Communication control method and user device

Abstract

A communication control method according to one aspect of the present invention is a method for controlling communication in a mobile communication system. The communication control method includes a step in which a user device acquires model identification information for identifying an AI / ML model and additional information of the AI / ML model. The communication control method also includes a step in which the user device transmits the model identification information to a network device. Here, said step for transmission includes a step in which the user device transmits the model identification information to the network device when it is determined on the basis of the additional information that inference with respect to the AI / ML model is possible.

Description

Communication control method and user device

[0001] The present disclosure relates to a communication control method and a user device.

[0002] In recent years, the Third Generation Partnership Project (3GPP) (registered trademark; the same applies hereinafter), a standardization project for mobile communication systems, has been studying the application of artificial intelligence (AI) technology, particularly machine learning (ML) technology, to wireless communication (air interface) in mobile communication systems.

[0003] 3GPP TR 38.843 V2.0.0 (2023-12)

[0004] A communication control method according to a first aspect is a communication control method in a mobile communication system. The communication control method includes a step in which a user device acquires model identification information that identifies an AI / ML model and additional information of the AI / ML model. The communication control method also includes a step in which the user device transmits the model identification information to a network device. Here, the transmitting step includes a step in which the user device transmits the model identification information to the network device when the user device determines, based on the additional information, that inference for the AI / ML model is possible.

[0005] A communication control method according to a second aspect is a communication control method in a mobile communication system. The communication control method includes a step in which a user device trains an AI / ML model. The communication control method also includes a step in which the user device transmits parameters used for training and additional information of the AI / ML model to a network device. Here, the network device trains the AI / ML model using the parameters.

[0006] A communication control method according to a third aspect is a communication control method in a mobile communication system. The communication control method includes a step in which a network device trains an AI / ML model. The communication control method also includes a step in which the network device transmits identification information of the AI / ML model and additional information of the AI / ML model to a user device. Here, the additional information is used in the user device to determine whether to execute inference of the AI / ML model and whether to transmit parameters used in the inference to the network device.

[0007] FIG. 1 is a diagram showing an example of the configuration of a mobile communication system according to the first embodiment. FIG. 2 is a diagram showing an example of the configuration of a UE (user equipment) according to the first embodiment. FIG. 3 is a diagram showing an example of the configuration of a gNB (base station) according to the first embodiment. FIG. 4 is a diagram showing an example of the configuration of a protocol stack according to the first embodiment. FIG. 5 is a diagram showing an example of the configuration of a protocol stack according to the first embodiment. FIG. 6 is a diagram showing an example of the configuration of functional blocks of AI / ML technology according to the first embodiment. FIG. 7 is a diagram showing an example of operation in AI / ML technology according to the first embodiment. FIG. 8 is a diagram showing an example of the arrangement of functional blocks of AI / ML technology according to the first embodiment. FIG. 9 is a diagram showing an example of operation according to the first embodiment. FIG. 10 is a diagram showing an example of operation according to the first embodiment. FIG. 11 is a diagram showing an example of a setting message according to the first embodiment. FIGS. 12(A) to 12(C) are diagrams showing an example of operation according to the first embodiment. FIG. 13 is a diagram showing an example of operation according to the first embodiment. FIG. 14 is a diagram showing an example of operation according to the first embodiment. FIG. 15 is a diagram showing an example of operation according to the first embodiment. Fig. 16 is a diagram illustrating an example of operation according to the first embodiment. Fig. 17 is a diagram illustrating an example of operation according to the first embodiment. Fig. 18 is a diagram illustrating an example of operation according to the first embodiment. Fig. 19 is a diagram illustrating an example of operation according to the first embodiment. Fig. 20 is a diagram illustrating an example of operation according to the first embodiment. Fig. 21 is a diagram illustrating an example of operation according to the first embodiment. Fig. 22 is a diagram illustrating an example of operation according to the first embodiment. Fig. 23 is a diagram illustrating an example of operation according to the first embodiment. Fig. 24 is a diagram illustrating an example of operation according to the first embodiment.

[0008] The present disclosure aims to efficiently perform LCM operations.

[0009] [First embodiment] A mobile communication system according to a first embodiment will be described with reference to the drawings. In the description of the drawings, the same or similar parts are denoted by the same or similar reference numerals.

[0010] (Configuration of mobile communication system) The configuration of a mobile communication system according to the first embodiment will be described. FIG. 1 is a diagram showing an example of the configuration of a mobile communication system 1 according to the first embodiment. The mobile communication system 1 conforms to the 5th Generation System (5GS) of the 3GPP standard. Although 5GS will be described below as an example, the mobile communication system may also be at least partially applied to an LTE (Long Term Evolution) system. The mobile communication system may also be at least partially applied to a sixth generation (6G) system or later system.

[0011] The mobile communication system 1 includes a user equipment (UE) 100, a 5G radio access network (NG-RAN: Next Generation Radio Access Network) 10, and a 5G core network (5GC: 5G Core Network) 20. Hereinafter, the NG-RAN 10 may be simply referred to as the RAN 10. Furthermore, the 5GC 20 may be simply referred to as the core network (CN) 20. Furthermore, devices included in the core network CN may be referred to as core network devices.

[0012] The UE 100 is a mobile wireless communication device. The UE 100 may be any device that is used by a user. For example, the UE 100 may be a mobile phone terminal (including a smartphone) and / or a tablet terminal, a notebook PC, a communication module (including a communication card or a chipset), a sensor or a device provided in a sensor, a vehicle or a device provided in a vehicle (Vehicle UE), or an aircraft or a device provided in an aircraft (Aerial UE).

[0013] The NG-RAN 10 includes a base station (called a "gNB" in a 5G system) 200. The gNBs 200 are connected to each other via an Xn interface, which is an interface between base stations. The gNB 200 manages one or more cells. The gNB 200 performs wireless communication with a UE 100 that has established a connection with its own cell. The gNB 200 has a radio resource management (RRM) function, a routing function for user data (hereinafter simply referred to as "data"), a measurement control function for mobility control and scheduling, and the like. The term "cell" is used to indicate the smallest unit of a wireless communication area. The term "cell" is also used to indicate a function or resource for wireless communication with the UE 100. One cell belongs to one carrier frequency (hereinafter simply referred to as "frequency").

[0014] In addition, gNBs can also be connected to the Evolved Packet Core (EPC), which is the core network of LTE. LTE base stations can also be connected to 5GC. LTE base stations and gNBs can also be connected via an inter-base station interface.

[0015] The 5GC20 includes an AMF (Access and Mobility Management Function) and a UPF (User Plane Function) 300. The AMF performs various mobility controls for the UE 100. The AMF manages the mobility of the UE 100 by communicating with the UE 100 using NAS (Non-Access Stratum) signaling. The UPF controls data forwarding. The AMF and UPF 300 are connected to the gNB 200 via an NG interface, which is an interface between a base station and a core network. The AMF and UPF 300 may be core network devices included in the CN 20. The core network device and the gNB 200 may be collectively referred to as a network device.

[0016] 2 is a diagram showing an example of the configuration of a UE 100 (user equipment) according to the first embodiment. The UE 100 includes a receiving unit 110, a transmitting unit 120, and a control unit 130. The receiving unit 110 and the transmitting unit 120 constitute a communication unit that performs wireless communication with the gNB 200. The UE 100 is an example of a communication device.

[0017] The receiving unit 110 performs various types of reception under the control of the control unit 130. The receiving unit 110 includes an antenna and a receiver. The receiver converts a radio signal received by the antenna into a baseband signal (received signal) and outputs the baseband signal to the control unit 130.

[0018] The transmitting unit 120 performs various transmissions under the control of the control unit 130. The transmitting unit 120 includes an antenna and a transmitter. The transmitter converts a baseband signal (transmission signal) output by the control unit 130 into a radio signal and transmits it from the antenna.

[0019] The control unit 130 performs various controls and processes in the UE 100. Such processes include processes of each layer described below. The control unit 130 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in the processes by the processor. The processor may include a baseband processor and a CPU (Central Processing Unit). The baseband processor performs modulation / demodulation, encoding / decoding, etc. of baseband signals. The CPU executes programs stored in the memory to perform various processes. Note that the processes or operations performed in the UE 100 may be performed in the control unit 130.

[0020] 3 is a diagram showing an example of the configuration of a gNB 200 (base station) according to the first embodiment. The gNB 200 includes a transmitter 210, a receiver 220, a controller 230, and a backhaul communication unit 250. The transmitter 210 and the receiver 220 constitute a communication unit that performs wireless communication with the UE 100. The backhaul communication unit 250 constitutes a network communication unit that communicates with the CN 20. The gNB 200 is another example of a communication device.

[0021] The transmitting unit 210 performs various transmissions under the control of the control unit 230. The transmitting unit 210 includes an antenna and a transmitter. The transmitter converts a baseband signal (transmission signal) output by the control unit 230 into a radio signal and transmits it from the antenna.

[0022] The receiving unit 220 performs various types of reception under the control of the control unit 230. The receiving unit 220 includes an antenna and a receiver. The receiver converts a radio signal received by the antenna into a baseband signal (received signal) and outputs the baseband signal to the control unit 230.

[0023] The control unit 230 performs various controls and processes in the gNB 200. Such processes include processes for each layer described below. The control unit 230 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in the processes by the processor. The processor may include a baseband processor and a CPU. The baseband processor performs modulation / demodulation, encoding / decoding, etc. of baseband signals. The CPU executes programs stored in the memory to perform various processes. Note that the processes or operations performed in the gNB 200 may be performed by the control unit 230.

[0024] The backhaul communication unit 250 is connected to adjacent base stations via an Xn interface, which is an interface between base stations. The backhaul communication unit 250 is connected to the AMF / UPF 300 via an NG interface, which is an interface between a base station and a core network. The gNB 200 may be configured (i.e., functionally divided) with a central unit (CU) and a distributed unit (DU), and the two units may be connected via an F1 interface, which is a fronthaul interface.

[0025] FIG. 4 is a diagram showing an example of the configuration of a protocol stack of a radio interface of a user plane that handles data.

[0026] The user plane air interface protocol includes a physical (PHY) layer, a medium access control (MAC) layer, a radio link control (RLC) layer, a packet data convergence protocol (PDCP) layer, and a service data adaptation protocol (SDAP) layer.

[0027] The PHY layer performs encoding / decoding, modulation / demodulation, antenna mapping / demapping, and resource mapping / demapping. Data and control information are transmitted between the PHY layer of UE100 and the PHY layer of gNB200 via a physical channel. The PHY layer of UE100 receives downlink control information (DCI) transmitted from gNB200 on a physical downlink control channel (PDCCH). Specifically, UE100 performs blind decoding of the PDCCH using a radio network temporary identifier (RNTI) and acquires successfully decoded DCI as DCI addressed to the UE. The DCI transmitted from gNB200 has a CRC (Cyclic Redundancy Code) parity bit scrambled by the RNTI added.

[0028] In NR, UE100 can use a bandwidth narrower than the system bandwidth (i.e., the cell bandwidth). The gNB200 configures the UE100 with a bandwidth portion (BWP) consisting of contiguous PRBs (Physical Resource Blocks). The UE100 transmits and receives data and control signals in the active BWP. For example, up to four BWPs may be configurable for the UE100. Each BWP may have a different subcarrier spacing. The BWPs may overlap in frequency. When multiple BWPs are configured for the UE100, the gNB200 can specify which BWP to apply by controlling the downlink. This allows the gNB200 to dynamically adjust the UE bandwidth according to the amount of data traffic of the UE100, etc., thereby reducing UE power consumption.

[0029] The gNB 200 can configure, for example, up to three control resource sets (CORESETs) for each of up to four BWPs on the serving cell. The CORESET is a radio resource for control information to be received by the UE 100. Up to 12 or more CORESETs may be configured on the serving cell for the UE 100. Each CORESET may have an index of 0 to 11 or more. The CORESET may consist of six resource blocks (PRBs) and one, two, or three consecutive Orthogonal Frequency Division Multiplex (OFDM) symbols in the time domain.

[0030] The MAC layer performs data priority control, retransmission processing using Hybrid Automatic Repeat reQuest (HARQ), random access procedures, etc. Data and control information are transmitted between the MAC layer of the UE 100 and the MAC layer of the gNB 200 via a transport channel. The MAC layer of the gNB 200 includes a scheduler. The scheduler determines the uplink and downlink transport format (transport block size, modulation and coding scheme (MCS)) and the resource blocks to be allocated to the UE 100.

[0031] The RLC layer transmits data to the receiving RLC layer using the functions of the MAC layer and PHY layer. Data and control information are transmitted between the RLC layer of the UE 100 and the RLC layer of the gNB 200 via a logical channel.

[0032] The PDCP layer performs header compression / decompression, encryption / decryption, and the like.

[0033] The SDAP layer maps IP flows, which are units for Quality of Service (QoS) control by the core network, to radio bearers, which are units for QoS control by the access stratum (AS). Note that if the RAN is connected to the EPC, SDAP may not be required.

[0034] FIG. 5 is a diagram showing the configuration of a protocol stack of a radio interface of a control plane that handles signaling (control signals).

[0035] The protocol stack of the radio interface of the control plane includes a radio resource control (RRC) layer and a non-access stratum (NAS) instead of the SDAP layer shown in FIG.

[0036] RRC signaling for various settings is transmitted between the RRC layer of UE100 and the RRC layer of gNB200. The RRC layer controls logical channels, transport channels, and physical channels according to the establishment, re-establishment, and release of radio bearers. When there is a connection (RRC connection) between the RRC of UE100 and the RRC of gNB200, UE100 is in an RRC connected state. When there is no connection (RRC connection) between the RRC of UE100 and the RRC of gNB200, UE100 is in an RRC idle state. When the connection between the RRC of UE100 and the RRC of gNB200 is suspended, UE100 is in an RRC inactive state.

[0037] The NAS, which is located above the RRC layer, performs session management, mobility management, etc. NAS signaling is transmitted between the NAS of the UE 100 and the NAS of the AMF 300. Note that the UE 100 has an application layer and the like in addition to the radio interface protocol. Also, the layer below the NAS is called an Access Stratum (AS).

[0038] (AI / ML Technology) Next, the AI / ML technology according to the embodiment will be described. Fig. 6 is a diagram showing an example of the configuration of functional blocks of the AI / ML technology in the mobile communication system 1 according to the first embodiment.

[0039] The functional block configuration example shown in FIG. 6 includes a data collection unit (Data Collection) A1, a model training unit (Model Training) A2, a model inference unit (Inference) A3, a management unit (Management) A5, and a model storage unit (Model Storage) A6.

[0040] The functional block configuration example shown in FIG. 6 represents a functional framework of a general AI / ML technology. Therefore, depending on a hypothetical use case, some of the functional block configuration example (e.g., model recording unit A6, etc.) may not be included in the functional block configuration example. The functional block configuration example shown in FIG. 6 may also be distributed between the UE 100 and a network-side device. Alternatively, some functions of the functional block configuration example (e.g., model learning unit A2 or model inference unit A3, etc.) may be located in both the UE 100 and the network-side device.

[0041] The data collection unit A1 provides input data to the model learning unit A2, the model inference unit A3, and the management unit A5. The input data includes training data for the model learning unit A2, inference data for the model inference unit A3, and monitoring data for the management unit A5.

[0042] The training data is data required as input when the AI / ML model is learning. The inference data is data required as input when the AI / ML model is inferring. The monitoring data is data required as input when the AI / ML model is managing.

[0043] In addition, data collection may refer to the process of collecting data at a network node, a management entity, or a UE 100, for example, to train an AI / ML model, manage an AI / ML model, and perform inference on an AI / ML model.

[0044] The model learning unit A2 performs AI / ML model training, AI / ML model validation, and AI / ML model testing.

[0045] AI / ML model learning is the process of learning an AI / ML model from input / output relationships to obtain a trained AI / ML model to be used for inference. For example, considering y = ax + b, AI / ML model learning may be the process of optimizing a (slope) and b (intercept) by providing input (x) and output (y) (i.e., providing learning data).

[0046] Generally, machine learning includes supervised learning, unsupervised learning, and reinforcement learning. Supervised learning is a method that uses correct answer data as training data. Unsupervised learning is a method that does not use correct answer data as training data. For example, unsupervised learning memorizes feature points from a large amount of training data and determines the correct answer (estimates the range). Reinforcement learning is a method that assigns a score to the output result and learns how to maximize the score. Although supervised learning will be described below, either unsupervised learning or reinforcement learning may be applied as machine learning.

[0047] The model learning unit A2 outputs a trained AI / ML model (Trained Model) obtained by AI / ML model learning to the model recording unit A6, and also outputs an updated AI / ML model (Updated Model) obtained by relearning the trained AI / ML model to the model recording unit A6.

[0048] In the following, AI / ML model learning may be referred to as "model learning" or "learning."

[0049] The model inference unit A3 performs AI / ML model inference. Specifically, the model inference unit A3 applies the inference data provided by the data collection unit A1 to the trained AI / ML model (or updated AI / ML model) to obtain inference output data. For example, in the equation y = ax + b, x corresponds to the inference data and y corresponds to the inference output data. Note that "y = ax + b" is an AI / ML model. A model with optimized slope and intercept, for example, "y = 5x + 3," is a trained AI / ML model. There are various model approaches, including linear regression analysis, neural networks, and decision tree analysis. The above "y = ax + b" can also be considered a type of linear regression analysis.

[0050] The model inference unit A3 outputs inference output data to the management unit A5. The model inference unit A3 also receives management instructions from the management unit A5. For example, the management instructions include selection of an AI / ML model, activation (deactivation) of an AI / ML model, switching of an AI / ML model, and fallback (performing inference without using an AI / ML model). The model inference unit A3 performs model inference in accordance with the management instructions.

[0051] Note that AI / ML model inference is, for example, a process of obtaining a set of outputs from a set of inputs using a trained AI / ML model (or an updated AI / ML model). Alternatively, model inference may be a process of obtaining inference output data from inference data using a trained AI / ML model (or an updated AI / ML model). Hereinafter, AI / ML model inference may be referred to as "model inference" or "inference."

[0052] In the following, an AI / ML model that is currently being trained (or updated) may be referred to as a training AI / ML model (or an updating AI / ML model). In the following, when there is no need to distinguish between a training (or updating) AI / ML model and a trained (or updated) AI / ML model, they may be simply referred to as an "AI / ML model."

[0053] The management unit A5 supervises operations on the AI / ML model (selection, activation, deactivation, switching, fallback, etc.). The management unit A5 also supervises monitoring of the AI / ML model. The management unit A5 can also perform operations to ensure appropriate inference operations based on monitoring data and inference output data. To this end, the management unit A5 outputs a model transfer and / or model delivery request (Model Transfer / Delivery Request) to the model recording unit A6, and causes the trained (or updated) AI / ML model recorded in the model recording unit A6 to be output to the model inference unit A3. The management unit A5 also outputs management instructions to the model inference unit A3 and supervises operations on the AI / ML model. Furthermore, the management unit A5 can output performance feedback and a re-learning request to the model learning unit A2, causing the model learning unit A2 to re-learn the AI / ML model (i.e., update the learned AI / ML model).

[0054] FIG. 7 is a diagram illustrating an example of operation in the AI / ML technique according to the first embodiment.

[0055] In Fig. 7, the transmitting entity TE is an entity capable of performing model inference and transmitting inference output data to the receiving entity RE. Meanwhile, the receiving entity RE is an entity capable of receiving inference output data from the transmitting entity TE. Model training may be performed in the transmitting entity TE. The model training may also be performed in the receiving entity RE. If the model training is performed in the receiving entity RE, the trained AI / ML model may be transmitted from the receiving entity RE to the transmitting entity TE.

[0056] The entity may be, for example, a device, a functional block included in the device, or a hardware block included in the device.

[0057] For example, the transmitting entity TE may be the UE 100, and the receiving entity RE may be the gNB 200 or a core network device. Alternatively, the transmitting entity TE may be the gNB 200 or a core network device, and the receiving entity RE may be the UE 100.

[0058] As shown in Fig. 7, in step S1, the transmitting entity TE transmits control data related to AI / ML technology to the receiving entity RE and receives the control data from the receiving entity RE. The control data may be an RRC message, which is signaling of the RRC layer (i.e., Layer 3). The control data may be a MAC Control Element (CE), which is signaling of the MAC layer (i.e., Layer 2). The control data may be Downlink Control Information (DCI), which is signaling of the PHY layer (i.e., Layer 1). The downlink signaling may be UE-specific signaling. The downlink signaling may be broadcast signaling. The control data may be a control message in a control layer (e.g., an AI / ML layer) specialized for artificial intelligence or machine learning. Alternatively, the control data may be a NAS message in the NAS layer.

[0059] (Layout Examples and Use Cases) Next, a description will be given of how the functional blocks shown in Fig. 6 are arranged in the mobile communication system 1. Below, layout examples of the functional blocks will be described along with specific use cases.

[0060] For example, there are three use cases in which AI / ML technology is applied:

[0061] (X1.1) "CSI (Channel State Information) Feedback Enhancement"

[0062] (X1.2) "Beam management"

[0063] (X1.3) “Positioning accuracy enhancement”

[0064] (X1.1) Example of functional block arrangement in "CSI feedback improvement" "CSI feedback improvement" represents a use case in which AI / ML technology is applied to CSI fed back from UE100 to gNB200, for example. CSI is information about the channel state in the downlink between UE100 and gNB200. The CSI includes at least one of a channel quality indicator (CQI), a precoding matrix indicator (PMI), and a rank indicator (RI). The gNB200 performs, for example, downlink scheduling based on the CSI feedback from UE100.

[0065] 8 is a diagram showing an example of the arrangement of each functional block in "CSI feedback improvement". In the example of "CSI feedback improvement" shown in FIG. 8, a data collection unit A1, a model learning unit A2, and a model inference unit A3 are included in the control unit 130 of the UE 100. On the other hand, a data processing unit A4 is included in the control unit 230 of the gNB 200. That is, model learning and model inference are performed in the UE 100. FIG. 8 shows an example in which the transmitting entity TE is the UE 100 and the receiving entity RE is the gNB 200.

[0066] In "CSI feedback improvement", the gNB 200 transmits a reference signal for the UE 100 to estimate the downlink channel state. As the reference signal, a CSI reference signal (CSI-RS) will be described as an example below, but the reference signal may be a demodulation reference signal (DMRS).

[0067] First, in model learning, UE100 (receiving unit 110) receives a first reference signal from gNB200 using a first resource. Then, UE100 (model learning unit A2) derives a learned model for inferring CSI from the reference signal using learning data including the first reference signal and CSI. Such a first reference signal is sometimes referred to as a full CSI-RS.

[0068] For example, the CSI generation unit 131 performs channel estimation using the received signal (CSI-RS) received by the receiving unit 110 to generate CSI. The transmitting unit 120 transmits the generated CSI to the gNB 200. The model learning unit A2 performs model learning using a set of the received signal (CSI-RS) and the CSI as learning data, and derives a learned model for inferring the CSI from the received signal (CSI-RS).

[0069] Second, in model inference, the receiver 110 receives a second reference signal from the gNB 200 using a second resource that is less than the first resource. Then, the model inference unit A3 uses the trained model to infer CSI as inference result data using the second reference signal as inference data. Hereinafter, such a second reference signal may be referred to as a partial CSI-RS or a punctured CSI-RS.

[0070] For example, the model inference unit A3 inputs the partial CSI-RS received by the receiving unit 110 as inference data into the trained model, and infers CSI from the CSI-RS. The transmitting unit 120 transmits the inferred CSI to the gNB 200.

[0071] This enables UE 100 to feed back (or transmit) accurate (complete) CSI to gNB 200 from the small amount of CSI-RS (partial CSI-RS) received from gNB 200. For example, gNB 200 can reduce (puncture) CSI-RS when intended to reduce overhead. In addition, UE 100 can respond to situations where the radio conditions deteriorate and some CSI-RS cannot be received normally.

[0072] FIG. 9 is a diagram illustrating an example of operation in "CSI feedback improvement" according to the first embodiment.

[0073] 9, in step S10, the gNB 200 may notify or set the CSI-RS transmission pattern (puncture pattern) in the inference mode to the UE 100 as control data. For example, the gNB 200 transmits to the UE 100 the antenna port and / or time-frequency resource that transmits or does not transmit the CSI-RS in the inference mode.

[0074] In step S11, gNB200 may send a switching notification to UE100 to start learning mode.

[0075] In step S12, the UE 100 starts a learning mode.

[0076] In step S13, the gNB 200 transmits the full CSI-RS. The receiver 110 of the UE 100 receives the full CSI-RS, and the CSI generator 131 generates (or estimates) CSI based on the full CSI-RS. In the learning mode, the data collector A1 collects the full CSI-RS and CSI. The model learning unit A2 uses the full CSI-RS and the CSI as learning data to create a learned AI / ML model.

[0077] In step S14, UE100 transmits the generated CSI to gNB200.

[0078] Thereafter, in step S15, when the model learning is completed, the UE 100 transmits a completion notification indicating that the model learning is completed to the gNB 200. The UE 100 may transmit a completion notification when the creation of the learned model is completed.

[0079] In step S16, in response to receiving the completion notification, gNB200 sends a switching notification to UE100 to switch UE100 from learning mode to inference mode.

[0080] In step S17, in response to receiving the switching notification, the UE 100 switches from the learning mode to the inference mode.

[0081] In step S18, the gNB 200 transmits a partial CSI-RS. The receiver 110 of the UE 100 receives the partial CSI-RS. In the inference mode, the data collector A1 collects the partial CSI-RS. The model inference unit A3 inputs the partial CSI-RS as inference data into the trained model, and obtains CSI as the inference result.

[0082] In step S19, the UE 100 feeds back (or transmits) the CSI, which is the inference result, to the gNB 200 as inference result data. In the UE 100, by repeating model learning in the learning mode, a trained model with a predetermined accuracy or higher can be generated. It is expected that the inference result using the trained model generated in this way will also have a predetermined accuracy or higher.

[0083] In addition, in step S20, if UE100 determines that model learning is necessary, it may send a notification indicating that model learning is necessary to gNB200 as control data.

[0084] In the example shown in Fig. 9, an example has been described in which the training data is "(full) CSI-RS" and "CSI", and the inference data is "(partial) CSI-RS". Hereinafter, the training data and / or the inference data may be referred to as a "dataset".

[0085] In "improving CSI feedback," in addition to "CSI-RS" and "CSI," at least one of the following data or information may be used as a data set:

[0086] (Y1) RSRP (Reference Signals Received Power), RSRQ (Reference Signal Received Quality), SINR (Signal-to-interference-plus-noise ratio), or AD converter output waveform (These measurements may be CSI-RS. These measurements may also be other received signals received from gNB200.)

[0087] (Y2) Bit Error Rate (BER) or Block Error Rate (BLER) (The total number of transmitted bits (or the total number of transmitted blocks) is known, and the BER (or BLER) may be measured based on the CSI-RS.)

[0088] (Y3) The movement speed of UE100 (which may be measured by a speed sensor within UE100). The data set to be used for machine learning may be set. For example, the following processing may be performed. That is, UE100 transmits capability information indicating which type of input data it can handle in machine learning to gNB200 as control data. The capability information may represent, for example, any of the data or information shown in (Y1) to (Y3). The capability information may be information in which learning data and inference data are separately specified. Then, gNB200 transmits data type information to be used as the data set to UE100 as control data. The data type information may represent, for example, any of the data or information shown in (Y1) to (Y3). Furthermore, the data type information may specify separately data type information to be used as learning data and data type information to be used as inference data.

[0089] An example of the arrangement of functional blocks in (X1.1) "CSI feedback" has been described above. The above-mentioned arrangement example is just one example, and in 3GPP, the arrangement example of functional blocks is still in the process of being studied. Similarly, (X1.2) "Beam management" and (X1.3) "Position accuracy improvement" are also still in the process of being studied.

[0090] (X1.4) Example of Model Transfer Next, we will explain the transfer of an AI / ML model (Model Transfer). Note that the terms "transfer of an AI / ML model" and "delivery of an AI / ML model" have the same meaning.

[0091] (X1.4.1) First operation pattern related to model forwarding Figure 10 is a diagram showing an example of an operation of the first operation pattern related to model forwarding according to the first embodiment. In the example shown in Figure 10, the receiving entity RE will be described as mainly being the UE 100, but the receiving entity RE may be the gNB 200 or the AMF 300. Also, in the example shown in Figure 10, the transmitting entity TE will be described as being the gNB 200, but the transmitting entity TE may be the UE 100 or the AMF 300.

[0092] 10, in step S25, the gNB 200 transmits a capability inquiry message to the UE 100 to request transmission of a message including an information element (IE) indicating the execution capability for the learning process. The UE 100 receives the capability inquiry message. However, the gNB 200 may transmit the capability inquiry message when it executes the learning process (when it determines that it will execute the learning process).

[0093] In step S26, the UE 100 transmits to the gNB 200 a message including an information element indicating execution capabilities for the learning process (in another respect, the execution environment for the learning process). The gNB 200 receives the message. The message may be an RRC message (for example, a "UE Capability" message or a newly defined message (for example, a "UE AI Capability" message, etc.). Alternatively, the transmitting entity TE may be the AMF 300, and the message may be a NAS message. Alternatively, if a new layer is defined for performing or controlling the learning process (AI / ML process), the message may be a message of the new layer.

[0094] The information element indicating the execution capability related to the learning process may be an information element indicating the capability of a processor for executing the learning process and / or an information element indicating the capability of a memory for executing the learning process. Specifically, the information element indicating the processor capability may be an information element indicating the product number (or model number) of the AI ​​processor. Specifically, the information element indicating the memory capability may be information indicating the memory capacity.

[0095] Alternatively, the information element indicating the execution capability of the learning process may be an information element indicating the execution capability of the inference process (model inference). Specifically, the information element indicating the execution capability of the inference process may be an information element indicating whether a deep neural network model is supported. The information element may also be an information element indicating the time (or response time) required to execute the inference process.

[0096] Alternatively, the information element indicating the execution capability related to the learning process may be an information element indicating the execution capability of the learning process (model learning). Specifically, the information element indicating the execution capability of the learning process may be an information element indicating the number of concurrent executions of the learning process. The information element may be an information element indicating the processing capacity of the learning process.

[0097] In step S27, gNB200 determines the model to be configured (or deployed) in UE100 based on the information elements contained in the message received in step S26.

[0098] In step S28, gNB200 transmits a message including the model determined in step S27 to UE100. UE100 receives the message and performs a learning process (i.e., a model learning process and / or a model inference process) using the model included in the message. A specific example of step S28 will be described in the following second operation pattern.

[0099] (X1.4.2) Second Operation Pattern Related to Model Transfer FIG. 11 is a diagram showing an example of a configuration message including a model and additional information according to the first embodiment. The configuration message may be an RRC message transmitted from the gNB 200 to the UE 100 (for example, an "RRC Reconfiguration" message, or a newly defined message (for example, an "AI Deployment" message or an "AI Reconfiguration" message, etc.). Alternatively, the configuration message may be a NAS message transmitted from the AMF 300 to the UE 100. Alternatively, when a new layer for performing or controlling machine learning processing (AI / ML processing) is defined, the message may be a message of the new layer.

[0100] In the example of FIG. 11, the setting message includes three models (Model #1 to #3). Each model is included as a container in the setting message. However, the setting message may include only one model. The setting message further includes, as additional information, three individual additional information (Info #1 to #3) provided individually corresponding to each of the three models (Model #1 to #3), and common additional information (Meta-Info) commonly associated with the three models (Model #1 to #3). Each of the individual additional information (Info #1 to #3) includes information unique to the corresponding model. The common additional information (Meta-Info) includes information common to all models in the setting message.

[0101] The individual additional information may be a model index indicating an index (index number) assigned to each model, or may be a model execution condition indicating the performance (e.g., processing delay) required to apply (execute) the model.

[0102] The individual additional information or the common additional information may be a model usage that specifies a function to which a model is to be applied (e.g., "CSI feedback," "beam management," "positioning," etc.). The individual additional information or the common additional information may be a model selection criterion that applies (executes) a corresponding model depending on whether a specified criterion (e.g., a moving speed) is satisfied.

[0103] (LCM according to the first embodiment) 3GPP specifies LCM (Life Cycle Management) for AI / ML technology. LCM is, for example, management of an AI / ML model from its generation to its termination. Through such management, the mobile communication system 1 can appropriately operate the AI / ML model and appropriately support various use cases that use the AI / ML model. LCM includes, for example, the following elements:

[0104] (Z1) Data collection

[0105] (Z2) Model learning

[0106] (Z3) Functionality / Model Identification

[0107] (Z4) Model delivery or model transfer

[0108] (Z5) Model inference operation

[0109] (Z6) Feature or Model Selection, Activation, Deactivation, Switching, and Fallback

[0110] (Z7) Function or model monitoring

[0111] (Z8) Model Update

[0112] (Z9) UE Capabilities These elements may basically be executed over time. However, this does not necessarily mean that the elements are executed strictly in this order. For example, model identification (Z3) may be used to determine which model is the target in each of model training (Z2), model distribution (Z4), and model inference (Z5).

[0113] In the following description, each element included in the LCM may be referred to as an LCM operation. An LCM operation may be the execution of each element.

[0114] Regarding LCM, 3GPP specifies function-based LCM and model ID-based LCM. In function-based LCM, for example, an AI / ML model can be specified on a function-by-function basis. For example, specifying one function can specify multiple AI / ML models. In function-based LCM, activation of an AI / ML function can be instructed using 3GPP signaling (e.g., RRC, MAC-CE, or DCI) from a network device. On the other hand, in model ID-based LCM, each AI / ML model can be specified using a model ID. For example, in model ID-based LCM, each AI / ML model is identified in the network device, and operations for each AI / ML model can be instructed via the model ID.

[0115] Note that the only difference between the function-based LCM and the model ID-based LCM is how the AI / ML model is identified. After the AI / ML model is identified, other operations related to the LCM may be performed using the same process.

[0116] In the first embodiment, a description will be given assuming that a model ID-based LCM is applied to the mobile communication system 1. However, in the first embodiment, a function-based LCM may also be applied to the mobile communication system 1.

[0117] (Model Identification According to the First Embodiment) In the first embodiment, attention is focused on model identification (Z3) in LCM. Model identification refers to, for example, identifying (or specifying) an AI / ML model. For example, in model learning, model identification is performed to identify which model is the learning target. Alternatively, in model inference, model identification may be performed to identify which model is the inference target. Alternatively, model identification may be performed to identify the AI / ML model that is the target in each operation, such as a re-learning request, a management instruction, or a model transfer. Model identification is used in model ID-based LCM. Therefore, in model identification, a model ID is used to distinguish an AI / ML model from other AI / ML models.

[0118] Regarding model identification, 3GPP mainly classifies them into Type A and Type B.

[0119] Type A is a type in which the AI / ML model is identified in the network device or UE 100 without radio signaling. For example, Type A is applicable when the learned AI / ML model is applicable to any scenario and is rarely updated. In this case, both the network device and the UE 100 can identify the AI / ML model without performing model identification. Note that in Type A, although a model ID is assigned, the model ID is not used during model identification, but is used after model identification is performed.

[0120] On the other hand, Type B is a type in which an AI / ML model is identified using radio signaling. For example, when a trained AI / ML model is frequently updated, Type B can be applied to the trained AI / ML model. Alternatively, when an initial type is Type A but is updated for some reason, Type B can be applied. 3GPP further classifies Type B into Type B1 and Type B2.

[0121] In Type B1, model identification is initiated by the UE 100, and the network device assists in the remaining steps of model identification. On the other hand, in Type B2, model identification is initiated by the network device, and the UE 100 responds to the remaining steps of model identification. There are two types depending on where model identification is initiated. Note that a model ID is assigned to both Type B1 and Type B2, and the model ID is used during model identification.

[0122] (Example of Proposal Regarding Model Identification) In response to the above discussion on model identification, the following contribution (R1-2310999, 3GPP TSG-RAN WG1 Meeting #115, Chicago, USA, November 13th-November 17th, 2023) was submitted to 3GPP.

[0123] That is, the above-mentioned contribution proposes that Type A (identification of an AI / ML model without radio signaling) be performed in the following order:

[0124] (A1) Model IDs are assigned offline.

[0125] (A2) Model ID and meta information are provided to the network device (if possible) without over-the-air signaling.

[0126] (A3) UE capabilities including model ID are updated.

[0127] FIG. 12(A) is a diagram illustrating an example of operation illustrating the above proposal of Type A. In FIG. 12(A), an external server 500 is used as an example of offline. Offline means, for example, that wireless signaling defined by 3GPP is not used. Offline may also mean a state where there is no connection to a 3GPP cellular network system. Therefore, an example of offline may be an external server 500 located outside the 3GPP cellular network system.

[0128] 12A, the external server 500 transmits the model ID to the UE 100 (step S30), and then transmits the model ID and meta information to the network device 400 (step S31). The UE 100 can acquire the model ID offline without using wireless signaling, so step S30 corresponds to the above (A1). Similarly, step S31 corresponds to the above (A2) because the network device 400 can acquire the model ID offline without using wireless signaling. Furthermore, step S32 corresponds to the above (A3).

[0129] Note that when the UE 100 and the network device 400 receive the model ID (steps S30 and S31), they are able to perform model identification using the model ID. Model identification in the UE 100 and the network device 400 is possible at the time of steps S30 and S31, respectively. Therefore, the condition of Type A, that model identification is performed without using radio signaling, is satisfied.

[0130] Furthermore, the above-mentioned contribution proposes that Type B1 (in which the AI / ML model is identified using radio signaling and model identification is initiated by the UE 100) be performed in the following order:

[0131] (B1-1) Learning of the AI / ML model is performed on the UE 100 side.

[0132] (B1-2) The existence of a new AI / ML model as a result of the learning is provided to the UE 100 and the network device together with meta information.

[0133] (B1-3) The network device assigns a model ID and provides it to the UE 100.

[0134] FIG. 12(B) shows an example of operation illustrating the above proposal for Type B1. The UE 100 trains an AI / ML model (step S40, (B1-1)) and transmits the existence of a new AI / ML model and meta information to the network device 400 (step S41, (B1-2)). The network device 400 then transmits a model ID to the UE 100 (step S42, (B1-3)). Type B1 differs from the other types in that model identification is initiated by the UE 100. The model identification shown in FIG. 12(B) is initiated from model learning in the UE 100 (step S40), and therefore satisfies the conditions of Type B1.

[0135] Furthermore, the above-mentioned contribution proposes that Type B2 (in which the AI / ML model is identified using radio signaling and model identification is initiated from the network device 400) be performed in the following order:

[0136] (B2-1) A new AI / ML model is learned on the network device side.

[0137] (B2-2) The network device transfers the new AI / ML model along with the model ID to the UE 100.

[0138] 12(C) shows an example of operation illustrating the above proposal for Type B2. The network device 400 learns the AI / ML model (step S50, (B2-1)) and transmits the learned AI / ML model and model ID to the UE 100 (step S51, (B2-2)). The model identification in FIG. 12(C) also satisfies the conditions of Type B2 because it starts with model learning in the network device 400 (step S50).

[0139] (Communication Control Method According to First Embodiment) Here, the model identification proposed in the above-mentioned contribution focuses on model identification in the LCM and clearly indicates how operations are performed for each type.

[0140] However, the model identification proposed in the above contribution may not necessarily be an efficient operation when the operation of the entire LCM is taken into consideration, and therefore the LCM motion may not be executed efficiently.

[0141] Therefore, the first embodiment aims to perform LCM operation efficiently. Below, specific operation examples will be described to explain how the LCM operation is performed efficiently. Specifically, an operation example related to Type A (first operation example) will be described, then an operation example related to Type B1 (second operation example), and finally an operation example related to Type B2 (third operation example) will be described, and the efficiency of the LCM operation will be explained in each operation example.

[0142] (1) First Operation Example (Operation Example of Type A) In the first operation example, an operation example related to Type A will be described. As described above, Type A is a type in which model identification is performed without using radio signaling. The above contribution proposes that Type A be performed by the operations shown in (A1) to (A3) above (or FIG. 12(A)).

[0143] 12(A) again, UE 100 only receives the model ID from external server 500 (step S30). When UE 100 performs model inference by itself with only the model ID, UE 100 may not actually be able to perform model inference, taking into account its own capabilities. Furthermore, when model inference is performed in network device 400, UE 100 may not be able to acquire inference data by itself and may not be able to transmit the inference data to network device 400, depending on the type of inference data used in model inference.

[0144] Therefore, in a first operation example, the UE 100 acquires meta information along with a model ID. Specifically, first, the user device (e.g., the UE 100) acquires model identification information (e.g., a model ID) that identifies the AI / ML model and additional information (e.g., meta information) of the AI / ML model. Second, the user device transmits the model identification information to a network device (e.g., the network device 400). Third, when transmitting the model identification information, the user device transmits the model identification information when it determines, based on the additional information, that inference for the AI / ML model is possible.

[0145] As described above, in the first operation example, UE 100 acquires meta information together with the model ID, and when it determines that inference for the AI / ML model is possible based on the meta information, it transmits the model ID to network device 400. As a result, in the first operation example, UE 100 can transmit the model ID if model inference is possible, and can not transmit the model ID if model inference is not possible. Therefore, UE 100 can efficiently perform subsequent LCM operations (e.g., model inference).

[0146] The first operation example will be specifically described below. In the first operation example, the operation example will be divided into three operation examples, namely, a UE-side model, a network-side model, and a two-sided model, taking into consideration the scenario in which inference is performed.

[0147] The UE-side model is a model in which model inference is performed by the UE 100. The network-side model is a model in which model inference is performed by the network device 400. The two-side model is a model in which model inference is performed by both the UE 100 and the network device 400.

[0148] (1.1) (1-1) Operation Example In this operation example, an operation example in the case where model inference is performed on the UE 100 side (UE side model) for Type A will be described.

[0149] FIG. 13 is a diagram showing a (1-1) example of operation according to the first embodiment. In FIG. 13, in accordance with the above contribution, an external server 500 is shown as an offline example. The external server may be, for example, an OTT (Over The Top) server. Note that instead of the external server 500, a wireless LAN (Local Area Network) access point or a portable memory (for example, a memory card) may be used.

[0150] 13 , in step S50, the external server 500 performs model learning. This is performed in the external server 500 in consideration of, for example, resources in the UE 100. Since the UE 100 does not need to perform model learning, it is possible to reduce resources compared to when model learning is performed.

[0151] In step S51, UE 100 acquires a trained AI / ML model, a model ID of the trained AI / ML model, and meta information of the trained AI / ML model. In the example of Fig. 13, receiver 110 of UE 100 acquires the trained AI / ML model, the model ID, and the meta information by receiving them from external server 500. That is, UE 100 acquires the trained AI / ML model, the model ID, and the meta information offline.

[0152] 12A with this step S51, the UE 100 receives meta information in addition to the model ID. The meta information will now be described.

[0153] The meta information may include the type of inference data and / or the type of inference output data used for model inference. The control unit 130 of the UE 100 can confirm whether the inference data can be acquired and determine whether model inference can be performed by checking the type of the inference data and / or the inference output data based on the meta information.

[0154] Alternatively, the meta information may include application conditions under which the trained AI / ML model used for model inference is applied.

[0155] First, the application condition may represent a location where model inference is possible. The location may be represented by a cell, a RAN-based Notification Area (RNA), a Tracking Area (TA), a Public Land Mobile Network (PLMN), or the like. Alternatively, the location may be represented by latitude and longitude acquired by a Global Navigation Satellite System (GNSS) function. The control unit 130 of the UE 100 compares its current location with the location where model inference is possible, thereby determining whether the UE 100 is located in a location where model inference is possible and whether model inference is possible.

[0156] Second, the application condition may be expressed as the computational power required when model inference is executed. The control unit 130 can determine whether model inference is possible based on the computational power.

[0157] Third, the application condition may be expressed as the memory capacity required when model inference is executed. The control unit 130 can determine whether model inference is executable based on the memory capacity.

[0158] Alternatively, the meta information may include information about a use case. As described above, examples of the use case include CSI feedback, position accuracy improvement, beam management, etc. Since the control unit 130 of the UE 100 may not be able to execute model inference depending on the use case, the control unit 130 can determine whether or not model inference is executable based on the information about the use case.

[0159] Alternatively, the meta information may include version information of the trained AI / ML model used for model inference. For example, if model inference using a specific version of an AI / ML model was not possible in the past, the control unit 130 can determine whether or not to perform model inference based on the version information.

[0160] Alternatively, the meta information may include model pairing information. For example, in the case of a two-sided model, model inference may be performed jointly in the UE 100 and the network device 400 using different trained AI / ML models. In this case, information indicating which AI / ML model is used in the UE 100 and the network device 400 is represented as model pairing information. The model pairing information may be represented by linking the model ID of the AI / ML model used in the UE 100 with the model ID of the AI / ML model used in the network device 400. The control unit 130 can determine whether or not to perform model inference by determining whether or not it has the AI / ML model based on the pairing information.

[0161] In step S52, the network device 400 (for example, the receiver 220 of the gNB 200) acquires a model ID and meta information. The model ID and meta information may be the same as the model ID and meta information of step S51. In the example of FIG. 13, when the network device 400 is a gNB 200, the receiver 220 of the gNB 200 receives the model ID and meta information from the external server 500, thereby acquiring the model ID and meta information.

[0162] In step S53, the network device 400 (e.g., the control unit 230 of the gNB 200) records the model ID and meta information in a database (or memory).

[0163] In step S54, when the control unit 130 of the UE 100 determines, based on the meta information, that model inference is executable, the transmission unit 120 of the UE 100 transmits the model ID to the network device 400. The transmission unit 120 of the UE 100 may transmit control data (FIG. 7) including the model ID to the gNB 200. The network device 400 (e.g., the reception unit 220 of the gNB 200) receives the model ID.

[0164] In step S55, in response to receiving the model ID (step S54), the network device 400 (e.g., the transmitter 210 of the gNB 200) specifies the model ID and transmits an instruction to the UE 100 to execute model inference using the trained AI / ML model. The network device 400 may transmit control data (FIG. 7) including the execution instruction to the UE 100. The receiver 110 of the UE 100 receives the execution instruction.

[0165] In step S56, the control unit 130 of the UE 100 performs model inference in accordance with the execution instruction.

[0166] In step S57, the control unit 130 of the UE 100 detects that the result of the model inference is below the expected value. The control unit 130 may determine that the result of the model inference is below the expected value when the inference output data is below the threshold value. The network device 400 (e.g., the control unit 230 of the gNB 200) may perform this detection.

[0167] In step S58, the control unit 130 of the UE 100 decides to perform re-learning, and the transmission unit 120 transmits the learning data to the external server 500 for the re-learning.

[0168] The external server 500 re-learns the trained AI / ML model based on the training data (step S59).

[0169] In step S60, the receiving unit 110 of the UE 100 receives (or acquires) the updated AI / ML model updated by relearning, the model ID of the updated AI / ML model, and meta information of the updated AI / ML model from the external server 500. As in step S51, the control unit 130 of the UE 100 determines whether model inference using the updated AI / ML model is possible based on the meta information. The meta information may be the same as the meta information of step S51, except that the target is the updated AI / ML model.

[0170] In step S62, the network device 400 receives (or acquires) the model ID of the updated AI / ML model and meta information of the updated AI / ML model from the external server 500. The meta information may be the same as the meta information of step S60.

[0171] Thereafter, steps S63 to S66 are the same as steps S53 to S56, respectively, except that the target of model inference is the updated AI / ML model.

[0172] (1.2) Operation Example (1-2) In the operation example (1-2), an operation example will be described for Type A in which model inference is performed on the network device 400 side (network side model). This operation example will be described focusing on the differences from the operation example (1-1) (FIG. 13).

[0173] FIG. 14 is a diagram illustrating this operation example according to the first embodiment.

[0174] Steps S70 to S73 are the same as steps S50 to S53 in the (1-1) operation example, respectively.

[0175] In step S71, the control unit 130 of the UE 100 determines whether inference for the trained AI / ML model is possible in response to receiving the meta-information, as in the first operation example. However, in this case, the network device 400 actually performs the model inference. Therefore, in this operation example, the control unit 130 determines whether parameters (i.e., inference data) used for model inference can be acquired based on the meta-information. The network device 400 performs model inference using the model parameters. Here, the meta-information includes the type of data used in the trained AI / ML model. Therefore, the control unit 130 of the UE 100 can grasp the type of model parameters based on the meta-information. Therefore, the control unit 130 determines whether the model parameters can actually be acquired based on the meta-information. If it is determined that the model parameters can be acquired, the transmission unit 120 of the UE 100 transmits the model ID (step S74).

[0176] In this example, the control unit 130 can determine whether the model parameters used in model inference can be acquired based on the meta information. If the UE 100 determines that the model parameters can be acquired, the control unit 130 transmits the model ID (step S74). Therefore, the UE 100 can efficiently perform the subsequent LCM operation (e.g., model inference).

[0177] In step S74, when the transmitter 120 of the UE 100 determines that it is possible to acquire the model parameters (or that model inference is possible), it transmits the model ID to the network device 400. The transmission of the model ID itself may be performed using control data, as in the (1-1) operation example.

[0178] In step S75, the network device 400 (for example, the transmitter 210 of the gNB 200) determines that the model parameters can be acquired in the UE 100 in response to receiving the model ID, and specifies the model ID and transmits an instruction to execute model inference to the UE 100. The execution instruction may also be transmitted using control data. The receiver 110 of the UE 100 receives the execution instruction.

[0179] In step S76, the control unit 130 of the UE 100 acquires model parameters in response to receiving the execution instruction, and the transmission unit 120 of the UE 100 transmits the acquired model parameters to the network device 400 (step S76). The model parameters may be transmitted using control data. The model parameters may be transmitted using a user plane (user data). The network device 400 (for example, the reception unit 220 of the gNB 200) receives the model parameters.

[0180] In step S77, the network device 400 (e.g., the control unit 230 of the gNB 200) performs model inference using the model parameters. Then, in step S78, if the network device 400 (e.g., the control unit 230 of the gNB 200) detects that the inference result is lower than expected, in step S79, the network device 400 (e.g., the transmission unit 210 of the gNB 200) transmits learning data for re-learning to the external server 500. The subsequent processing (steps S80 to S87) is the same as steps S70 to S77, respectively, except that the target is the updated AI / ML model.

[0181] (1.3) (1-3) Operation Example In this operation example, an operation example will be described for Type A in which model inference is performed on both the UE 100 side and the network device 400 side (two-sided model). However, in this third operation example, the operation will be divided into two examples: a case in which inference is performed using the same AI / ML model in the UE 100 and the network device 400, and a case in which inference is performed using different AI / ML models in the UE 100 and the network device 400.

[0182] (1.3.1) (1-3-1) Operation Example In this operation example, among the (1-3) operation examples, an operation example in which inference is performed using the same AI / ML model in UE 100 and network device 400 will be described.

[0183] 15 is a diagram illustrating this operation example according to the first embodiment. The differences between this operation example and the (1-1) operation example (FIG. 13) are as follows.

[0184] First, model inference is performed (steps S96 and S98) by both the UE 100 and the network device 400. In this case, model inference is first performed in the UE 100 (step S96), and the inference result (inference output data) is transmitted from the UE 100 to the network device 400 (step S97), and then model inference is performed in the network device 400 using the inference result (step S98).

[0185] Second, this example differs from the operation example (1-1) in that the same trained AI / ML model is used for the UE 100 and the network device 400 (step S90). In Fig. 15, the same AI / ML model is used, and is therefore represented as "UE_NW." Other than that, this example is the same as the operation example (1-1).

[0186] (1.3.2) (1-3-2) Operation Example Next, this operation example will be described. This operation example is an operation example of the (1-3) operation example in which inference is performed using different AI / ML models for the UE 100 and the network device 400.

[0187] 16 is a diagram illustrating this operation example according to the first embodiment. This operation example will be described, focusing on the differences from the (1-3-1) operation example (FIG. 15).

[0188] First, this operation example differs from the first (1-3-1) operation example in that the model training targets are the AI / ML model used on the UE 100 side (step S120) and the AI / ML model used on the network device 400 (step S121). Therefore, the receiver 110 of the UE 100 receives the trained AI / ML model used on the UE 100 side, the model ID of the trained AI / ML model, and meta information of the trained AI / ML model (step S122). Furthermore, the network device 400 (for example, the receiver 220 of the gNB 200) receives the trained AI / ML model used on the network device 400 side, the model ID of the trained AI / ML model, and meta information of the trained AI / ML model (step S123).

[0189] Second, regarding the meta information, the content of the meta information received by the UE 100 is different from that of the meta information received by the network device 400. That is, the meta information received by the UE 100 is meta information related to the trained AI / ML model for the UE 100, and the meta information received by the network device 400 is the trained AI / ML model for the network device 400. The type of information included in the meta information may be the same.

[0190] Third, this operation example differs in that the targets of relearning are the trained AI / ML model (step S124) used on the UE 100 side and the trained AI / ML model (step S125) used on the network device 400 side. The subsequent processing (steps S126 to S110) is the same as steps S122 to S98 before the update, except that the targets are the updated AI / ML models.

[0191] Other than that, this operation example is the same as the operation example (1-3-1).

[0192] (2) Second Operation Example (Operation Example of Type B1) Next, a second operation example will be described. In the second operation example, an operation example related to Type B1 will be described. As described above, Type B1 is a type in which a model ID is identified using radio signaling, and model identification is initiated by the UE 100. The above contribution proposes that Type B1 be performed by the operations shown in (B1-1) to (B1-3) (FIG. 12(B)) above.

[0193] Here, referring again to FIG. 12(B), UE 100 simply transmits to network device 400 the existence of a trained AI / ML model newly created by model training (step S40) and meta information of the trained AI / ML model (step S41). For example, as seen in a two-sided model, when model training is performed not only in UE 100 but also in network device 400, training data is required for model training, and model existence and meta information alone may not be enough to properly train the model. Furthermore, even if UE 100 receives a model ID from network device 400, the model ID may be inappropriate if model training is not properly performed in network device 400. Therefore, UE 100 may notify the network device 400 of an additional message. In other words, FIG. 12(B) may not be an appropriate model identification operation that takes into account the entire LCM operation.

[0194] Therefore, in a second operation example, instead of transmitting the existence of a model, UE 100 transmits parameters for learning (i.e., learning data) to network device 400. That is, first, user equipment (e.g., UE 100) trains an AI / ML model. Second, the user equipment transmits parameters used for learning and additional information (e.g., meta information) of the AI / ML model to network device (e.g., network device 400). Third, the network device uses the parameters to train the AI / ML model.

[0195] As described above, in the second operation example, the UE 100 transmits the learning parameters to the network device 400, and therefore, it becomes possible to appropriately perform model learning in the network device 400. Then, the UE 100 can receive an appropriate model ID from the network device 400. Therefore, the UE 100 does not transmit an additional message to the network device 400, and therefore, it becomes possible to efficiently perform the LCM operation.

[0196] The second operation example will be specifically described below. As with the first operation example, the second operation example will be described separately as a UE-side model and a two-side model, taking into consideration the situation in which inference is performed.

[0197] (2.1) (2-1) Example of Operation In this example of operation, for type B1 (model ID is identified using radio signaling and model identification is initiated from UE100), an example of operation will be described in which model inference is performed on the UE100 side (i.e., UE-side model).

[0198] FIG. 17 is a diagram illustrating this operation example according to the first embodiment.

[0199] 17 , in step S130, the control unit 130 of the UE 100 performs model learning. The control unit 130 generates a trained AI / ML model through the model learning. Step S130 represents Type B1 in which model identification is initiated from the UE 100 side.

[0200] In step S131, the transmitter 120 of the UE 100 transmits meta information of the learned AI / ML model to the network device 400. The transmitter 120 may transmit the meta information using the control data (FIG. 7). Compared to FIG. 12(B), the UE 100 does not need to transmit information indicating the existence of the learned AI / ML model to the network device 400, thereby enabling effective use of radio resources. The content of the meta information may be the same as that of the first embodiment. The network device 400 (for example, the receiver 220 of the gNB 200) receives the meta information.

[0201] In step S132, the network device 400 (e.g., the control unit 230 of the gNB 200) records the meta information in memory.

[0202] In step S133, the network device 400 (e.g., the transmitter 210 of the gNB 200) transmits the model ID to the UE 100. The model ID may be transmitted using the control data (FIG. 7).

[0203] In step S134, the network device 400 specifies the model ID and transmits an instruction to execute model inference using the trained AI / ML model to the UE 100. For example, the transmitter 210 of the gNB 200 may transmit control data including the execution instruction to the UE 100. The receiver 110 of the UE 100 receives the execution instruction.

[0204] In step S135, the control unit 130 of the UE 100 performs model inference in accordance with the execution instruction.

[0205] In step S136, the control unit 130 of the UE 100 detects that the result of the model inference is lower than expected. The detection itself may be the same as in the first operation example.

[0206] In step S137, the control unit 130 of the UE 100 determines to perform re-learning and performs re-learning of the trained AI / ML model. The control unit 130 generates an updated AI / ML model through the re-learning.

[0207] Thereafter, steps S138 to S142 are the same as steps S131 to S135, respectively, except that the target is the updated AI / ML model.

[0208] (2.2) Operational Examples of Type B1 and Two-Side Model Next, an operational example of the two-side model for Type B1 will be described. In this operational example, a total of four operational examples will be described. Two operational examples will be described, one in which model learning is performed on the UE 100 side and not on the network device 400 side (hereinafter, this may be referred to as "a case where model learning is performed only on the UE 100 side"), and two operational examples will be described, one in which model learning is performed on both the UE 100 side and the network device 400 side.

[0209] (2.2.1) Operational Examples When Model Learning is Performed Only on the UE 100 Side First, two operational examples will be described when model learning is performed only on the UE 100 side. In these operational examples, two operational examples will be described: one in which the trained AI / ML model used for model inference is the same (or common) between the UE 100 and the network device 400, and one in which the trained AI / ML model used for model inference is different between the UE 100 and the network device 400.

[0210] (2.2.1.1) When the same trained AI / ML model is used Figure 18 shows an example of operation when the same trained AI / ML model is used. That is, Figure 18 shows an example of operation when a type B1, 2-side model (the same trained AI / ML model) is used and model learning is performed only on the UE 100 side. Hereinafter, for convenience, this example of operation may be referred to as the (2-2-1-1) example of operation.

[0211] The differences between this operation example and the (2-1) operation example (FIG. 17) are as follows.

[0212] First, this operation example differs from the (2-1) operation example in that it is a two-sided model. That is, model inference is performed in the control unit 130 of the UE 100 (step S155). Then, the transmission unit 120 of the UE 100 transmits the inference result (inference output data) to the network device 400, and the network device 400 (for example, the control unit 230 of the gNB 200) performs model inference using the inference result (step S157).

[0213] Second, in this operation example, the transmitter 120 of the UE 100 transmits the trained AI / ML model together with meta information to the network device 400 (step S151). The trained AI / ML model may be transmitted via the user plane. The trained AI / ML model may be transmitted using control data. The network device 400 (e.g., the control unit 230 of the gNB 200) performs model inference using the trained AI / ML model (step S157).

[0214] Third, the target of model learning, model inference, and re-learning is the same AI / ML model ("UE_NW") in UE100 and network device 400. Therefore, the control unit 130 of UE100 performs model learning and generates a trained AI / ML model ("UE_NW") used by both UE100 and network device 400 (step S150). Also, UE100 and gNB200 perform model inference using the same trained AI / ML model (steps S155 and S157). Furthermore, the control unit 130 of UE100 performs re-learning and generates an updated AI / ML model ("UE_NW") used by both UE100 and network device 400 (step S160), and model inference is performed using the updated AI / ML model (steps S165 and S167).

[0215] Other than the above, this operation example is the same as the (2-1) operation example.

[0216] (2.2.1.2) When different trained AI / ML models are used Figure 19 shows an example of operation when different trained AI / ML models are used. That is, Figure 19 shows an example of operation when a type B1, 2-side model (different AI / ML models) is used and model learning is performed only on the UE 100 side. Hereinafter, for convenience, this example of operation may be referred to as the (2-2-1-2) example of operation.

[0217] The differences between this operation example and the (2-1-1-1) operation example (FIG. 18) are as follows.

[0218] First, the difference is that the targets of model training and relearning are separated into the AI / ML model for UE 100 and the AI / ML model for network device 400. Therefore, control unit 130 of UE 100 separately performs model training of the AI / ML model for UE 100 and the AI / ML model for network device 400 (steps S190 and S191). Furthermore, transmission unit 120 of UE 100 transmits the trained AI / ML model for network device 400 ("model_NW") to network device 400 (step S192). Furthermore, control unit 130 separately performs relearning of the trained AI / ML model for UE 100 and relearning of the trained AI / ML model for network device 400 (steps S193 and S194).

[0219] Second, the present embodiment differs in that the processing targets are the AI / ML model for UE 100 and the AI / ML model for network device 400. For example, network device 400 specifies a model ID and transmits an instruction to execute model inference to UE 100 (step S153), but the target of the instruction to execute is the trained AI / ML model for UE 100.

[0220] (2.2.2) Cases where model learning is performed by both the UE 100 and the network device 400 Next, two operation examples will be described for cases where model learning is performed by both the UE 100 and the network device 400. In these operation examples, two operation examples will be described: a case where the trained AI / ML model used for model inference is the same (or common) between the UE 100 and the network device 400, and a case where the trained AI / ML model used for model inference is different between the UE 100 and the network device 400.

[0221] (2.2.2.1) When the same trained AI / ML model is used Figure 20 shows an example of operation when the same trained AI / ML model is used. That is, Figure 20 shows an example of operation when a type B1, two-side model (same AI / ML model) is used and model learning is performed by both the UE 100 and the network device 400. Hereinafter, for convenience, this example of operation may be referred to as the (2-2-2-1) example of operation.

[0222] The differences between this operation example and the (2-2-1-1) operation example (FIG. 18) are as follows.

[0223] First, in this operation example, model learning is also performed in the network device 400 (step S172). Therefore, the transmitter 120 of the UE 100 transmits learning parameters (learning data) and meta information to the network device 400 (step S171). Because the same trained AI / ML model is used in the UE 100 and the network device 400, the UE 100 may transmit the learning parameters used when performing model learning. The transmitter 120 may transmit the learning parameters and meta information using control data. Re-learning is also performed in the network device 400 (step S182), and the transmitter 120 of the UE 100 transmits the learning parameters and meta information for re-learning to the network device 400 (step S181). As the learning parameters for re-learning, the same learning parameters as those used by the UE 100 itself for re-learning may be transmitted to the network device 400.

[0224] Second, in this operation example, the network device 400 does not transmit the learning data to the UE 100. Since the model learning is also performed in the network device 400, the learning data does not need to be transmitted to the UE 100.

[0225] Other than the above, this operation example is the same as the (2-2-1-1) operation example (FIG. 18).

[0226] (2.2.2.2) When Different Trained AI / ML Models are Used Figure 21 shows an example of operation when different AI / ML models are used. That is, Figure 21 shows an example of operation when a Type B1, two-side model (different AI / ML models) is used and model learning is performed by both the UE 100 and the network device 400. Hereinafter, for convenience, this example of operation may be referred to as the (2-2-2-2) example of operation.

[0227] The differences between this operation example and the (2-2-2-1) operation example (FIG. 20) are as follows.

[0228] First, the targets for model learning are different between UE 100 and network device 400. That is, UE 100 performs model learning of an AI / ML model for UE 100 (step S200), and network device 400 performs model learning of an AI / ML model for network device 400 (step S201). Therefore, transmitter 120 of UE 100 may transmit to network device 400 learning parameters used in network device 400, rather than learning parameters used by itself (step S171). The targets for re-learning are also different AI / ML models for UE 100 and network device 400 (steps S202 and S203).

[0229] Secondly, the processing target is the AI / ML model for UE 100 and the AI / ML model for network device 400. For example, network device 400 specifies a model ID and transmits an instruction to execute model inference to UE 100 (step S174), but the target of the instruction is the trained AI / ML model for UE 100. The same applies to re-learning (steps S184 and S185).

[0230] Other than the above, this operation example and the (2-1-2-1) operation example are basically the same.

[0231] (3) Third Operation Example (Operation Example of Type B2) Next, a third operation example will be described. In the third operation example, an operation example related to Type B2 will be described. As described above, Type B2 is a type in which the model ID is identified using wireless signaling, and model identification is initiated by the network device 400. The above contribution proposes that Type B2 be performed by the operations shown in (B2-1) to (B2-3) (FIG. 12(C)) above.

[0232] Here, referring again to FIG. 12(C), the network device 400 transmits the trained AI / ML model itself generated by model learning to the UE 100 (step S51). The transmission of the AI / ML model itself consumes significantly more resources than other data. In addition, if model inference is not performed in the UE 100, the transmission of the trained AI / ML model in the network device 400 is wasted. Even if the UE 100 receives the trained AI / ML model, considering its own capabilities, it may not be able to perform model inference using the AI / ML model. In other words, like the first and second operation examples, the model identification operation example shown in FIG. 12(C) focuses only on model identification and does not consider the entire LCM. Therefore, the UE 100 may not be able to perform LCM operation efficiently.

[0233] Therefore, in a third operation example, the network device 400 transmits meta information to the UE 100 instead of transmitting the trained AI / ML model. Specifically, first, the network device (e.g., the network device 400) trains the AI / ML model. Second, the network device transmits identification information (e.g., a model ID) of the AI / ML model and additional information (e.g., meta information) of the AI / ML model to the user device (e.g., the UE 100). Third, the additional information is used in the user device to determine whether or not to execute inference for the AI / ML model and whether or not to transmit parameters used for inference (e.g., inference data) to the network device.

[0234] In this way, the UE 100 can determine whether or not model inference is executable based on the additional information. Also, the UE 100 can determine whether or not model parameters can be transmitted to the network device 400 based on the additional information. Therefore, the UE 100 can efficiently execute an LCM operation (e.g., model inference) by utilizing the additional information.

[0235] The third operation will be described in detail below. As with the first and second operation examples, the third operation example will be described separately as a network-side model and a two-side model, taking into consideration the situation in which inference is performed.

[0236] (3.1) (3-1) Example of Operation In this example of operation, for Type B2 (where the model ID is identified using wireless signaling and model identification is initiated from the network device 400), an example of operation will be described in which model inference is performed on the network device 400 side (i.e., network-side model).

[0237] FIG. 22 is a diagram illustrating this operation example according to the first embodiment.

[0238] 22, in step S210, the network device 400 (e.g., the control unit 230 of the gNB 200) performs model learning and generates a learned AI / ML model. Step S210 represents Type B2, in which model identification is initiated by the network device 400.

[0239] In step S211, the network device 400 (e.g., the control unit 230 of the gNB 200) may record in memory the model ID of the trained AI / ML model and the meta information used when generating the trained AI / ML model.

[0240] In step S212, the network device 400 (for example, the transmitter 210 of the gNB 200) transmits a model ID and meta information to the UE 100. The network device 400 may transmit the model ID and meta information using control data. Comparing this step with FIG. 12 (C), meta information is transmitted instead of transmitting the AI / ML model. In this operation example, since the AI / ML model is not transmitted, it is also possible to effectively utilize radio resources. The receiver 110 of the UE 100 receives the model ID and meta information. Note that the meta information may be the same as the meta information described in the first operation example.

[0241] In step S213, the network device 400 (e.g., the transmitter 210 of the gNB 200) specifies a model ID and transmits to the UE 100 an instruction to execute model inference of the trained AI / ML model specified by the model ID. However, the execution instruction may not be an execution instruction for model inference executed by the UE 100, but may be an instruction indicating that model inference is to be performed by the network device 400. Alternatively, the instruction may be a transmission instruction instructing the network device 400 to transmit model parameters (e.g., inference data) necessary for model inference, since model inference is performed by the network device 400. The network device 400 may transmit the transmission instruction using control data. The receiver 110 of the UE 100 receives the transmission instruction.

[0242] In step S214, in response to receiving the execution instruction, the control unit 130 of the UE 100 uses the meta information to determine whether to transmit the model parameters to the network device 400. As described in the first operation example, the meta information may include, as an application condition, information regarding the location where the trained AI / ML model can be executed. In a trained AI / ML model, model parameters acquired in a specific region may be used to perform model inference. Therefore, model parameters acquired in regions other than the specific region may not be necessary for model inference. The control unit 130 of the UE 100, for example, may use a GNSS function to acquire location information when the model parameters were acquired and compare it with the location information included in the meta information to determine whether the model parameters are necessary for the trained AI / ML model. For example, if the location information when the model parameters were acquired matches the location information included in the meta information, the control unit 130 may determine to transmit the model parameters to the network device 400, and if they do not match, may determine not to transmit the model parameters to the network device 400. In the following description, it is assumed that the control unit 130 has determined to transmit the model parameters. The determination of whether or not to transmit the model parameters may be made at the timing when the UE 100 receives the model ID in step S212. In step S214, the transmission unit 120 of the UE 100 transmits the model parameters. The network device 400 (for example, the reception unit 220 of the gNB 200) receives the model parameters.

[0243] In step S215, the network device 400 (e.g., the control unit 230 of the gNB 200) uses the model parameters to perform model inference using the trained AI / ML model.

[0244] In step S216, the network device 400 (e.g., the control unit 230 of the gNB 200) detects that the result of the model inference is below expectations.

[0245] In step S217, the network device 400 (e.g., the control unit 230 of the gNB 200) retrains the trained AI / ML model and generates an updated AI / ML model.

[0246] The subsequent steps S218 to S222 are the same as steps S211 to S215, respectively, except that the target is the updated AI / ML model.

[0247] (3.2) Operational Example of Type B2 and Two-Side Model Next, an operational example of the two-side model for Type B2 will be described. In this operational example, two operational examples will be described in total. That is, the case where the same trained AI / ML model is used for model inference in the UE 100 and the network device 400, and the case where different trained AI / ML models are used for model inference in the UE 100 and the network device 400 will be described.

[0248] (3.2.1) When the same trained AI / ML model is used Figure 23 shows an example of operation when the same trained AI / ML model ("UE_NW") is used. That is, Figure 23 shows an example of operation of Type B2 and a two-side model (the same trained AI / ML model). Hereinafter, for convenience, this example of operation may be referred to as the (3-2-1) example of operation.

[0249] The differences between this operation example and the (3-1) operation example (FIG. 22) are as follows.

[0250] First, the target of model learning, model inference, and re-learning is the AI / ML model ("UE_NW") that is commonly used by UE 100 and network device 400. Therefore, the target of model learning, model inference, and re-learning is the same AI / ML model for UE 100 and network device 400 (steps S230, S234, S236, S238, S242, and S244).

[0251] Second, this operation example is a two-sided model. Therefore, the network device 400 (e.g., the transmitter 210 of the gNB 200) transmits the trained AI / ML model in addition to the model ID and meta information (step S232). Then, the control unit 130 of the UE 100 performs model inference using the trained AI / ML model (step S234). The trained AI / ML model that is the subject of model inference is the same as the model used for model inference in the network device 400. The updated AI / ML model is also transmitted to the UE 100 (step S240) and used for model inference (step S242), similar to the trained AI / ML model.

[0252] Third, the transmitter 120 of the UE 100 transmits the inference result (inference result data) to the network device 400 (step S235), and the network device 400 (e.g., the control unit 230 of the gNB 200) performs model inference using the inference result (step S236).

[0253] (3.2.2) When Different Trained AI / ML Models are Used Figure 24 shows an example of operation when different trained AI / ML models are used. That is, Figure 23 shows an example of operation of Type B2 and a two-side model (different trained AI / ML models). Hereinafter, for convenience, this example of operation may be referred to as the (3-2-2) example of operation.

[0254] The differences between this operation example and the (3-2-1) operation example (FIG. 23) are as follows.

[0255] First, the difference is that the targets of model learning and re-learning are divided into the AI / ML model for UE100 and the AI / ML model for network device 400. Therefore, network device 400 (e.g., control unit 230 of gNB200) performs model learning of the AI / ML model for UE100 and model learning of the AI / ML model for network device 400 separately (steps S250 and S251). In addition, network device (e.g., transmission unit 210 of gNB200) transmits the learned AI / ML model for UE100 ("model_UE") to UE100 (step S252). The same applies to re-learning (steps S253 to S255).

[0256] Secondly, the processing target is an AI / ML model for UE 100 and an AI / ML model for network device 400. For example, network device 400 (e.g., transmitter 210 of gNB 200) specifies a model ID and transmits an instruction to execute model inference to UE 100 (step S252), but the target of the execution instruction is the trained AI / ML model for UE 100. The same applies to re-learning (step S255).

[0257] [Other Embodiments] In the first embodiment described above, supervised learning has been mainly described, but the present invention is not limited to this. For example, unsupervised learning or reinforcement learning may be applied to the first embodiment.

[0258] The above-described operational flows are not limited to being implemented independently, but can also be implemented by combining two or more operational flows. For example, some steps of one operational flow may be added to another operational flow, or some steps of one operational flow may be replaced with some steps of another operational flow. In each flow, it is not necessary to execute all steps, and only some steps may be executed.

[0259] In the above-described embodiments and examples, an example in which the base station is an NR base station (gNB) has been described, but the base station may be an LTE base station (eNB) or a 6G base station. The base station may also be a relay node such as an IAB (Integrated Access and Backhaul) node. The base station may also be a DU of the IAB node. The UE 100 may also be an MT (Mobile Termination) of the IAB node.

[0260] That is, the UE 100 may be a terminal function unit (a type of communication module) for a base station to control a repeater that relays signals. Such a terminal function unit is referred to as an MT. Examples of the MT include, in addition to the IAB-MT, an NCR (Network Controlled Repeater)-MT and a RIS (Reconfigurable Intelligent Surface)-MT.

[0261] The term "network node" primarily refers to a base station, but may also refer to a core network device or a part of a base station (CU, DU, or RU). A network node may also be configured by a combination of at least a part of a core network device and at least a part of a base station.

[0262] A program may be provided that causes a computer to execute each process performed by the UE 100, the gNB 200, or the network device 400. The program may be recorded on a computer-readable medium. Using a computer-readable medium, the program can be installed on a computer. Here, the computer-readable medium on which the program is recorded may be a non-transitory recording medium. The non-transitory recording medium is not particularly limited, and may be, for example, a recording medium such as a CD-ROM and / or a DVD-ROM. Furthermore, circuits that execute each process performed by the UE 100, the gNB 200, or the network device 400 may be integrated, and at least a portion of the UE 100, the gNB 200, or the network device 400 may be configured as a semiconductor integrated circuit (chip set, SoC: System on a chip).

[0263] The functions performed by the UE 100, the gNB 200, or the network device 400 may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may also be a programmed processor that executes a program stored in memory. In this specification, circuitry, unit, or means refers to hardware that is programmed to perform the described functions or hardware that executes them. The hardware may be any hardware disclosed herein or any hardware known to be programmed or capable of performing the described functions. If the hardware is a processor, the circuitry, means, or unit is a combination of hardware and software used to configure the hardware and / or processor.

[0264] As used in this disclosure, the terms "based on" and "depending on / in response to" do not mean "based only on" or "depending only on," unless expressly stated otherwise. The term "based on" means both "based only on" and "based at least in part on." Similarly, the term "depending on" means both "depending only on" and "depending at least in part on." The terms "include," "comprise," and variations thereof do not mean including only the listed items, but may mean including only the listed items or may include additional items in addition to the listed items. Additionally, the term "or," as used in this disclosure, is not intended to mean an exclusive or. Furthermore, any reference to elements using designations such as "first," "second," etc., as used in this disclosure does not generally limit the quantity or order of those elements. These designations may be used herein as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed therein or that the first element must precede the second element in some way. In this disclosure, where articles are added by translation, such as a, an, and the in English, these articles shall include the plural unless the context clearly indicates otherwise.

[0265] The functions performed by the UE 100, the gNB 200, or the network device 400 may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and / or other circuits and is considered to be circuitry or processing circuitry. A processor may be a programmed processor that executes a program stored in a memory.

[0266] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0267] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of hardware and software used to configure the hardware and / or processor.

[0268] Although one embodiment has been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design changes can be made within the scope of the gist. Furthermore, the embodiments, operation examples, and processes can be appropriately combined within the scope of not being inconsistent.

[0269] This application claims priority from Japanese Patent Application No. 2023-217328 (filed December 22, 2023), the entire contents of which are incorporated herein by reference.

[0270] (Additional Note) The above can be summarized as follows:

[0271] (Supplementary Note 1) A communication control method in a user device of a mobile communication system, comprising: a step in which the user device acquires model identification information that identifies an AI / ML model and additional information of the AI / ML model; and a step in which the user device transmits the model identification information to a network device, wherein the transmitting step includes a step in which the user device transmits the model identification information to the network device when it determines, based on the additional information, that inference for the AI / ML model is possible.

[0272] (Supplementary Note 2) The communication control method according to Supplementary Note 1, wherein the acquiring step includes a step in which the user device acquires the AI / ML model, and further includes a step in which the user device performs inference using the AI / ML model.

[0273] (Supplementary Note 3) The communication control method according to Supplementary Note 1 or Supplementary Note 2, wherein the transmitting step includes a step of determining that the inference is possible and transmitting the model identification information to the network device when the user device determines, based on the additional information, that parameters used for the inference can be obtained.

[0274] (Supplementary Note 4) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 3, wherein inference for the AI / ML model is performed by the network device.

[0275] (Supplementary Note 5) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 4, wherein the additional information includes at least any of the type of data used for the inference, application conditions under which the AI / ML model is applied, use cases under which the AI / ML model is applied, and version information of the AI / ML model.

[0276] (Supplementary Note 6) The communication control method according to any one of Supplementary Notes 1 to 5, wherein the acquiring step includes a step of the user device receiving the model identification information and the additional information from an external server.

[0277] (Supplementary Note 7) A communication control method in a user device of a mobile communication system, comprising: a step in which the user device causes an AI / ML model to learn; and a step in which the user device transmits parameters used for learning and additional information of the AI / ML model to a network device, wherein the network device learns the AI / ML model using the parameters.

[0278] (Supplementary Note 8) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 7, wherein the inference of the AI / ML model is performed by the user device and the network device.

[0279] (Supplementary Note 9) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 8, wherein the additional information includes at least any of data used in the inference, conditions under which the AI / ML model is applied, use cases under which the AI / ML model is applied, and version information of the AI / ML model.

[0280] (Supplementary Note 10) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 9, wherein the AI / ML model learned in the user device and the AI / ML model learned in the network device are different AI / ML models.

[0281] (Supplementary Note 11) A communication control method in a network device of a mobile communication system, comprising: a step in which the network device trains an AI / ML model; and a step in which the network device transmits identification information of the AI / ML model and additional information of the AI / ML model to a user device, wherein the additional information is used in the user device to determine whether or not to execute inference of the AI / ML model and whether or not to transmit parameters used for inference to the network device.

[0282] (Supplementary Note 12) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 11, further comprising: the network device receiving the parameters from the user device; and the network device performing inference using the AI / ML model.

[0283] (Supplementary Note 13) The communication control method according to any one of Supplementary Note 1 to Supplementary Note 12, further comprising the steps of: the network device receiving an inference result of the AI / ML model from the user device; and the network device executing inference of the AI / ML model using the inference result.

[0284] 1: Mobile communication system 20: 5GC (CN) 100: UE 110: Receiving unit 120: Transmitting unit 130: Control unit 200: gNB 210: Transmitting unit 220: Receiving unit 230: Control unit 400: Network device 500: External server

Claims

1. A communication control method in a user device of a mobile communication system, comprising: the user device acquiring model identification information that identifies an AI (Artificial Intelligence) / ML (Machine Learning) model and additional information of the AI / ML model; and the user device transmitting the model identification information to a network device, wherein the transmitting step includes transmitting the model identification information to the network device when the user device determines, based on the additional information, that inference on the AI / ML model is possible.

2. The communication control method according to claim 1, wherein said acquiring includes said user device acquiring said AI / ML model, and further comprising said user device performing inference using said AI / ML model.

3. The communication control method of claim 1, wherein the transmitting step includes, when the user device determines based on the additional information that parameters used in the inference can be obtained, determining that the inference is possible and transmitting the model identification information to the network device.

4. A communication control method according to claim 2 or claim 3, wherein inference for said AI / ML model is performed in said network device.

5. The communication control method according to claim 1, wherein the additional information includes at least any one of the following: a type of data used in the inference, application conditions under which the AI / ML model is applied, a use case to which the AI / ML model is applied, and version information of the AI / ML model.

6. The communication control method according to claim 1, wherein said obtaining includes said user device receiving said model identification information and said additional information from an external server.

7. A communication control method in a user device of a mobile communication system, comprising: the user device causes an AI / ML model to learn; and the user device transmits parameters used for the learning and additional information of the AI / ML model to a network device, wherein the network device learns the AI / ML model using the parameters.

8. The communication control method according to claim 7, wherein the inference of the AI / ML model is performed in the user device and the network device.

9. The communication control method according to claim 7, wherein the additional information includes at least any one of data used in the inference, conditions under which the AI / ML model is applied, use cases to which the AI / ML model is applied, and version information of the AI / ML model.

10. The communication control method according to claim 7, wherein the AI / ML model trained in the user device and the AI / ML model trained in the network device are different AI / ML models.

11. A communication control method in a network device of a mobile communication system, comprising: the network device learning an AI / ML model; and the network device transmitting identification information of the AI / ML model and additional information of the AI / ML model to a user device, wherein the additional information is used in the user device to determine whether or not to execute inference of the AI / ML model and whether or not to transmit parameters used in the inference to the network device.

12. The method of claim 11, further comprising: said network device receiving said parameters from said user device; and said network device performing inference using said AI / ML model.

13. The communication control method according to claim 12, further comprising: the network device receiving an inference result of the AI / ML model from the user device; and the network device executing inference of the AI / ML model using the inference result.

Images