Method and apparatus for monitoring artificial intelligence model in radio access network

Abstract

Methods and apparatus for monitoring an artificial intelligence (AI) model in a radio access network (RAN) are disclosed. The apparatus includes: a receiver that receives first configuration signaling for monitoring performance of one or more activated artificial intelligence (AI) models, and second configuration signaling for reporting a result of the performance monitoring; the processor is used for generating a performance report of the AI model according to the first configuration signaling and the second configuration signaling; and a transmitter that transmits the performance report.

Description

Technical Field

[0001] The subject matter disclosed herein relates generally to wireless communications, and more particularly, but not limited to, methods and apparatus for monitoring artificial intelligence (AI) models in a radio access network (RAN). Background Art

[0002] The following abbreviations and acronyms are defined herein, at least some of which are referred to within the specification:

[0003] 3rd Generation Partnership Project (3GPP), fifth generation (5G), new radio (NR), 5G Node B (gNB), long term evolution (LTE), LTE advanced (LTE-A), E-UTRAN Node B (eNB), universal mobile telecommunications system (UMTS), world-wide interoperability for microwave access (WiMAX), evolved UMTS terrestrial radio access network (E-UTRAN), wireless local area network (WLAN), orthogonal frequency division multiplexing (OFDM), single carrier frequency division multiple access (SC-FDMA), downlink (DL), uplink (UL), user equipment (UE), network equipment (NE), radio access technology (RAT), reception or receiver (RX, or Rx), transmission or transmitter (TX, Tx), physical uplink control Control channel (PUCCH), physical uplink shared channel (PUSCH), hybrid automatic repeat request (HARQ), acknowledgement (ACK), hybrid automatic repeat request acknowledgement (HARQ-ACK), physical downlink shared channel (PDSCH), physical uplink control channel (PUCCH), physical uplink shared channel (PUSCH), physical broadcast channel (PBCH), block error rate (BLER), control element (CE), channel state information (CSI), channel state information reference signal (CSI-RS), downlink control information (DCI), frequency division multiple access (FDMA), index / identifier (ID), information element (IE), medium access control (MAC), medium access control-control element (MAC-CE or MAC CE), Multiple-Input Multiple-Output (MIMO), Physical Layer (PHY), Radio Access Network (RAN), Radio Resource Control (RRC), Reference Signal (RS), Reference Signal Received Power (RSRP), Received Signal Strength Indicator (RSSI), Subcarrier Spacing (SCS), Sounding Reference Signal (SRS), Synchronization Signal Block (SSB), Transmission and Reception Point (TRP), Frequency Range 1 (FR1), Frequency Range 2 (FR2), Layer 1 Reference Signal Received Power (L1-RSRP), Synchronization Signal (SS), Layer 1 / Physical Layer (L1), Channel Busy Rate (CBR), Synchronization Signal and Physical Broadcast Channel (SS / PBCH), Artificial Intelligence (AI), Machine Learning (ML), Computer Vision (CV), Natural Language Processing (NLP), Neural Network (NN), Non Line of Sight (NLOS), Line of Sight (LOS), High Speed ​​Train (HST), Cross Link Interference (CLI).

[0004] In wireless communications such as the 3rd Generation Partnership Project (3GPP) mobile network, the wireless mobile network can provide seamless wireless communication services to wireless communication terminals with mobility, i.e., user equipment (UE). The wireless mobile network can be formed by a plurality of base stations, and the base stations can perform wireless communications with the UE.

[0005] 5G New Radio (NR) is the latest in the 3GPP family of standards and supports very high data rates with lower latency than its predecessor LTE (4G) technology. Two types of frequency ranges (FR) are defined in 3GPP. Frequencies in the sub-6GHz range (from 450 to 6000MHz) are referred to as FR1, and the millimeter wave range (from 24.25GHz to 52.6GHz) is referred to as FR2. 5G NR supports both FR1 and FR2 frequency bands.

[0006] Enhancements to multi-TRP / panel transmissions are studied, which include improved reliability and robustness with respect to both ideal and non-ideal backhaul between these TRPs (Transmit Receiving Points). A TRP is a device that transmits and receives signals and is controlled by the gNB through the backhaul between the gNB and the TRPs.

[0007] It is important to identify and specify the necessary enhancements for both downlink and uplink MIMO to facilitate the use of large antenna arrays not only for FR1 but also for FR2 to meet the requirements of the evolution of NR deployments in Release 18.

[0008] Artificial Intelligence (AI) / Machine Learning (ML) is used to learn and perform certain tasks by training neural networks with large amounts of data, and has been successfully applied in the fields of Computer Vision (CV) and Natural Language Processing (NLP). As a subset of ML, Deep Learning (DL) uses multi-layer neural networks (NN) as "AI models" to learn to solve problems and optimize performance from large amounts of data. Given the promising benefits presented in many academic papers and field test results, AI / ML-based methods may achieve better performance than traditional methods if well trained. In 3GPP, the introduction of AI / ML into the air interface in NR Release 18 is being discussed, including potential use cases, evaluation methods, and frameworks.

[0009] The characteristics of lifecycle management of AI / ML models can be studied based on investigation and evaluation of selected use cases, namely, CSI feedback enhancement, beam management, and positioning accuracy improvement.

[0010] Some evaluation results show that AI-based methods with well-trained AI models may perform better than non-AI-based methods. However, the benefits of AI methods mostly rely on the datasets constructed for training. If the training input data does not have the same characteristics as the actual dataset, the benefits of AI methods are questionable. Therefore, for reasonable deployment in RAN, the performance of AI-based methods should be monitored and enabled.

[0011] In this disclosure, a set of signaling including both MAC CE and RRC signaling is proposed to support relevant behaviors to monitor AI / ML models deployed to enhance air interface performance in RAN. Summary of the invention

[0012] Methods and apparatus for monitoring artificial intelligence (AI) models in a radio access network (RAN) are disclosed.

[0013] According to a first aspect, a device is provided, comprising: a receiver that receives a first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and a second configuration signaling for reporting a result of the performance monitoring; a processor that generates a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; and a transmitter that sends the performance report.

[0014] According to a second aspect, a device is provided, comprising: a transmitter that sends a first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and a second configuration signaling for reporting a result of the performance monitoring; a receiver that receives a performance report of the AI ​​model generated and reported according to the first configuration signaling and the second configuration signaling.

[0015] According to a third aspect, a method is provided, comprising: receiving, by a receiver, a first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and a second configuration signaling for reporting a result of the performance monitoring; generating, by a processor, a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; and sending, by a transmitter, the performance report.

[0016] According to a fourth aspect, a method is provided, comprising: sending, by a transmitter, a first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and a second configuration signaling for reporting the results of the performance monitoring; and receiving, by a receiver, a performance report of the AI ​​model generated and reported according to the first configuration signaling and the second configuration signaling. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] A more particular description of the embodiments will be presented by reference to specific embodiments illustrated in the accompanying drawings. Given that these drawings depict only some embodiments and are therefore not to be considered limiting in scope, these embodiments will be described and explained with additional specificity and detail through the use of the accompanying drawings, in which:

[0018] Figure 1 is a schematic diagram illustrating a wireless communication system according to some embodiments of the present disclosure;

[0019] Figure 2is a schematic block diagram illustrating components of a user equipment (UE) according to some embodiments of the present disclosure;

[0020] Figure 3 is a schematic block diagram illustrating components of a network equipment (NE) according to some embodiments of the present disclosure;

[0021] Figure 4A is a diagram illustrating an example of using an AI / ML method to compress CSI to reduce CSI feedback overhead according to some embodiments of the present disclosure.

[0022] Figure 4B is a schematic diagram illustrating an example of using an AI / ML approach for beam measurement with less overhead according to some embodiments of the present disclosure.

[0023] Figure 4C is a schematic diagram illustrating an example of using an AI / ML method to enhance positioning accuracy according to some embodiments of the present disclosure.

[0024] Figure 5A is a schematic diagram illustrating an example of replacing a communication module with an AI model according to some embodiments of the present disclosure.

[0025] Figure 5B is a schematic diagram illustrating an example of utilizing an AI model to assist a communication module according to some embodiments of the present disclosure.

[0026] FIG. 6A to FIG. 6C 2 is a schematic diagram illustrating examples of AI model deployment scenarios at a gNB side, at a UE side, and with an AI model at both sides, respectively, according to some embodiments of the present disclosure.

[0027] Figure 7 is a schematic diagram illustrating an example of an AI model performance monitoring process according to some embodiments of the present disclosure.

[0028] Fig. 8A is a schematic diagram illustrating an example of an AI model activation process with scenario identification at a gNB or network side according to some embodiments of the present disclosure.

[0029] Figure 8B is a schematic diagram illustrating an example of an AI model activation process with scenario identification at a UE side according to some embodiments of the present disclosure.

[0030] Fig. 9 is a schematic diagram illustrating an example of a measurement trigger for monitoring an activated AI model according to some embodiments of the present disclosure.

[0031] Fig.10is a flow chart illustrating steps of monitoring an AI model in a RAN by a UE according to some embodiments of the present disclosure; and

[0032] Fig.11 is a flow chart illustrating steps for monitoring an AI model in a RAN by a gNB according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0033] As will be appreciated by those skilled in the art, aspects of the embodiments may be embodied as a system, apparatus, method, or program product. Thus, the embodiments may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware aspects.

[0034] Furthermore, one or more embodiments may take the form of a program product embodied in one or more computer-readable storage devices storing machine-readable code, computer-readable code, and / or program code, hereinafter referred to as “code.” The storage devices may be tangible, non-transitory, and / or non-transmitting.

[0035] References to "one embodiment," "an embodiment," "an example," "some embodiments," "some examples," or similar language throughout this specification mean that the particular features, structures, or characteristics being described are included in at least one embodiment or example. Therefore, the phrases "in one embodiment," "in an example," "in some embodiments," and instances of similar language throughout the specification may, but need not, all refer to the same embodiment. It may or may not include all disclosed embodiments. Unless expressly specified otherwise, the features, structures, elements, or characteristics described in conjunction with one or some embodiments also apply to other embodiments. Unless expressly specified otherwise, the terms "include," "comprise," "have," and variations thereof mean "including but not limited to."

[0036] Unless expressly specified otherwise, the enumerated list of items does not imply that any or all of the items are mutually exclusive. Unless expressly specified otherwise, the terms "a", "an", and "the" also refer to "one or more", and similar items expressed in the plural also include reference to one or more instances of the item.

[0037] Throughout this disclosure, unless otherwise expressly specified, the terms "first", "second", "third", etc. are all used as nomenclatures for referencing related devices, components, procedural steps, etc. only, without implying any spatial or temporal order. For example, "first device" and "second device" may refer to two separately formed devices, or two parts or components of the same device. In some cases, for example, "first device" and "second device" may be the same and may be arbitrarily named. Similarly, the "first step" of a method or process may be performed or carried out after or simultaneously with the "second step".

[0038] It should be understood that the term "and / or" as used herein refers to and includes any and all possible combinations of one or more associated listed items. For example, "A and / or B" may refer to any of the following three combinations: only A exists, only B exists, and both A and B exist together. The character " / " generally indicates an "or" relationship of associated items. However, this may also include an "and" relationship of associated items. For example, "A / B" means "A or B", which may also include the presence of both A and B, unless the context indicates otherwise.

[0039] In addition, the features, structures or characteristics of the described embodiments may be combined in any suitable manner. In the following description, a number of specific details are provided, such as examples of programming, software modules, user selections, network transactions, database queries, database structures, hardware modules, hardware circuits, hardware chips, etc., to provide a thorough understanding of the embodiments. However, those skilled in the relevant art will recognize that the embodiments may be practiced without one or more of the specific details or using other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid confusing aspects of the embodiments.

[0040] Aspects of various embodiments are described below with reference to schematic flow charts and / or schematic block diagrams of methods, devices, systems, and program products. It will be understood that each box in the schematic flow charts and / or schematic block diagrams, and the combination of boxes in the schematic flow charts and / or schematic block diagrams, can be implemented by code. The code can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine, so that instructions executed by the processor of the computer or other programmable data processing device can create a device for implementing the functions or actions specified in the schematic flow charts and / or schematic block diagrams.

[0041] The code may also be stored in a storage device that can direct a computer, other programmable data processing apparatus, or other device to function in a specific manner so that the instructions stored in the storage device produce an article of manufacture including instructions that implement the functions or actions specified in the schematic flowchart and / or schematic block diagram.

[0042] The schematic flow charts and / or schematic block diagrams in the accompanying drawings illustrate the architecture, functions and operations of possible implementations of different devices, systems, methods and program products according to various embodiments. In this regard, each box in the schematic flow chart and / or schematic block diagram may represent a module, a fragment or a portion of a code, which includes one or more executable instructions of a code for implementing (one or more) specified logical functions. However, those skilled in the relevant art will recognize that the flow chart does not have to be practiced in the sequence shown and can be practiced without one or more specific steps or with other steps not shown.

[0043] It should also be noted that in some alternative implementations, the functions annotated in the identified blocks may not occur in the order annotated in the drawings. For example, depending on the functions involved, two blocks shown in succession may actually be executed substantially simultaneously, or these blocks may sometimes be executed in reverse order.

[0044] Figure 1 1 is a schematic diagram illustrating a wireless communication system. It depicts an embodiment of a wireless communication system 100. In one embodiment, the wireless communication system 100 may include a user equipment (UE) 102 and a network equipment (NE) 104. Figure 1 A specific number of UEs 102 and NEs 104 are depicted in FIG. 1 , but those skilled in the art will recognize that any number of UEs 102 and NEs 104 may be included in the wireless communication system 100 .

[0045] UE 102 may be referred to as a remote device, a remote unit, a subscriber unit, a mobile device, a mobile station, a user, a terminal, a mobile terminal, a fixed terminal, a subscriber station, a user terminal, an apparatus, a device, a user equipment or by other terminology used in the art.

[0046] In one embodiment, UE 102 may be an autonomous sensor device, an alarm device, an actuator device, a remote control device, etc. In some other embodiments, UE 102 may include a computing device such as a desktop computer, a laptop computer, a personal digital assistant (PDA), a tablet computer, a smart phone, a smart TV (e.g., a TV connected to the Internet), a set-top box, a game console, a security system (including security cameras), a vehicle-mounted computer, a network device (e.g., a router, a switch, a modem), etc. In some embodiments, UE 102 includes a wearable device such as a smart watch, a fitness band, an optical head-mounted display, etc. UE 102 may communicate directly with one or more of NEs 104.

[0047] NE 104 may also be referred to as a base station, access point, access terminal, base, node-B, eNB, gNB, home node-B, relay node, device, equipment, or any other term used in the art. Throughout this specification, references to base stations may refer to any of the above reference types of network devices 104 such as eNBs and gNBs.

[0048] The NEs 104 may be distributed over a geographic area. The NEs 104 are typically part of a radio access network that includes one or more controllers communicatively coupled to one or more corresponding NEs 104. The radio access network is typically communicatively coupled to one or more core networks, which may be coupled to other networks, such as the Internet and a public switched telephone network. These and other elements of the radio access and core networks are not illustrated, but are generally known to those of ordinary skill in the art.

[0049] In one embodiment, the wireless communication system 100 complies with 3GPP 5G New Radio (NR). In some embodiments, the wireless communication system 100 complies with the 3GPP protocol, wherein the NE 104 transmits on the DL using an OFDM modulation scheme, and the UE 102 transmits on the uplink (UL) using an SC-FDMA scheme or an OFDM scheme. However, more generally, the wireless communication system 100 may implement some other open or proprietary communication protocols, such as WiMAX. The present disclosure is not intended to be limited to the implementation of any particular wireless communication system architecture or protocol.

[0050] The NE 104 can serve multiple UEs 102 within a service area of, for example, a cell (or cell sector) or more cells via wireless communication links. The NE 104 sends DL communication signals to serve the UEs 102 in the time, frequency, and / or spatial domains.

[0051] A communication link is provided between the NE 104 and the UE 102a, 102b, which may be, for example, a NR UL or DL ​​communication link. Some UEs 102 may communicate using different radio access technologies (RATs) such as NR and LTE simultaneously. A direct or indirect communication link may be provided between two or more NEs 104.

[0052] The NE 104 may also include one or more transmit receive points (TRPs) 104a. In some embodiments, the network device may be a gNB 104 that controls multiple TRPs 104a. In addition, there is a backhaul between two TRPs 104a. In some other embodiments, the network device may be a TRP 104a controlled by a gNB.

[0053] A communication link is provided between the NE 104, 104a and the UE 102, 102a, respectively, which may be, for example, a NR UL / DL communication link. Some UEs 102, 102a may communicate simultaneously using different radio access technologies (RATs), such as NR and LTE.

[0054] In some embodiments, a UE 102a may be able to simultaneously communicate with two or more TRPs 104a utilizing non-ideal or ideal backhaul. A TRP may be a transmission point of a gNB. The UE and / or the TRP may use multiple beams. The two or more TRPs may be TRPs of different gNBs or the same gNB. That is, different TRPs may have the same cell-ID or different cell-IDs. The terms "TRP" and "transmit-receive identity" may be used interchangeably throughout this disclosure.

[0055] Figure 2 2 is a schematic block diagram illustrating components of a user equipment (UE) according to one embodiment. UE 200 may include a processor 202, a memory 204, an input device 206, a display 208, and a transceiver 210. In some embodiments, the input device 206 and the display 208 are combined into a single device, such as a touch screen. In some embodiments, UE 200 may not include any input device 206 and / or display 208. In various embodiments, UE 200 may include one or more processors 202 and may not include an input device 206 and / or display 208.

[0056] In one embodiment, the processor 202 may include any known controller capable of executing computer-readable instructions and / or capable of performing logical operations. For example, the processor 202 may be a microcontroller, a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), an auxiliary processing unit, a field programmable gate array (FPGA), or a similar programmable controller. In some embodiments, the processor 202 executes instructions stored in the memory 204 to perform the methods and routines described herein. The processor 202 is communicatively coupled to the memory 204 and the transceiver 210.

[0057] In one embodiment, the memory 204 is a computer-readable storage medium. In some embodiments, the memory 204 includes a volatile computer storage medium. For example, the memory 204 may include a RAM, which includes a dynamic RAM (DRAM), a synchronous dynamic RAM (SDRAM), and / or a static RAM (SRAM). In some embodiments, the memory 204 includes a non-volatile computer storage medium. For example, the memory 204 may include a hard drive, a flash memory, or any other suitable non-volatile computer storage device. In some embodiments, the memory 204 includes both volatile and non-volatile computer storage media. In some embodiments, the memory 204 stores data related to the triggering conditions for sending the measurement report to the network device. In some embodiments, the memory 204 also stores program code and related data.

[0058] In one embodiment, input device 206 may include any known computer input device including a touch panel, buttons, keyboard, stylus, microphone, etc. In some embodiments, input device 206 may be integrated with display 208, for example, as a touch screen or similar touch-sensitive display.

[0059] In one embodiment, the display 208 may include any known electronically controllable display or display device. The display 208 may be designed to output visual, audio, and / or tactile signals.

[0060] In one embodiment, the transceiver 210 is configured to communicate wirelessly with a network device. In some embodiments, the transceiver 210 includes a transmitter 212 and a receiver 214. The transmitter 212 is used to send UL communication signals to the network device, and the receiver 214 is used to receive DL communication signals from the network device.

[0061] The transmitter 212 and the receiver 214 may be any suitable type of transmitter and receiver. Although only one transmitter 212 and one receiver 214 are illustrated, the transceiver 210 may have any suitable number of transmitters 212 and receivers 214. For example, in some embodiments, the UE 200 includes multiple transmitter 212 and receiver 214 pairs for communicating on multiple wireless networks and / or radio frequency bands, wherein each of the transmitter 212 and receiver 214 pairs is configured to communicate on a different wireless network and / or radio frequency band.

[0062] Figure 3 3 is a schematic block diagram illustrating components of a network device (NE) 300 according to one embodiment. The NE 300 may include a processor 302, a memory 304, an input device 306, a display 308, and a transceiver 310. As can be appreciated, the processor 302, the memory 304, the input device 306, the display 308, and the transceiver 310 may be similar to the processor 202, the memory 204, the input device 206, the display 208, and the transceiver 210 of the UE 200, respectively.

[0063] In some embodiments, the processor 302 controls the transceiver 310 to transmit a DL signal or data to the UE 200. The processor 302 may also control the transceiver 310 to receive a UL signal or data from the UE 200. In another example, the processor 302 may control the transceiver 310 to transmit a DL signal containing various configuration data to the UE 200.

[0064] In some embodiments, the transceiver 310 includes a transmitter 312 and a receiver 314. The transmitter 312 is used to send DL communication signals to the UE 200, and the receiver 314 is used to receive UL communication signals from the UE 200.

[0065] The transceiver 310 can communicate with multiple UEs 200 simultaneously. For example, the transmitter 312 can send DL communication signals to the UE 200. As another example, the receiver 314 can simultaneously receive UL communication signals from the UE 200. The transmitter 312 and the receiver 314 can be any suitable type of transmitter and receiver. Although only one transmitter 312 and one receiver 314 are shown, the transceiver 310 can have any suitable number of transmitters 312 and receivers 314. For example, the NE 300 can serve multiple cells and / or cell sectors, wherein the transceiver 310 includes a transmitter 312 and / or a receiver 314 for each cell or cell sector.

[0066] In 3GPP technical specification TS38.331, it is defined that the network or gNB can configure an RRC_CONNECTED UE, i.e. a UE in RRC_CONNECTED state, to perform measurements. The network can configure the UE to report measurement results according to the measurement configuration, or to perform conditional reconfiguration evaluation according to conditional reconfiguration. The measurement configuration is provided by means of dedicated signaling, i.e. using the RRCReconfiguration or RRCResume information elements.

[0067] The network may configure the UE to report measurement information based on SS / PBCH blocks, CSI-RS resources, SRS resources, or CLI-RSSI, or perform CBR measurements on the sidelink.

[0068] The measurement configuration includes the following parameters:

[0069] - Measurement objects, as a list of objects on which the UE should perform measurements (e.g. carrier frequency, reference signal (RS) frequency / time position);

[0070] - Report configurations, as a list of report configurations (e.g., reporting standard / format, RS type), where each measurement object may have one or more report configurations;

[0071] - measurement identity, which links a measurement object with a reporting configuration, to be included in the measurement report, used as a reference to the network;

[0072] - a quantity configuration which defines the measurement filter configuration used for all event evaluations and related reporting, as well as for periodic reporting of that measurement; and

[0073] - Measurement interval, which is a period that the UE can use to perform measurements.

[0074] A UE in RRC_CONNECTED state maintains a list of measurement objects, a list of reporting configurations and a list of measurement identities according to signaling and procedures. The list of measurement objects may include NR measurement objects, cross-link interference (CLI) measurement objects and inter-RAT objects. Similarly, the list of reporting configurations includes NR and inter-RAT reporting configurations. Any measurement object can be linked to any reporting configuration of the same RAT type.

[0075] FIG. 4A to FIG. 4C Figure 2 shows three use cases with AI / ML approaches being investigated in Release 18, namely: 1) CSI feedback enhancement, 2) beam management, and 3) positioning accuracy improvement.

[0076] CSI Feedback Enhancement

[0077] Figure 4Ais a diagram illustrating an example of using an AI / ML method to compress CSI to reduce CSI feedback overhead according to some embodiments of the present disclosure.

[0078] The potential goals for CSI enhancements are overhead reduction, as well as improved accuracy and prediction. Figure 4A As shown in , an AI-based method for reducing CSI feedback overhead may include an autoencoder (constructed by encoder 410a and decoder 410b) trained to compress downlink CSI. The encoder 410a constructed with NN is deployed at the UE side 200, and the decoder 410b constructed with paired NN is deployed at the gNB side 300.

[0079] With this structure, DL CSI 412 is compressed by encoder 410a using AI model 1, and then the compressed CSI 414 as the output of AI model 1 is sent over the air, whose overhead is assumed to be less than the traditional non-AI method, i.e., the method based on Release 16 Type II codebook. The compressed CSI 414 is decoded by decoder 410b using AI model 2 at gNB 300 to obtain recovered CSI 416. In general, the input 412 of AI model 1 at UE 200 can be the original channel or the feature vector with preprocessing, while the output 416 of AI model 2 at gNB 300, i.e., the recovered CSI 416, can be the reconstructed channel or feature vector, respectively. The paired models (referred to as autoencoders in the AI ​​community) are assumed to be trained with unsupervised learning, i.e., minimizing the difference between the input and the output.

[0080] Beam management

[0081] Figure 4B is a schematic diagram illustrating an example of using an AI / ML approach for beam measurement with less overhead according to some embodiments of the present disclosure.

[0082] The key issue for beam management enhancement use cases may be beam prediction in the temporal and / or spatial domains to reduce overhead and latency and improve beam selection accuracy. A typical deployment using AI / ML approaches is to apply AI models to assist in optimal beam selection with fewer resources and potentially lower latency, such as Figure 4B As shown in the figure.

[0083] In this example, it is assumed that there are eight Tx beams, each of which is represented by a Tx beam index from #0 to #7; and there are two Rx beams, each of which is represented by an Rx beam index from #0 to #1. Each Tx beam and each Rx beam are formed in Figure 4BThe four beam pairs shaded may be measured and the L1-RSRP measurement results 422 may be input into the AI ​​model 420. With the trained AI model 420, using measurements from some resources, e.g., from 4 of the 16 resources as illustrated in the example, the best Tx beam index and corresponding L1-RSRP 424 may be obtained and reported.

[0084] In this case, if the AI ​​model 420 is trained and deployed at the UE side, the gNB can potentially configure fewer CSI-RS resources, or the UE can use fewer resources for beam training or beam tracking, which can reduce RS overhead and processing latency at the UE side.

[0085] Enhanced positioning accuracy

[0086] Figure 4C is a schematic diagram illustrating an example of using an AI / ML method to enhance positioning accuracy according to some embodiments of the present disclosure.

[0087] The key issue in this case is to enhance the positioning accuracy using AI / ML methods in scenarios with severe non-line-of-sight (NLOS) conditions, as the positioning accuracy using traditional methods is not good enough. Figure 4C A typical deployment utilizing AI / ML approaches is illustrated in Figure .

[0088] In this example, based on the input of measurement results 432, and optionally some side information 434, AI model 430 can provide an estimated position 436.

[0089] Using this approach, the measurements and possible side information can be used to estimate the location, e.g., the location of the gNB.

[0090] In 3GPP Release 18, it is expected that AI / ML methods will be introduced to improve and / or enhance system performance, especially air interface performance. In some examples, AI / ML models can be used as a "replacement" or "additional", such as Figure 5A and Figure 5B As shown in the figure.

[0091] Figure 5A 504 is a schematic diagram illustrating an example of using an AI model to replace a communication module according to some embodiments of the present disclosure. Because input data 512 can be input to the communication module 502 or the AI ​​model 504 to obtain output data 514, the AI ​​model 504 is used as a substitute for the communication module 502. Figure 5B504 is a schematic diagram illustrating an example of using an AI model to assist a communication module according to some embodiments of the present disclosure. Because input data 512 is input to both the communication module 502 and the AI ​​model 504, the AI ​​model 504 is used together with the communication module 502.

[0092] Under the current 3GPP specification, a communication module including a signal processing module and a protocol implementation that meets the specification can always work to support radio connections between nodes. Therefore, the deployment of AI models for air interfaces can be enabled or disabled.

[0093] In general, AI functions can be deployed only at the gNB side, only at the UE side, or at both sides. FIG. 6A to FIG. 6C 1 is a schematic diagram illustrating examples of AI model deployment scenarios with AI models at the gNB side, at the UE side, and at both sides, respectively, according to some embodiments of the present disclosure. Fig. 6A In the embodiment, three AI models 611 including AI model 1, AI model 2 and AI model 3 are deployed at gNB 300; Figure 6B In , two AI models are deployed at the UE side, where AI model 1' 621 is deployed at UE1 200a and AI model 3' 622 is deployed at UE2 200b; and Figure 6C In the figure, three AI models 631 are deployed at gNB 300, and at the same time, AI model 1' 632 is deployed at UE1 200a, and AI model 3' 633 is deployed at UE2 200b.

[0094] To monitor the performance of AI models, there are already two simple methods: one (i.e., method A) is to directly compare the results of the AI ​​model with non-AI methods (method A-1) or ground truth (method A-2); the other (i.e., method B) is to monitor link performance indirectly using the AI ​​model. These methods are explained as follows.

[0095] - Method A-1: ​​Compare the inference results of the AI ​​model with those of the non-AI method.

[0096] In this method, the results of the AI ​​method and the non-AI method are compared; and if the results of the AI ​​method are better than the non-AI method, the AI ​​model can be considered to be an appropriate and healthy model. Otherwise, the AI ​​model is considered to be underperforming. However, in this method, it may be necessary to repeat signal processing, i.e., one using the AI ​​method and the other using the non-AI method, which results in additional processing overhead.

[0097] - Method A-2: Compare the AI ​​model’s inference results with the ground truth (if available).

[0098] In this method, if the ground truth is available, the results of the AI ​​method are compared with the ground truth. If the difference, for example, the mean squared error (MSE) is greater than a threshold, the AI ​​model can be considered as a poorly performing model. However, in this method, the ground truth needs to be available at the node, which may require additional transmission overhead.

[0099] - Method B: Monitoring the AI ​​model enhanced radio link performance such as throughput and BLER.

[0100] In this approach, conventional measurement methods can be used, and if the measurement results using the AI ​​method become worse, i.e., the expected degradation is detected, the AI ​​model can be considered as an underperforming model. However, a radio link may degrade due to many factors such as fading and interference.

[0101] The above-mentioned typical methods for monitoring deployed AI models may require different signaling over the air interface.

[0102] In this disclosure, a set of signaling to support the AI ​​model performance monitoring process is proposed, including signaling for registration, configuration, event triggering and result reporting. This disclosure focuses on the case where the AI ​​model is deployed at the UE side for reasoning, that is, Figure 6B and Figure 6C The situation shown in the figure.

[0103] Figure 7 is a schematic diagram illustrating an example of an AI model performance monitoring process according to some embodiments of the present disclosure. The proposed AI model performance monitoring process includes five main steps with corresponding signaling: AI capability registration, AI model activation, AI model monitoring configuration, AI model monitoring result reporting, and AI model deactivation.

[0104] In some examples, these five steps may not be as follows: Figure 7 The steps are performed in the order presented in the figure, and some steps may even be omitted.

[0105] Typically, a UE with AI capabilities may first be registered in the network to indicate the deployed AI model, i.e., AI capability registration. The signaling that may be used in this step is AI_Capability_Register 711 and AI_Model_Registered 712.

[0106] Once the scenario and configuration meet the conditions for activation of the AI ​​model, the corresponding AI model can be activated and applied to subsequent operations such as compressed CSI and predicted beam, that is, AI model activation. The signaling that can be used in this step is AI_Model_Activation_Req 721, AI_Model_Activation 722, and AI_Model_Activation_Ack723.

[0107] With the model activated, the performance can then be monitored from time to time; and the monitoring behavior can be configured in the newly designed configuration, i.e., the AI ​​Model Monitor Configuration. The signaling that can be used in this step is AI_Model_Monitor_Config 731 .

[0108] According to further configuration, related interactions can be further defined, including auxiliary information and reports, and the AI ​​model monitoring results can be reported accordingly, that is, the AI ​​model monitoring result report. The signaling that can be used in this step is AI_Model_Monitor_ReportConfig 741, AI_Model_Monitor_Trigger 742, and AI_Model_Monitor_AssistInfo 743.

[0109] If the activated model is found to perform poorly, it will be deactivated, ie, AI model deactivation. Signaling that may be used in this step is AI_Model_Deactivation_Req 751 , AI_Model_Deactivation 752 , and AI_Model_Deactivation_Ack 753 .

[0110] The signaling may be sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

[0111] In detail, refer to Figure 7 Signaling within the design and interpretation steps. In this disclosure, the message or signaling sent by the gNB to configure the UE may also be referred to as "configuration signaling".

[0112] AI capability registration: AI_Capability_Register

[0113] If the UE, e.g., UE 200, has the capability to support AI-based methods, a message may be used to indicate such capability to the network or gNB 300 and cause the deployed AI model to be registered in the network. The message or signaling, i.e., AI_Capability_Register 711, may include one or more of the following information:

[0114] A description of the hardware that supports the AI ​​model (i.e., neural network, NN), such as the number of floating point operations per second (FLOPS), memory size, and bit width for the AI ​​model to be registered;

[0115] Description of software that accelerates operations on NNs, such as algorithms optimized for convolutional NNs or dense NNs;

[0116] Description and identification of deployed AI models that enhance communication modules, such as models for CSI feedback compression or beam management;

[0117] Description of the scenarios and / or configurations of the deployed AI models that assist the network in deciding model selection, such as non-line-of-sight (NLOS) / line-of-sight (LOS), indoor / outdoor; and

[0118] Input and / or output information required for AI model training (e.g., offline / online) and inference. For example, some AI models for beam prediction require auxiliary information such as UE location and measured beams as input.

[0119] That is, the UE may send a message indicating a set of AI models to be registered with the network, and the message may include a model identification (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

[0120] After receiving this message, the model deployed in the UE will be registered at the network with the corresponding identity. The network 300 can confirm the registered AI model to the UE using signaling such as AI_Model_Registered 712 through appropriate selection processing.

[0121] The model for UE registration may be managed in a table, an example of which is illustrated in Table 1, which may be configured by the network according to UE capabilities.

[0122] surface . Registered AI models

[0123] PHY module 1 (e.g., CSI reporting) PHY module 2 (e.g., beam management) PHY module 3 (e.g., positioning) Scene 1 (e.g., indoors) Model 1-1 Model 1-2 Model 1-3 Scenario 2 (e.g., city) Model 2-1 Model 2-2 Model 2-3 Scenario 3 (e.g., HST) Model 3-1 Model 3-2 Model 3-3

[0124] In the example of Table 1, the registered AI models are specified for three different scenarios for each of the three use cases discussed above.

[0125] In some examples, information of registered AI capabilities may also be reported as part of the current UE capability report.

[0126] AI model activation: AI_Model_Activation

[0127] The signal, ie, AI_Model_Activation 722, is used to activate one or more AI models registered in the network when certain conditions (eg, the scenario and / or configuration of the registered models) are met.

[0128] The signal may include one or more of the following information:

[0129] The model identifier to be activated for the UE, which is selected from the models registered with the network in the previous step of AI capability registration;

[0130] The associated information of whether the UE activates the model, such as the computing resources required for the AI ​​operation. An example of computing resources is the inference time (e.g., processing latency), which is defined as the period from the time slot in which the UE receives the input of the AI ​​model to the first UL symbol to obtain the corresponding AI output, which can be defined as the number of symbols per SCS;

[0131] The duration of model activation, which is used to indicate how long the model is activated after receiving the signal, which is defined as the time between the UE receiving the activation command and the first symbol / slot in which the model is ready to be used; and

[0132] The duration of model activation, which indicates how long the model is activated before it is deactivated.

[0133] That is, configuration signaling for activating one or more AI models registered in the network may be sent to the UE and include a model ID for each AI model to be activated and associated requirements regarding operation.

[0134] By default, the scenario and configuration of the UE are estimated and decided by the network, which may have more environmental information than the UE. Fig. 8A is a schematic diagram illustrating an example of an AI model activation process with scenario identification at the gNB or network side according to some embodiments of the present disclosure. Fig. 8AAs shown in , after the network or gNB 300 identifies or estimates the scenario and configuration of the UE and the associated requirements regarding AI operation 801, the network or gNB 300 indicates to the UE 200 the target model to be activated 811 in the signal AI_Model_Activation 722. If the requirements (e.g., latency) can be met for the model being activated (i.e., the target model), the UE 200 will confirm the selection of the model 812 by sending an acknowledgment back to the network or gNB via AI_Model_Activation_Ack 723 to indicate whether it agrees to activate or can activate the model, as the UE may not be able to support the model in some cases, such as those requiring a high computational load.

[0135] In the example, if the gNB or network estimates that the registered UE is in a scenario defined in Table 1, for example, scenario 2, it will attempt to activate model 2-1 to enhance the UE's CSI reporting module. Because the network does not have information about the computational load in the UE, the recommended requirements for AI operations are also associated for reference.

[0136] In some cases, the UE itself is able to identify the scenario and configuration. Figure 8B is a schematic diagram illustrating an example of an AI model activation process with scenario identification at the UE side according to some embodiments of the present disclosure. Figure 8B In the example illustrated in , UE 200 identifies the scenario and configuration 821, and it requests the network or gNB 300 to activate the corresponding AI model 831 via AI_Model_Activation_Req 721. The signal may include the following information:

[0137] - The identification of the proposed model, which has been registered at the network.

[0138] In some examples, it is assumed that the network will decide whether to activate the model, while the UE may only provide a recommendation. Therefore, after receiving AI_Model_Activation_Req 721 from UE 200, network 300 identifies 822 the requirement for AI operation and indicates 832 the target model to UE 200 via AI_Model_Activation 722.

[0139] In case the model is requested or recommended by the UE via AI_Model_Activation_Req 721 , confirmation information (ie, AI_Model_Activation_Ack 723 ) is not necessary because the network decides whether to activate the model according to the overhead reported in AI_Model_Activation_Req 721 .

[0140] The activation message may be sent using RRC IE, MAC-CE or DCI.

[0141] If the activation message is sent via RRC, for example, an RRC reconfiguration message, the confirmation signal or message may be an RRC reconfiguration complete message.

[0142] If the activation message is sent via MAC CE, the confirmation message may be a HARQ-ACK corresponding to the PDSCH carrying the MAC CE and received from the The activated model will be applied starting from the first time slot after n, where n is the time slot in which the HARQ-ACK is sent and is the number of slots per subframe for subcarrier spacing configuration μ.

[0143] If the activation message is sent via DCI with DL assignment, the confirmation message can be a dedicated MAC CE in response to receiving or applying a model activation command; or the confirmation message is a HARQ-ACK corresponding to the scheduled PDSCH, and the activation is applied starting from the first time slot after time slot n+M, where M is the number of OFDM symbols per SCS reported or configured according to the UE capability.

[0144] In some other examples, the UE may decide to activate the AI ​​model when it identifies the scenario and configuration, and report the activated AI model to the gNB.

[0145] AI model monitoring configuration: AI_Model_Monitor_Config

[0146] After one or more AI models are activated at the UE side, the performance of the activated models will be monitored according to a monitoring configuration called AI_Model_Monitor_Config 731 in this example, which may include one or more of the following information for monitoring UE behavior:

[0147] -For the model identifier to be monitored by the UE, it is selected from the activated models in the previous step of AI model activation;

[0148] -Timing information for performance monitoring. In the time domain, AI model performance monitoring can be configured to be performed in a periodic, semi-persistent, or aperiodic manner.

[0149] oFor periodic monitoring, configure the periodicity value.

[0150] oFor semi-persistent monitoring, a specific periodicity is configured via MAC CE.

[0151] o For aperiodic monitoring, no periodicity is configured and the UE is explicitly triggered by each monitoring operation, such as by means of DCI signaling.

[0152] - UE triggers performance monitoring event definition, such as thresholds and / or timers to be reported, which are associated with the AI_Model_Monitor_Trigger message;

[0153] - Identification of candidate AI models to be monitored by the UE, which enables the UE to further derive results about the candidate AI models for subsequent deactivation and activation;

[0154] -Indications on reference signals to measure when required by the model, such as CSI-RS / SSB for AI models for beam management.

[0155] In the present disclosure, AI_Model_Monitor_Config 731 may also be referred to as configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models. The configuration signaling for monitoring the performance of the activated AI model may include a model ID for each AI model to be monitored, and timing information and / or trigger events for monitoring. It may also include timing information and / or trigger events for reporting the results of performance monitoring.

[0156] The UE follows the configuration in AI_Model_Monitor_Config 731 to perform measurements on monitoring occasions to monitor model performance and report the results accordingly.

[0157] AI model monitoring result report: AI_Model_Monitor_ReportConfig

[0158] In the step of AI model monitoring result reporting, the model monitoring result associated with an AI_Model_Monitor_Config 731 will be reported as a measurement result according to the report configuration. AI_Model_Monitor_ReportConfig 741 may include one or more items of the following information:

[0159] - Model identifier of the monitored model, which should be the same as the identifier in the corresponding monitoring configuration, i.e., AI_Model_Monitor_Config.

[0160] -Measurements of the model to be monitored, such as the output of the AI ​​model and the non-AI approach, the performance gap between the monitored AI model and the non-AI approach (i.e., the baseline), and / or the performance gap between the monitored AI model and the candidate AI model (if configured). Fig. 9is a schematic diagram illustrating an example of a measurement trigger for monitoring an activated AI model according to some embodiments of the present disclosure. In this example, three different measurement results generated from the same input 912 are available based on three different methods / models. There are: 1) a non-AI result 922 based on a non-AI method 902, 2) an AI result 924 of model 1 based on AI model 1 904, and 3) an AI result 926 of model 2 based on AI model 2 906. For example, in the case where the performance of AI model 1 904 is to be monitored, the performance gap between the monitored AI model and the non-AI method can be the gap between results 924 and 922; and the performance gap between the monitored AI model and the candidate AI model can be the gap between results 924 and 926.

[0161] - Monitoring report configuration, such as the type of content (e.g., outputs of AI models and non-AI methods or the gap between them) and transmission format (e.g., PUSCH or PUCCH).

[0162] - Indication for the UE to enable a baseline - ie, non-AI method - to derive results to report as a reference for performance monitoring.

[0163] o Alternatively, a baseline functionality without AI functionality, e.g., a CSI-ReportConfig without an AI model, can be associated with a CSI-ReportConfig with an AI model for monitoring.

[0164] - An indication of the computational resource consumption of the model, such as processing latency and / or power consumption.

[0165] In the present disclosure, AI_Model_Monitor_ReportConfig 741 may also be referred to as configuration signaling for reporting the results of performance monitoring. The configuration signaling for reporting the results of performance monitoring may include a model ID, content, and transmission format for the performance report.

[0166] Accordingly, the monitoring report or performance report may include measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

[0167] The content of the monitoring report can be multiplexed in PUSCH or PUCCH according to the configuration. It is worth noting that the report can be periodic, semi-persistent, or aperiodic, which follows the configuration of its corresponding monitoring operation.

[0168] As mentioned in the step of AI model monitoring configuration, if the UE is configured to trigger reporting by local measurement, then if the pre-configured condition is met, i.e., the output of the monitored AI model performs poorly compared to non-AI methods or candidate AI models according to the configured threshold, then a report will be triggered. Signaling of the event trigger called AI_Model_Monitor_Trigger 742 will be sent from the UE, which may include one or more of the following information:

[0169] - Indications of underperformance events for activated AI models;

[0170] - Model identification of underperforming AI models, which should be used as an indicator for performance monitoring;

[0171] - Identification of candidate models for measurement.

[0172] As event-triggered signaling, this message can only be a pre-configured MAC CE for event-triggered monitoring or a DCI for aperiodic monitoring.

[0173] Optionally, signaling indicating auxiliary information AI_Model_Monitor_AssistInfo 743 to assist monitoring may be sent to the UE, which may include the following information:

[0174] - Ground truth data transmission for monitoring of target models;

[0175] In general, enabling more methods (non-AI methods and other AI models, i.e., candidate AI models) to monitor model performance may require more power consumption and complexity. Therefore, there may be some trade-offs between appropriate model selection and beneficial performance from the model.

[0176] AI model deactivation: AI_Model_Deactivation

[0177] If the configured activation duration expires or the performance of the activated AI model is several times worse than the non-AI method or other AI models with respect to a threshold, where the threshold and the number of times are defined in the monitoring configuration, the model will be deactivated accordingly via the signal AI_Model_Deactivation 752, which may include one or more of the following information:

[0178] - the identifier of the model to be deactivated, which should be one of the activated models; and

[0179] - The identification of the model recommended for the next activation (if any).

[0180] That is, the deactivation signal or message from the network to the UE includes one or more model IDs to be deactivated; and it may further include one or more model IDs for the recommended AI models.

[0181] In some cases, if the UE is able to identify the scenario or configuration and / or poor performance by itself, it may request the network to deactivate the AI ​​model via AI_Model_Deactivation_Req 751, which includes the identification of the model to be deactivated, and a value indicating the reason for deactivation, which may be added to the deactivation request sent by the UE, for example, deactivation may be requested due to degraded performance or lack of processing power, etc. It is up to the network to ultimately decide whether to deactivate the AI ​​model, and to send the deactivation signal AI_Model_Deactivation 752 to the UE 200.

[0182] Upon receiving the deactivation signal from the network, the UE 200 may send an acknowledgement back to the network to indicate whether the model is deactivated via AI_Model_Deactivation_Ack 753. After deactivation, the model will not be used to derive results until it is activated again.

[0183] Fig.10 is a flow chart illustrating steps of monitoring an AI model in the RAN by the UE 200 according to some embodiments of the present disclosure.

[0184] At step 1002 , the receiver 214 of the UE 200 receives first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and second configuration signaling for reporting a result of the performance monitoring.

[0185] At step 1004, the processor 202 of the UE 200 generates a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling.

[0186] At step 1006, the transmitter 212 of the UE 200 sends a performance report.

[0187] Fig.11 is a flowchart illustrating steps for monitoring an AI model in the RAN by a gNB 300 according to some embodiments of the present disclosure.

[0188] At step 1102, the transmitter 312 of the gNB 300 sends a first configuration signaling for monitoring the performance of one or more activated artificial intelligence (AI) models, and a second configuration signaling for reporting the results of the performance monitoring.

[0189] At step 1104, the receiver 314 of the gNB 300 receives a performance report of the AI ​​model generated and reported according to the first configuration signaling and the second configuration signaling.

[0190] In one aspect, some items as examples of the present disclosure regarding a UE may be summarized as follows:

[0191] 1. A device comprising:

[0192] a receiver that receives first configuration signaling for monitoring performance of one or more artificial intelligence (AI) models that are activated, and second configuration signaling for reporting a result of the performance monitoring;

[0193] a processor, wherein the processor generates a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; and

[0194] A transmitter sends the performance report.

[0195] 2. The apparatus of item 1, wherein the transmitter further sends a first message indicating a set of AI models to be registered.

[0196] 3. An apparatus according to item 2, wherein the receiver further receives a third configuration signaling for activating a subset of the AI ​​model.

[0197] 4. An apparatus according to item 2, wherein the first message includes: a model identifier (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

[0198] 5. An apparatus according to item 3, wherein the third configuration signaling includes: a model ID for each activated AI model in the subset of the AI ​​models and associated requirements regarding operation.

[0199] 6. The apparatus according to item 5, wherein the transmitter further sends a request message for triggering transmission of the third configuration signaling.

[0200] 7. An apparatus according to item 1, wherein the first configuration signaling includes: a model ID for each of the AI ​​models to be monitored, and timing information and / or trigger events for monitoring.

[0201] 8. The apparatus according to item 7, wherein the first configuration signaling further includes timing information and / or triggering events for reporting the results of the performance monitoring.

[0202] 9. The apparatus according to item 1, wherein the second configuration signaling comprises: a model ID, content, and transmission format for the performance report.

[0203] 10. An apparatus according to item 1 or 9, wherein the performance report includes: measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

[0204] 11. An apparatus according to item 1, wherein the transmitter further sends a second message for deactivating one or more of the activated AI models.

[0205] 12. An apparatus according to item 11, wherein the second message includes one or more model IDs to be deactivated.

[0206] 13. An apparatus according to item 12, wherein the second message further includes one or more model IDs for the recommended AI model.

[0207] 14. An apparatus according to item 11, wherein the receiver further receives a fourth configuration signaling for deactivating one or more AI models.

[0208] 15. The apparatus of any one of items 1 to 14, wherein the configuration signaling and the message are sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

[0209] In another aspect, some items as examples of the present disclosure regarding gNB may be summarized as follows:

[0210] 16. An apparatus comprising:

[0211] a transmitter that sends a first configuration signaling for monitoring performance of one or more artificial intelligence (AI) models that are activated, and a second configuration signaling for reporting a result of the performance monitoring;

[0212] A receiver that receives a performance report of the AI ​​model generated and reported based on the first configuration signaling and the second configuration signaling.

[0213] 17. The apparatus of claim 16, wherein the receiver further receives a first message indicating a set of AI models to be registered.

[0214] 18. An apparatus according to item 17, wherein the transmitter further sends a third configuration signaling for activating a subset of the AI ​​model.

[0215] 19. An apparatus according to item 17, wherein the first message includes: a model identifier (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

[0216] 20. An apparatus according to item 18, wherein the third configuration signaling includes: a model ID for each activated AI model in the subset of AI models and associated requirements regarding operation.

[0217] 21. An apparatus according to item 20, wherein the receiver further receives a request message for triggering transmission of the third configuration signaling.

[0218] 22. An apparatus according to item 16, wherein the first configuration signaling includes: a model ID for each of the AI ​​models to be monitored, and timing information and / or trigger events for monitoring.

[0219] 23. An apparatus according to item 22, wherein the first configuration signaling further includes timing information and / or triggering events for reporting the results of the performance monitoring.

[0220] 24. The apparatus according to item 16, wherein the second configuration signaling includes: a model ID, content, and transmission format for the performance report.

[0221] 25. An apparatus according to item 16 or 24, wherein the performance report includes: measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

[0222] 26. An apparatus according to item 16, wherein the receiver further receives a second message for deactivating one or more of the activated AI models.

[0223] 27. An apparatus according to item 26, wherein the second message includes one or more model IDs to be deactivated.

[0224] 28. An apparatus according to item 27, wherein the second message further includes one or more model IDs for the recommended AI model.

[0225] 29. An apparatus according to item 26, wherein the transmitter further sends a fourth configuration signaling for deactivating one or more AI models.

[0226] 30. An apparatus according to any one of items 16 to 29, wherein the configuration signaling and the message are sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

[0227] In yet another aspect, some items as examples of the present disclosure regarding the UE method may be summarized as follows:

[0228] 31. A method comprising:

[0229] Receiving, by a receiver, first configuration signaling for monitoring performance of one or more activated artificial intelligence (AI) models, and second configuration signaling for reporting a result of the performance monitoring;

[0230] generating, by a processor, a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; and

[0231] The performance report is sent by the transmitter.

[0232] 32. A method according to item 31, wherein the transmitter further sends a first message indicating a set of AI models to be registered.

[0233] 33. A method according to item 32, wherein the receiver further receives a third configuration signaling for activating a subset of the AI ​​model.

[0234] 34. A method according to item 32, wherein the first message includes: a model identifier (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

[0235] 35. A method according to item 33, wherein the third configuration signaling includes: a model ID for each activated AI model in the subset of AI models and associated requirements regarding operation.

[0236] 36. A method according to item 35, wherein the transmitter further sends a request message for triggering the transmission of the third configuration signaling.

[0237] 37. A method according to item 31, wherein the first configuration signaling includes: a model ID for each of the AI ​​models to be monitored, and timing information and / or trigger events for monitoring.

[0238] 38. A method according to item 37, wherein the first configuration signaling further includes timing information and / or triggering events for reporting the results of the performance monitoring.

[0239] 39. A method according to item 31, wherein the second configuration signaling includes: a model ID, content, and transmission format for the performance report.

[0240] 40. A method according to item 31 or 39, wherein the performance report includes: measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

[0241] 41. A method according to item 31, wherein the transmitter further sends a second message for deactivating one or more of the activated AI models.

[0242] 42. A method according to item 41, wherein the second message includes one or more model IDs to be deactivated.

[0243] 43. A method according to item 42, wherein the second message further includes one or more model IDs for the recommended AI model.

[0244] 44. A method according to item 41, wherein the receiver further receives a fourth configuration signaling for deactivating one or more AI models.

[0245] 45. A method according to any one of items 31 to 44, wherein the configuration signaling and the message are sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

[0246] In yet another aspect, some items as examples of the method of the present disclosure regarding gNB may be summarized as follows:

[0247] 46. ​​A method comprising:

[0248] Sending, by the transmitter, first configuration signaling for monitoring performance of one or more activated artificial intelligence (AI) models, and second configuration signaling for reporting a result of the performance monitoring;

[0249] A performance report of the AI ​​model generated and reported according to the first configuration signaling and the second configuration signaling is received by a receiver.

[0250] 47. The method of claim 46, wherein the receiver further receives a first message indicating a set of AI models to be registered.

[0251] 48. A method according to item 47, wherein the transmitter further sends a third configuration signaling for activating a subset of the AI ​​model.

[0252] 49. A method according to item 47, wherein the first message includes: a model identifier (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

[0253] 50. A method according to item 48, wherein the third configuration signaling includes: a model ID for each activated AI model in the subset of AI models and associated requirements regarding operation.

[0254] 51. A method according to item 50, wherein the receiver further receives a request message for triggering the transmission of the third configuration signaling.

[0255] 52. A method according to item 46, wherein the first configuration signaling includes: a model ID for each of the AI ​​models to be monitored, and timing information and / or trigger events for monitoring.

[0256] 53. A method according to item 52, wherein the first configuration signaling further includes timing information and / or triggering events for reporting the results of the performance monitoring.

[0257] 54. A method according to item 46, wherein the second configuration signaling includes: a model ID, content, and transmission format for the performance report.

[0258] 55. A method according to item 46 or 54, wherein the performance report includes: measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

[0259] 56. A method according to item 46, wherein the receiver further receives a second message for deactivating one or more of the activated AI models.

[0260] 57. A method according to item 56, wherein the second message includes one or more model IDs to be deactivated.

[0261] 58. A method according to item 57, wherein the second message further includes one or more model IDs for the recommended AI model.

[0262] 59. A method according to item 56, wherein the transmitter further sends a fourth configuration signaling for deactivating one or more AI models.

[0263] 60. A method according to any one of items 46 to 59, wherein the configuration signaling and the message are sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

[0264] Various embodiments and / or examples are disclosed to provide exemplary and explanatory information to enable those skilled in the art to put the present disclosure into practice. Unless otherwise specifically indicated, features or components disclosed with reference to one embodiment or example also apply to all embodiments or examples.

[0265] Embodiments may be practiced in other specific forms. The described embodiments should be considered in all respects as merely illustrative and not restrictive. Therefore, the scope is indicated by the appended claims rather than by the foregoing description. All changes that fall within the meaning and range of equivalents of the claims should be included within their scope.

Claims

1. A device comprising: a receiver that receives first configuration signaling for monitoring performance of one or more artificial intelligence (AI) models that are activated, and second configuration signaling for reporting results of the performance monitoring; A processor, wherein the processor generates a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; as well as A transmitter sends the performance report.

2. The device according to claim 1, wherein: The transmitter further sends a first message indicating a set of AI models to be registered.

3. The device according to claim 2, wherein: The receiver further receives third configuration signaling for activating the subset of the AI ​​model.

4. The device according to claim 2, wherein: The first message includes: a model identifier (ID) for each corresponding AI model in the set of AI models, one or more deployment scenarios, and configuration parameters.

5. The device according to claim 3, wherein: The third configuration signaling includes: a model ID and associated requirements on operation for each activated AI model in the subset of AI models.

6. The device according to claim 5, wherein: The transmitter further sends a request message for triggering transmission of the third configuration signaling.

7. The device according to claim 1, wherein: The first configuration signaling includes: a model ID for each of the AI ​​models to be monitored, and timing information and / or a trigger event for the monitoring.

8. The device according to claim 7, wherein: The first configuration signaling further includes timing information and / or a triggering event for reporting the result of the performance monitoring.

9. The device according to claim 1, wherein: The second configuration signaling includes: a model ID, content, and transmission format for the performance report.

10. The device according to claim 1 or 9, wherein: The performance report includes: measurement results of the output of the monitored AI model, measurement results of the output of the indicated non-AI method, measurement results of the performance gap between the monitored AI model and the indicated non-AI method and / or between the monitored AI model and other AI models, and computational overhead.

11. The device according to claim 1, wherein: The transmitter further sends a second message for deactivating one or more of the activated AI models.

12. The device according to claim 11, wherein The second message includes one or more model IDs to be deactivated; the second message further includes one or more model IDs for the recommended AI model; and / or the receiver further receives a fourth configuration signaling for deactivating one or more AI models.

13. The device according to any one of claims 1 to 12, wherein: The configuration signaling and the message are sent using a radio resource control (RRC) information element (IE), a medium access control-control element (MAC-CE), and / or downlink control information (DCI).

14. An apparatus comprising: a transmitter that sends a first configuration signaling for monitoring performance of one or more artificial intelligence (AI) models that are activated, and a second configuration signaling for reporting a result of the performance monitoring; A receiver that receives a performance report of the AI ​​model generated and reported based on the first configuration signaling and the second configuration signaling.

15. A method comprising: Receiving, by a receiver, first configuration signaling for monitoring performance of one or more activated artificial intelligence (AI) models, and second configuration signaling for reporting a result of the performance monitoring; Generating, by a processor, a performance report of the AI ​​model according to the first configuration signaling and the second configuration signaling; as well as The performance report is sent by the transmitter.

Images