an ASIC, etc. The processor 728 may include memory. The processor 728 may process the data 730 and the instructions 732 received from the memory 734, and information transmitted and received via the transceiver 720, the baseband communications module, and / or the network communications module. The processor 728 may also process information to send to the transceiver 720 for transmission via the antenna 736 to the network communications module for transmission to a CN.

[0209] One or more presentation components 738 may present data indications to a person or another device. Examples of presentation components 738 may include a display device, a speaker, a printing component, a vibrating component, etc.

[0210] In view of the present disclosure, it is obvious that various techniques may be used for implementing the disclosed concepts without departing from the scope of those concepts. Moreover, while the concepts have been disclosed with specific reference to certain implementations, a person of ordinary skill in the art may recognize that changes may be made in form and detail without departing from the scope of those concepts. As such, the disclosed implementations are to be considered in all respects as illustrative and not restrictive. It should also be understood that the present disclosure is not limited to the particular implementations disclosed and many rearrangements, modifications, and substitutions are possible without departing from the scope of the present disclosure.

Claims

1. A User Equipment (UE) for enabling Artificial Intelligence (AI) / Machine Learning (ML) functionalities, the UE comprising: 　　　　at least one processor; and 　　　　at least one non-transitory computer-readable medium coupled to the at least one processor and storing one or more computer-executable instructions that, when executed by the at least one processor, cause the UE to: 　　　　　　　　receive, from a Base Station (BS), a request for an AI / ML-based UE capability report; 　　　　　　　　transmit, to the BS, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; 　　　　　　　　receive, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; 　　　　　　　　select, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE; 　　　　　　　　transmit, to the BS, information of the at least one available functionality; 　　　　　　　　receive, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality; and 　　　　　　　　activate the at least one designated functionality.

2. The UE of claim 1, wherein the one or more computer-executable instructions, when executed by the at least one processor, further cause the UE to: 　　　　transmit, to the BS, a deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE; 　　　　receive, from the BS, a confirmation message corresponding to the deactivation request; and 　　　　deactivate the at least one functionality to be deactivated.

3. The UE of claim 1, wherein the one or more computer-executable instructions, when executed by the at least one processor, further cause the UE to: 　　　　transmit, to the BS, Hybrid Automatic Repeat reQuest-Acknowledgment (HARQ-ACK) information corresponding to the activation command, 　　　　wherein the at least one designated functionality is activated no later than a time duration after the last symbol of a Physical Uplink Control Channel (PUCCH) or a Physical Uplink Shared Channel (PUSCH) with the HARQ-ACK information.

4. The UE of claim 2, wherein the deactivation request comprises UE assistance information (UAI) or an output of an AI / ML model executed by the UE.

5. The UE of claim 4, wherein the output indicates that a confidence level is below a preset threshold.

6. The UE of claim 4, wherein the UAI comprises a deactivation reason of the at least one functionality to be deactivated.

7. The UE of claim 6, wherein the deactivation reason comprises a remaining battery capacity being below a preset threshold.

8. A Base Station (BS) for supporting Artificial Intelligence (AI) / Machine Learning (ML) functionalities for a User Equipment (UE), the BS comprising: 　　　　at least one processor; and 　　　　at least one non-transitory computer-readable medium coupled to the at least one processor and storing one or more computer-executable instructions that, when executed by the at least one processor, cause the BS to: 　　　　　　　　transmit, to the UE, a request for an AI / ML-based UE capability report; 　　　　　　　　receive, from the UE, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; 　　　　　　　　transmit, to the UE, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; 　　　　　　　　receive, from the UE, information of at least one available functionality selected from the at least one AI / ML functionality supported by the UE; and 　　　　　　　　transmit, to the UE, an activation command for activating at least one designated functionality of the at least one available functionality.

9. The BS of claim 8, wherein the one or more computer-executable instructions, when executed by the at least one processor, further cause the BS to: 　　　　receive, from the UE, a deactivation request corresponding to at least one functionality to be deactivated that has been activated by the UE; and 　　　　transmit, to the UE, a confirmation message corresponding to the deactivation request.

10. The BS of claim 8, wherein the one or more computer-executable instructions, when executed by the at least one processor, further cause the BS to: 　　　　receive, from the UE, Hybrid Automatic Repeat reQuest-Acknowledgment (HARQ-ACK) information corresponding to the activation command, 　　　　wherein the at least one designated functionality is activated by the UE no later than a time duration after the last symbol of a Physical Uplink Control Channel (PUCCH) or a Physical Uplink Shared Channel (PUSCH) with the HARQ-ACK information.

11. The BS of claim 9, wherein the deactivation request comprises UE assistance information (UAI) or an output of an AI / ML model executed by the UE.

12. The BS of claim 11, wherein the output indicates that a confidence level is below a preset threshold.

13. The BS of claim 11, wherein the UAI comprises a deactivation reason of the at least one functionality to be deactivated.

14. The BS of claim 13, wherein the deactivation reason comprises a remaining battery capacity being below a preset threshold.

15. A method performed by a User Equipment (UE) for enabling Artificial Intelligence (AI) / Machine Learning (ML) functionalities, the method comprising: 　　　　receiving, from a Base Station (BS), a request for an AI / ML-based UE capability report; 　　　　transmitting, to the BS, the AI / ML-based capability report, the AI / ML-based capability report indicating at least one AI / ML functionality supported by the UE; 　　　　receiving, from the BS, a list of configurations corresponding to the at least one AI / ML functionality supported by the UE; 　　　　selecting, based on the list of configurations, at least one available functionality from the at least one AI / ML functionality supported by the UE; 　　　　transmitting, to the BS, information of the at least one available functionality; 　　　　receiving, from the BS, an activation command for activating at least one designated functionality of the at least one available functionality; and 　　　　activating the at least one designated functionality.

Images