Communication method and user device

Abstract

A communication method according to the present invention, in which a machine learning technology is applied to wireless communication between a user device and a base station in a mobile communication system, includes: a step in which one communication device of the user device and the base station transmits, to the other communication device of the user device and the base station, a notification indicating at least one of the fact that the one communication device has an unlearned model, the fact that the one communication device has a model being learned, and the fact that the one communication device has a learned model the inspection of which has been completed; and a step in which the one communication device receives a response corresponding to the notification from the other communication device.

Description

Communication Method

[0001] The present disclosure relates to a communication method for use in a mobile communication system.

[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 contribution: RP-213599, “New SI: Study on Artificial Intelligence (AI) / Machine Learning (ML) for NR Air Interface”

[0004] A communication method according to a first aspect is a method for applying machine learning technology to wireless communication between a user device and a base station in a mobile communication system, the communication method including the steps of: transmitting a notification indicating at least one of that one of the user device and the base station has an untrained model, that one of the user device and the base station has a model being trained, and that one of the user device and the base station has a trained model for which testing has been completed, to the other of the user device and the base station;

[0005] A communication method according to a second aspect is a method of applying machine learning technology to wireless communication between a user device and a base station in a mobile communication system, the communication method including the steps of: one of the user device and the base station performing inference processing using a trained model obtained by training a model; determining, by the one of the communication devices, whether re-training of the model is necessary by monitoring performance of the trained model; and, in response to determining that the re-training is necessary, transmitting a notification indicating the necessity of the re-training to the other of the user device and the base station.

[0006] A communication method according to a third aspect is a method of applying machine learning technology to wireless communication between a user device and a base station in a mobile communication system, the communication method including the steps of: receiving, by one of the user device and the base station, configuration information from the other of the user device and the base station, the configuration information including information indicating a time when a dataset for monitoring performance of a trained model is provided; and receiving, by the one of the user device and the base station, the dataset from the other of the user device and performing a monitoring process for monitoring performance of the trained model using the dataset at the time.

[0007] 1 is a diagram showing a configuration of a mobile communication system according to an embodiment. FIG. 1 is a diagram showing a configuration of a UE (user equipment) according to an embodiment. FIG. 2 is a diagram showing a configuration of a gNB (base station) according to an embodiment. FIG. 3 is a diagram showing a protocol stack configuration of a radio interface of a user plane that handles data. FIG. 4 is a diagram showing a protocol stack configuration of a radio interface of a control plane that handles signaling (control signals). FIG. 5 is a diagram showing a functional block configuration of AI / ML technology (machine learning technology) in a mobile communication system according to an embodiment. FIG. 6 is a diagram showing an overview of operations related to each operation scenario according to an embodiment. FIG. 7 is a diagram showing a first operation scenario according to an embodiment. FIG. 8 is a diagram showing a first example of reducing CSI-RS according to an embodiment. FIG. 9 is a diagram showing a second example of reducing CSI-RS according to an embodiment. FIG. 10 is an operation flow diagram showing a first operation pattern related to the first operation scenario according to an embodiment. FIG. 11 is an operation flow diagram showing a second operation pattern related to the first operation scenario according to an embodiment. FIG. 12 is a diagram showing a second operation scenario according to an embodiment. FIG. 13 is an operation flow diagram showing an operation example related to the second operation scenario according to an embodiment. FIG. 14 is a diagram showing a third operation scenario according to an embodiment. FIG. 15 is an operation flow diagram showing an operation example related to the third operation scenario according to an embodiment. FIG. 16 is a diagram showing a first operation pattern related to model transfer according to an embodiment. 1 is a diagram showing an example of a setting message including a model and additional information according to an embodiment. FIG. 2 is a diagram showing a second operation pattern related to model transfer according to an embodiment. FIG. 3 is a diagram showing a third operation pattern related to model transfer according to an embodiment. FIG. 4 is a diagram showing an example of model management according to an embodiment. FIG. 5 is a diagram showing details of model management according to an embodiment. FIG. 6 is a diagram showing a first operation pattern related to model monitoring according to an embodiment. FIG. 7 is a diagram showing an example of applying AI / ML technology to CSI feedback as an example of a second operation pattern related to model monitoring according to an embodiment. FIG. 8 is a diagram showing an example of applying AI / ML technology to positioning as another example of the second operation pattern related to model monitoring according to an embodiment. FIG. 9 is a diagram showing CSI feedback and beam management applying AI / ML technology according to an embodiment. FIG. 10 is a diagram for explaining beam management applying AI / ML technology according to an embodiment.1A and 1B are diagrams illustrating a specific example of CSI feedback to which the AI / ML technique according to the embodiment is applied, and a specific example of beam management to which the AI / ML technique according to the embodiment is applied.

[0008] When attempting to apply machine learning technology to a mobile communication system, the technology for specifically utilizing machine learning processing has not yet been established.

[0009] Therefore, an object of the present disclosure is to enable the utilization of machine learning processing in mobile communication systems.

[0010] A mobile communication system according to an 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.

[0011] (Configuration of mobile communication system) First, the configuration of a mobile communication system according to an embodiment will be described. FIG. 1 is a diagram showing the configuration of a mobile communication system 1 according to an embodiment. The mobile communication system 1 conforms to the 5th Generation System (5GS) of the 3GPP standard. In the following description, 5GS will be used as an example, but an LTE (Long Term Evolution) system may also be applied at least in part to the mobile communication system. A sixth generation (6G) system may also be applied at least in part to the mobile communication system.

[0012] 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.

[0013] 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), 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).

[0014] 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").

[0015] 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.

[0016] 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 the UPF are connected to the gNB 200 via an NG interface, which is an interface between a base station and a core network.

[0017] 2 is a diagram showing the configuration of a UE 100 (user equipment) according to an 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.

[0018] The receiving unit 110 performs various reception operations 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.

[0019] 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.

[0020] 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 and encoding / decoding of baseband signals. The CPU executes programs stored in the memory to perform various processes.

[0021] 3 is a diagram showing the configuration of a gNB 200 (base station) according to an embodiment. The gNB 200 includes a transmitter 210, a receiver 220, a controller 230, and a backhaul communication unit 240. The transmitter 210 and the receiver 220 constitute a communication unit that performs wireless communication with the UE 100. The backhaul communication unit 240 constitutes a network communication unit that communicates with the CN 20. The gNB 200 is another example of a communication device.

[0022] 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.

[0023] 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.

[0024] The control unit 230 performs various controls and processes in the gNB 200. Such processes include processes for each layer, which will be described later. 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 and encoding / decoding of baseband signals. The CPU executes programs stored in the memory to perform various processes.

[0025] The backhaul communication unit 240 is connected to adjacent base stations via an Xn interface, which is an interface between base stations. The backhaul communication unit 240 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.

[0026] FIG. 4 is a diagram showing the configuration of a protocol stack of a radio interface of a user plane that handles data.

[0027] 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.

[0028] 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 the successfully decoded DCI as DCI addressed to the UE. The DCI transmitted from gNB200 has a CRC parity bit scrambled by the RNTI added.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] The PDCP layer performs header compression / decompression, encryption / decryption, and the like.

[0034] 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.

[0035] 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).

[0036] 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.

[0037] 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.

[0038] 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 300A. 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).

[0039] (Overview of AI / ML Technology) Next, the AI / ML technology according to the embodiment will be described. Fig. 6 is a diagram showing a functional block configuration of the AI / ML technology (also referred to as "machine learning technology") in the mobile communication system 1 according to the embodiment.

[0040] The functional block configuration shown in FIG. 6 includes a data collection unit A1, a model learning unit A2, a model inference unit A3, and a data processing unit A4.

[0041] The data collection unit A1 collects input data, specifically, learning data and inference data, outputs the learning data to the model learning unit A2, and outputs the inference data to the model inference unit A3. The data collection unit A1 may acquire data in the device on which the data collection unit A1 is provided as input data. The data collection unit A1 may also acquire data in another device as input data.

[0042] The model learning unit A2 performs model learning. Specifically, the model learning unit A2 optimizes parameters of a learning model (hereinafter also referred to as a "model" or an "AI / ML model") through machine learning using learning data, derives (generates and updates) a learned model, and outputs the learned model to the model inference unit A3. The model is a data-driven algorithm that applies AI / ML technology to generate a set of outputs based on a set of inputs. For example, considering y = ax + b, a (slope) and b (intercept) are parameters, and optimizing these corresponds to machine learning. Generally, machine learning is classified into supervised learning, unsupervised learning, and reinforcement learning. Supervised learning is a method that uses correct answer data as learning data. Unsupervised learning is a method that does not use correct data for training data. For example, in unsupervised learning, feature points are memorized from a large amount of training data and the correct answer is determined (range estimation). Reinforcement learning is a method that assigns a score to the output result and learns how to maximize the score.

[0043] The model inference unit A3 performs model inference. Specifically, the model inference unit A3 infers an output from the inference data using a trained model and outputs the inference result data to the data processing unit A4. For example, in the case of y = ax + b, x corresponds to the inference data and y corresponds to the inference result data. Note that "y = ax + b" is a model. A model in which the slope and intercept are optimized, for example, "y = 5x + 3", is a trained model. There are various modeling techniques, including linear regression analysis, neural networks, and decision tree analysis. The above "y = ax + b" can be considered a type of linear regression analysis. The model inference unit A3 may provide model performance feedback to the model learning unit A2.

[0044] The data processing unit A4 receives the inference result data and performs processing that utilizes the inference result data.

[0045] 7 is a diagram showing an overview of the operation related to each operation scenario according to the embodiment. In FIG. 7, one of the UE 100 and the gNB 200 corresponds to the first communication device, and the other corresponds to the second communication device.

[0046] In step S1, the UE 100 transmits control data related to machine learning techniques to the gNB 200 or receives it from the gNB 200. 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.

[0047] (First operation scenario) Figure 8 is a diagram showing a first operation scenario according to the embodiment. In the first operation scenario, the data collection unit A1, the model learning unit A2, and the model inference unit A3 are arranged in the UE 100 (e.g., the control unit 130), and the data processing unit A4 is arranged in the gNB 200 (e.g., the control unit 230). In other words, model learning and model inference are performed on the UE 100 side.

[0048] In the first operation scenario, machine learning technology is introduced into channel state information (CSI) feedback from UE100 to gNB200. The CSI (CSI feedback information) transmitted (feedback) from UE100 to gNB200 is information regarding the downlink channel state 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.

[0049] The gNB 200 transmits a reference signal for the UE 100 to estimate the downlink channel state. Such a reference signal may be, for example, a CSI reference signal (CSI-RS). Such a reference signal may be a demodulation reference signal (DMRS). For example, the reference signal is assumed to be a CSI-RS.

[0050] 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. Such a first reference signal is sometimes referred to as a full CSI-RS.

[0051] For example, UE 100 (CSI generation unit 131) performs channel estimation using a received signal (CSI-RS) received by receiving unit 110 from gNB 200, and generates CSI. UE 100 (transmission unit 120) transmits the generated CSI to gNB 200. Model learning unit A2 performs model learning using multiple sets of received signals (CSI-RS) and CSI as learning data, and derives a learned model for inferring CSI from the received signal (CSI-RS).

[0052] Second, in model inference, UE100 (receiving unit 110) receives a second reference signal from gNB200 using second resources that are less than the first resources. Then, UE100 (model inference unit A3) uses the learned model to infer CSI from inference data including the second reference signal as inference result data. In the description of the first operating scenario, such a second reference signal may be referred to as a partial CSI-RS or a punctured CSI-RS.

[0053] For example, the UE 100 (model inference unit A3) uses the received signal (CSI-RS) received by the receiver 110 from the gNB 200 as inference data, and infers CSI from the received signal (CSI-RS) using a trained model. The UE 100 (transmitter 120) transmits the inferred CSI to the gNB 200.

[0054] This enables UE 100 to feed back 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.

[0055] FIG. 9 is a diagram illustrating a first example of reducing CSI-RS according to the embodiment. In the first example, the gNB 200 reduces the number of antenna ports that transmit the CSI-RS. For example, in a mode in which the UE 100 performs model learning, the gNB 200 transmits the CSI-RS from all antenna ports of the antenna panel. On the other hand, in a mode in which the UE 100 performs model inference, the gNB 200 reduces the number of antenna ports that transmit the CSI-RS and transmits the CSI-RS from half the antenna ports of the antenna panel. Note that the antenna port is an example of a resource. This reduces overhead, improves antenna port utilization efficiency, and reduces power consumption.

[0056] 10 is a diagram showing a second example of reducing CSI-RS according to the embodiment. In the second example, the gNB 200 reduces the number of radio resources, specifically, time-frequency resources, that transmit the CSI-RS. For example, in a mode in which the UE 100 performs model learning, the gNB 200 transmits the CSI-RS using predetermined time-frequency resources. On the other hand, in a mode in which the UE 100 performs model inference, the gNB 200 transmits the CSI-RS using a smaller amount of time-frequency resources than the predetermined time-frequency resources. This reduces overhead, improves the utilization efficiency of radio resources, and reduces power consumption.

[0057] Next, a first operation pattern related to the first operation scenario will be described. In this first operation pattern, the gNB 200 transmits a switching notification to the UE 100 as control data, notifying the UE 100 of mode switching between a mode for performing model learning (hereinafter also referred to as "learning mode") and a mode for performing model inference (hereinafter also referred to as "inference mode"). The UE 100 receives the switching notification and switches between the learning mode and the inference mode. This makes it possible to appropriately switch between the learning mode and the inference mode. The switching notification may be setting information for setting a mode in the UE 100. The switching notification may also be a switching command for instructing the UE 100 to switch modes.

[0058] In this first operation pattern, when the model learning is completed, the UE 100 transmits a completion notification indicating that the model learning is completed to the gNB 200 as control data. The gNB 200 receives the completion notification. This allows the gNB 200 to understand that the model learning has been completed on the UE 100 side.

[0059] 11 is an operation flow diagram showing a first operation pattern according to a first operation scenario of the embodiment. This flow may be performed after the UE 100 establishes an RRC connection with a cell of the gNB 200. In the following operation flow diagram, optional steps are indicated by dashed lines.

[0060] In step S101, the gNB 200 may notify or set the input data pattern in the inference mode, for example, the transmission pattern (puncture pattern) of the CSI-RS in the inference mode, as control data to the UE 100. For example, the gNB 200 notifies the UE 100 of the antenna port and / or time-frequency resource from which the CSI-RS is or is not transmitted in the inference mode.

[0061] In step S102, gNB200 may send a switching notification to UE100 to start the learning mode.

[0062] In step S103, the UE 100 starts the learning mode.

[0063] In step S104, the gNB 200 transmits the full CSI-RS. The UE 100 receives the full CSI-RS and generates CSI based on the received CSI-RS. In the learning mode, the UE 100 can perform supervised learning using the received CSI-RS and the corresponding CSI. The UE 100 may derive and manage learning results (trained models) for each of its communication environments, for example, for each reception quality (RSRP, RSRQ, SINR) and / or movement speed.

[0064] In step S105, UE100 transmits (feeds back) the generated CSI to gNB200.

[0065] Thereafter, in step S106, when the model learning is completed, the UE 100 transmits a completion notification to the gNB 200 indicating that the model learning is completed. The UE 100 may transmit a completion notification to the gNB 200 when the derivation (generation, update) of the learned model is completed. Here, the UE 100 may notify that the learning is completed for each of its own communication environments (e.g., movement speed, reception quality). In this case, the UE 100 includes information indicating which communication environment the completion notification is for in the notification.

[0066] In step S107, gNB200 sends a switching notification to UE100 to switch from learning mode to inference mode.

[0067] In step S108, in response to receiving the switching notification in step S107, the UE 100 switches from the learning mode to the inference mode.

[0068] In step S109, the gNB 200 transmits a partial CSI-RS. When the UE 100 receives the partial CSI-RS, it infers CSI from the received CSI-RS using a trained model. The UE 100 may select a trained model corresponding to its own communication environment from the trained models managed for each communication environment, and infer CSI using the selected trained model.

[0069] In step S110, UE100 transmits (feeds back) the inferred CSI to gNB200.

[0070] In step S111, if UE 100 determines that model learning is necessary, it may transmit a notification that model learning is necessary as control data to gNB 200. For example, when UE 100 moves, when its movement speed changes, when its reception quality changes, when the cell in which it is located changes, or when the bandwidth portion (BWP) used for communication changes, it assumes that the accuracy of the inference result can no longer be guaranteed, and transmits the notification to gNB 200.

[0071] Next, a second operation pattern related to the first operation scenario will be described. This second operation pattern may be used in combination with the above-mentioned operation pattern. In this second operation pattern, the gNB 200 transmits a completion condition notification indicating the completion condition of model learning to the UE 100 as control data. The UE 100 receives the completion condition notification and determines the completion of model learning based on the completion condition notification. This allows the UE 100 to appropriately determine the completion of model learning. The completion condition notification may be setting information that sets the completion condition of model learning in the UE 100. The completion condition notification may be included in a switching notification that notifies (instructs) the UE 100 to switch to the learning mode.

[0072] FIG. 12 is an operation flow diagram showing a second operation pattern according to a first operation scenario according to the embodiment.

[0073] In step S201, the gNB 200 transmits a completion condition notification indicating the completion condition of the model learning as control data to the UE 100. The completion condition notification may include at least one of the following completion condition information.

[0074] - Acceptable error range for correct answer data: For example, this is the acceptable range of error between the CSI generated using a normal CSI feedback calculation method and the CSI inferred by model inference. When learning has progressed to a certain extent, the UE 100 infers the CSI using the trained model at that time, compares it with the correct CSI, and can determine that learning is complete based on whether the error is within the acceptable range.

[0075] Number of training data: The number of data used for training, for example, the number of received CSI-RSs corresponds to the number of training data. The UE 100 can determine that training is complete when the number of received CSI-RSs in the training mode reaches the notified (set) number of training data.

[0076] Number of learning attempts: The number of times model learning is performed using learning data. The UE 100 can determine that learning is complete when the number of learning attempts in the learning mode reaches a notified (set) number of times.

[0077] Output score threshold: For example, a score in reinforcement learning. The UE 100 can determine that the learning is completed when the score reaches a notified (set) score.

[0078] The UE 100 continues learning based on the full CSI-RS until it is determined that the learning is completed (steps S203 and S204).

[0079] In step S205, when UE100 determines that model learning has been completed, it may send a completion notification to gNB200 indicating that model learning has been completed.

[0080] Next, a third operation pattern related to the first operation scenario will be described. This third operation pattern may be used in conjunction with the above-described operation pattern. When it is desired to improve the accuracy of CSI feedback, not only CSI-RS but also other types of data, such as the reception characteristics of the physical downlink shared channel (PDSCH), can be used as training data and inference data. In this third operation pattern, the gNB 200 transmits data type information specifying at least the type of data to be used as training data to the UE 100 as control data. In other words, the gNB 200 specifies to the UE 100 what type of training data and inference data to use (type of input data). The UE 100 receives the data type information and performs model training using the specified type of data. This allows the UE 100 to perform appropriate model training.

[0081] FIG. 13 is an operation flow diagram showing a third operation pattern according to the first operation scenario according to the embodiment.

[0082] In step S301, the UE 100 may transmit capability information indicating which type of input data the UE 100 can handle by machine learning as control data to the gNB 200. Here, the UE 100 may further notify accompanying information such as the accuracy of the input data.

[0083] In step S302, the gNB 200 transmits data type information to the UE 100. The data type information may be setting information for setting the type of input data to the UE 100. Here, the type of input data may be reception quality and / or UE movement speed for CSI feedback. The reception quality may be reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-interference-and-noise ratio (SINR), bit error rate (BER), block error rate (BLER), analog-to-digital converter output waveform, etc.

[0084] In addition, assuming UE positioning, which will be described later, the type of input data may be location information (latitude, longitude, and altitude) of the GNSS (Global Navigation Satellite System), an RF fingerprint (cell ID and its reception quality, etc.), the angle of arrival (AoA) of the received signal, the reception level, reception phase, and reception time difference (OTDOA) for each antenna, round trip time, wireless LAN (Local Area Network) or other short-range wireless reception information.

[0085] The gNB200 may specify the type of input data independently for each of the training data and the inference data. The gNB200 may specify the type of input data independently for each of the CSI feedback and the UE positioning.

[0086] (Second Operation Scenario) Next, the second operation scenario will be described, focusing on differences from the first operation scenario. In the first operation scenario, the downlink reference signal (i.e., downlink CSI estimation) has been mainly described. In the second operation scenario, the uplink reference signal (i.e., uplink CSI estimation) will be described. In the description of the second operation scenario, the uplink reference signal is a sounding reference signal (SRS), but it may also be an uplink DMRS or the like.

[0087] 14 is a diagram showing a second operation scenario according to the embodiment. In the second operation scenario, the data collection unit A1, the model learning unit A2, the model inference unit A3, and the data processing unit A4 are arranged in the gNB 200 (e.g., the control unit 230). In other words, model learning and model inference are performed on the gNB 200 side.

[0088] In the second operation scenario, machine learning technology is introduced into the CSI estimation performed by the gNB200 based on the SRS from the UE100. Therefore, the gNB200 (e.g., the control unit 230) has a CSI generation unit 231 that generates CSI based on the SRS received by the receiving unit 220 from the UE100. This CSI is information indicating the channel state of the uplink between the UE100 and the gNB200. The gNB200 (e.g., the data processing unit A4) performs, for example, uplink scheduling based on the CSI generated based on the SRS.

[0089] First, in model learning, gNB200 (receiving unit 220) receives a first reference signal from UE100 using a first resource. Then, gNB200 (model learning unit A2) derives a learned model for inferring CSI from a reference signal (SRS) using learning data including the first reference signal. In the description of the second operating scenario, such a first reference signal may be referred to as a full SRS.

[0090] For example, gNB200 (CSI generation unit 231) performs channel estimation using the received signal (SRS) received by receiver 220 from UE100 to generate CSI. Model learning unit A2 performs model learning using multiple sets of received signals (SRS) and CSI as learning data, and derives a learned model for inferring CSI from the received signal (SRS).

[0091] Second, in model inference, gNB200 (receiving unit 220) receives a second reference signal from UE100 using a second resource that is less than the first resource. Then, UE100 (model inference unit A3) uses the learned model to infer CSI as inference result data from inference data including the second reference signal. In the description of the second operating scenario, such a second reference signal may be referred to as a partial SRS or a punctured SRS. For the SRS puncturing pattern, a pattern similar to that in the first operating scenario can be used (see Figures 9 and 10).

[0092] For example, gNB200 (model inference unit A3) uses the received signal (SRS) received by the receiving unit 220 from UE100 as inference data, and infers CSI from the received signal (SRS) using a learned model.

[0093] This enables the gNB 200 to generate accurate (complete) CSI from the small number of SRSs (partial SRSs) received from the UE 100. For example, the UE 100 can reduce (puncture) the SRS when intended to reduce overhead. In addition, the gNB 200 can respond to situations where the radio conditions deteriorate and some SRSs cannot be received normally.

[0094] In such an operating scenario, the "CSI-RS" in the operation of the first operating scenario described above can be replaced with "SRS," "gNB200" with "UE100," and "UE100" with "gNB200."

[0095] In addition, in the second operation scenario, gNB200 transmits reference signal type information indicating the type of reference signal to be transmitted by UE100, between the first reference signal (full SRS) and the second reference signal (partial SRS), as control data to UE100. UE100 receives the reference signal type information and transmits the SRS specified by gNB200 to gNB200. This allows UE100 to transmit an appropriate SRS.

[0096] FIG. 15 is an operational flow diagram showing an example of an operation according to a second operation scenario according to the embodiment.

[0097] In step S501, gNB200 configures UE100 for SRS transmission.

[0098] In step S502, gNB200 starts learning mode.

[0099] In step S503, UE100 transmits the full SRS to gNB200 according to the setting in step S501. gNB200 receives the full SRS and performs model learning for channel estimation.

[0100] In step S504, gNB200 identifies an SRS transmission pattern (puncture pattern) to be input as inference data into the learned model, and sets the identified SRS transmission pattern to UE100.

[0101] In step S505, gNB200 transitions to inference mode and starts model inference using the trained model.

[0102] In step S506, UE100 transmits a partial SRS in accordance with the SRS transmission setting in step S504. When gNB200 inputs the SRS as inference data into the trained model to obtain a channel estimation result, it uses the channel estimation result to perform uplink scheduling of UE100 (for example, control of uplink transmission weight, etc.). Note that if the inference accuracy using the trained model deteriorates, gNB200 may reconfigure UE100 to transmit a full SRS.

[0103] (Third Operation Scenario) Next, the third operation scenario will be described, mainly focusing on the differences from the first and second operation scenarios. The third operation scenario is an embodiment in which position estimation of the UE 100 (so-called UE positioning) is performed using federated learning. Fig. 16 is a diagram showing the third operation scenario according to the embodiment. In an application example of such federated learning, for example, the following procedure is performed.

[0104] First, the location server 400 transmits the model to the UE 100 .

[0105] Secondly, the UE 100 performs model learning on the UE 100 (model learning unit A2) side using data stored in the UE 100. The data stored in the UE 100 is, for example, a positioning reference signal (PRS) received by the UE 100 from the gNB 200 and / or output data of the GNSS receiver 140. The data stored in the UE 100 may include location information (including latitude and longitude) generated by the location information generation unit 132 based on the reception result of the PRS and / or the output data of the GNSS receiver 140.

[0106] Third, the UE 100 applies the learned model, which is the learning result, in the UE 100 (model inference unit A3), and transmits variable parameters (hereinafter also referred to as "learned parameters") included in the learned model to the location server 400. In the above example, the optimized a (slope) and b (intercept) correspond to the learned parameters.

[0107] Fourth, location server 400 (associated learning unit A5) collects learned parameters from multiple UEs 100 and integrates them. Location server 400 may transmit the learned model obtained by the integration to UE 100. Location server 400 can estimate the location of UE 100 based on the learned model obtained by the integration and measurement reports from UE 100.

[0108] In the third operation scenario, gNB200 transmits trigger setting information to UE100 as control data, which sets a transmission trigger condition for UE100 to transmit the learned parameters. UE100 receives the trigger setting information, and transmits the learned parameters to gNB200 (location server 400) when the set transmission trigger condition is satisfied. This allows UE100 to transmit the learned parameters at an appropriate timing.

[0109] FIG. 17 is an operational flow diagram showing an example of an operation according to a third operation scenario according to the embodiment.

[0110] In step S601, the gNB 200 may notify the UE 100 of the base model to be learned. Here, the base model may be a model that has been learned in the past. As described above, the gNB 200 may transmit data type information indicating what the input data is to be to the UE 100.

[0111] In step S602, the gNB 200 instructs the UE 100 to perform model learning and sets the timing (trigger condition) for reporting the learned parameters. The set reporting timing may be periodic. The set reporting timing may be triggered (i.e., an event trigger) when the learning proficiency level meets the condition.

[0112] In the case of periodic timing, the gNB 200 sets, for example, a timer value to the UE 100. The UE 100 starts the timer when learning starts (step S603), and when it expires, reports the learned parameters to the gNB 200 (location server 400) (step S604). Alternatively, the gNB 200 may specify the radio frame or time to report to the UE 100. The radio frame may be specified as an absolute value, for example, SFN = 512. The radio frame may be calculated by modulo arithmetic. For example, the gNB 200 sets N as a setting value, and the UE 100 reports the learned parameters with an SFN such that "SFN mod N = 0" (step S604).

[0113] In the case of an event trigger, the completion condition as described above is set in the UE 100. When the completion condition is satisfied, the UE 100 reports the learned parameters to the gNB 200 (location server 400) (step S604). The UE 100 may trigger the reporting of the learned parameters, for example, when the accuracy of the model inference becomes better than the previously transmitted model. The UE 100 may introduce an offset here and trigger when "current accuracy > previous accuracy + offset". The UE 100 may trigger the reporting of the learned parameters, for example, when the learning data has been input (learned) N or more times. Such an offset and / or the value of N may be set to the UE 100 by the gNB 200.

[0114] In step S604, when the reporting timing conditions are met, UE100 reports the learned parameters at that time to the network (gNB200).

[0115] In step S605, the network (location server 400) integrates the learned parameters reported from the multiple UEs 100.

[0116] (First operation pattern related to model transfer) Figure 18 is a diagram showing a first operation pattern related to model transfer according to the embodiment. In the drawings referred to in the following embodiments, non-essential processes are indicated by dashed lines. Furthermore, in the following embodiments, the communication device 501 is primarily assumed to be the UE 100, but the communication device 501 may be the gNB 200 or the AMF 300A. Furthermore, the communication device 502 is primarily assumed to be the gNB 200, but the communication device 502 may be the UE 100 or the AMF 300A.

[0117] As shown in FIG. 18, in step S701, the gNB 200 transmits a capability inquiry message to the UE 100 to request transmission of a message including an information element indicating execution capability for machine learning processing. The capability inquiry message is an example of a transmission request requesting transmission of a message including an information element indicating execution capability for machine learning processing. The UE 100 receives the capability inquiry message. However, the gNB 200 may transmit the capability inquiry message when executing the machine learning processing (when determining that the processing will be executed).

[0118] In step S702, the UE 100 transmits to the gNB 200 a message including an information element indicating execution capabilities for machine learning processing (or, from another perspective, the execution environment for machine learning processing). The gNB 200 receives the message. The message may be an RRC message, for example, a "UE Capability" message defined in the RRC technical specifications, or a newly defined message (for example, a "UE AI Capability" message, etc.). Alternatively, the communication device 502 may be an AMF 300A, and the message may be a NAS message. Alternatively, if a new layer is defined for executing or controlling machine learning processing (AI / ML processing), the message may be a message of the new layer. The new layer will be referred to as the "AI / ML layer" as appropriate.

[0119] The information element indicating the execution capability for machine learning processing is at least one of the following information elements (A1) to (A3).

[0120] Information Element (A1) The information element (A1) is an information element indicating the processor's capability for executing machine learning processing and / or an information element indicating the memory's capability for executing machine learning processing.

[0121] The information element indicating the processor's capability for executing machine learning processing may be an information element indicating whether the UE 100 has an AI processor. If the UE 100 has the processor, the information element may include the AI ​​processor product number (model number). The information element may be an information element indicating whether the UE 100 can use a GPU (Graphics Processing Unit). The information element may be an information element indicating whether the machine learning processing must be executed by a CPU. By transmitting the information element indicating the processor's capability for executing machine learning processing from the UE 100 to the gNB 200, the network side can determine, for example, whether the UE 100 can use a neural network model as a model. The information element indicating the processor's capability for executing machine learning processing may be an information element indicating the clock frequency and / or the number of parallel executions of the processor.

[0122] The information element indicating the memory capacity for executing the machine learning process may be an information element indicating the memory capacity of a volatile memory (e.g., RAM: Random Access Memory) among the memories of the UE 100. The information element may also be an information element indicating the memory capacity of a non-volatile memory (e.g., ROM: Read Only Memory) among the memories of the UE 100. The information element may be both of these. The information element indicating the memory capacity for executing the machine learning process may be specified for each type, such as a memory for storing a model, a memory for an AI processor, or a memory for a GPU.

[0123] The information element (A1) may be defined as an information element for inference processing (model inference). The information element (A1) may be defined as an information element for learning processing (model learning). The information element (A1) may be defined as both an information element for inference processing and an information element for learning processing.

[0124] Information element (A2) Information element (A2) is an information element indicating the execution capability of the inference process. Information element (A2) may be an information element indicating a model supported in the inference process. The information element may be an information element indicating whether a deep neural network model is supported. In this case, the information element may include at least one of information indicating the number of layers (stages) of a neural network that can be supported, information indicating the number of neurons that can be supported (which may be the number of neurons per layer), and information indicating the number of synapses that can be supported (which may be the number of input or output synapses per layer or per neuron).

[0125] The information element (A2) may be an information element indicating the execution time (response time) required to execute the inference process. The information element (A2) may be an information element indicating the number of inference processes that can be executed simultaneously (e.g., how many inference processes can be executed in parallel). The information element (A2) may be an information element indicating the processing capacity of the inference process. For example, if the processing load of a certain standard model (standard task) is determined to be 1 point, the information element indicating the processing capacity of the inference process may be information indicating how many points its own processing capacity is.

[0126] Information element (A3) Information element (A3) is an information element indicating the execution capability of the learning process. Information element (A3) may be an information element indicating a learning algorithm supported in the learning process. The learning algorithm indicated by the information element includes supervised learning (e.g., linear regression, decision tree, logistic regression, k-nearest neighbor method, support vector machine, etc.), unsupervised learning (e.g., clustering, k-means method, principal component analysis, etc.), reinforcement learning, and deep learning. When UE 100 supports deep learning, the information element may include at least one of information indicating the number of layers (stages) of a neural network that can be supported, information indicating the number of neurons that can be supported (which may be the number of neurons per layer), and information indicating the number of synapses that can be supported (which may be the number of input or output synapses per layer or per neuron).

[0127] The information element (A3) may be an information element indicating the execution time (response time) required to execute the learning process. The information element (A3) may be an information element indicating the number of concurrent executions of the learning process (e.g., how many learning processes can be executed in parallel). The information element (A3) may be an information element indicating the processing capacity of the learning process. For example, if the processing load of a certain standard model (standard task) is determined to be 1 point, the information element indicating the processing capacity of the learning process may be information indicating how many points its own processing capacity is. Note that, with regard to the number of concurrent executions, since the learning process generally has a higher processing load than the inference process, the information may be information indicating the number of concurrent executions with the inference process (e.g., two inference processes and one learning process).

[0128] In step S703, gNB200 determines a model to be configured (deployed) in UE100 based on the information elements included in the message received in step S702. The model may be a trained model used by UE100 in the inference process. The model may also be an untrained model used by UE100 in the learning process.

[0129] In step S704, gNB200 transmits a message including the model determined in step S703 to UE100. UE100 receives the message and performs machine learning processing (learning processing and / or inference processing) using the model included in the message. A specific example of step S704 will be described in the following second operation pattern.

[0130] (Second operation pattern related to model transfer) Figure 19 is a diagram showing an example of a configuration message including a model and additional information according to the embodiment. The configuration message may be an RRC message transmitted from the gNB 200 to the UE 100, for example, an "RRC Reconfiguration" message defined in the RRC technical specifications, or a newly defined message (for example, an "AI Deployment" message or an "AI Reconfiguration" message). Alternatively, the configuration message may be a NAS message transmitted from the AMF 300A 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.

[0131] In the example of FIG. 19, 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.

[0132] FIG. 20 is a diagram illustrating a second operation pattern regarding model transfer according to the embodiment.

[0133] In step S711, gNB200 transmits a configuration message including a model and additional information to UE100. UE100 receives the configuration message. The configuration message includes at least one of the following information elements (B1) to (B6).

[0134] (B1) Model The "model" may be a trained model used by UE100 in the inference process. The "model" may be an untrained model used by UE100 in the training process. In the configuration message, the "model" may be encapsulated (containerized). If the "model" is a neural network model, the "model" may be expressed by the number of layers (stages), the number of neurons per layer, and the synapses (weighting) between each neuron. For example, a trained (or untrained) neural network model may be expressed by a combination of matrices.

[0135] A single setting message may contain multiple "models." In this case, the multiple "models" may be included in the setting message in list form. The multiple "models" may be set for the same purpose, or may be set for different purposes. Details of the use of the models will be described later.

[0136] (B2) Model Index A "model index" is an example of additional information (for example, individual additional information). A "model index" is an index (index number) assigned to a model. In the activation command and deletion message described below, a model can be specified by the "model index." When changing the model settings, the model can also be specified by the "model index."

[0137] (B3) Model Use "Model Use" is an example of additional information (individual additional information or common additional information). "Model Use" specifies the function to which the model is applied. For example, functions to which the model is applied include CSI feedback, beam management (beam estimation, overhead / latency reduction, beam selection accuracy improvement), positioning, modulation / demodulation, encoding / decoding (CODEC), and packet compression. The content of the model use and its index (identifier) ​​may be pre-defined in the 3GPP technical specifications, and the "model use" may be specified by an index. For example, the model use and its index (identifier) ​​are defined such that CSI feedback is specified as use index #A and beam management is specified as use index #B. The UE 100 deploys a model for which a "model use" is specified in a functional block corresponding to the specified use. Note that the "model use" may also be an information element specifying the input data and output data of the model.

[0138] (B4) Model Execution Requirements "Model execution requirements" are an example of additional information (for example, individual additional information). "Model execution requirements" are information elements that indicate the performance (required performance) required to apply (execute) the model, for example, the processing delay (required latency).

[0139] (B5) Model Selection Criteria The "model selection criteria" is an example of additional information (individual additional information or common additional information). UE 100 applies (executes) a corresponding model in response to the criteria specified in the "model selection criteria" being satisfied. The "model selection criteria" may be the movement speed of UE 100. In this case, the "model selection criteria" may be specified as a speed range such as "low speed movement" or "high speed movement." The "model selection criteria" may be specified as a movement speed threshold. The "model selection criteria" may be radio quality (e.g., RSRP / RSRQ / SINR) measured by UE 100. In this case, the "model selection criteria" may be specified as a radio quality range. The "model selection criteria" may be specified as a radio quality threshold. The "model selection criteria" may be the location (latitude / longitude / altitude) of UE 100. The "model selection criteria" may be set to follow sequential notifications from the network (an activation command, described later). The "model selection criteria" may specify an autonomous selection of the UE 100.

[0140] (B6) Necessity of Learning Process "Necessity of Learning Process" is an information element indicating whether or not a learning process (or re-learning) is necessary or possible for the corresponding model. If a learning process is necessary, the type of parameter to be used in the learning process may be further set. For example, in the case of CSI feedback, it is set so that CSI-RS and UE movement speed are used as parameters. If a learning process is necessary, the method of the learning process, for example, supervised learning, unsupervised learning, reinforcement learning, or deep learning, may be further set. It may also be set whether or not the learning process is to be performed immediately after the model is set. If it is not to be performed immediately, the execution of the learning may be controlled by an activation command described below. For example, in the case of federated learning, it may also be set whether or not the result of the learning process of the UE100 is to be notified to the gNB200. When it is necessary to notify the gNB200 of the results of the learning process of the UE100, the UE100 may, after executing the learning process, encapsulate the learned model or learned parameters and transmit them to the gNB200 by an RRC message, etc. The information element indicating "whether or not learning process is required" may be an information element indicating, in addition to whether or not learning process is required, whether or not the corresponding model is to be used only for model inference.

[0141] In step S712, the UE 100 determines whether the model set in step S711 can be deployed (executed). The UE 100 may make this determination when activating the model, which will be described later, and step S713, which will be described later, may be a message notifying an error at the time of activation. Furthermore, the UE 100 may make this determination while the model is in use (during execution of the machine learning process), rather than at the time of deployment or activation. If it is determined that the model cannot be deployed (step S712: NO), that is, if an error occurs, in step S713, the UE 100 transmits an error message to the gNB 200. The error message may be an RRC message transmitted from the UE 100 to the gNB 200, for example, a "Failure Information" message specified in the RRC technical specifications, or a newly defined message (for example, an "AI Deployment Failure Information" message). The error message may be UCI (Uplink Control Information) defined in the physical layer or MAC CE (Control Element) defined in the MAC layer. Alternatively, the error message may be a NAS message transmitted from the UE 100 to the AMF 300A. Alternatively, when a new layer (AI / ML layer) for performing machine learning processing (AI / ML processing) is defined, the message may be a message of the new layer.

[0142] The error message includes at least one of the following information elements (C1) to (C3):

[0143] (C1) Model Index This is the model index of the model that has been determined to be undeployable.

[0144] (C2) Usage Index This is the usage index of the model that has been determined to be undeployable.

[0145] (C3) Error Cause This is an information element regarding the cause of the error. The "error cause" may be, for example, an "unsupported model," "exceeding processing capacity," "error occurrence phase," or "other error." The "unsupported model" may be, for example, the UE 100 being unable to support a neural network model, or being unable to support machine learning processing (AI / ML processing) of a specified function. The "exceeding processing capacity" may be, for example, an overload (processing load and / or memory load exceeding capacity), an inability to satisfy the requested processing time, interrupt processing or priority processing of an application (upper layer), etc. The "error occurrence phase" is information indicating when the error occurred. The "error occurrence phase" may be classified as during deployment (configuration), activation, or operation. The "error occurrence phase" may be classified as during inference processing or learning processing. The "other errors" are other causes.

[0146] When an error occurs, the UE 100 may automatically delete the corresponding model. The UE 100 may delete the model when it confirms that an error message has been received by the gNB 200, for example, when it receives an ACK in a lower layer. When the gNB 200 receives an error message from the UE 100, it may recognize that the model has been deleted.

[0147] On the other hand, if it is determined that the model set in step S711 can be deployed (step S712: YES), that is, if no error occurs, in step S714, the UE 100 deploys the model according to the setting. "Deployment" may mean making the model applicable. "Deployment" may mean actually applying the model. In the former case, the model is not applied simply by being deployed, but is applied when the model is activated by an activation command described below. In the latter case, once the model is deployed, the model becomes in use.

[0148] In step S715, the UE 100 transmits a response message to the gNB 200 in response to the completion of the model deployment. The gNB 200 receives the response message. The UE 100 may transmit the response message when the activation of the model is completed by an activation command described below. The response message may be an RRC message transmitted from the UE 100 to the gNB 200, for example, an "RRC Reconfiguration Complete" message defined in the RRC technical specifications, or a newly defined message (for example, an "AI Deployment Complete" message). The response message may be a MAC CE defined in the MAC layer. Alternatively, the response message may be a NAS message transmitted from the UE 100 to the AMF 300A. Alternatively, if a new layer for performing machine learning processing (AI / ML processing) is defined, the message may be a message of the new layer.

[0149] In step S716, the UE 100 may transmit a measurement report message, which is an RRC message including measurement results of the radio environment, to the gNB 200. The gNB 200 receives the measurement report message.

[0150] In step S717, the gNB 200 selects a model to be activated, for example, based on the measurement report message, and transmits an activation command (selection command) to the UE 100 to activate the selected model. The UE 100 receives the activation command. The activation command may be a DCI, MAC CE, RRC message, or an AI / ML layer message. The activation command may include a model index indicating the selected model. The activation command may include information specifying whether the UE 100 performs an inference process or a learning process.

[0151] The gNB 200 selects a model to be deactivated, for example, based on a measurement report message, and transmits a deactivation command (selection command) to the UE 100 to deactivate the selected model. The UE 100 receives the deactivation command. The deactivation command may be a DCI, MAC CE, RRC message, or a message of the AI ​​ / ML layer. The deactivation command may include a model index indicating the selected model. When the UE 100 receives the deactivation command, the UE 100 may deactivate (discontinue application of) the specified model without deleting it.

[0152] In step S718, in response to receiving the activation command, the UE 100 applies (activates) the designated model. The UE 100 performs an inference process and / or a learning process using the activated model from among the deployed models.

[0153] Then, in step S719, the gNB 200 transmits a deletion message to the UE 100 to delete the model. The UE 100 receives the deletion message. The deletion message may be a MAC CE, an RRC message, a NAS message, or a message of the AI ​​ / ML layer. The deletion message may include a model index of the model to be deleted. When the UE 100 receives the deletion message, it deletes the specified model.

[0154] (Third Operation Pattern Regarding Model Transfer) In this third operation pattern, the UE 100 notifies the network of the load status of the machine learning process (AI / ML process). As a result, the network (e.g., gNB 200) can determine how many more models can be deployed (or activated) in the UE 100 based on the notified load status. This third operation pattern does not need to be based on the first operation pattern regarding the above-mentioned model transfer. This third operation pattern may be based on the first operation pattern.

[0155] FIG. 21 is a diagram illustrating a third operation pattern regarding model transfer according to the embodiment.

[0156] In step S751, the gNB 200 transmits a message to the UE 100 including a request for information provision of the AI / ML processing load status or a setting for reporting the AI / ML processing load status. The UE 100 receives the message. The message may be a MAC CE, an RRC message, a NAS message, or a message of the AI / ML layer. The setting for reporting the AI / ML processing load status may include information for setting a report trigger (transmission trigger), for example, "Periodic" or "Event triggered." "Periodic" sets the reporting period, and the UE 100 reports at that period. "Event triggered" sets a threshold value to be compared with a value indicating the AI / ML processing load status in the UE 100 (processing load value and / or memory load value), and the UE 100 reports when that value satisfies the condition of the threshold. Here, the threshold value may be set for each model. For example, the message may associate a model index with a threshold value.

[0157] In step S752, the UE 100 transmits a message (report message) including information indicating the AI / ML processing load status to the gNB 200. The message may be an RRC message, for example, a "UE Assistance Information" message or a "Measurement Report" message. The message may be a newly defined message (for example, an "AI Assistance Information" message). The message may be a NAS message or an AI / ML layer message.

[0158] The message includes a "processing load status" and / or a "memory load status." The "processing load status" may indicate what percentage of the processing capacity (processor capacity) is being used or what percentage is remaining available. Alternatively, the "processing load status" may express the load in points as described above, and notify how many points are being used and how many points are remaining available. The UE 100 may notify the "processing load status" for each model. For example, the UE 100 may include at least one set of a "model index" and a "processing load status" in the message. The "memory load status" may be memory capacity, memory usage, or remaining memory. The UE 100 may notify the "memory load status" for each type, such as memory for storing models, memory for the AI ​​processor, memory for the GPU, etc.

[0159] In step S752, if the UE 100 wants to stop using a particular model, for example, because of a high processing load or inefficiency, the UE 100 may include information (model index) indicating the model whose setting is to be deleted or deactivated in the message. When the UE 100's processing load becomes critical, the UE 100 may include alert information in the message and transmit it to the gNB 200.

[0160] In step S753, gNB200 determines a model setting change, etc. based on the message received from UE100 in step S752, and transmits a message for the model setting change to UE100. The message may be a MAC CE, an RRC message, a NAS message, or an AI / ML layer message. gNB200 may transmit the above-mentioned activation command or deactivation command to UE100.

[0161] (Model Management) FIG. 22 is a diagram illustrating an example of model management according to the embodiment.

[0162] In step S801, the communication device 501 executes AI / ML processing (machine learning processing). The machine learning processing is one of the steps shown in FIG. 23, which will be described later.

[0163] In step S802, the communication device 501 transmits a notification regarding the machine learning process as control data to the communication device 502. The communication device 502 receives the notification.

[0164] In step S802, the communication device 501 sends a notification to the communication device 502 indicating at least one of, for example, that it has an untrained model, that it has a model in training, and that it has a trained model that has been inspected.

[0165] In step S803, the communication device 502 transmits a response corresponding to the notification in step S802 as control data to the communication device 501. The communication device 501 receives the response.

[0166] The notification in step S802 may be a notification indicating that the communication device 501 has an unlearned model. In this case, step S803 may include at least one of a dataset and setting parameters used for model learning.

[0167] The notification in step S802 may be a notification indicating that the communication device 501 has a model that is being trained. In this case, the response in step S803 may include a dataset for continuing model training.

[0168] The notification in step S802 may be a notification indicating that the communication device 501 has a trained model for which inspection has been completed. The response in step S803 may include information for starting use of the trained model for which inspection has been completed.

[0169] The notification in step S802 and the response in step S803 may each include an index of the corresponding model and / or identification information for identifying the type or use of the corresponding model (e.g., for CSI feedback, for beam management, for positioning, etc.). Hereinafter, this information is also referred to as "model use information, etc."

[0170] FIG. 23 is a diagram showing details of model management according to this embodiment, specifically, step S801 in FIG. 22.

[0171] In step S811, the communication device 501 executes a model deployment process. Here, the communication device 501 notifies the communication device 502 that it has an untrained model, i.e., a model that requires training. For example, the untrained model may be pre-installed when the communication device 501 is shipped. The communication device 501 may acquire the untrained model from the communication device 502. The communication device 501 may notify the communication device 502 that it has an untrained model if model training is not complete, for example, if a certain level of quality is not satisfied. For example, this may apply to a case where model training was once completed, but the quality of the model can no longer be guaranteed in monitoring due to movement to a different environment (e.g., from indoors to outdoors). The communication device 502 may provide a training dataset to the communication device 501 based on the notification. The communication device 502 may also perform associated settings on the communication device 501. The communication device 502 may deactivate, for example, discard, deconfigure, or deactivate the model.

[0172] In step S812, the communication device 501 executes a model learning process. The communication device 501 notifies the communication device 502 that model learning is in progress. The notification may include, as described above, model usage information, etc. The communication device 502 continues to provide the communication device 501 with a learning dataset based on the notification. Note that, when the communication device 502 receives a notification indicating that learning is in progress or before learning, the communication device 502 may recognize that the communication device 501 is applying a conventional method that does not apply a model.

[0173] In step S813, the communication device 501 executes a model verification process. The model verification process is a sub-process of the model training process. The model verification process is a process for evaluating the quality of an AI / ML model using a dataset different from the dataset used for model training and selecting (adjusting) model parameters. The communication device 501 may notify the communication device 502 that model training is in progress or that model verification has been completed.

[0174] In step S814, the communication device 501 executes a model checking process. The model checking process is a sub-process of the model learning process. In the model checking process, a dataset different from the datasets used in model learning and model validation is used to evaluate the performance of the final AI / ML model. Unlike model validation, model checking does not involve adjusting the model. The communication device 501 notifies the communication device 502 that it has a validated model (i.e., a model that can guarantee a certain level of quality). The notification may include, as described above, information on the use of the model. Based on the notification, the communication device 502 performs a process to start using the model, such as configuring or activating the model. The communication device 502 may decide to provide an inference dataset and perform the necessary settings on the communication device 501.

[0175] In step S815, the communication device 501 executes a model sharing process. For example, the communication device 501 transmits (uploads) the trained model to the communication device 502.

[0176] In step S816, the communication device 501 executes a model activation process. The model activation process is a process for activating (enabling) a model for a specific function. The communication device 501 may notify the communication device 502 that the model has been activated. The notification may include, as described above, information about the use of the model.

[0177] In step S817, the communication device 501 executes a model inference process. The model inference process is a process of generating a set of outputs based on a set of inputs using a trained model. The communication device 501 may notify the communication device 502 that model inference has been executed. The notification may include, as described above, information about the use of the model.

[0178] In step S818, the communication device 501 executes a model monitoring process. The model monitoring process is a process for monitoring the inference performance of the AI / ML model. The communication device 501 may transmit a notification regarding the model monitoring process to the communication device 502. The notification may include, as described above, model usage information, etc. Specific examples of the notification will be described later.

[0179] In step S819, the communication device 501 executes a model deactivation process. The model deactivation process is a process of deactivating (disabling) a model for a specific function. The communication device 501 may notify the communication device 502 that the model has been deactivated. The notification may include information about the model's use, as described above. The model deactivation process may be a process of deactivating a currently active model and activating another model. This process is also called model switching.

[0180] (First Operation Pattern Related to Model Monitoring) In this first operation pattern, the communication device 501 performs inference processing using a trained model obtained by training the model. The communication device 501 monitors the performance of the trained model to determine the need for relearning of the model. In response to determining that the relearning is necessary, the communication device 501 transmits a notification indicating the need for the relearning to the communication device 502. This enables the communication device 502 to provide the communication device 501 with training data to be used for the relearning.

[0181] FIG. 24 is a diagram illustrating a first operation pattern related to model monitoring according to the embodiment.

[0182] 24, in step S821, the communication device 502 provides training data (training data set) to the communication device 501. The training data may be, for example, the full CRI-RS.

[0183] In step S822, the communication device 501 generates a trained model by performing model training using the training data.

[0184] In step S823, the communication device 501 may activate the model.

[0185] In step S824, the communication device 502 transmits setting information including parameters related to monitoring of model performance (monitoring process) as control data to the communication device 501. The communication device 501 receives the setting information. The communication device 501 performs model monitoring process to monitor the performance of the trained model using the parameters included in the setting information (step S826).

[0186] The setting information includes at least one of a model index (or application) of the monitored object, a monitoring threshold, and a condition for sending a monitoring result report. The threshold (monitoring threshold) is a threshold that is compared with a value indicating the inference performance of the model in model monitoring and is used to determine whether the model's performance satisfies a certain standard. The condition for sending a monitoring result report may be, for example, a periodic report or an event trigger (such as when the monitoring threshold is no longer satisfied).

[0187] In step S825, the communication device 501 performs model inference using the model.

[0188] In step S826, the communication device 501 starts monitoring the model.

[0189] In step S827, the communication device 501 determines whether the performance of the model satisfies a certain standard, that is, whether re-learning of the model is necessary.

[0190] If it is determined that the performance of the model does not satisfy a certain standard, that is, that re-training of the model is necessary (step S827: YES), in step S828, the communication device 501 transmits a notification as control data to the communication device 502. The notification is a notification requesting the provision of a training dataset and may include, for example, identification information for identifying the type of dataset requested to be provided (e.g., full CRI-RS). The notification is a notification that model re-training is necessary and may include, for example, identification information for identifying a model index and / or a model type / application (CSI inference, beamforming inference, position inference, etc.).

[0191] In step S829, the communication device 502 provides the requested training data set to the communication device 501. For example, the communication device 502 starts transmitting the full CSI-RS.

[0192] In step S830, the communication device 501 performs model learning (re-learning) using the learning data to correct the learned model (for example, adjust the model parameters).

[0193] (Second Operation Pattern Related to Model Monitoring) In this second operation pattern, the communication device 501 receives configuration information including monitoring parameters for monitoring the performance of a trained model from the communication device 502. The communication device 501 performs model monitoring processing using a monitoring dataset based on the configuration information. This second operation pattern may be implemented in combination with the above-described first operation pattern related to model monitoring.

[0194] Here, the monitoring parameters may include time information indicating the time at which the monitoring dataset is provided from the communication device 502 (i.e., the monitoring period). The communication device 501 may receive the monitoring dataset from the communication device 502 and perform model monitoring processing at the time indicated by the time information. The monitoring parameters may include use condition information indicating conditions under which the monitoring dataset is reduced and used to monitor model performance in the model monitoring processing. The monitoring parameters may include a performance evaluation threshold for monitoring model performance in the model monitoring processing.

[0195] The communication device 501 may transmit request information requesting the provision of a monitoring dataset to the communication device 502. The communication device 501 may receive the monitoring dataset transmitted from the communication device 502 based on the request information.

[0196] The communication device 501 may perform the model monitoring process as follows. Specifically, the communication device 501 derives correct data (e.g., CSI feedback information) from a monitoring dataset (e.g., full CSI-RS) without using a trained model. Furthermore, the communication device 501 acquires inference result data (e.g., inferred CSI feedback information) output by a trained model using partial data (e.g., partial CSI-RS) obtained by reducing the monitoring dataset as input.

[0197] The communication device 501 then evaluates the performance of the trained model by comparing the correct answer data with the inference result data. For example, the communication device 501 may determine that the inference performance of the model meets the standard if the error of the inference result data relative to the correct answer data is within a predetermined threshold range. On the other hand, the communication device 501 may determine that the inference performance of the model does not meet the standard if the error of the inference result data relative to the correct answer data is outside the predetermined threshold range.

[0198] In addition, when the communication device 501 is the UE 100 and the communication device 502 is the gNB 200, the monitoring data set may include a reference signal transmitted from the gNB 200 to the UE 100. The reference signal may be a CSI-RS or a positioning reference signal (PRS).

[0199] FIG. 25 is a diagram illustrating an example of a second operation pattern related to model monitoring according to an embodiment, in which AI / ML technology is applied to CSI feedback.

[0200] In step S851, the communication device 502 provides training data (training data set) to the communication device 501. The training data may be, for example, the full CRI-RS.

[0201] In step S852, the communication device 501 generates a trained model by performing model training using the training data.

[0202] In step S853, the communication device 501 may activate the model.

[0203] In step S854, if a monitoring dataset is required, the communication device 501 may transmit a request for the dataset as control data to the communication device 502. The request may include requested values ​​of the monitoring parameters set in step S855.

[0204] In step S855, the communication device 502 transmits the setting information including the monitoring parameters as control data to the communication device 501. The communication device 501 receives the setting information.

[0205] The setting information includes at least one of the following setting parameters (D1) to (D5).

[0206] (D1) Time information indicating a monitoring period This information indicates at least one of a monitoring cycle, monitoring timing, and a period for evaluating model performance. This information may be a bitmap of slots (or the like). For example, the communication device 502 transmits a full CSI-RS once per radio frame (one slot) and transmits a partial CSI-RS at other times (other timings). The monitoring period may be dynamically configured. If the communication device 502 is a gNB, the start and / or stop of monitoring data transmission (or model monitoring processing by the communication device 501) may be notified by DCI (or MAC CE).

[0207] (D2) Resource information indicating a monitoring resource This information may be information indicating a frequency resource to which a monitoring data set is provided, such as a resource block start position and the number of resource blocks, etc. The resource providing the monitoring data set may be set independently of the resource providing the inference data.

[0208] (D3) Information indicating the use of the monitoring dataset This information indicates, for example, one of CSI feedback, beam management, and positioning as the use of the monitoring dataset. Here, it is assumed that CSI feedback is set as the use of the monitoring dataset.

[0209] (D4) Information indicating the conditions for using the monitoring dataset. This information indicates, for example, CSI-RS resources (which may be multiple patterns) that may be punctured by the communication device 502. When monitoring the model, the communication device 501 (simulates) puncturing of a specified pattern to evaluate the performance of the model.

[0210] (D5) Threshold for performance evaluation of model monitoring This threshold is, for example, a threshold that indicates the error range between the inference result data and the correct answer data.

[0211] In step S856, the communication device 502 provides monitoring data (monitoring data set) during the monitoring period to the communication device 501. The monitoring data may be, for example, a full CRI-RS.

[0212] In step S857, the communication device 501 performs model inference and model monitoring. Here, the communication device 501 derives correct data (e.g., CSI feedback information) from the monitoring dataset without using a trained model. The communication device 501 also acquires inference result data (e.g., inferred CSI feedback information) output by the trained model using partial data (e.g., partial CSI-RS) obtained by reducing the monitoring dataset as input. The communication device 501 then evaluates the performance of the trained model by comparing the correct data with the inference result data.

[0213] Here, an example has been described in which the communication device 501 is the UE 100, the communication device 502 is the gNB 200, and the AI / ML technology is applied to CSI feedback. However, the communication device 501 may be the gNB 200, the communication device 502 may be the UE 100, and the AI / ML technology may be applied to SRS transmission. In this case, the gNB 200 performs model inference and model monitoring. For example, the gNB 200 sets the timing (i.e., the monitoring period) for transmitting the full SRS to the communication device 501. At that time, the gNB 200 may transmit at least one of the above-mentioned (D1) to (D3) to the UE 100. The UE 100 transmits the full SRS at that timing and transmits the punctured SRS at other timings. Here, the UE 100 may periodically transmit the full SRS like a configured grant. The UE 100 may transmit the full SRS in one shot by DCI (or MAC CE) like a PDCCH order. The gNB 200 may cause the UE 100 to transmit the full SRS by notifying the UE 100 of the start and / or stop of transmission of the full SRS by DCI (or MAC CE).

[0214] Figure 26 is a diagram showing an example of applying AI / ML technology to positioning (specifically, generating location information for the communication device 501) as another example of the second operation pattern related to model monitoring in the embodiment.

[0215] In step S871, the communication device 502 provides training data (training data set) to the communication device 501. The training data may be, for example, a full PRS. The communication device 501 may derive position information from the full PRS and generate a trained model that derives position information from the PRS using the full PRS and the position information as training data (step S872). The training data may be a general reference signal or PRS. If the communication device 501 has a GNSS receiver, the communication device 501 may use the reception state of a general reference signal or PRS (so-called RF fingerprint) and GNSS position information as training data to generate a trained model that derives position information from the RF fingerprint (step S872). Here, if the communication device 501 does not have a GNSS receiver, position information provided by a location server may be used instead of the GNSS position information.

[0216] In step S873, the communication device 501 may activate the model.

[0217] In step S874, if a monitoring dataset is required, the communication device 501 may transmit a request for the dataset as control data to the communication device 502. The request may include requested values ​​of the monitoring parameters set in step S875.

[0218] In step S875, the communication device 502 transmits setting information including the monitoring parameters as control data to the communication device 501. The communication device 501 receives the setting information. The setting information may include at least one of the setting parameters (D1) to (D5) described above, as well as parameters for setting a data source of the ground truth data (e.g., a GNSS receiver, a location server, etc.) during model monitoring.

[0219] In step S876, the communication device 501 provides the communication device 501 with training data (training data set) during the monitoring period. The training data may be, for example, a full PRS or location information provided by a location server.

[0220] In step S877, the communication device 501 performs model inference and model monitoring. As a first example of model monitoring, when position information (GNSS position information) from the GNSS receiver of the communication device 501 is used as ground truth data, the communication device 501 performs model inference using the PRS reception state as input data, and evaluates the performance of the model by comparing the error between the inference result data and the GNSS position information with a threshold value.

[0221] As a second example of model monitoring, when location information provided by a location server (server-provided location information) is used as ground truth data, the communication device 501 performs model inference using the PRS reception state as input data, and evaluates the performance of the model by comparing the error between the inference result data and the server-provided location information with a threshold. Here, the communication device 501 may obtain the server-provided location information from the location server by notifying the location server of the PRS reception state.

[0222] As a third example of model monitoring, when location information derived from a full PRS is used as ground truth data, the communication device 501 derives location information as ground truth data from the full PRS without using a trained model. The communication device 501 also acquires inference result data (i.e., inferred location information) output by a trained model using a partial PRS obtained by reducing the full PRS as input. The communication device 501 then evaluates the performance of the trained model by comparing the ground truth data with the inference result data.

[0223] (CSI Feedback and Beam Management Applying AI / ML Technology) FIG. 27 is a diagram illustrating CSI feedback and beam management applying AI / ML technology according to the embodiment.

[0224] In step S901, the communication device 501 receives radio signals transmitted by each of a plurality of communication resources of the communication device 502. When CSI feedback is assumed, the plurality of communication resources are a plurality of antenna ports of the communication device 502 (see FIG. 9). When beam management is assumed, the plurality of communication resources are a plurality of beams formed by the communication device 502 (e.g., gNB 200) (see FIG. 28).

[0225] In step S902, the communication device 501 communicates (i.e., transmits and / or receives) information indicating a combination of communication resources having a predetermined correlation among the plurality of communication resources with the communication device 502. When CSI feedback is assumed, the information indicating the combination includes identification information of each antenna port constituting the combination. When beam management is assumed, the information indicating the combination includes identification information of each beam constituting the combination.

[0226] In step S903, the communication device 501 executes AI / ML processing (machine learning processing) using the combination.

[0227] In this way, the communication device 501 communicates information indicating a combination of communication resources having a predetermined correlation with the communication device 502, thereby enabling efficient AI / ML processing (machine learning processing) using the combination. Hereinafter, a combination of communication resources having a predetermined correlation may be referred to as a combination of antenna ports having a high correlation or a combination of beams having a high correlation.

[0228] In step S902, the communication device 501 may receive, as control data, a notification including information indicating the combination from the communication device 502. In step S902, the communication device 501 may transmit, to the communication device 502, a notification including information indicating the combination and / or information regarding communication resources for which transmission of wireless signals can be stopped.

[0229] The communication device 501 may identify a combination of communications and acquire a trained model based on the identified combination and a wireless signal received from the communication device 502. The trained model may be a model for deriving inference result data for another communication resource constituting the combination based on reception status data of the wireless signal of one communication resource constituting the combination. The communication device 501 may derive inference result data for the other communication resource based on reception status data of the wireless signal of the one communication resource.

[0230] FIG. 29 is a diagram illustrating a specific example of CSI feedback to which the AI / ML technique according to the embodiment is applied.

[0231] In step S911, the communication device 502 may transmit a notification including information indicating a combination of antenna ports with a high correlation as control data to the communication device 501. The communication device 501 receives the notification. The combination may be a combination of antenna ports that the communication device 501 targets for model learning.

[0232] In step S912, the communication device 502 transmits full CSI-RS from multiple antenna ports. The communication device 501 receives full CSI-RS. If there is no notification in step S911, the communication device 501 may identify a combination of antenna ports with high correlation. For example, the communication device 501 performs model training using CSI-RS (full CSI-RS) transmitted from two antenna ports to generate a trained model (step S913). Then, the communication device 501 inputs the CSI-RS (partial CSI-RS) of one antenna port obtained by puncturing the full CSI-RS as input data to the trained model, and if the inference result output from the trained model falls within a certain error range with respect to the correct data, the communication device 501 may determine that the two antenna ports are highly correlated.

[0233] In step S914, the communication device 501 generates a trained model using a combination of highly correlated antenna ports. For example, the communication device 501 generates a trained model for estimating the CSI of antenna port #2 from the CSI-RS reception result of antenna port #1 using a combination of highly correlated antenna ports #1 and #2.

[0234] In step S915, the communication device 501 transmits, as control data, to the communication device 502, a notification indicating that model learning has been completed or a notification including information on antenna ports with high correlation. The communication device 501 may perform the notification of step S915 when generation of the trained model is completed in step S914 or when inspection of the trained model is completed. The notification of step S915 may include at least one of information indicating a combination of antenna ports with high correlation (e.g., antenna port numbers of each antenna port with high correlation), information indicating a combination of antenna ports for which generation of the trained model has been completed, information indicating antenna ports for which CSI-RS transmission can be stopped (e.g., antenna port numbers), and information indicating antenna ports for which CSI-RS transmission should be continued.

[0235] In step S916, the communication device 502 identifies the antenna port that will stop transmitting the CSI-RS based on the notification in step S915.

[0236] In step S917, the communication device 502 stops transmitting some of the CSI-RS, thereby entering a state in which it transmits partial CSI-RS. The communication device 501 receives the partial CSI-RS.

[0237] In step S918, the communication device 501 derives CSI as inference result data based on the partial CSI-RS by model inference using the trained model generated in step S914.

[0238] In step S919, the communication device 501 transmits the CSI feedback information, which is the inference result data, to the communication device 502. The communication device 502 receives the CSI feedback information.

[0239] FIG. 30 is a diagram showing a specific example of beam management to which AI / ML technology according to an embodiment is applied. Here, as shown in FIG. 28, it is assumed that a communication device 502 (in the illustrated example, gNB 200) forms multiple beams (e.g., beams #1 to #3). The beam management may be beam management for SSB-based beamforming. Such beam management includes, for example, SSB selection in the RRC idle state and beam monitoring and recovery in the RRC connected state. The beam management may be beam management for CSI-RS-based beamforming (precoding). Such beam management includes, for example, management of PDSCH beamforming in the RRC connected state.

[0240] In this specific example, the communication device 501 uses AI / ML technology to estimate a beam (e.g., beam #2) by measuring a certain beam (e.g., beam #1), specifically, to estimate the measurement results of the other beam. For such estimation to be possible, the beam controls of beam #1 and beam #2 must be linked. This is because if the correlation between the beams is low, for example, if the precoding is different for each slot, the above-mentioned estimation cannot be performed.

[0241] 30 , in step S931, the communication device 502 transmits a notification including information indicating a combination of two or more beams to be controlled in a coordinated manner to the communication device 501. The notification may be a notification indicating a correspondence between a beam to be used as inference data and a beam to be inferred. The communication device 501 receives the notification. The notification includes at least one of the following information (E1) to (E3):

[0242] (E1) Beam identifier association information This is a set of two or more beam identifiers (beam indexes). For example, communication device 502 notifies communication device 501 that beam #1 and beam #2 are associated (i.e., linked). Communication device 502 may notify communication device 501 that beam #1 and beam #2 are highly correlated in the current location and propagation environment of communication device 501. It may also notify that estimation is possible between beam #1 and beam #2.

[0243] (E2) Timing information for interlocking control For example, this is information associating time such as beam #1 and beam #2 in slot #1, and beam #1 and beam #3 in slot #2. One piece of time information may be individually associated with one set of beam identifiers. One piece of time information may be commonly associated with two or more sets of beam identifiers.

[0244] (E3) Precoding (weight) information of linked beams For example, communication device 501 infers the measurement results when received via beam #2 using the precoding (weight) information of beam #2 and the measurement information of beam #1.

[0245] In step S932, the communication device 502 forms a plurality of beams and transmits a radio signal (for example, SSB or PDSCH, etc.) The communication device 501 receives and measures each beam.

[0246] In step S933, the communication device 501 performs model learning based on the notification in step S931. For example, the communication device 501 performs model learning to infer the measurement results of one beam from the measurement information (measurement results) of another beam.

[0247] In step S934, the communication device 501 generates a trained model using a combination of highly correlated beams. For example, the communication device 501 generates a trained model for estimating the measurement results of beam #2 from the measurement results of beam #1 using a combination of highly correlated beams #1 and #2.

[0248] In step S935, the communication device 502 forms a plurality of beams and transmits a radio signal (for example, SSB or PDSCH, etc.) The communication device 501 receives and measures the beams.

[0249] In step S936, the communication device 501 performs beam measurements and infers other beams. The communication device 501 may use the inference results to estimate, for example, that the other beams have higher quality than the current beam (precoding).

[0250] In step S937, the communication device 501 may feed back the inference result, for example, as CSI, to the communication device 502. Here, the communication device 501 may transmit CSI feedback information to the communication device 502 together with information indicating that the CSI is the inferred CSI.

[0251] (Other Embodiments) In the above-described embodiments, communication between the UE100 and the gNB200 has been mainly described, but the operations according to the above-described embodiments may be applied to communication between the gNB200 and the AMF300A (i.e., communication between a base station and a core network). The above-described control data may be transmitted from the gNB200 to the AMF300A over the NG interface. The above-described control data may be transmitted from the AMF300A to the gNB200 over the NG interface. A request to execute federated learning and / or learning results of federated learning may be exchanged between the AMF300A and the gNB200. The operations of each of the above-described operational scenarios may be applied to communication between the gNB200 and another gNB200 (i.e., communication between base stations). The above-described control data may be transmitted from the gNB200 to another gNB200 over the Xn interface. A request to perform federated learning and / or a learning result of federated learning may be exchanged between the gNB200 and another gNB200. Each of the above operations may be applied to communication between the UE100 and another UE100 (i.e., communication between user equipments). The above control data may be transmitted from the UE100 to another UE100 over a side link. A request to perform federated learning and / or a learning result of federated learning may be exchanged between the UE100 and another UE100.

[0252] 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. 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.

[0253] In the above embodiment, an example in which the base station is an NR base station (gNB) has been described, but the base station may also be an LTE base station (eNB). 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 (Distributed Unit) of the IAB node. The user equipment (terminal device) may also be a relay node such as an IAB node, or an MT (Mobile Termination) of the IAB node.

[0254] A program may be provided that causes a computer to execute each process performed by a communication device (for example, UE 100 or gNB 200). 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 or DVD-ROM. Furthermore, circuits that execute each process performed by the communication device may be integrated, and at least a part of the communication device may be configured as a semiconductor integrated circuit (chip set, SoC: System on a chip).

[0255] As used in this disclosure, the terms "based on" and "depending on" 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 "based only on" and "at least in part on." Furthermore, "obtain" may mean obtaining information from stored information, obtaining information from information received from another node, or obtaining information by generating the information. The terms "include" and "comprise" do not mean including only the listed items, but may mean including only the listed items or including additional items in addition to the listed items. Furthermore, as used in this disclosure, the term "or" is not intended to mean an exclusive or. Furthermore, any reference to an element using a designation 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, 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, when articles are added by translation, such as a, an, and the in English, these articles are intended to include the plural unless the context clearly dictates otherwise.

[0256] The above describes the embodiments in detail with reference to the drawings, but the specific configuration is not limited to that described above, and various design changes can be made within the scope that does not deviate from the gist of the invention.

[0257] This application claims priority from Japanese Patent Application No. 2022-117532 (filed July 22, 2022), the entire contents of which are incorporated herein by reference.

[0258] (Additional Notes) Additional notes will be given regarding the features of the above-described embodiment.

[0259] (Supplementary Note 1) A communication method for applying machine learning technology to wireless communication between a user device and a base station in a mobile communication system, the communication method comprising the steps of: transmitting a notification to the other communication device of the user device and the base station indicating at least one of that the communication device has an untrained model, that the communication device has a model being trained, and that the communication device has a trained model that has been inspected; and receiving, by the one communication device, a response corresponding to the notification from the other communication device.

[0260] (Supplementary Note 2) The communication method according to Supplementary Note 1, wherein the sending step is a step of sending the notification indicating that the untrained model is present, and the receiving step includes a step of receiving at least one of a dataset and configuration parameters used for model training as the response.

[0261] (Supplementary Note 3) The communication method according to Supplementary Note 1, wherein the sending step is a step of sending the notification indicating that the model is being trained, and the receiving step includes a step of receiving as the response a dataset for continuing model training.

[0262] (Supplementary Note 4) The communication method according to Supplementary Note 1, wherein the sending step is a step of sending the notification indicating that the trained model for which the inspection has been completed is present, and the receiving step includes a step of receiving, as the response, information for starting use of the trained model for which the inspection has been completed.

[0263] (Supplementary Note 5) The communication method according to any one of Supplementary Notes 1 to 4, wherein the notification includes an index of the model and / or identification information for identifying a type or use of the model.

[0264] (Supplementary Note 6) A communication method that applies machine learning technology to wireless communication between a user device and a base station in a mobile communication system, comprising: a step in which one of the user device and the base station performs inference processing using a trained model obtained by training a model; a step in which the one of the communication devices determines whether re-learning of the model is necessary by monitoring performance of the trained model; and a step in which the one of the communication devices, in response to determining that the re-learning is necessary, sends a notification indicating the need for the re-learning to the other of the user device and the base station.

[0265] (Supplementary Note 7) The communication method according to Supplementary Note 6, further comprising: a step in which the one communication device receives configuration information including parameters related to monitoring of the performance from the other communication device; and a step in which the one communication device performs a monitoring process to monitor the performance of the trained model using the parameters included in the configuration information.

[0266] (Supplementary Note 8) The communication method according to Supplementary Note 6 or 7, wherein the notification includes information requesting provision of learning data to be used for the re-learning.

[0267] (Supplementary Note 9) The communication method according to Supplementary Note 8, wherein the notification includes identification information for identifying a type of the learning data.

[0268] (Supplementary Note 10) The communication method according to any one of Supplementary Notes 6 to 9, wherein the notification includes an index of the model and / or identification information for identifying a type or use of the model.

[0269] (Supplementary Note 11) A communication method for applying machine learning technology to wireless communication between a user device and a base station in a mobile communication system, the communication method comprising: a step in which one of the user device and the base station receives configuration information including monitoring parameters for monitoring performance of a trained model from the other of the user device and the base station; and a step in which the one of the communication devices performs the monitoring process using a monitoring dataset based on the configuration information.

[0270] (Supplementary Note 12) The communication method according to Supplementary Note 11, wherein the monitoring parameters include time information indicating a time at which the monitoring dataset is provided from the other communication device, and the step of performing the monitoring processing includes a step of receiving the monitoring dataset from the other communication device and performing the monitoring processing at the time indicated by the time information.

[0271] (Supplementary Note 13) The communication method described in Supplementary Note 11 or 12, further comprising a step in which the one communication device transmits request information requesting the provision of the monitoring dataset to the other communication device, and the step of performing the monitoring process includes a step of receiving the monitoring dataset transmitted from the other communication device based on the request information.

[0272] (Supplementary Note 14) The communication method according to any one of Supplementary Notes 11 to 13, wherein the monitoring parameters further include information indicating a condition under which the monitoring dataset is reduced and used to monitor the performance in the monitoring process.

[0273] (Supplementary Note 15) The communication method according to any one of Supplementary Notes 11 to 14, wherein the monitoring parameters further include a performance evaluation threshold for monitoring the performance in the monitoring process.

[0274] (Supplementary Note 16) The communication method described in any one of Supplementary Notes 11 to 15, wherein the monitoring step includes: a step of deriving correct answer data from the monitoring dataset without using the trained model; a step of acquiring inference result data output by the trained model using partial data obtained by reducing the monitoring dataset as input; and a step of evaluating performance of the trained model by comparing the correct answer data with the inference result data.

[0275] (Supplementary Note 17) The communication method according to any one of Supplementary Notes 11 to 16, wherein the one communication device is the user equipment and the other communication device is the base station, and the monitoring dataset includes a reference signal transmitted from the base station to the user equipment.

[0276] 1: Mobile communication system 100: UE 110: Receiving unit 120: Transmitting unit 130: Control unit 131: CSI generating unit 132: Location information generating unit 140: GNSS receiver 200: gNB 210: Transmitting unit 220: Receiving unit 230: Control unit 231: CSI generating unit 240: Backhaul communication unit 400: Location server 501: Communication device 502: Communication device A1: Data collection unit A2: Model learning unit A3: Model inference unit A4: Data processing unit A5: Federated learning unit

Claims

1. A communication method for applying machine learning techniques to wireless communication between a user equipment and a network node in a mobile communication system, comprising: sending a notification to the other communication device of the user equipment and the network node indicating at least one of that the communication device has an untrained model, that the communication device has a model being trained, and that the communication device has a trained model that has been checked; and receiving, by the one communication device, a response corresponding to the notification from the other communication device. Communication methods.

2. The sending includes sending the notification indicating that the untrained model is present; The receiving includes receiving as the response at least one of a data set and configuration parameters for use in model training. The communication method according to claim 1 .

3. the sending includes sending the notification indicating that the model is being trained; The receiving includes receiving a data set in response for continuing model training. The communication method according to claim 1 .

4. The sending includes sending the notification indicating that the inspection has a completed trained model; The receiving includes receiving, as the response, information for starting use of the trained model for which the inspection has been completed. The communication method according to claim 1 .

5. The notification includes an index of the model and / or identification information for identifying the type or use of the model. A communication method according to any one of claims 1 to 4.

6. A communication method for applying machine learning techniques to wireless communication between a user equipment and a network node in a mobile communication system, comprising: One of the user device and the network node performs inference processing using a trained model obtained by training a model; The one communication device monitors the performance of the trained model to determine the need to retrain the model; and and in response to determining that the re-learning is necessary, transmitting a notification indicating the need for the re-learning to the other communication device of the user equipment and the network node. Communication methods.

7. receiving, from the other communication device, configuration information including parameters related to monitoring the performance; The one communication device performs a monitoring process to monitor the performance of the trained model using the parameters included in the setting information. The communication method according to claim 6.

8. The notification includes information requesting provision of learning data to be used for the re-learning. A communication method according to claim 6 or 7.

9. The notification includes identification information for identifying a type of the learning data. The communication method according to claim 8.

10. The notification includes an index of the model and / or identification information for identifying the type or use of the model. A communication method according to claim 6 or 7.

11. A communication method for applying machine learning techniques to wireless communication between a user equipment and a network node in a mobile communication system, comprising: One of the communication devices, the user equipment and the network node, receives configuration information including monitoring parameters for monitoring performance of a trained model from the other communication device, the user equipment and the network node; The one communication device performs a monitoring process using a monitoring data set based on the setting information. Communication methods.

12. the monitoring parameters include time information indicating a time at which the monitoring data set is provided from the other communication device; Performing the monitoring process includes receiving the monitoring data set from the other communication device and performing the monitoring process at the time indicated by the time information. The communication method according to claim 11.

13. The method further includes the step of: transmitting, from the one communication device, request information requesting provision of the monitoring data set to the other communication device; The performing of the monitoring process includes receiving the monitoring data set transmitted from the other communication device based on the request information. The communication method according to claim 11.

14. The monitoring parameters further include information indicating a condition under which the monitoring data set is reduced and used to monitor the performance in the monitoring process. The communication method according to claim 11.

15. The monitoring parameters further include a performance evaluation threshold for monitoring the performance in the monitoring process. The communication method according to claim 11.

16. The monitoring comprises: Deriving correct answer data from the monitoring dataset without using the trained model; Obtaining inference result data output by the trained model using partial data obtained by reducing the monitoring dataset as input; and evaluating the performance of the trained model by comparing the correct answer data with the inference result data. The communication method according to claim 11.

17. the one communication device being the user equipment and the other communication device being the network node; The monitoring data set includes a reference signal transmitted from the network node to the user equipment. A communication method according to any one of claims 11 to 16.

18. A user device that performs wireless communication with a network node to which machine learning technology is applied, a transmitter configured to transmit a notification to the network node indicating at least one of that the user equipment has an untrained model, that the user equipment has a model being trained, and that the user equipment has a trained model that has been checked; a receiving unit for receiving a response corresponding to the notification from the network node; User equipment.

19. A user device that performs wireless communication with a network node to which machine learning technology is applied, Perform inference processing using the trained model obtained by training the model, A control unit that determines the need to re-learn the model by monitoring the performance of the trained model; a transmission unit configured to transmit a notification indicating the need for the relearning to the network node in response to determining that the relearning is necessary. User equipment.

20. A user device that performs wireless communication with a network node to which machine learning technology is applied, A receiving unit that receives configuration information including monitoring parameters for monitoring performance of a trained model from the network node; a control unit that performs the monitoring process using a monitoring data set based on the setting information. User equipment.