Communication device and communication method

Abstract

A communication device that performs wireless communication with other communication devices, a processor that controls a wireless transceiver for performing wireless communication, wherein the processor controls the wireless transceiver to transmit, to the other communication device, information on an AI / ML model applied to at least one of: a first transmission process of the communication device on a wireless signal transmitted from the communication device to the other communication device, a first reception process of the other communication device with respect to a wireless signal transmitted from the communication device to the other communication device, a second transmission process of the other communication device with respect to a wireless signal transmitted from the other communication device to the communication device, and a third reception process of the other communication device with respect to a wireless signal transmitted from the other communication device to the communication device, or a second reception process of the communication device with respect to a wireless signal transmitted from the other communication device to the communication device.

Description

Technical Field

[0001] This disclosure relates to communication equipment and communication methods. Background Technology

[0002] The development of technologies related to wireless communication is actively underway. In recent years, research has been conducted on techniques to improve communication performance by applying artificial intelligence / machine learning technologies to wireless access networks.

[0003] [List of Citations]

[0004] [Non-patent literature]

[0005] [Non-Patent Literature 1] R1-2210997, "Discussions on AI / ML framework", vivo, 3GPPTSG-RAN WG1 Meeting #111, Toulouse, France, November 14-18, 2022

[0006] [Non-Patent Document 2] TR22.876 V19.0.0, "Study on AI / ML Model Transfer Phase 2", 3GPP Technical Specification Group Services and System Aspects, June 2023 Summary of the Invention

[0007] [Technical Issues]

[0008] However, simply applying artificial intelligence / machine learning techniques to wireless access networks does not necessarily achieve high communication performance (e.g., improvements in frequency utilization efficiency, capacity, speed, low latency, high reliability, power saving, or low processing load).

[0009] Therefore, this disclosure proposes a communication device and a communication method capable of achieving high communication performance.

[0010] Note that the above-described problems or objectives are merely one of many problems or objectives that can be solved or achieved by the various embodiments disclosed in this specification.

[0011] [Problem Solution]

[0012] To address the aforementioned problems, a communication device according to an embodiment of this disclosure is a communication device that performs wireless communication with another communication device, and has a processor that controls a wireless transceiver for performing wireless communication, wherein the processor controls the wireless transceiver to send information to the other communication device about an AI / ML model applied to at least one of the following processes: a first transmission process of the communication device regarding wireless signals transmitted from the communication device to the other communication device, a first reception process of the other communication device regarding wireless signals transmitted from the communication device to the other communication device, a second transmission process of the other communication device regarding wireless signals transmitted from the other communication device to the communication device, or a second reception process of the communication device regarding wireless signals transmitted from the other communication device to the communication device. Attached Figure Description

[0013] Figure 1A This is a diagram illustrating an example of functionality in use cases related to the transmission and reception of CSI information.

[0014] Figure 1B This is a diagram illustrating an example of functionality in use cases related to the transmission and reception of CSI information.

[0015] Figure 1C This is a diagram illustrating an example of functionality in use cases related to the transmission and reception of CSI information.

[0016] Figure 2A This is a diagram illustrating an example of functionality in use cases related to beam ID identification.

[0017] Figure 2B This is a diagram illustrating an example of functionality in use cases related to beam ID identification.

[0018] Figure 2C This is a diagram illustrating an example of functionality in use cases related to beam ID identification.

[0019] Figure 3A This is a diagram illustrating examples of functionality in usage scenarios related to positioning.

[0020] Figure 3B This is a diagram illustrating examples of functionality in usage scenarios related to positioning.

[0021] Figure 3C This is a diagram illustrating examples of functionality in usage scenarios related to positioning.

[0022] Figure 4 This is a diagram illustrating the configuration of the communication system according to this embodiment.

[0023] Figure 5This is a diagram illustrating the configuration of the management device according to this embodiment.

[0024] Figure 6 This is a diagram illustrating the configuration of the base station in this embodiment.

[0025] Figure 7 This is a diagram illustrating the configuration of a relay station according to this embodiment.

[0026] Figure 8 This is a diagram illustrating the configuration of the terminal device in this embodiment.

[0027] Figure 9 This is a diagram illustrating an example of a 5G architecture.

[0028] Figure 10 This is a diagram used to illustrate LCM.

[0029] Figure 11 This is a diagram used to illustrate the fallback process.

[0030] Figure 12A This is a diagram illustrating an example of the basic sequence of communication processing when the transmitting device is an LCM control entity.

[0031] Figure 12B This is a diagram illustrating a specific sequence example of communication processing when the transmitting device is an LCM control entity.

[0032] Figure 13A This is a diagram illustrating an example of the basic sequence of communication processing when the receiving device is an LCM control entity.

[0033] Figure 13B This is a diagram illustrating a specific sequence example of communication processing when the receiving device is an LCM control entity.

[0034] Figure 14 This is a diagram illustrating an example of a sequence of communication processes related to model transformation.

[0035] Figure 15 This is a diagram illustrating an example sequence of communication processes related to model initiation.

[0036] Figure 16 This is a diagram illustrating an example of a signal processing sequence using an AI / ML model.

[0037] Figure 17 This is a diagram illustrating an example sequence of communication processes related to model monitoring.

[0038] Figure 18 This is a diagram illustrating an example of a sequence of communication processes related to model rollback. Detailed Implementation

[0039] In the following, embodiments of the present disclosure will be described in detail based on the accompanying drawings. Note that in the following embodiments, the same reference numerals are assigned to the same parts to omit redundant descriptions.

[0040] Furthermore, in this specification and accompanying drawings, multiple components having substantially the same functional configuration can be distinguished by adding different letters or numbers after the same reference numerals. For example, multiple configurations having substantially the same functional configuration may be distinguished as terminal devices 401, 402, and 403 as needed. However, when there is no particular need to distinguish each of the multiple components having substantially the same functional configuration, only the same reference numerals are assigned. For example, when there is no specific need to distinguish terminal devices 401, 402, and 403, they are simply referred to as terminal device 40.

[0041] The one or more embodiments (including examples and modifications) described below can be implemented independently. On the other hand, the various embodiments described below can be implemented by appropriately combining at least a portion of them with at least a portion of other embodiments. These various embodiments may include novel features that differ from each other. Therefore, these various embodiments can help solve different purposes or problems and can achieve different effects.

[0042] << 1. Overview>>

[0043] Before describing this embodiment in detail, an overview of this embodiment will be explained.

[0044] <1-1. Question>

[0045] Within 3GPP (registered trademark), research is underway on NR (New Radio), Beyond 5G, and 6G. NR is a radio access technology that can support a wide range of use cases, including enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). NR was standardized in Rel-15 as a technical framework corresponding to the use cases, requirements, and deployment scenarios in those use cases, and technology expansion is ongoing. Beyond 5G and 6G are the next-generation mobile communication technologies following 5G. Further improvements in enhanced mobile broadband, massive simultaneous connections, and ultra-reliable low latency are expected in Beyond 5G and 6G.

[0046] To achieve this, technologies for using artificial intelligence (AI) / machine learning (ML) to improve communication performance are being researched. For example, within 3GPP, research is underway to improve frequency efficiency by applying AI / ML models to the radio access network. For instance, Non-Patent Document 1 (3GPP R1-2210997) describes AI / ML model management when using machine learning. Furthermore, Non-Patent Document 2 (3GPP TR22.876) describes use cases related to applying machine learning to sidelink communication. Sidelink communication is characterized by communication between multiple different terminal devices, which differs from traditional communication (uplink and / or downlink communication) between base stations and terminal devices.

[0047] However, simply using artificial intelligence / machine learning does not necessarily achieve high communication performance (e.g., improved frequency utilization efficiency, large capacity, high speed, low latency, high reliability, low power consumption, or low processing load).

[0048] For example, it is desirable to use AI / ML models in various use cases, but the means of applying AI / ML models can vary depending on the use case. Therefore, it is necessary to change the method of managing AI / ML models for each use case as needed. On the other hand, as the number of use cases applying AI / ML models increases, problems arise such as increased processing load and increased memory requirements for managing AI / ML models. Therefore, efficient management of AI / ML models becomes essential in communications utilizing AI / ML models. That is, in order to achieve high communication performance in communications using AI / ML, communication devices need to appropriately implement AI / ML model management.

[0049] Note that in this embodiment, the use of AI / ML models is not limited to uplink communication, downlink communication, and sidelink communication. For example, the use of AI / ML models can be in communication using non-cellular communication such as wireless LAN communication. Furthermore, the use cases for using AI / ML models can be more specific use cases using these communications (e.g., at least one of uplink communication, downlink communication, and sidelink communication).

[0050] <1-2. Solution>

[0051] In this embodiment, the above-mentioned problems are solved as follows.

[0052] Before explaining the overview of the solution, we will explain the functionality and AI / ML model management that are important elements in this embodiment.

[0053] <1-2-1. Functionality>

[0054] First, we will explain functionality.

[0055] Functionality is a concept that represents the communication requirements needed for AI communication, the use cases that use AI / ML models, or the functions implemented using AI / ML models in the characteristics of use cases such as applying AI / ML models. Figures 1A to 3C These are diagrams illustrating examples of functionality.

[0056] Figures 1A to 1C Functional features 1-A, 1-B, and 1-C are shown under use cases (features) related to the transmission and reception of CSI information. Figures 1A to 1C In the example, auxiliary information (e.g., information about the AI / ML model) is transmitted from the sending device ( Figures 1A to 1C In the example, the terminal device (terminal) sends to the receiving device ( Figures 1A to 1C (Network nodes in the example). In Figures 1A to 1C In the example, auxiliary information is information necessary for the receiving device to perform wireless communication-related processing, or information that assists the receiving device in performing wireless communication-related processing. For example, auxiliary information could be information necessary for the operation of the terminal device, such as downlink communication, uplink communication, sidelink communication, unauthorized communication, handover, etc. (e.g., control information). When sending and receiving CSI information, one or more communication requirements are defined. For example, suppose three communication requirements are defined, namely Functionality 1-A, Functionality 1-B, and Functionality 1-C. In Functionality 1-A, 1-B, and 1-C, both the sending and receiving devices use an AI / ML model based on the communication requirements defined in each functionality to achieve efficient information transmission from the sending device to the receiving device.

[0057] Figures 2A to 2C Functional examples 2-A, 2-B, and 2-C are shown for use cases (features) related to beam ID identification. Figures 2A to 2C In the example, the receiving device ( Figures 2A to 2C The terminal in the example) is based on the sending device ( Figures 2A to 2C The information / signals sent by the network nodes in the example are used to identify the beam ID (e.g., reference signal, synchronization signal ID). When identifying the beam ID, one or more communication requirements are defined. For example, suppose three communication requirements are defined, namely Functionality 2-A, Functionality 2-B, and Functionality 2-C. In Functionality 2-A, 2-B, and 2-C, the receiving device uses the received information / signals and an AI / ML model to identify the beam ID based on the communication requirements defined in each function.

[0058] Figures 3A to 3C Functionalities 3-A, 3-B, and 3-C are shown in usage scenarios (features) related to positioning. Figures 3A to 3C In the example, the receiving device ( Figures 3A to 3CThe terminal in the example) is based on the sending device ( Figures 3A to 3C The network node in the example sends information / signals to generate location information. When generating location information, one or more communication requirements are defined. For example, suppose three communication requirements are defined, namely Functionality 3-A, Functionality 3-B, and Functionality 3-C. In Functionality 3-A, 3-B, and 3-C, the receiving device uses the received information / signals and an AI / ML model to generate location information based on the communication requirements defined in each function.

[0059] Note that functional descriptions appearing in the following descriptions can be replaced with other descriptions. For example, functionalities appearing in the following descriptions can be replaced with descriptions such as use cases, features, functions, function indexes, function IDs, AI / ML indexes, AI indexes, ML indexes, or model indexes. Of course, functional descriptions are not limited to these and can be replaced with other terms that represent, for example, use cases using AI / ML models or functions implemented using AI / ML models.

[0060] <1-2-2. AI / ML Model Management>

[0061] Next, we will explain AI / ML model management.

[0062] In 3GPP, AI / ML model management is being studied (e.g., RP-221347). In this embodiment, AI / ML model management may be referred to as LCM (Lifecycle Management).

[0063] AI / ML model management (e.g., LCM) may consist of one or more elements (processes / methods / procedures / functions) from (M1) to (M25). For example, AI / ML model management (e.g., LCM) may be one of (M1) to (M25), or it may be a combination of multiple elements including at least one of (M1) to (M25).

[0064] (M1) Data Collection

[0065] (M2) model training

[0066] (M3) Function Identification

[0067] (M4) model recognition

[0068] (M5) model delivery

[0069] (M6) Model Transfer

[0070] (M7) Model Download

[0071] (M8) model upload

[0072] (M9) Model Reasoning

[0073] (M10) Model Validation

[0074] (M11) Model Testing

[0075] (M12) model activation

[0076] (M13) Functional activation

[0077] (M14) model deactivation

[0078] (M15) Functional deactivation

[0079] (M16) Mode Switching

[0080] (M17) Function Switching

[0081] (M18) model rollback

[0082] (M19) Model Monitoring

[0083] (M20) Model Update

[0084] (M21) Model Registration

[0085] (M22) model deployment

[0086] (M23) Model Configuration

[0087] (M24) Model Selection

[0088] (M25) UE capability

[0089] Note that (M1) through (M25) are merely examples. The AI / ML model management (e.g., LCM (or elements constituting LCM)) of this embodiment may include management (or elements) other than those described above (M1) through (M25). (M1) - (M25) will be described in detail later.

[0090] <1-2-3: Overview of the Solution>

[0091] Based on the above, an overview of the solution in this embodiment will be provided.

[0092] The communication device (hereinafter referred to as the first communication device) in this embodiment is configured to communicate wirelessly with another communication device (hereinafter referred to as the second communication device).

[0093] Here, the first communication device and the second communication device can be a base station or a terminal device. For example, the first communication device can be a base station, and the second communication device can be a terminal device. Alternatively, the first communication device can be a terminal device, and the second communication device can be a base station. Alternatively, the first communication device can be a terminal device, and the second communication device can be another terminal device. Alternatively, the first communication device can be a base station, and the second communication device can be another base station. Note that the first communication device and the second communication device are not limited to base stations and / or terminal devices. At least one of the first communication device and the second communication device can be a relay station, a device constituting the core network, server equipment, or other control equipment.

[0094] At least one of the first and second communication devices is configured to perform one or more of the functional processes described above or below. For example, at least one of the first and second communication devices may be configured to perform at least one of the following: processing related to the transmission or reception of auxiliary information, processing related to beam ID identification, and processing related to positioning. The functional processing is implemented using AI / ML models. One or more AI / ML models are associated with a functionality.

[0095] The first communication device transmits information relating to an AI / ML model applied to at least one of the following processes (P1) to (P4) (first transmission process, first reception process, second transmission process, and second reception process) to the second communication device.

[0096] (P1) Regarding the transmission process of the wireless signal from the first communication device to the second communication device (first transmission process)

[0097] (P2) Regarding the reception processing of wireless signals transmitted from the first communication device to the second communication device (first reception processing)

[0098] (P3) Regarding the transmission processing of wireless signals from the second communication device to the first communication device (second transmission processing)

[0099] (P4) Regarding the reception processing of wireless signals transmitted from the second communication device to the first communication device (second reception processing)

[0100] Here, the first and second transmission processes described above can indicate the processing performed on the wireless signal to be transmitted. In other words, the first and second transmission processes can be performed on the wireless signal before it is transmitted. Alternatively, the first and second reception processes described above can indicate the processing performed on the wireless signal to be received. In other words, the first and second reception processes can be performed on the wireless signal after it is received.

[0101] The wireless signals in this disclosure can be Layer 1 physical signals or physical channels, or combinations thereof. For example, a Layer 1 physical signal can be at least one of the following: SS (synchronization signal), SSB (SS / PBCH block), CSI-RS (channel state information-reference signal), SRS (sound reference signal), PT-RS (phase tracking reference signal), or DMRS (demodulation reference signal). SS can be PSS (primary synchronization signal), SSS (secondary synchronization signal), SPSS (sidelink primary synchronization signal), or SSSS (sidelink secondary synchronization signal), or a combination of at least two of these signals. DMRS can be PDCCH DMRS, PDSCHDMRS, PBCH DMRS, PUCCH DMRS, PUSCH DMRS, PSCCH DMRS, or PSSCH DMRS, or a combination of at least two of these. For example, a Layer 1 physical channel can be at least one of the following: PBCH (Physical Broadcast Channel), PDCCH (Physical Downlink Control Channel), PDSCH (Physical Downlink Shared Channel), PRACH (Physical Random Access Channel), PUCCH (Physical Uplink Control Channel), PUSCH (Physical Uplink Shared Channel), PSBCH (Physical Sidelink Broadcast Channel), PSCCH (Physical Sidelink Control Channel), PSSCH (Physical Sidelink Shared Channel), or PSFCH (Physical Sidelink Feedback Channel), or a combination of at least two of these.

[0102] Alternatively or concurrently, the radio signals in this disclosure may be Layer 1, Layer 2, or Layer 3 messages. For example, the radio signals in this disclosure may include at least one of the following: DCI (Downlink Control Information), UCI (Uplink Control Information), SCI (Sidelink Control Information), MAC CE (Media Access Control - Control Element), or RRC (Radio Resource Control) messages. The RRC message may be at least one of the following: MIB (Master Information Block), SIB (System Information Block) (e.g., SIB1, SIB2, SIB3, SIB4, ..., SIBX (X is an integer)), RRCRequest message, RRCSetup message, RRCSetupComplete message, RRCReconfiguration message, RRCRelease message, SecurityModeCommand message, SecurityModeComplete message, SecurityModeFailure message, CounterCheck message, CounterCheckResponse message, Paging message, RRCReestablishmentRequest message, or RRCReestablishment message.

[0103] Alternatively or additionally, at least one of the first transmission process, the first reception process, the second transmission process, or the second reception process described above may include at least one corresponding process and / or procedure described in at least one of the processes and / or procedures described below (K1) to (K6).

[0104] (K1) Sequence generation, scrambling, layer mapping, antenna port mapping, mapping to radio resources, precoding, modulation, cyclic shifting, or extension of a sequence of bits carried by a radio signal or a radio signal, or a combination of at least two of these.

[0105] (K2) CRC calculation, codebook segmentation, polarity coding, LDPC (low-density parity-check) coding, rate matching, or code block concatenation for a wireless signal or a bit sequence carried by a wireless signal, or a combination of at least two of these.

[0106] (K3) Cell search, transmission timing adjustment, radio link monitoring, radio link recovery, power control of radio signals, random access procedure, HARQ-ACK codebook determination or UCI report (HARQ-ACK, SR (scheduling request)) or a combination of at least two of these.

[0107] (K4) Resource allocation on the time or frequency axis, modulation order, TBS (transmit block size) determination, processing related to quasi-co-location of antenna ports between radio signals, CSI reporting, CSI feedback, CSI calculation time, radio signal processing time, radio signal preparation time, or uplink switching, or a combination of at least two of these.

[0108] (K5) Random access procedure in MAC, HARQ operation of transport block carried by radio signal, buffer status report, discontinuous reception (DRX), SCell (secondary cell) activation / deactivation, PDCP copy activation / deactivation, listen-before-tell (LBT), location or small data transfer (SDT), or a combination of at least two of these.

[0109] (K6) System information reception (acquisition) process, paging, RRC connection establishment process, initial AS security activation process, RRC reconfiguration process, counter check process, RRC connection re-establishment process, radio link failure related operation, RRC connection recovery, idle mode mobility (cell selection, cell reselection), connection mode mobility (handover, beam-level mobility) or measurement (L1 measurement, L2 measurement, L3 measurement) or at least a combination of these.

[0110] As a representative example, at least one of the first transmission process, the first reception process, the second transmission process, or the second reception process described above may be at least one corresponding process and / or process in at least one of the processes and / or procedures described below (H1) to (H3).

[0111] (H1)CSI Feedback

[0112] At least one of the aforementioned first transmission process, first reception process, second transmission process, or second reception process can be a process related to CSI feedback. Here, CSI feedback can be at least one of type 1 single-panel codebook, type 1 multi-panel codebook, type 2 codebook, type 2 port selection codebook, enhanced type 2 port selection codebook, and further enhanced type 2 port selection codebook.

[0113] (H2) Beam Management

[0114] At least one of the aforementioned first transmission process, first reception process, second transmission process, or second reception process can be a process related to beam management. Here, beam management can be the operation of receiving an SSB (Synchronization Signal Block) and transmitting a PRACH preamble on the PRACH resource corresponding to the SSB. Alternatively, beam management can be the operation of receiving a CSI-RS and reporting a CSI report.

[0115] (H3) Positioning

[0116] At least one of the aforementioned first transmission process, first reception process, second transmission process, or second reception process can be a positioning-related process. Here, positioning can be an operation of transmitting a PRS (Positioning Reference Signal) and performing position estimation on the downlink. Alternatively, positioning can be an operation of transmitting an SRS (Sound Reference Signal) and performing position estimation on the uplink.

[0117] Here, the information about the AI / ML model can be information about the LCM. For example, suppose the first communication device is the control entity of the LCM, and the second communication device is the controlled object of the LCM. In this case, the first communication device can send at least one of the following information as information about the AI / ML model.

[0118] (Model Transfer)

[0119] The first communication device can transmit data from an AI / ML model used in at least one of the first receiving process and the second transmitting process as information about the AI / ML model to the second communication device. The second communication device uses the AI / ML model data obtained from the first communication device to perform either the first receiving process or the second transmitting process.

[0120] (Model activation)

[0121] The first communication device can send information indicating the activation of an AI / ML model for at least one of the first receiving process and the second sending process as information about the AI / ML model. When the second communication device receives the activation indication information from the first communication device, it performs the activation process of the AI / ML model associated with the indication.

[0122] (Model deactivation)

[0123] The first communication device can send information indicating the deactivation of an AI / ML model for at least one of the first receiving process and the second sending process as information about the AI / ML model. When the second communication device receives activation indication information from the first communication device, it performs deactivation processing of the AI / ML model associated with the indication.

[0124] (Model rollback)

[0125] The first communication device can send information about the AI / ML model to cause at least one of the first receiving process and the second sending process to revert from a process using the AI / ML model to a process not using the AI / ML model. When the second communication device receives the revert information from the first communication device, it reverts the process associated with the instruction from a process using the AI / ML model to a process not using the AI / ML model.

[0126] Note that the above describes the processing of the first communication device when the LCM is model transfer, model activation, model deactivation, or model rollback. However, the LCM in which the first communication device acts as a control entity is not limited to this. In other LCM scenarios, the first communication device can also be used as a control entity.

[0127] Furthermore, the second communication device can be the control entity of the LCM, while the first communication device can be the control object of the LCM. In this case, the first communication device can send at least one of the following information as information about the AI / ML model.

[0128] (Terminal capabilities)

[0129] The first communication device may send its own capability information regarding at least one of the first transmission process and the second reception process to the second communication device. For example, suppose an AI / ML model is associated with a functionality involving at least one of the first transmission process and the second reception process. In this case, the first communication device may send information as capability information to the second communication device indicating which of a plurality of functionalities it corresponds to in relation to at least one of the first transmission process and the second reception process. Alternatively, the first communication device may send information as capability information to the second communication device indicating which of a plurality of AI / ML models it corresponds to in relation to at least one of the first transmission process and the second reception process. The second communication device may perform LCM based on the capability information of the first communication device.

[0130] (Model Training)

[0131] The first communication device can send training information about an AI / ML model for at least one of the first transmission process and the second reception process to the second communication device. For example, suppose the AI / ML model is associated with a functionality involving at least one of the first transmission process and the second reception process. In this case, the first communication device can send information to the second communication device indicating which of the plurality of functionalities has completed AI / ML model training as training information. The second communication device can then perform LCM based on the training information.

[0132] Note that the above illustrates the processing of the first communication device when the LCM is a terminal capability or model training. However, the use of the first communication device as a control object in an LCM is not limited to this. In other LCM scenarios, the first communication device can also act as a control object.

[0133] Furthermore, although the above description is divided into two cases: the case where the first communication device is the control entity of the LCM and the second communication device is the control object of the LCM, and the case where the second communication device is the control entity of the LCM and the first communication device is the control object of the LCM, examples of information about the AI / ML model explained in each case can be implemented regardless of the two cases. That is, at least two examples of information about the AI / ML model explained in each case can be implemented simultaneously. In this case, even when spanning the above two cases, at least two examples of information about the AI / ML model can be implemented simultaneously.

[0134] Therefore, in this embodiment, the first communication device sends information about the AI / ML model to the second communication device when implementing LCM. Alternatively, the first communication device sends information about the AI / ML model to the second communication device implementing LCM. The second communication device uses or manages the AI / ML model based on the information received from the first communication device. This enables efficient use or management of the AI / ML model, thus allowing the first and / or second communication devices to achieve high communication performance. Note that the purpose of sending or receiving information about the AI / ML model is not limited to implementing LCM. In other words, sending or receiving information about the AI / ML model may not involve implementing LCM.

[0135] The above describes an overview of this embodiment, and the communication system 1 of this embodiment will be described in detail below.

[0136] <<Configuration of Communication Systems>>

[0137] First, the configuration of communication system 1 will be described.

[0138] Figure 4 This diagram illustrates the configuration of a communication system 1 according to this embodiment. The communication system 1 includes a management device 10, a base station 20, a relay station 30, and a terminal device 40. The communication system 1 provides a wireless network that enables users to perform mobile communications through the coordinated operation of each wireless communication device constituting the communication system 1. The wireless network in this embodiment consists of a radio access network (RAN) and a core network (CN). In this embodiment, the wireless communication device is a device with wireless communication capabilities, and... Figure 4 In the example, these devices correspond to base station 20, relay station 30, and terminal device 40.

[0139] The communication system 1 may include multiple management devices 10, base stations 20, relay stations 30, and terminal devices 40. Figure 4In the example, communication system 1 includes management devices 101 and 102 as management devices 10, and base stations 201, 202, and 203 as base stations 20. Furthermore, communication system 1 includes relay stations 301 and 302 as relay stations 30, and terminal devices 401, 402, and 403 as terminal devices 40. Note that, depending on the communication system, relay station 30 may not be present.

[0140] Terminal device 40 can be configured to connect to a network using radio access technologies (RATs) such as LTE (Long Term Evolution), NR (New Radio), 6G, Wi-Fi, Bluetooth (registered trademark), etc. In this case, terminal device 40 can be configured to use different radio access technologies (wireless communication methods). For example, terminal device 40 can be configured to use NR and Wi-Fi. Furthermore, terminal device 40 can be configured to use different cellular communication technologies (e.g., LTE, NR, or 6G). In the following description, terminal device 40 may be referred to as UE (User Equipment) 40.

[0141] LTE and NR are types of cellular communication technologies that enable mobile communication for terminal devices by deploying multiple areas covered by base stations in a cellular pattern. Furthermore, it is assumed that 6G is a cellular communication technology, and that mobile communication for terminal devices can also be achieved by deploying multiple areas covered by base stations in a cellular pattern.

[0142] Note that in the following description, "LTE" should include LTE-A (LTE-Advanced), LTE-A Pro (LTE-Advanced Pro), and EUTRA (Evolved Universal Terrestrial Radio Access). Furthermore, NR should include NRAT (New Radio Access Technology) and FEUTRA (Further EUTRA). Additionally, NR may include 5G Advanced. Note that a single base station 20 can manage multiple cells. In the following description, cells corresponding to LTE may be referred to as LTE cells, and cells corresponding to NR may be referred to as NR cells.

[0143] NR is the next-generation (fifth-generation) radio access technology following LTE (including LTE Advanced and LTE Advanced Pro, the fourth generation of communication). NR is a radio access technology that can support a variety of use cases, including eMBB (enhanced Mobile Broadband), mMTC (massive Machine-Type Communications), and URLLC (Ultra-Reliable and Low-Latency Communications). NR is standardized in 3GPP (registered trademark) Rel-15 as a technical framework corresponding to the use cases, requirements, and deployment scenarios in these use cases. Furthermore, to simultaneously achieve high speed, high capacity, low latency, high reliability, multi-axis connectivity, and massive simultaneous connections, 5G and 6G technologies are required.

[0144] 6G is the next-generation mobile communication technology following NR and 5GS (5G systems), the fifth-generation mobile communication technology. 6G can also be a cellular communication technology similar to 5G (NR). 6G includes radio access technologies and network technologies between base stations, core networks, and data networks. Furthermore, 6G includes advanced (extreme connectivity) technologies for each of eMBB, mMTC, and URLLC, which are major use cases or requirements in NR. In addition, 6G includes new technologies in new aspects. For example, 6G may include technologies related to AI (cognitive networks, AI local air interfaces), sensing (including radar sensing, networks as sensors), and terahertz communication.

[0145] Note that the wireless networks described above or below may support at least one of radio access technologies (RATs) such as LTE (Long Term Evolution), NR (New Radio), 6G, etc. Note that the radio access method used by communication system 1 is not limited to LTE, NR, and 6G, and may be other radio access methods such as W-CDMA (Wideband Code Division Multiple Access) and CDMA2000 (Code Division Multiple Access 2000).

[0146] Furthermore, base station 20 and relay station 30 can be ground stations or non-ground stations. That is to say, Figure 4 The communication system can be a non-terrestrial network. The non-terrestrial station can be a satellite station or an aircraft station. If the non-terrestrial station is a satellite station, then the wireless network can be a bend-tube (transparent) type mobile satellite communication system.

[0147] Note that in this embodiment, ground station and ground base station refer to base stations and relay stations installed on the ground. Here, "ground" is used in a broad sense, including not only land but also underground, above water, and underwater. Note that in the following description, the term "ground station" can be replaced by "gateway".

[0148] Note that LTE base stations can be referred to as eNodeB (Evolved Node B) or eNB. Similarly, NR base stations can be referred to as gNodeB or gNB. Furthermore, 6G base stations can be referred to as 6GNodeB (6GNB). LTE RAN can be referred to as EUTRAN. NRRAN can be referred to as NGRAN. 6G RAN can be referred to as 6GRAN. Additionally, in LTE, NR, and 6G, the terminal equipment (also called a mobile station or terminal) can be referred to as UE (User Equipment). Note that a terminal equipment is a communication device, also referred to as a mobile station or terminal.

[0149] Note that terminal device 40 may be able to connect to the network using radio access technologies (wireless communication methods) other than LTE, NR, 6G, Wi-Fi, and Bluetooth. For example, terminal device 40 may be able to connect to the network using LPWA (Low Power Wide Area) communication. Furthermore, terminal device 40 may be able to connect to the network using proprietary wireless communication.

[0150] Here, LPWA communication refers to wireless communication capable of low-power wide-area communication. For example, LPWA wireless refers to IoT (Internet of Things) wireless communication using specific low-power wireless (e.g., the 920MHz band) or ISM (Industrial, Scientific and Medical) bands. Note that the LPWA communication used by terminal device 40 may conform to LPWA standards. LPWA standards may be at least one of ELTRES, ZETA, SIGBOX, LoRaWAN, LTE-M, and NB-IoT. Of course, LPWA standards are not limited to these and may be other LPWA standards.

[0151] Figure 4 Each wireless communication device shown can be considered a device in a logical sense. That is, a part of each wireless communication device can be implemented by a virtual machine (VM) or a container such as Docker, and they can be physically implemented on the same hardware.

[0152] In this embodiment, the concept of a wireless communication device includes not only portable mobile devices (terminal devices) such as mobile terminals, but also devices mounted on a structure or mobile body. The structure or mobile body itself can be considered a wireless communication device. Furthermore, the concept of a wireless communication device includes not only terminal device 40, but also base station 20 and relay station 30. A wireless communication device is a processing device or information processing device. A wireless communication device can also be referred to as a transmitting device or a receiving device.

[0153] Note that in this embodiment, "resource" can refer to at least one of frequency, time, resource element (including REG, CCE, CORESET), resource block, bandwidth portion, component carrier, symbol, sub-symbol, time slot, mini-time slot, sub-time slot, subframe, frame, PRACH timing, timing, code, multiple access physical resource, multiple access signature, and subcarrier spacing (digital). That is, the term "resource" above or below can be replaced by at least any of the above examples.

[0154] The configuration of each wireless communication device constituting communication system 1 will be described in detail below. Note that the configuration of each wireless communication device shown below is merely an example. The configuration of each wireless communication device may differ from the configuration shown below.

[0155] <2-1. Configuration of Management Equipment>

[0156] Management device 10 is an information processing device (computer) that manages the wireless network. For example, management device 10 is an information processing device that manages the communications of base station 20.

[0157] Management device 10 can be an apparatus constituting the core network (CN). For example, management device 10 can be a device with MME (Mobility Management Entity) functionality. Furthermore, management device 10 can be a device with AMF (Access and Mobility Management Function) and / or SMF (Session Management Function) functionality. MME, AMF, and SMF are control plane network function nodes in the core network (CN). Management device 10 can also be a device with 6G control plane network function (6G CPNF) functionality. A 6G CPNF can consist of one or more logical nodes.

[0158] Of course, the functions of management device 10 are not limited to MME, AMF, SMF, and 6G CPNF. Management device 10 can be a device with NSSF (Network Slice Selection Function), AUSF (Authentication Server Function), PCF (Policy Control Function), and UDM (Unified Data Management) functions. In addition, management device 10 can be a device with HSS (Home Subscriber Server) functions.

[0159] Note that management device 10 may have gateway functionality. For example, management device 10 may function as both an S-GW (Serving Gateway) and a P-GW (Packet Data Network Gateway). Furthermore, management device 10 may have UPF (User Plane Function) functionality. In this case, management device 10 may have multiple UPFs. Additionally, management device 10 may be a device with 6G User Plane Network Function (6G UPNF) functionality.

[0160] The core network (CN) consists of multiple network functions, each of which can be integrated into a single physical device or distributed across multiple physical devices. In other words, the management device 10 can be distributed across multiple devices. Furthermore, this distributed deployment can be dynamically controlled. The base station 20 and the management device 10 form a network and provide wireless communication services to the terminal device 40. The management device 10 is connected to the Internet, and the terminal device 40 can use various services provided via the Internet through the base station 20.

[0161] Note that management device 10 need not be a device constituting the core network CN. For example, suppose the core network CN is a W-CDMA (Wideband Code Division Multiple Access) or CDMA2000 (Code Division Multiple Access 2000) core network. In this case, management device 10 can be a device used as an RNC (Radio Network Controller).

[0162] Figure 5 This is a diagram illustrating the configuration of the management device 10 according to this embodiment. The management device 10 includes a communication unit 11, a storage unit 12, and a control unit 13. Figure 5 The configuration shown is a functional configuration, and the hardware configuration may differ. Furthermore, the functionality of management device 10 can be implemented by statically or dynamically distributing it across multiple physically separate configurations. Management device 10 may consist of multiple server devices.

[0163] Communication unit 11 is a communication interface for communicating with wireless communication devices (e.g., base station 20). Communication unit 11 can be a network interface or a device connection interface. Communication unit 11 can be a LAN (Local Area Network) interface, such as a NIC (Network Interface Card), or a USB interface configured by a USB (Universal Serial Bus) host controller or USB port. Communication unit 11 can be a wired interface or a wireless interface. Communication unit 11 is controlled by control unit 13.

[0164] Storage unit 12 is a read / write storage device, such as DRAM, SRAM, flash memory, or hard disk. Storage unit 12 stores, for example, the connection status of terminal device 40. Storage unit 12 stores the RRC (Radio Resource Control) status and ECM (Electronic Power Management) status of terminal device 40, or the CM (Connectivity Management) status of 5G system. Storage unit 12 can be used as a local storage device for storing the location information of terminal device 40.

[0165] Control unit 13 is a controller that controls the various parts of management device 10. Control unit 13 can be implemented by a processor such as a CPU (Central Processing Unit) or MPU (Microprocessor Unit). For example, control unit 13 can be implemented by a processor that executes various programs stored in the internal storage of management device 10 using RAM (Random Access Memory) as its working area. Control unit 13 can be implemented by integrated circuits, such as ASIC (Application-Specific Integrated Circuit) or FPGA (Field-Programmable Gate Array). Furthermore, control unit 13 can be implemented by GPU (Graphics Processing Unit). CPU, MPU, ASIC, FPGA, and GPU can all be considered controllers. Note that control unit 13 can consist of multiple physically separate objects. For example, control unit 13 can be composed of multiple semiconductor chips.

[0166] <2-2. Base Station Configuration>

[0167] Base station 20 is a wireless communication device that performs wireless communication with other wireless communication devices (e.g., relay station 30, terminal device 40, or other base stations 20). Base station 20 can perform wireless communication with terminal device 40 via relay station 30, or it can perform wireless communication with terminal device 40 directly.

[0168] Base station 20 is a device corresponding to a radio base station (base station, node B, eNB, gNB, or 6GNB, etc.) or a radio access point (access point). Base station 20 can be a radio relay station. Base station 20 can be an optical extension device called RRH (Remote Radio Head). Base station 20 can be a receiving station such as an FPU (Field Pickup Unit). Base station 20 can be an IAB (Integrated Access and Backhaul) donor node or IAB relay node, which provides radio access lines and radio backhaul lines through time division multiplexing, frequency division multiplexing, or space division multiplexing.

[0169] The radio access technology used by base station 20 can be cellular communication technology. The radio access technology used by base station 20 can be wireless LAN technology. The radio access technology used by base station 20 can be LPWA (Low Power Wide Area) communication technology. However, the radio access technology used by base station 20 is not limited to these and can be other radio access technologies. The wireless communication used by base station 20 can be wireless communication using millimeter waves or wireless communication using terahertz waves. The wireless communication used by base station 20 can be wireless communication using radio waves or wireless communication using infrared or visible light (optical wireless). Furthermore, base station 20 is capable of NOMA (Non-Orthogonal Multiple Access) communication with terminal device 40. Here, NOMA communication refers to communication using non-orthogonal resources (transmission, reception, or both). Note that base station 20 can be capable of NOMA communication with other base stations 20.

[0170] Note that base stations 20 can communicate with each other via base station-core network interfaces (e.g., NG interface, S1 interface, etc.). These interfaces can be wired or wireless. Furthermore, base stations can communicate with other base stations via inter-base station interfaces (e.g., Xn interface, X2 interface, F1 interface, etc.). These interfaces can also be wired or wireless.

[0171] The concept of a base station (also known as "base station equipment") includes not only donor base stations but also relay base stations (also known as "relay stations"). A relay base station can be any of an RF repeater, a smart repeater, and a smart surface. The concept of a base station includes not only the structure equipped with base station functions but also the equipment installed on the structure.

[0172] Structures include, for example, high-rise buildings, houses, iron towers, railway station facilities, airport facilities, port facilities, office buildings, school buildings, hospitals, factories, commercial facilities, and stadiums. The concept of structure includes not only buildings but also non-building structures such as tunnels, bridges, dams, walls, and steel beams, as well as equipment such as cranes, gates, and windmills. The concept of structure includes not only land-based (ground-level) or underground structures but also water-based structures such as bridge piers or large floating structures, and underwater structures such as marine observation equipment. A base station can also be referred to as information processing equipment.

[0173] Base station 20 can be a donor station or a relay station. Furthermore, base station 20 can be a fixed station or a mobile station. A mobile station is a wireless communication device configured to be mobile (e.g., a base station). In this case, base station 20 can be a device installed on a mobile body, or it can be the mobile body itself. For example, a mobile relay station can be considered as base station 20 acting as a mobile station. Additionally, vehicles, drones, smartphones, and other devices that are inherently mobile and possess base station functionality (at least some base station functionality) are also equivalent to base station 20 acting as a mobile station.

[0174] Here, a mobile body can be a mobile terminal such as a smartphone or mobile phone. A mobile body can be one that moves on land (in a narrow sense, the ground) (e.g., a vehicle such as a car, bicycle, bus, truck, motorcycle, train, or linear motor vehicle), or one that moves underground (e.g., in a tunnel) (e.g., a subway). Furthermore, a mobile body can be one that moves on water (e.g., a ship such as a passenger ship, cargo ship, or hovercraft), or one that moves underwater (e.g., a submarine such as a submarine, unmanned underwater vehicle). Additionally, a mobile body can be one that moves in the atmosphere (e.g., an aircraft such as an airplane, airship, or drone).

[0175] Base station 20 can be a ground-based base station (Ground station) installed on the ground. Base station 20 can be a base station deployed on a ground structure, or it can be a base station installed on a mobile body moving on the ground. Base station 20 can be an antenna installed on a structure such as a building and signal processing equipment connected to that antenna. Base station 20 can be the structure or the mobile body itself. "Ground" is the ground in a broad sense, including not only land (ground in a narrow sense), but also underground, above water, and underwater. Base station 20 is not limited to ground-based base stations. When communication system 1 is a satellite communication system, base station 20 can be an aircraft station. From the perspective of a satellite station, an aircraft station located on Earth is a ground station.

[0176] Base station 20 is not limited to ground stations. Base station 20 can be a non-terrestrial base station (non-land station) that can float in the air or space. Base station 20 can be an aircraft station or a satellite station.

[0177] A satellite station is a satellite station capable of floating outside the atmosphere. A satellite station can be a device mounted on a spacecraft, such as an artificial satellite, or it can be the spacecraft itself. A spacecraft is a moving body that moves within the atmosphere. A spacecraft can be at least one of an artificial satellite, spacecraft, space station, and probe. Of course, a spacecraft can also be an artificial celestial body other than these. Note that a satellite used as a satellite station can be any of a low Earth orbit (LEO) satellite, a medium Earth orbit (MEO) satellite, a geostationary Earth orbit (GEO) satellite, or a highly elliptical orbit (HEO) satellite. A satellite station can be a device mounted on a low Earth orbit (LEO), medium Earth orbit (MEO), geostationary, or highly elliptical orbit satellite.

[0178] An aircraft station is a wireless communication device capable of floating in the atmosphere, such as an aircraft. An aircraft station can be a device mounted on an aircraft or the aircraft itself. The concept of an aircraft includes not only aircraft heavier than air, such as airplanes or gliders, but also aircraft lighter than air, such as balloons or airships. The concept of an aircraft includes not only aircraft heavier or lighter than air, but also rotorcraft, such as helicopters or autogyros. An aircraft station, or an aircraft equipped with an aircraft station, can be an unmanned aerial vehicle, such as a drone.

[0179] The concept of unmanned aerial vehicles (UAVs) includes unmanned aerial vehicle systems (UAS) and tethered UASs. UAV concepts also include lighter-than-air UASs (LTA) and heavier-than-air UASs (HTA). Furthermore, the concept of high-altitude UAS platforms (HAPs) is also included.

[0180] The coverage area of ​​base station 20 can be relatively large, such as a macrocell, or relatively small, such as a picocell. The coverage area of ​​base station 20 can also be very small, like a femtocell. Base station 20 can have beamforming capabilities. Base station 20 can have cells or service areas formed for each beam. Alternatively, base station 20 can have the function of delivering the desired beam to a predetermined point with precise targeting accuracy by further considering distance information from the antenna of base station 20 in addition to beamforming, which provides beam directionality. This function can be referred to as beam focusing or point formation.

[0181] Figure 6 This is a diagram illustrating the configuration of base station 20 according to this embodiment. Base station 20 includes a wireless communication unit 21, a storage unit 22, and a control unit 23. However, Figure 6 The configuration shown is a functional configuration, and the hardware configuration may differ. Furthermore, the functionality of base station 20 can be implemented by distributing it across multiple physically separate configurations.

[0182] The wireless communication unit 21 is a signal processing unit for wireless communication with other wireless communication devices (e.g., at least one of relay station 30, terminal device 40, and other base stations 20). The wireless communication unit 21 may be referred to as a wireless transceiver. The wireless communication unit 21 is controlled by the control unit 23. The wireless communication unit 21 supports one or more wireless access methods. The wireless communication unit 21 may support at least one of NR, LTE, and 6G. In addition to NR, LTE, and 6G, the wireless communication unit 21 may also support W-CDMA and CDMA2000. The wireless communication unit 21 may support automatic repeater technology, such as HARQ (Hybrid Automatic Repeat Request). Some or all of these processes performed by the wireless communication unit 21 can be considered as the transmission process (the transmission process described above or below (first transmission process or second transmission process)) or the reception process (the reception process described above or below (first reception process or second reception process)) of this embodiment. Some or all of these processes (the transmission process described above or the reception process described below) may be performed by the control unit 23.

[0183] The wireless communication unit 21 may indicate at least one of the transmitting processing unit 211, the receiving processing unit 212, and the antenna 213. The wireless communication unit 21 may include multiple transmitting processing units 211, receiving processing units 212, and antennas 213. When the wireless communication unit 21 supports multiple wireless access methods, each part of the wireless communication unit 21 can be configured separately for each wireless access method. The transmitting processing unit 211 and the receiving processing unit 212 may be configured for LTE, NR, and 6G, respectively. The antenna 213 may be configured with multiple antenna elements, such as multiple patch antennas. The wireless communication unit 21 may have beamforming capabilities. For example, the wireless communication unit 21 may have polarized beamforming capabilities using vertical polarization (V-polarization) and horizontal polarization (H-polarization) (or dual-polarized beamforming capabilities using polarization directions at 45 degrees and -45 degrees to the vertical direction).

[0184] The transmission processing unit 211 performs transmission processing of downlink control information and downlink data. For example, the transmission processing unit 211 encodes the downlink control information and downlink data input from the control unit 23 using encoding methods such as block coding, convolutional coding, and Turbo coding. Here, polar codes or LDPC (low-density parity-check) codes can be used to perform encoding. Then, the transmission processing unit 211 modulates the coded bits using a predetermined modulation scheme (e.g., BPSK, QPSK, 16QAM, 64QAM, 256QAM, or a higher-order multi-level modulation scheme). In this case, the signal points on the constellation do not necessarily need to be equidistant. The constellation can be a non-uniform constellation (NUC). Then, the transmission processing unit 211 multiplexes the modulation symbols and downlink reference signals of each channel and configures them in the specified resource elements. Then, the transmission processing unit 211 performs various signal processing on the multiplexed signals. For example, the transmission processing unit 211 performs frequency domain transformation using discrete Fourier transform, addition of guard interval (cyclic prefix), generation of baseband digital signal, analog signal transformation, quadrature modulation, up-conversion, removal of unwanted frequency components, power amplification, and other processes. The signal generated by the transmission processing unit 211 is transmitted from the antenna 213. Some or all of these processes performed by the transmission processing unit 211 can be considered as the transmission processes of this embodiment (the transmission processes described above or below (first transmission process or second transmission process)).

[0185] The receiving processing unit 212 processes the uplink signal received via antenna 213. For example, the receiving processing unit 212 performs down-conversion, removes unwanted frequency components, controls amplification levels, performs quadrature demodulation, converts to a digital signal, removes guard intervals (cyclic prefixes), and extracts the frequency domain signal through discrete Fourier transform. Then, the receiving processing unit 212 separates the uplink channel (PUSCH, Physical Uplink Shared Channel), PUCCH (Physical Uplink Control Channel), and uplink reference signal from the processed signal. Furthermore, the receiving processing unit 212 demodulates the received signal using modulation schemes such as BPSK (Binary Phase Shift Keying) or QPSK (Quadrature Phase Shift Keying) for the modulation symbols used for the uplink channel. The modulation scheme used for demodulation can be 16QAM (Quadrature Amplitude Modulation), 64QAM, or 256QAM. In this case, the signal points on the constellation do not necessarily need to be equidistant. The constellation can be a non-uniform constellation (NUC). Then, the receiving processing unit 212 decodes the coded bits of the demodulated uplink channel. The decoded uplink data and uplink control information are output to the control unit 23. Some or all of these processes performed by the receiving processing unit 212 can be considered as the receiving processes of this embodiment (the receiving processes described above or below (first receiving process or second receiving process)).

[0186] Antenna 213 is an antenna device that converts current and radio waves to each other. Antenna 213 can be configured with a single antenna element, such as a patch antenna. Antenna 213 can also be configured with multiple antenna elements, such as multiple patch antennas. When antenna 213 is configured with multiple antenna elements, wireless communication unit 21 can have beamforming functionality. Wireless communication unit 21 can be configured to generate a directional beam by controlling the directivity of the wireless signal using multiple antenna elements. Antenna 213 can be a dual-polarized antenna. When antenna 213 is a dual-polarized antenna, wireless communication unit 21 can use vertical polarization (V-polarization) and horizontal polarization (H-polarization) (or dual polarization in polarization directions at 45 degrees and -45 degrees to the vertical direction) when transmitting wireless signals. Wireless communication unit 21 can control the directivity of wireless signals transmitted using vertical polarization and horizontal polarization (or dual polarization in polarization directions at 45 degrees and -45 degrees to the vertical direction). Furthermore, wireless communication unit 21 can transmit and receive spatially multiplexed signals via multiple layers configured with multiple antenna elements.

[0187] Storage unit 22 is a read / write storage device, such as DRAM, SRAM, flash memory, or hard disk. Storage unit 22 can store AI / ML models used in the processing of each function. The AI / ML models stored in base station 20 may be the same as or different from the AI / ML models stored in relay station 30 and terminal device 40, which will be described later.

[0188] Control unit 23 is a controller that controls each part of base station 20. Control unit 23 can be implemented by a processor such as a CPU or MPU. For example, control unit 23 can be implemented by a processor that executes various programs stored in the internal storage device of base station 20, using RAM as the working area. Control unit 23 can be implemented by an integrated circuit such as an ASIC or FPGA. Furthermore, control unit 23 can be implemented by a GPU. CPU, MPU, ASIC, FPGA, and GPU can all be considered as controllers. Note that control unit 23 can consist of multiple physically separate objects. For example, control unit 23 can be composed of multiple semiconductor chips.

[0189] The control unit 23 includes at least one of an acquisition unit 231, a management unit 232, a determination unit 233, and a transmission unit 234. Each block constituting the control unit 23 (acquisition unit 231 to transmission unit 234) is a functional block representing the function of the control unit 23. These functional blocks can be software blocks or hardware blocks. For example, each of the above functional blocks can be a software module implemented through software (including microprograms) or a circuit block on a semiconductor chip (die). Of course, each functional block can be a processor or an integrated circuit. The control unit 23 can be composed of functional units different from the above functional blocks. The configuration method of the functional blocks is arbitrary. The operation of each functional block of the control unit 23 can be the same as the operation of each functional block of the relay station 30 or the terminal device 40.

[0190] Note that in some embodiments, base station 20 may be configured as a collection of multiple physical or logical devices. As an example, base station 20 in this embodiment may be distinguished as multiple devices, such as BBU (Baseband Unit) and RU (Radio Unit). Base station 20 can be interpreted as a collection of these multiple devices. Furthermore, the base station may be a BBU or an RU, or both. BBUs and RUs may be connected via a predetermined interface such as ePRI (Enhanced General Public Radio Interface).

[0191] The RU can be referred to as an RRU (Remote Radio Unit) or RD (Radio DoT). The RU may correspond to the gNB-DU (gNB Distributed Unit) described later. The BBU may correspond to the gNB-CU (gNB Central Unit) described later. The RU can be a device integrally formed with the antenna. The antenna of base station 20, such as an antenna integrally formed with the RU, can employ an advanced antenna system and support MIMO such as FD-MIMO or beamforming MIMO. The antenna of base station 20 may include, for example, 64 transmit antenna ports and 64 receive antenna ports.

[0192] An antenna mounted on an RU can be an antenna panel consisting of one or more antenna elements, and the RU can mount one or more antenna panels. The RU can mount two types of antenna panels: horizontally polarized antenna panels and vertically polarized antenna panels. The RU can also mount two types of antenna panels: right-hand circularly polarized antenna panels and left-hand circularly polarized antenna panels, or antenna panels with a polarization direction at 45 degrees to the vertical direction and antenna panels with a polarization direction at -45 degrees. Multiple antennas with these multiple polarization directions can be implemented in a single antenna panel. The RU can form and control independent beams for each antenna panel.

[0193] Multiple base stations 20 can be interconnected. One or more base stations 20 may be included in a radio access network (RAN). In this case, base station 20 may be simply referred to as RAN, RAN node, AN (access network), or AN node, etc. The RAN in LTE may be called EUTRAN (Enhanced Universal Terrestrial RAN). The RAN in NR may be called NGRAN. In addition, the RAN in 6G may be called 6GRAN. The RAN in W-CDMA (UMTS) may be called UTRAN.

[0194] The LTE base station 20 can be referred to as an eNodeB (Evolved Node B) or eNB. In this case, the EUTRAN includes one or more eNodeBs (eNBs). The NR base station 20 can be referred to as a gNodeB or gNB. In this case, the NGRAN includes one or more gNBs. The 6G base station can be referred to as a 6GNodeB, 6gNodeB, 6GNB, or 6gNB. In this case, the 6GRAN includes one or more 6GNBs. The EUTRAN may include gNBs (en-gNBs) connected to the core network (EPC) in the LTE communication system (EPS). The NGRAN may include ng-eNBs connected to the core network 5GC in the 5G communication system (5GS).

[0195] When base station 20 is an eNB, gNB, 6GNB, etc., base station 20 can be referred to as a 3GPP access point. When base station 20 is a radio access point (access point), base station 20 can be referred to as a non-3GPP access point. Base station 20 can be an optical extension device called an RRH (Remote Radio Header). When base station 20 is a gNB, base station 20 can be a combination of the above-mentioned gNB-CU and gNB-DU, or it can be either gNB-CU or gNB-DU.

[0196] Here, the gNB-CU manages several upper layers of the access layer (e.g., RRC (Radio Resource Control), SDAP (Service Data Adaptation Protocol), and PDCP (Packet Data Convergence Protocol)) for communication with the UE. On the other hand, the gNB-DU manages several lower layers of the access layer (e.g., RLC (Radio Link Control), MAC (Media Access Control), and PHY (Physical Layer)). That is, in the messages / information described later, RRC signaling (semi-static notification) is generated by the gNB-CU, while MAC CE and DCI (dynamic notification) can be generated by the gNB-DU. Alternatively, in RRC configuration (semi-static notification), some configurations, such as IE:cellGroupConfig, can be generated by the gNB-DU, while the remaining configurations can be generated by the gNB-CU. These configurations can be sent and received via the F1 interface described later.

[0197] Base station 20 can be configured to communicate with other base stations. When multiple base stations 20 are eNBs or a combination of eNB and en-gNB, these base stations 20 can be connected via the X2 interface. When multiple base stations 20 are gNBs or a combination of gn-eNB and gNB, these base stations 20 can be connected via the Xn interface. When multiple base stations 20 are a combination of gNB-CU and gNB-DU, these base stations 20 can be connected via the aforementioned F1 interface. Messages / information described later (e.g., RRC signaling, MAC CE (MAC control element), or DCI (downlink control information), etc.) can be transmitted between the multiple base stations 20 via these inter-base station interfaces (e.g., X2 interface, Xn interface, or F1 interface, etc.).

[0198] The cell provided by base station 20 can be referred to as the serving cell. The concept of a serving cell includes PCell (primary cell) and SCell (secondary cell). When providing dual connectivity to terminal device 40, the PCell provided by MN (master node) and zero or more SCells can be referred to as a primary cell group. Dual connectivity can be at least one of EUTRA-EUTRA dual connectivity, EUTRA-NR dual connectivity (ENDC), EUTRA-NR dual connectivity with 5GC, NR-EUTRA dual connectivity (NEDC), NR-NR dual connectivity, NR-6G dual connectivity, and 6G-NR dual connectivity. Of course, dual connectivity is not limited to these.

[0199] A serving cell may include a PSCell (primary / secondary cell or primary SCG cell). When providing dual connectivity to terminal device 40, the PSCell provided by the SN (secondary node) and zero or more SCells can be referred to as an SCG (secondary cell group). Unless specially configured (e.g., PUCCH on the SCell), the Physical Uplink Control Channel (PUCCH) is transmitted by the PCell and PSCell, not by the SCell. Radio link failures are detected by the PCell and PSCell, but not by the SCell (it is not necessary to detect them). Therefore, the PCell and PSCell play a special role between serving cells, and thus they are also referred to as SpCells (special cells).

[0200] A cell can be associated with one downlink component carrier and one uplink component carrier. The system bandwidth corresponding to a cell can be divided into multiple BWPs (bandwidth portions). At this time, one or more BWPs are configured for terminal device 40, and terminal device 40 can use one BWP as the active BWP. The radio resources that terminal device 40 can use, such as frequency bands, digitization (subcarrier spacing), or time slot format (time slot configuration), can be different for each cell, each component carrier, or each BWP.

[0201] <2-3. Relay Station Configuration>

[0202] Relay station 30 is a wireless communication device that acts as a repeater for base station 20. Relay station 30 is a type of base station. Relay station 30 is an information processing device. Relay station 30 can also be referred to as a relay base station. Note that relay station 30 can be a device referred to as a repeater (e.g., an RF repeater, smart terrestrial). Relay station 30 is a wireless communication device that performs wireless communication with other wireless communication devices (e.g., base station 20, other relay stations 30, or terminal device 40).

[0203] Relay station 30 can communicate with terminal device 40 via NOMA. Relay station 30 relays communication between base station 20 and terminal device 40. Relay station 30 can also communicate wirelessly with other relay stations 30 and base station 20. Relay station 30 can be a ground station or a non-ground station. Relay station 30, together with base station 20, constitutes a Radio Access Network (RAN).

[0204] The repeater station 30 can be a fixed device, a mobile device, or a floating device. The coverage area of ​​the repeater station 30 is not limited to a specific size. The cells covered by the repeater station 30 can be macrocells, microcells, or small cells.

[0205] The relay station 30 can be a device installed on other equipment, as long as it meets the relay function. For example, the relay station 30 can be installed on terminal equipment such as smartphones, on mobile vehicles such as cars, trains, balloons, airplanes or drones, or on household appliances such as televisions, game consoles, air conditioners, refrigerators or lighting equipment.

[0206] The configuration of relay station 30 can be the same as that of base station 20. Like base station 20, relay station 30 can be a device mounted on a mobile vehicle or the mobile vehicle itself. The mobile vehicle can be a mobile terminal, such as a smartphone or mobile phone as described above. The mobile vehicle can move on land (in a narrow sense, the ground) or underground. The mobile vehicle can move on water or underwater. The mobile vehicle can move in the atmosphere or outside the atmosphere. Relay station 30 can be ground station equipment or non-ground station equipment. Relay station 30 can be an aircraft station or a satellite station, etc.

[0207] The coverage area of ​​relay station 30 (such as base station 20) can be as large as a macrocell or as small as a picocell. The coverage area of ​​relay station 30 can be as small as a femtocell. Relay station 30 can have beamforming capabilities. Relay station 30 can have cells or service areas formed for each beam.

[0208] Figure 7 This is a diagram illustrating the configuration of the relay station 30 according to this embodiment. The relay station 30 includes a wireless communication unit 31, a storage unit 32, and a control unit 33; however, Figure 7 The configuration shown is a functional configuration, and the hardware configuration may differ. Furthermore, the functionality of relay station 30 can be implemented by distributing it across multiple physically separate configurations.

[0209] The wireless communication unit 31 is a signal processing unit for wireless communication with other wireless communication devices (e.g., at least one of base station 20, terminal device 40, and other relay stations 30). The wireless communication unit 31 may be referred to as a wireless transceiver. The wireless communication unit 31 supports one or more wireless access methods. The wireless communication unit 31 may support at least one of NR, LTE, and 6G. In addition to supporting NR, LTE, and 6G, the wireless communication unit 31 may also support W-CDMA and CDMA2000. Part or all of the processing performed by the wireless communication unit 31 can be considered as the transmission processing (the transmission processing described above or below (first transmission processing or second transmission processing)) or reception processing (the reception processing described above or below (first reception processing or second reception processing)) of this embodiment. Part or all of these processes (the transmission processing described above or the reception processing described later) may be performed by the control unit 33.

[0210] The wireless communication unit 31 may represent at least one of the following: a transmitting processing unit 311, a receiving processing unit 312, and an antenna 313. The wireless communication unit 31 may include multiple transmitting processing units 311, receiving processing units 312, and antennas 313. When the wireless communication unit 31 supports multiple wireless access methods, each part of the wireless communication unit 31 may be configured individually for each wireless access method. The transmitting processing unit 311 and the receiving processing unit 312 may be configured for LTE, NR, and 6G, respectively. The configuration of the transmitting processing unit 311, the receiving processing unit 312, and the antenna 313 may be the same as the configuration of the transmitting processing unit 211, the receiving processing unit 212, and the antenna 213 of the base station 20 described above. Similar to the wireless communication unit 21 of the base station 20, the wireless communication unit 31 may also have a beamforming function, using part or all of the processing performed by the transmitting processing unit 311 as the transmitting processing in this embodiment (the transmitting processing described above or the following transmitting processing (first transmitting processing or second transmitting processing)). Furthermore, part or all of the processing performed by the receiving processing unit 312 can be regarded as the receiving processing of this embodiment (the receiving processing described above or below (first receiving processing or second receiving processing)).

[0211] Storage unit 32 is a read / write storage device, such as DRAM, SRAM, flash memory, or hard disk. Storage unit 32 serves as a storage device for relay station 30, and its configuration and function can be the same as those of storage unit 22 in base station 20 described above. Storage unit 32 can store AI / ML models used in each functional process. The AI / ML models stored in relay station 30 can be the same as or different from the AI / ML models stored in base station 20 and terminal device 40 described later.

[0212] Control unit 33 is a controller that controls the various parts of relay station 30. Control unit 33 can be implemented by a processor such as a CPU or MPU. Specifically, control unit 33 can be implemented by the processor executing various programs stored in the internal storage device of relay station 30, using RAM or similar memory as the working area. Control unit 33 can be implemented by an integrated circuit such as an ASIC or FPGA. CPU, MPU, ASIC, and FPGA can all be considered controllers. Furthermore, control unit 33 can be implemented by a GPU. CPU, MPU, ASIC, FPGA, and GPU can all be considered controllers. Note that control unit 33 can be composed of multiple physically separate components. For example, control unit 33 can be composed of multiple semiconductor chips. The configuration and function of control unit 33 can be the same as those of control unit 23 of base station 20 described above.

[0213] The control unit 33 includes at least one of an acquisition unit 331, a management unit 332, a determination unit 333, and a transmission unit 334. Each block constituting the control unit 33 (acquisition unit 331 to transmission unit 334) is a functional block representing the function of the control unit 33. These functional blocks can be software blocks or hardware blocks. For example, each of the above functional blocks can be a software module implemented through software (including microprograms) or a circuit block on a semiconductor chip (die). Of course, each functional block can be a processor or an integrated circuit. The control unit 33 can also be configured as a functional unit different from the above functional blocks. The configuration method of the functional blocks is arbitrary. The operation of each functional block of the control unit 33 can also be the same as the operation of each functional block of the base station 20 or the terminal device 40.

[0214] Note that relay station 30 can be an IAB relay node. Relay station 30 operates as an IAB-MT (Mobile Terminal) providing backhaul as an IAB donor node and as an IAB-DU (Distributed Unit) providing access as a terminal device 40. The IAB donor node can be, for example, base station 20, and operates as an IAB-CU (Central Unit).

[0215] <2-4. Configuration of Terminal Equipment>

[0216] Terminal device 40 is a wireless communication device that performs communication with other wireless communication devices (e.g., base station 20, relay station 30, or other terminal devices 40, etc.). Terminal device 40 may be referred to as UE (user equipment) 40.

[0217] Terminal device 40 can be any form of information processing device (computer). For example, terminal device 40 can be a mobile terminal, such as a mobile phone, smart device (smartphone or tablet), PDA (personal digital assistant), or laptop PC. Furthermore, terminal device 40 can be a communication module that connects to an information processing device (e.g., an imaging device without wireless communication capabilities) and provides wireless communication capabilities to the information processing device. Alternatively, terminal device 40 can be an imaging device equipped with wireless communication capabilities (e.g., a camera).

[0218] Furthermore, terminal device 40 can be a motorcycle or mobile relay vehicle equipped with communication devices such as an FPU (Field Pickup Unit). Additionally, terminal device 40 can be an M2M (Machine-to-Machine) device or an IoT (Internet of Things) device. Furthermore, terminal device 40 can be a wearable device such as a smartwatch.

[0219] Furthermore, the terminal device 40 can be an XR (Extended Reality) device, such as an AR (Augmented Reality) device, a VR (Virtual Reality) device, or an MR (Mixed Reality) device. In this case, the XR device can be an eyeglass-type device, such as AR glasses or MR glasses, or a head-mounted device, such as a VR head-mounted display. When the terminal device 40 is an XR device, it can be a standalone device consisting only of a user-worn portion (e.g., glasses). Alternatively, the terminal device 40 can also be a terminal linkage device consisting of a user-worn portion (e.g., glasses) and a terminal portion (e.g., a smart device) linked to that portion.

[0220] Terminal device 40 can communicate with base station 20 via NOMA. When communicating with base station 20, terminal device 40 can use automatic repeater technology, such as HARQ. Terminal device 40 can also communicate with other terminal devices 40 via sidelink, again using automatic repeater technology, such as HARQ. When performing sidelink communication with other terminal devices 40, terminal device 40 can perform NOMA communication. Terminal device 40 can also communicate with other wireless communication devices, such as base station 20, via LPWA. The wireless communication used by terminal device 40 can be millimeter-wave wireless communication. The wireless communication used by terminal device 40 can be wireless communication using radio waves, including sidelink communication, or wireless communication using infrared or visible light (i.e., optical wireless).

[0221] Terminal device 40 can be a mobile wireless communication device, i.e., a mobile device. Terminal device 40 can be a wireless communication device mounted on a mobile body, or it can be the mobile body itself. Terminal device 40 can be a vehicle moving on a road, such as a car, bus, truck, or motorcycle, or a train running on rails, or it can be a wireless communication device mounted on these vehicles. The mobile body can be a mobile terminal, or it can be a mobile body moving on land (narrowly defined as the ground), underground, on water, or underwater. Furthermore, the mobile body can be a mobile body moving in the atmosphere, such as an airplane, spacecraft, balloon, or helicopter, or it can be a mobile body moving outside the atmosphere, such as a satellite. The mobile body can be a UAV (unmanned aerial vehicle), such as a drone. Additionally, terminal device 40 can be a wireless communication device mounted on the mobile body.

[0222] Terminal device 40 can simultaneously connect to and communicate with multiple base stations 20 or multiple cells. When a base station 20 supports a communication area via multiple cells (e.g., pCell or sCell), these multiple cells can be bundled using carrier aggregation (CA), dual connectivity (DC), or multiple connectivity (MC) technologies to perform communication between base station 20 and terminal device 40. Alternatively, communication between terminal device 40 and these multiple base stations 20 can be performed using cooperative multipoint transmission and reception (CoMP) technology via cells of different base stations 20.

[0223] Terminal device 40 can be a relay terminal that relays communication to a remote terminal.

[0224] Figure 8 This diagram illustrates the configuration of a terminal device 40 according to this embodiment. The terminal device 40 includes a wireless communication unit 41, a storage unit 42, and a control unit 43. However... Figure 8 The configuration shown is a functional configuration, and the hardware configuration may differ. Furthermore, the functions of the terminal device 40 can be implemented by distributing them across multiple physically separate configurations.

[0225] The wireless communication unit 41 is a signal processing unit for wireless communication with other wireless communication devices (e.g., at least one of base station 20, relay station 30, and other terminal devices 40). The wireless communication unit 41 may be referred to as a wireless transceiver. The wireless communication unit 41 is controlled by the control unit 43. The wireless communication unit 41 supports one or more wireless access methods. The wireless communication unit 41 may support at least one of NR, LTE, and 6G. In addition to supporting NR, LTE, and 6G, the wireless communication unit 41 may also support W-CDMA and CDMA2000. The wireless communication unit 41 may support automatic repeater technology, such as HARQ (Hybrid Automatic Repeat Request). Part or all of the processing performed by the wireless communication unit 41 can be considered as the transmission processing (the transmission processing described above or below (first transmission processing or second transmission processing)) or reception processing (the reception processing described above or below (first reception processing or second reception processing)) of this embodiment. Part or all of these processes (the transmission processing described above or the reception processing described below) may be performed by the control unit 43.

[0226] The wireless communication unit 41 may indicate at least one of the transmitting processing unit 411, the receiving processing unit 412, and the antenna 413. The wireless communication unit 41 may include multiple transmitting processing units 411, receiving processing units 412, and antennas 413. When the wireless communication unit 41 supports multiple wireless access methods, each part of the wireless communication unit 41 can be configured separately for each wireless access method. The transmitting processing unit 411 and the receiving processing unit 412 may be configured for LTE, NR, and 6G, respectively. The antenna 413 may be configured with multiple antenna elements, such as multiple patch antennas. The wireless communication unit 41 may have beamforming capabilities. For example, the wireless communication unit 41 may have polarized beamforming capabilities using vertical polarization (V-polarization) and horizontal polarization (H-polarization) (or dual-polarized beamforming capabilities using polarization directions at 45 degrees and -45 degrees to the vertical direction). Furthermore, antenna 413 is not necessarily part of terminal device 40 or wireless communication unit 41. For example, antenna 413 may not be included in wireless communication unit 41, but may be installed in terminal device 40 as an external antenna. Part or all of the processing performed by transmission processing unit 411 can be considered as the transmission processing of this embodiment (the transmission processing described above or below (first transmission processing or second transmission processing)). Similarly, part or all of the processing performed by reception processing unit 412 can be considered as the reception processing of this embodiment (the reception processing described above or below (first reception processing or second reception processing)).

[0227] Storage unit 42 is a read / write storage device, such as DRAM, SRAM, flash memory, or hard disk. Storage unit 42 serves as a storage device for terminal device 40, and can store AI / ML models used in the processing of each function. The AI / ML models stored in terminal device 40 may be the same as or different from the AI / ML models stored in base station 20 or relay station 30 described above.

[0228] Control unit 43 is a controller that controls various parts of terminal device 40. Control unit 43 can be implemented by a processor such as a CPU or MPU. For example, control unit 43 can be implemented by a processor that executes various programs stored in the internal storage of terminal device 40, which uses RAM or similar memory as its working area. Control unit 43 can be implemented by an integrated circuit such as an ASIC or FPGA. Control unit 43 can also be implemented by a GPU. CPU, MPU, ASIC, FPGA, and GPU can all be considered controllers. Note that control unit 43 can consist of multiple physically separate objects. For example, control unit 43 can be composed of multiple semiconductor chips.

[0229] The control unit 43 includes at least one of an acquisition unit 431, a management unit 432, a determination unit 433, and a transmission unit 434. Each block constituting the control unit 43 (acquisition unit 431 to transmission unit 434) is a functional block representing the function of the control unit 43. These functional blocks can be software blocks or hardware blocks. For example, each of the above functional blocks can be a software module implemented through software (including microprograms) or a circuit block on a semiconductor chip (die). Of course, each functional block can be a processor or an integrated circuit. The control unit 43 can be composed of functional units different from the above functional blocks. The configuration method of the functional blocks is arbitrary. The operation of each functional block of the control unit 43 can be the same as the operation of each functional block of the base station 20 or the relay station 30.

[0230] <2-5. AI / ML Models>

[0231] At least one of the storage unit 22 of base station 20, the storage unit 32 of relay station 30, and the storage unit 42 of terminal device 40 can store an AI / ML model. Furthermore, at least one of base station 20, relay station 30, and terminal device 40 can use the AI / ML model to perform various processes. For example, at least one of base station 20, relay station 30, and terminal device 40 can use the AI / ML model to generate information such as auxiliary information and transmission data. Note that in the following description, the AI / ML model may be referred to as a learning model or simply a model.

[0232] A learning model is, for example, a neural network model. A neural network model consists of layers called input layers, intermediate layers (or hidden layers), and output layers. Each layer contains multiple nodes, connected by edges. Each layer has a function called an activation function, and each edge is weighted. A learning model has one or more intermediate layers (or hidden layers). When the learning model is a neural network model, learning means, for example, setting the number of intermediate layers (or hidden layers), the number of nodes in each layer, or the weight of each edge.

[0233] Here, the neural network model can be a deep learning-based model. In this case, the neural network model can be a model called a DNN (Deep Neural Network). Furthermore, the neural network model can be a model called a CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), or LSTM (Long Short-Term Memory). Of course, the neural network model is not limited to these forms.

[0234] In a CNN, the hidden layers consist of layers called convolutional layers and pooling layers. In the convolutional layers, filtering is applied through convolution operations to extract data called feature maps. In the pooling layers, information from the feature maps output from the convolutional layers is compressed, and downsampling is applied. CNNs are used, for example, in image recognition applications, where information from each pixel (also called an image pixel) is input to the input layer, and information related to the recognized image can be obtained as the output layer.

[0235] RNNs have a network structure in which hidden layer values ​​are recursively input into the hidden layers, and are used, for example, to process short-term time series data.

[0236] In LSTM, the influence of distant past outputs can be preserved by introducing parameters called memory cells, which maintain the state of intermediate layers, into the intermediate layer output of an RNN. That is, LSTM takes longer to process time series data than RNN.

[0237] Furthermore, learning models are not limited to neural network models. For example, a learning model can be based on reinforcement learning. In reinforcement learning, the learner learns the action (set) that maximizes the value through repeated trials. Alternatively, a learning model can be a logistic regression model.

[0238] Note that a learning model can consist of multiple models. For example, a learning model can consist of multiple neural network models. More specifically, a learning model can consist of multiple neural network models selected from, for example, CNN, RNN, and LSTM. When a learning model consists of multiple neural network models, these multiple neural network models can be in a dependent or parallel relationship.

[0239] Note that a learning model can be rewritten as an AI (Artificial Intelligence) model, an ML (Machine Learning) model, or a training model. In the following description, a learning model may be simply referred to as a model. Hereinafter, it will be used as... Figure 1A The learning model M shown below is used as an example to illustrate the learning model, but the learning model in this embodiment is not limited to the learning model M shown below.

[0240] The learning model M is a learning model composed of a first model and a second model. The first model and the second model are each sub-model of the learning model M. The first model is the first half of the learning model M and is stored in the transmitting device. The second model is the second half of the learning model M and is stored in the receiving device. Note that a sub-model can also be considered a type of learning model. The learning model of this embodiment will be described in detail below. In the following description, the generation of auxiliary information will be used as an example to explain the learning model, but the functionality of the learning model is not limited to the generation of auxiliary information.

[0241] The learning model M is, for example, a learning model (training model) that learns by using auxiliary information T as input data and auxiliary information R as the correct label (teacher data). As mentioned above, auxiliary information is information necessary for the receiving device to perform processing related to wireless communication, or information that assists the receiving device in performing processing related to wireless communication. For example, auxiliary information is information (e.g., control information) required for operations such as uplink data transmission, sidelink transmission, unlicensed communication, and handover of the terminal device 40. Note that auxiliary information R is information related to auxiliary information T. Auxiliary information R can be information with content different from auxiliary information T, or it can be information with content the same as auxiliary information T.

[0242] When base station 20 or terminal device 40 inputs first information (e.g., auxiliary information T) into the first model of learning model M, the first model outputs second information (e.g., compressed information / feature information of auxiliary information T). Furthermore, when base station 20 or terminal device 40 inputs the second information into the second model of learning model M, the second model outputs third information (e.g., auxiliary information R).

[0243] In this context, the learning model M includes an input layer that takes first information (e.g., auxiliary information T) as input, an output layer that outputs third information (e.g., auxiliary information R), a first element belonging to any layer from the input layer to the output layer (excluding the output layer), and a second element whose value is calculated based on the first element and its weights. It can also be a learning model used to enable a computer to output third information from the output layer based on the first information input to the input layer by performing operations based on the first element and its weights (i.e., connection coefficients), wherein each element belonging to each layer other than the output layer serves as the first element of the information input to the input layer.

[0244] Here, we assume that the learning model M is implemented by a neural network with one or more intermediate layers, such as a DNN. In this case, the first element included in the learning model corresponds to any node in the input layer or intermediate layer. Furthermore, the second element corresponds to the next-level node, which is the node from which values ​​are sent from the node corresponding to the first element. Additionally, the weights of the first element correspond to connection coefficients, which are weights considered for the values ​​sent from the node corresponding to the first element to the node corresponding to the second element.

[0245] Furthermore, suppose the learning model M is implemented through a regression model represented by "y=a1*x1+a2*x2+...+ai*xi". In this case, the first element contained in the learning model M corresponds to the input data (xi) such as x1 and x2. Moreover, the weight of the first element corresponds to the coefficient ai corresponding to xi. Here, the regression model can be considered as a simple perceptron with an input layer and an output layer. When each model is considered as a simple perceptron, the first element corresponds to any node present in the input layer, and the second element can be considered as any node present in the output layer.

[0246] Base station 20 or terminal device 40 uses a model with any structure, such as a neural network or regression model, to compute output information. Specifically, the learning model M has coefficients configured to output third information (e.g., auxiliary information T) when first information (e.g., auxiliary information T) is input. For example, base station 20 or terminal device 40 sets the coefficients based on the similarity between the third information and the values ​​obtained by inputting the first information into the learning model. Base station 20 or terminal device 40 uses this sub-model (first model) of the learning model M to generate second information from the first information. Alternatively, base station 20 or terminal device 40 uses this sub-model (second model) of the learning model M to generate third information from the second information.

[0247] Note that in the above example, a model is shown that outputs third information when first information is input, as an example of a learning model M. However, the learning model M according to this embodiment can be a model generated based on the results obtained by repeatedly inputting and outputting data to the learning model.

[0248] Furthermore, when base station 20 or terminal device 40 uses GAN (Generative Adversarial Network) to perform the learning or generation of output information, the learning model can be a model that constitutes part of GAN.

[0249] Note that the learning device performing the learning of learning model M can be base station 20, relay station 30, or terminal device 40. Furthermore, the learning device can be another information processing device (e.g., management device 10 or a server device network-connected to management device 10). For example, suppose the server device performs the learning of learning model M. In this case, the server device performs the learning of learning model M and stores the learned learning model M in a storage unit. More specifically, the server device sets the connection coefficients of learning model M such that when first information (e.g., auxiliary information T) is input to learning model M, the learning model outputs third information (e.g., auxiliary information R).

[0250] For example, an information processing device (e.g., management device 10, base station 20, relay station 30, terminal device 40, or server device) inputs first information into the nodes of the input layer of the learning model M and propagates the data to the output layer of the learning model M through each intermediate layer, thereby causing the learning model M to output third information. Then, the information processing device corrects the connection coefficients of the learning model M based on the difference between the value actually output by the learning model M and the value used as the correct label (teacher data). At this point, the information processing device can use techniques such as backpropagation to correct the connection coefficients. Meanwhile, the server device can correct the connection coefficients based on the cosine similarity between the vector indicating the input value and the vector indicating the value actually output by the learning model.

[0251] Note that any learning algorithm can be used for learning. For example, information processing devices can use learning algorithms such as neural networks, support vector machines, clustering, reinforcement learning, random forests, decision trees, etc., to perform learning models.

[0252] The above describes the method for generating the learning model M, but the above embodiments can also be applied to learning models other than the learning model M.

[0253] Furthermore, the learning algorithm used in this embodiment can be an algorithm in which a single information processing device (e.g., management device 10, base station 20, relay station 30, terminal device 40, or server device) learns independently, or it can be an algorithm in which multiple information processing devices (e.g., multiple devices selected from management device 10, base station 20, relay station 30, terminal device 40, and server device) learn collaboratively. An example of a learning algorithm in which multiple information processing devices learn collaboratively is joint learning.

[0254] << 3. Network Architecture>>

[0255] The configuration of communication system 1 has been described above. Next, the network architecture that can be applied in communication system 1 in this embodiment will be described. Here, as an example of the core network CN of communication system 1, the architecture of a fifth-generation mobile communication system (5G) will be described.

[0256] Figure 9 This is a diagram illustrating an example of the 5G architecture. The 5G core network CN is also known as 5GC (5G Core) / NGC (Next Generation Core). In the following text, the 5G core network CN will also be referred to as 5GC / NGC. The core network CN is connected to the UE (User Equipment) 40 via AN 530.

[0257] Notice, Figure 8 The terminal device 40 shown is an example of UE 40. Figure 6 Base station 20 shown Figure 7 The relay station 30 shown is an example of RAN / AN 530. Furthermore, Figure 5 The management device 10 shown is an example of a device having the functions of, for example, AF 549 or AMF 541.

[0258] The (R)AN 530 has the capability to connect to a RAN (Radio Access Network) and to ANs (Access Networks) other than the RAN. The (R)AN 530 includes base stations referred to as gNBs or ng-eNBs.

[0259] The core network (CN) primarily handles connection authorization and session management when the UE 40 connects to the network. The core network (CN) can be configured to include user plane function group 520 and control plane function group 540.

[0260] User plane function group 520 includes UPF (User Plane Function) 521 and DN (Data Network) 522. UPF 521 provides user plane processing functions. UPF 521 includes routing / forwarding functions for data processed in the user plane. DN 522 provides functions for providing connectivity to operator-specific services such as MNO (Mobile Network Operator), providing Internet connectivity, or providing connectivity to third-party services. Therefore, user plane function group 520 acts as a gateway, serving as the boundary between the core network (CN) and the Internet.

[0261] Control plane function group 540 includes AMF (Access Management Function) 541, SMF (Session Management Function) 542, AUSF (Authentication Server Function) 543, NSSF (Network Slice Selection Function) 544, NEF (Network Exposure Function) 545, NRF (Network Repository Function) 546, PCF (Policy Control Function) 547, UDM (Unified Data Management) 548, and AF (Application Function) 549.

[0262] AMF 541 has functions such as UE 40 registration processing, connection management, and mobility management. SMF 542 has functions such as session management and IP allocation, as well as UE 40 management. AUSF 543 has authentication functions. NSSF 544 has functions related to network slice selection. NEF 545 has functions to provide network function capabilities and events to third-party AF 549, as well as edge computing functions.

[0263] NRF 546 has the function of discovering network functions and maintaining network function profiles. PCF 547 has policy control functions. UDM 548 has the function of generating 3GPP AKA authentication information and processing user IDs. AF 549 has the function of interacting with the core network to provide services.

[0264] For example, control plane function group 540 obtains information from UDM 548, which stores subscriber information of UE 40, and determines whether UE 40 can connect to the network. Control plane function group 540 uses the subscription information of UE 40 and the encryption key included in the information obtained from UDM 548 to make this determination. Furthermore, control plane function group 540 performs tasks such as generating encryption keys.

[0265] In other words, the control plane function group 540 determines network connectivity availability based on whether information of the UE 40 linked to a user number called the IMSI (International Mobile Subscriber Identity) is stored in the UDM 548. Note that the IMSI is stored, for example, in the SIM (Subscriber Identity Module) card within the UE 40.

[0266] Note that the core network (CN) may have LCM-related network functions. LCM-related network functions are network functions that provide LCM-related control functions. For example, LCM-related network functions have functions such as controlling at least one of (M1) to (M25) above, or providing information necessary to perform at least one of (M1) to (M25) above.

[0267] LCM-related network functions can be included in existing network functions. For example, LCM-related network functions can be included in MTLF (Model Training Logic Function) 551. Figure 5 The management device 10 shown is an example of a device with MTLF 551 functionality.

[0268] Furthermore, LCM-related network functions can be included in servers that constitute the core network CN. For example, LCM-related network functions can be included in servers connected to UPF 521 (e.g., MEC (Mobile Edge Computing) server 552). Also, LCM-related network functions can be included in servers connected to base station 20 ((R) AN 530) (e.g., MEC server 553). Figure 5 The management device 10 shown is an example of MEC server 552 or MEC server 553.

[0269] Furthermore, LCM-related network functions can be included in the core network (CN) as new network functions. For example, LCM-related network function 554 can be included in the core network (CN) as one of the control plane function groups 540. Figure 5 The management device 10 shown is an example of a device with the functionality of LCM-related network function 554.

[0270] Note that one or more LCM-related network functions may exist in the core network (CN).

[0271] <<4.LCM>>

[0272] The network architecture has been described above. Next, LCM (Lifecycle Management) will be described in detail.

[0273] As described above, LCM is the management of AI / ML models. LCM can be executed by the management unit of a communication device. For example, LCM can be executed by at least one of the management unit 232 of base station 20, the management unit 332 of relay station 30, and the management unit 432 of terminal device 40. Of course, LCM can be executed by management device 10, or it can be executed by server equipment on the Internet.

[0274] Note that in the following description, AI / ML models may be referred to as learning models or simply models.

[0275] The LCM may consist of one or more elements (processes / methods / procedures / functions) from (M1) to (M25). For example, the LCM may be one of (M1) to (M25), or it may be a combination of multiple elements selected from (M1) to (M25). Note that (M1) to (M25) shown below are merely examples. The LCM (or the elements constituting the LCM) of this embodiment may include elements other than (M1) to (M25). Hereinafter, (M1) to (M25) will be described respectively.

[0276] (M1) Data Collection

[0277] Data collection is the process by which network nodes (e.g., base station 20), management entities (e.g., management device 10), or UEs (user equipment) (e.g., terminal device 40) collect data for use in AI / ML model training, data analysis, and inference.

[0278] (M2) model training

[0279] Model training is a data-driven process that learns the input-output relationships of an AI / ML model and obtains a trained AI / ML model for inference.

[0280] (M3) Function Identification

[0281] Function identification is a process / method used to identify the functions or requirements to which AI / ML is applied.

[0282] (M4) model recognition

[0283] Model recognition is a process / method used to identify AI / ML models so that they can be understood jointly between the network (e.g., base station 20, management device 10) and the UE (e.g., terminal device 40).

[0284] (M5) model delivery

[0285] Model delivery is a general term that refers to the delivery of an AI / ML model from one entity (e.g., the communication device described above or below) to another entity (e.g., the communication device described above or below) by any method.

[0286] (M6) Model Transfer

[0287] Model transmission is the process of transmitting AI / ML models via an air interface, where the parameters of the model structure are known on the receiving side (e.g., the receiving device side), or it is a new model with parameters. Transmission can include the transmission of a complete model or a portion of a model.

[0288] (M7) Model Download

[0289] Model download is the process of transmitting a model from a network (e.g., base station 20, management device 10) to a UE (e.g., terminal device 40).

[0290] (M8) model upload

[0291] Model upload is the process of transmitting a model from a UE (e.g., terminal device 40) to a network (e.g., base station 20, management device 10).

[0292] (M9) Model Reasoning

[0293] Model inference is the process of using a trained AI / ML model to generate a series of outputs based on a series of inputs (e.g., information received from other communication devices).

[0294] (M10) Model Validation

[0295] Model validation is a training subprocess that uses a different dataset than the one used for model training to evaluate the quality of an AI / ML model.

[0296] (M11) Model Testing

[0297] Model testing is a training subprocess that uses a different dataset than that used for model training and validation to evaluate the performance of the final AI / ML model. Unlike AI / ML model validation, model testing does not assume subsequent model tuning.

[0298] (M12) model activation

[0299] Model activation is the process of enabling an AI / ML model for a specific function.

[0300] (M13) Functional activation

[0301] Functional activation is the process of enabling AI / ML functionality for specific features.

[0302] (M14) model deactivation

[0303] Model deactivation is the process of disabling specific features in AI / ML models.

[0304] (M15) Functional deactivation

[0305] Functional deactivation is the process of disabling AI / ML functionality for specific features.

[0306] (M16) Model Switching

[0307] Model switching is the process of deactivating the currently active AI / ML model and activating a different AI / ML model for a specific function.

[0308] (M17) Function Switching

[0309] Function switching is the process of deactivating the currently active AI / ML model and activating a different AI / ML model for a specific function.

[0310] (M18) model rollback

[0311] Model rollback is the process of deactivating the AI / ML model and switching to a process that does not use the AI / ML model.

[0312] (M19) Model Monitoring

[0313] Model monitoring is the process used to monitor the inference performance of AI / ML models.

[0314] (M20) Model Update

[0315] Model update is the process of updating the model parameters and / or model structure.

[0316] (M21) Model Registration

[0317] Model registration is the process of registering the model parameters and / or model structure of a model.

[0318] (M22) model deployment

[0319] Model deployment is a general term for deploying AI / ML models from one entity (e.g., a communication device described above or below) to another entity (e.g., a communication device described above or below) by any method.

[0320] (M23) Model Configuration

[0321] Model configuration is the process of configuring the model parameters and / or model structure.

[0322] (M24) Model Selection

[0323] Model selection is the process of choosing and activating an AI / ML model from among several currently available options.

[0324] (M25) UE capability

[0325] UE capability is the process of notifying another communication device (e.g., base station 20) of the capabilities of a terminal device (e.g., terminal device 40).

[0326] Note that in the following description, AI / ML model management based on functionality (e.g., AI / ML model management for each use case and / or function) may be referred to as functionality-based LCM.

[0327] Figure 10 This is a diagram used to explain LCM. More specifically, Figure 10 This is a diagram illustrating the relationships between the LCM (or the elements that constitute the LCM). The LCM can be like... Figure 10 Implement as shown in the diagram. Note that... Figure 10 The diagram shown can be implemented by any one of the base station 20, the transmitting device (e.g., terminal device 40), and the receiving device (e.g., terminal device 40), or a combination of at least two of them. In other words, any one of the base station 20, the transmitting device (e.g., terminal device 40), and the receiving device (e.g., terminal device 40) can implement the LCM mentioned above. Figure 10 At least a portion of the operations defined in the diagram. The relationships between LCMs (or the elements constituting an LCM) are not limited to... Figure 10 The relationship shown is as follows.

[0328] Data collection provides training data for model training. Additionally, data collection provides monitoring data for model management / performance monitoring. Furthermore, data collection provides inference data for model inference.

[0329] Model training stores the trained / updated model in the model storage.

[0330] Model management / performance monitoring performs control over model training (e.g., control over model maintenance and / or updates). Furthermore, model management / performance monitoring performs control over model inference (e.g., at least one of model activation, model deactivation, model selection, model switching, and model fallback).

[0331] Model storage delivers or transfers the trained / updated model to model inference.

[0332] The model inference process notifies the model management / performance monitoring system of the inference output.

[0333] <<4. Operation of the Communication System>>

[0334] Based on the above, the operation of communication system 1 related to LCM (Lifecycle Management) will be described. Note that the description of "LCM" in the following description can be replaced by "AI / ML model management".

[0335] The communication device (hereinafter referred to as the first communication device) provided in the communication system 1 is configured to communicate wirelessly with another communication device (hereinafter referred to as the second communication device).

[0336] Here, the first communication device and the second communication device can be a base station 20, a relay station 30, or a terminal device 40. For example, the first communication device can be a base station 20, and the second communication device can be a terminal device 40. Alternatively, the first communication device can be a terminal device 40, and the second communication device can be a base station 20. Alternatively, the first communication device can be a terminal device 40, and the second communication device can be another terminal device 40. Alternatively, the first communication device can be a base station 20, and the second communication device can be another base station 20. Note that the first and second communication devices are not limited to base station 20 and / or terminal device 40. At least one of the first and second communication devices can be a relay station 30 or a management device 10. Of course, at least one of the first and second communication devices can be a server device connected to the Internet or other communication devices.

[0337] Wireless communication between the first communication device and the second communication device can be cellular communication. For example, wireless communication between the first communication device and the second communication device can be uplink communication with the terminal device 40 as a transmitting device and the base station 20 as a receiving device, downlink communication with the base station 20 as a transmitting device and the terminal device 40 as a receiving device, or sidelink communication with the terminal device 40 as a transmitting device and another terminal device 40 as a receiving device.

[0338] In this embodiment, the first communication device sends information about an AI / ML model applied to at least one of the following processes (P1) to (P4) (first sending process, first receiving process, second sending process and second receiving process) to the second communication device.

[0339] (P1) Regarding the transmission process of the wireless signal from the first communication device to the second communication device (first transmission process)

[0340] (P2) Regarding the reception processing of wireless signals transmitted from the first communication device to the second communication device (first reception processing)

[0341] (P3) Regarding the transmission processing of wireless signals from the second communication device to the first communication device (second transmission processing)

[0342] (P4) Regarding the reception processing of wireless signals transmitted from the second communication device to the first communication device (second reception processing)

[0343] The transmission of information about the AI / ML model can be performed by the transmission unit of the communication device. Alternatively, the transmission of information about the AI / ML model can be performed by the control unit of the communication device that controls the transmission unit. For example, the transmission of information about the AI / ML model can be performed by at least one of the transmission unit 234 of base station 20, the transmission unit 334 of relay station 30, and the transmission unit 434 of terminal device 40. Of course, the transmission of information about the AI / ML model can also be performed by management device 10 or by server equipment on the Internet.

[0344] The second communication device obtains information about the AI / ML model from the first communication device. Then, the second communication device uses or manages the AI / ML model based on this information. For example, suppose the first communication device is the control entity of the LCM, and the second communication device is the controlled object of the LCM. In this case, the second communication device can perform processing related to its signal processing (first receiving processing or second transmitting processing) based on the information about the AI / ML model. Furthermore, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. In this case, the second communication device can perform LCM related to the first communication device's signal processing (first transmitting processing or second receiving processing) based on the information about the AI / ML model.

[0345] The acquisition of information about the AI / ML model can be performed by the acquisition unit of the communication device. Alternatively, the acquisition of information about the AI / ML model can be performed by the control unit of the communication device that controls the acquisition unit. For example, the acquisition of information about the AI / ML model can be performed by at least one of the acquisition unit 231 of base station 20, the acquisition unit 331 of relay station 30, and the acquisition unit 431 of terminal device 40. Of course, the acquisition of information about the AI / ML model can also be performed by management device 10 or by server equipment on the Internet.

[0346] In the above or below description, a communication device that performs transmission processing (first transmission processing / second transmission processing) may be referred to as a transmitting device, and a communication device that performs reception processing (first reception processing / second reception processing) may be referred to as a receiving device.

[0347] Here, the information about the AI / ML model can be information about LCM (information about AI / ML model management). The following description of the operation of communication system 1 will assume that the information about the AI / ML model is information about LCM (information about AI / ML model management).

[0348] <4-1. Control Entities for AI / ML Model Management>

[0349] First, the control entity of LCM (AI / ML model management) will be described.

[0350] The control entity of the LCM can be either a first communication device or a second communication device. In this case, the control entity of the LCM can be either the first or second communication device that acts as a transmitting device, or it can be either the first or second communication device that acts as a receiving device. Furthermore, the control entity of the LCM can be either the first or the second communication device.

[0351] Furthermore, the control entity of the LCM can be a communication device (hereinafter referred to as the communication control device) that controls at least one of the transmitting and receiving devices. For example, in the case of sidelink communication where terminal device 40 is the transmitting device and another terminal device 40 is the receiving device, the control entity of the LCM can be the base station 20 that controls the sidelink communication. Note that the communication between the transmitting and receiving devices is not limited to sidelink communication. The communication between the transmitting and receiving devices can be uplink communication or downlink communication. In addition, the communication control device that becomes the control entity of the LCM is not limited to the base station 20. For example, the communication control device can be the management device 10, the relay station 30, or a server device on the Internet. In addition, the communication control device can also be the terminal device 40. For example, the communication control device can be the master terminal (master terminal) that makes at least one of the transmitting and receiving devices a secondary terminal (slave terminal) (e.g., the master terminal device 40 (master terminal device 40) that controls the sidelink communication between the secondary terminals (slave terminals)).

[0352] <4-1-1. Specific Examples of Control Entities in LCM>

[0353] The wireless communication using the AI / ML model (wireless communication between the first and second communication devices (or related processing)) can be uplink communication (or related processing) with the terminal device 40 as the transmitting device and the base station 20 as the receiving device, or downlink communication (or related processing) with the base station 20 as the transmitting device and the terminal device 40 as the receiving device. The control entity of the LCM in the uplink / downlink communication can be any of the following (A1-1) to (A1-3).

[0354] (A1-1) Base Station 20

[0355] (A1-2) Terminal equipment 40

[0356] (A1-3) Both base station 20 and terminal equipment 40

[0357] Note that when both base station 20 and terminal device 40 become control entities (in the cases described above (A1-3)), base station 20 and terminal device 40 can cooperatively control the LCM, or base station 20 and terminal device 40 can independently control the LCM. Note that the control entity for the LCM is not limited to base station 20 and / or terminal device 40. The control entity for the LCM in uplink / downlink communication can be a communication control device (e.g., management device 10 or a server device on the Internet) that controls base station 20 and / or terminal device 40.

[0358] Furthermore, the wireless communication between the first and second communication devices can be sidelink communication, where terminal device 40 acts as the transmitter of sidelink data, and the other terminal device 40 acts as the receiver of sidelink data. In this case, the control entity of the LCM can be any one of the following (A2-1) to (A2-3).

[0359] (A2-1) A communication control device that controls at least one of the transmitting and receiving devices.

[0360] (A2-2) becomes the terminal device of the transmitting device 40

[0361] (A2-3) Becomes the terminal equipment of the receiving device 40

[0362] (A2-4) Both the terminal device 40 that becomes a transmitting device and the terminal device 40 that becomes a receiving device

[0363] Here, the communication control device in (A2-1) can be a base station 20 that controls the sidelink communication between the transmitting device and the receiving device, or it can be a master terminal (master terminal) that makes at least one of the transmitting device and the receiving device a secondary terminal (slave terminal). Furthermore, when both the terminal device 40 that becomes the transmitting device and the terminal device 40 that becomes the receiving device become control entities (in the case of (A2-4) above), the terminal device 40 that becomes the transmitting device and the terminal device 40 that becomes the receiving device can cooperate in controlling the LCM, or the terminal device 40 that becomes the transmitting device and the terminal device 40 that becomes the receiving device can independently control the LCM.

[0364] Figure 10 The LCM diagram shown can be implemented by either or both of the first and second communication devices. In other words, either the first or second communication device can implement the elements (processes / methods / procedures / functions) shown above (M1) to (M25), as well as... Figure 10 At least a portion of the operations defined in the diagram shown.

[0365] <4-1-2. Determination of the Control Entity of LCM>

[0366] Communication system 1 can identify (distinguish) the control entity of LCM.

[0367] The determination of the control entity of the LCM can be performed by at least one of the first and second communication devices. For example, the communication device that acts as the transmitting device in the first and second communication devices can determine the control entity of the LCM, or the communication device that acts as the receiving device can determine the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), then the base station 20 can determine the control entity of the LCM, or the terminal device 40 can determine the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), then the terminal device 40 that acts as the transmitting device can determine the control entity of the LCM, or the terminal device 40 that acts as the receiving device can determine the control entity of the LCM.

[0368] Furthermore, both the first communication device and the second communication device can determine the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), both the base station 20 and the terminal device 40 can determine the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), both the terminal device 40 acting as the transmitting device and the terminal device 40 acting as the receiving device can determine the control entity of the LCM.

[0369] Furthermore, a third device, distinct from both the transmitting and receiving devices, can determine the control entity of the LCM. For example, a communication control device that controls at least one of the transmitting and receiving devices can determine the control entity of the LCM. When the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), the device constituting the core network (e.g., management device 10) can determine the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), another device within the network coverage area (base station 20 or main terminal device 40 (main terminal device 40)) can determine the control entity of the LCM.

[0370] In the following text, the device that determines (distinguishes) the control entity of the LCM may be referred to as the determining device. As mentioned above, the determining device may be either the first communication device or the second communication device, or both. Furthermore, the determining device may be a communication device other than the first communication device or the second communication device.

[0371] Note that if the device is determined to be base station 20, the determination of the control entity of the LCM can be performed by the determination unit 233 of base station 20. If the device is determined to be relay station 30, the determination of the control entity of the LCM can be performed by the determination unit 333 of relay station 30. If the device is determined to be terminal device 40, the determination of the control entity of the LCM can be performed by the determination unit 433 of terminal device 40. Of course, the determined device is not limited to any one of base station 20, relay station 30, and terminal device 40. The determined device can be management device 10 or server device on the Internet.

[0372] The device can determine (distinguish) the control entity of the LCM based on at least one of the following determination methods (B1) to (B5). The means for determining the control entity of the LCM will be described below.

[0373] (B1) Determination based on standards

[0374] The determining device can identify the control entity of the LCM based on specifications (e.g., rules pre-shared between communication devices). These specifications can be technical standards defined by 3GPP (registered trademark). If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), the determining device can identify one or both of base station 20 and terminal device 40 as the control entity of the LCM based on the specifications. Furthermore, if the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the determining device can identify one or both of terminal device 40, which will become a transmitting device, and terminal device 40, which will become a receiving device, as the control entity of the LCM based on the specifications.

[0375] (B2) Determination based on configuration information

[0376] The device can determine the control entity of the LCM based on pre-stored configuration information (e.g., information pre-configured in the storage unit of the communication device). Here, the pre-configured information in the device can be information stored in the storage unit of the device when it is shipped from the factory. Note that the pre-configured LCM does not necessarily have to be an LCM. The pre-configured LCM can be only a part of the LCM (e.g., a part of (M1) to (M25) above).

[0377] (B3) Determination based on control information

[0378] The determining device can determine the control entity of the LCM based on control information received from other communication devices (including control information about the control entity of the LCM). When the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), the base station 20 can determine the control entity of the LCM based on the control information received from the terminal device 40, or the terminal device 40 can determine the control entity of the LCM based on the control information received from the base station 20. Furthermore, when the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), one or both of the terminal device 40 acting as the transmitting device and the terminal device 40 acting as the receiving device can determine the control entity of the LCM based on control information communicated between the terminal devices 40. The control information here can be at least one of RRC messages, MAC CE, and DCI.

[0379] (B4) Priority-based determination

[0380] Priorities can be provided to determine which AI / ML model from the first and second communication devices should be used preferentially. Then, the device that determines which communication device, between the first and second communication devices, holds the AI / ML model with the higher priority, can be identified as the control entity for the LCM. This will be explained using the determination of the control entity for model transmission as an example.

[0381] For example, suppose a rule is established that prioritizes the use of AI / ML models from communication devices that already possess trained AI / ML models. In this case, the determining device can identify the communication device with the trained AI / ML model among the first and second communication devices as the control entity of the LCM. If the wireless communication applying the AI / ML model (or its associated processing) is uplink / downlink communication (or its associated processing), then the communication device with the trained AI / ML model among the base station 20 and terminal device 40 can be identified as the control entity of the LCM. If the wireless communication applying the AI / ML model (or its associated processing) is sidelink communication (or its associated processing), then the terminal device 40 with the trained AI / ML model among the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device can be identified as the control entity of the LCM. The communication device identified as the control entity of the LCM (the communication device with the trained AI / ML model) can transmit the trained AI / ML model to communication devices that do not possess it.

[0382] For example, suppose a rule is established that prioritizes the use of AI / ML models from communication devices, where each communication device possesses a trained AI / ML model (or functionality) supported by both a first and a second communication device. In this case, the determining device can distinguish between the AI / ML models (or functionalities) supported by both the first and second communication devices based on capability information of the first and second communication devices. Then, the determining device can identify the communication device with the distinguished trained AI / ML model (or functionality) among the first and second communication devices as the control entity for the LCM. If the wireless communication applying the AI / ML model (or its associated processing) is uplink / downlink communication (or its associated processing), then the communication device with the trained AI / ML model in base station 20 and terminal device 40 can be identified as the control entity for the LCM. If the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), then among the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device, the terminal device 40 with the trained AI / ML model can be identified as the control entity of the LCM. The communication device identified as the control entity of the LCM (the communication device with the trained AI / ML model) can transmit the trained AI / ML model to communication devices that do not possess it.

[0383] (B5) Method for determining model type

[0384] The control entity of the LCM can be determined based on the type of AI / ML model applied to wireless communication. The type of AI / ML model can include at least one of a one-sided model and a two-sided model.

[0385] A one-sided model is the case where the AI / ML model is applied to either the first communication device or the second communication device. In the case where the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), in the one-sided model, the AI / ML model is applied to either the base station 20 or the terminal device 40. In the case where the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), the AI / ML model is applied to either the terminal device 40 that acts as a transmitting device or the terminal device 40 that acts as a receiving device. The one-sided model can be further described as a one-sided AI / ML model.

[0386] In the case of a one-sided model, the communication device with the AI / ML model among the first and second communication devices can become the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), then the communication device with the AI / ML model among the base station 20 and terminal device 40 can become the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the communication device with the AI / ML model in the terminal device 40 that is the transmitting device and the terminal device 40 that is the receiving device can become the control entity of the LCM.

[0387] The two-sided model refers to the application of an AI / ML model to both the first and second communication devices. When the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the two-sided model applies the AI / ML model to both the base station 20 and the terminal device 40. When the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), the two-sided model applies the AI / ML model to both the terminal device 40 acting as the transmitting device and the terminal device 40 acting as the receiving device. The two-sided model can be further described as a two-sided AI / ML model.

[0388] In the case of a two-sided model, the communication device with the AI / ML model among the first and second communication devices can become the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), either the base station 20 or the terminal device 40 can become the control entity of the LCM, or both the base station 20 and the terminal device 40 can become the control entity of the LCM. If the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), either the terminal device 40 that acts as a transmitting device or the terminal device 40 that acts as a receiving device can become the control entity of the LCM, or both the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device can become the control entity of the LCM.

[0389] <4-1-3. Case where one of the communication devices becomes the control entity of the LCM>

[0390] First, the operation of the control entity of the LCM will be described when either the first communication device or the second communication device becomes the control entity of the LCM.

[0391] The control entity of the LCM performs control related to the application of the AI / ML model (hereinafter also referred to as LCM control). For example, either the first communication device or the second communication device becomes the control entity of the LCM and performs LCM control of the other communication device (hereinafter also referred to as the controlled object). For example, if the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), one of the base station 20 and the terminal device 40 becomes the control entity of the LCM and performs LCM control of the other communication device. In the case where the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), either the terminal device 40 that is the transmitting device or the terminal device 40 that is the receiving device becomes the control entity of the LCM and performs LCM control of the other communication device.

[0392] Specifically, the control entity of the LCM may perform at least one of the following processes (C1) to (C10).

[0393] (C1) Model Transfer

[0394] When LCM control is model transmission, the LCM control entity sends data of the AI / ML model used for transmission or reception processing as information about the AI / ML model to the communication device (e.g., at least one of a first and a second communication device) that becomes the controlled object. For example, suppose the LCM control entity is the first communication device. In this case, the first communication device sends AI / ML model data for at least one of the first reception processing and the second transmission processing described above to the second communication device. The communication device that becomes the controlled object (e.g., the second communication device) performs at least one of the first reception processing and the second transmission processing using the received AI / ML model data.

[0395] For example, assuming the wireless communication (or related processing) using an AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, base station 20 sends data from its AI / ML model to terminal device 40 before or during the transmission of downlink data, or before receiving uplink data. Terminal device 40 uses the received AI / ML model data itself or through fine-tuning to receive downlink data or transmit uplink data.

[0396] For example, assuming the wireless communication (or related processing) using an AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is terminal device 40. In this case, terminal device 40 transmits data from its AI / ML model to base station 20 before receiving downlink data, or before or during transmitting uplink data. Base station 20 uses the received AI / ML model data itself or through fine-tuning for transmitting downlink data or receiving uplink data.

[0397] For example, assuming the wireless communication (or related processing) using an AI / ML model is sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which becomes the transmitting device. In this case, before or during the transmission of sidelink data, the terminal device 40, which becomes the transmitting device, sends the data of the AI / ML model it possesses to the terminal device 40, which becomes the receiving device. The terminal device 40, which becomes the receiving device, receives the sidelink data using the received AI / ML model data itself or using fine-tuned AI / ML model data.

[0398] For example, assuming the wireless communication (or related processing) using an AI / ML model is sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which becomes the receiving device. In this case, before or during the transmission of sidelink data, the terminal device 40, which becomes the transmitting device, sends the AI / ML model data possessed by the terminal device 40, which becomes the transmitting device, to the terminal device 40, which becomes the receiving device. The terminal device 40, which becomes the receiving device, uses the received AI / ML model data itself, or uses fine-tuned AI / ML model data, to receive the sidelink data.

[0399] The AI / ML model data sent by the communication device that becomes the control entity of the LCM need not be data that can be directly used by the communication device that becomes the control object it processes. For example, suppose the control entity of the LCM is a first communication device. In this case, the first communication device may transmit AI / ML model data for the processes performed by the first communication device (at least one of the first transmission process and the second reception process described above), rather than AI / ML model data directly for the processes performed by the second communication device (at least one of the first reception process and the second transmission process described above). In this case, the second communication device may generate an AI / ML model for its own use based on the AI / ML model data received from the first communication device.

[0400] (C2) Data Collection

[0401] When LCM control involves data collection, the LCM control entity sends control information, instructing the execution of data collection, as information about the AI / ML model to the communication device that becomes the controlled object (e.g., at least one of a first and a second communication device). For example, the LCM control entity sends control information to the communication device that becomes the controlled object, instructing the collection of data for model inference and / or model training. The communication device that becomes the controlled object performs data collection based on the control information.

[0402] For example, assuming the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, base station 20 sends control information to terminal device 40 instructing it to perform data collection. Terminal device 40 performs data collection based on the received control information.

[0403] For example, assuming the wireless communication (or related processing) using an AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is terminal device 40. In this case, terminal device 40 sends control information to base station 20 instructing it to perform data collection. Base station 20 performs data collection based on the received control information.

[0404] For example, assuming the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, sends control information to the terminal device 40, which acts as the receiving device, instructing it to perform data collection. The terminal device 40, acting as the receiving device, performs data collection based on the received control information.

[0405] For example, assuming the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which acts as the receiving device. In this case, the terminal device 40, acting as the receiving device, sends control information to the terminal device 40, which acts as the transmitting device, instructing it to perform data collection. The terminal device 40, acting as the transmitting device, performs data collection based on the received control information.

[0406] (C3) Function Identification

[0407] When LCM control is function identification, the LCM control entity notifies the control entity of information about the functionality of the communication device (e.g., at least one of a first and a second communication device) to which the AI / ML model is applied, as information about the AI / ML model. Here, the information about the functionality to which the AI / ML model is applied can be a functionality ID (including a list of IDs). The communication device to which the control is applied determines, based on the information related to the functionality to which the AI / ML model is applied, to apply the AI / ML model to that functionality.

[0408] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is base station 20. In this case, base station 20 notifies terminal device 40 of information regarding the functionality of applying the AI / ML model. Terminal device 40 determines, based on the received information, to apply the AI / ML model to the functionality related to the notification.

[0409] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is terminal device 40. In this case, terminal device 40 notifies base station 20 of information regarding the functionality of applying the AI / ML model. Base station 20 determines, based on the received information, to apply the AI / ML model to the functionality related to the notification.

[0410] For example, assuming the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, notifies the receiving device of information regarding the functionality of applying the AI / ML model. Based on the received information, the receiving device 40 determines how to apply the AI / ML model to the functionality related to the notification.

[0411] For example, assuming the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which acts as the receiving device. In this case, the terminal device 40, acting as the receiving device, notifies the transmitting terminal device 40 of information regarding the functionality of applying the AI / ML model. Based on the received information, the transmitting terminal device 40 determines how to apply the AI / ML model to the functionality related to the notification.

[0412] As described above, the communication device that becomes the controlled object determines to apply the AI / ML model to the notification-related functionality. In this case, the communication device that becomes the controlled object can always apply the AI / ML model to the notification-related functionality. Furthermore, the communication device that becomes the controlled object can apply the AI / ML model to the notification-related functionality when predetermined conditions are met. For example, the communication device that becomes the controlled object can combine a function identification operation and a model activation operation to make a determination related to the application of the AI / ML model. For example, when the control entity of the LCM activates an AI / ML model applicable to the notification-related functionality, the communication device that becomes the controlled object can apply the AI / ML model to that functionality.

[0413] (C4) Model Recognition

[0414] During LCM control and model identification, the control entity notifies the controlled communication device (e.g., at least one of a first and a second communication device) of information related to the AI / ML model applied by the controlled communication device as AI / ML model-related information. Here, the information regarding the AI / ML model to be applied by the controlled communication device may be an AI / ML model ID (including a list of IDs). Based on the received information, the controlled communication device determines the processing (e.g., sending and / or receiving) to be applied to itself using the AI / ML model.

[0415] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is base station 20. In this case, base station 20 notifies terminal device 40 of information about the AI / ML model to be applied. Based on the received information, terminal device 40 determines to apply the AI / ML model to the processing performed by itself.

[0416] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is terminal device 40. In this case, terminal device 40 notifies base station 20 of information about the AI / ML model that will be applied by base station 20. Based on the received information, base station 20 determines which AI / ML model to apply to the processing performed by itself.

[0417] For example, assuming the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the control entity of the LCM is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, notifies the receiving device 40 of the AI / ML model applied by the receiving device. The receiving device 40 then determines the AI / ML model to apply to the processing it performs based on the received information.

[0418] For example, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), and the control entity of the LCM is the terminal device 40 that becomes the receiving device. In this case, the terminal device 40 that becomes the receiving device notifies the terminal device 40 that the AI / ML model applied by the terminal device 40 that becomes the transmitting device is applying the AI / ML model. Based on the received information, the terminal device 40 that becomes the transmitting device determines the AI / ML model to apply to the processing it will perform.

[0419] As described above, when a communication device becoming the controlled object receives a notification from the LCM control entity, it determines to apply the AI / ML model to the processing it performs. In this case, the communication device becoming the controlled object can always apply the AI / ML model associated with the notification to the processing it performs. Furthermore, when predetermined conditions are met after receiving a notification from the LCM control entity, the communication device becoming the controlled object can apply the AI / ML model associated with that notification. For example, the communication device becoming the controlled object can combine model recognition operations and model activation operations to make a determination related to the application of the AI / ML model. For example, when the LCM control entity activates the AI / ML model associated with the notification, the communication device becoming the controlled object can apply the AI / ML model to the processing it performs.

[0420] (C5) Model Activation

[0421] When LCM control is model activation, the LCM control entity sends control information instructing the activation of the AI / ML model used for transmit or receive processing as information about the AI / ML model to the communication device that is the object of control (e.g., at least one of a first and a second communication device). For example, suppose the LCM control entity is the first communication device. In this case, the first communication device sends control information to the second communication device, which instructs the activation of the AI / ML model for at least one of the first receive processing and the second transmit processing described above. The communication device that is the object of control (e.g., the second communication device) performs the activation of the AI / ML model associated with the instruction based on the received control information.

[0422] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40, instructing the activation of the AI / ML model, and terminal device 40 performs the activation of the AI / ML model associated with the instruction based on the received control information.

[0423] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20 instructing the activation of the AI / ML model. Base station 20 then performs the activation of the AI / ML model associated with the received control information.

[0424] For example, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), and the LCM control entity is the terminal device 40 acting as the transmitting device. In this case, the terminal device 40 acting as the transmitting device sends control information to the terminal device 40 acting as the receiving device, which instructs the activation of the AI / ML model. Based on the received control information, the terminal device 40 acting as the receiving device executes the activation of the AI / ML model associated with the instruction.

[0425] For example, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), and the LCM control entity is terminal device 40 acting as a receiving device. In this case, terminal device 40 acting as a receiving device sends control information to terminal device 40 acting as a transmitting device, which instructs the activation of the AI / ML model. Based on the received control information, terminal device 40 acting as a transmitting device performs the activation of the AI / ML model associated with the instruction.

[0426] Note that the control information instructing AI / ML model activation may include information (e.g., a timer value) specifying the timing of AI / ML model activation for the communication device being controlled. For example, when the communication device being controlled receives a timer value, it can start a timer with that timer value. The communication device being controlled can then perform AI / ML model activation at the timing specified in the information (e.g., when the timer expires). Note that the information specifying the timing of AI / ML model activation may be information indicating the time period (window, duration, or period) until AI / ML model activation is performed, or it may be information indicating the period (periodic or cyclical) for performing AI / ML model activation.

[0427] Additionally or alternatively, the control information indicating AI / ML model activation may include information (e.g., a timer value) specifying the effective period of AI / ML model activation for the communication device being controlled. For example, when the communication device being controlled receives a timer value, it can start a timer with that timer value. The communication device being controlled can then perform deactivation of the AI / ML model when the effective period specified by the information expires (e.g., when the timer expires). Note that the information specifying the effective period of AI / ML model activation may be indicated by a window, duration, or time period, or it may be information indicating the period (periodic or cyclical) of the effective period of AI / ML model activation. Furthermore, deactivation here may be AI / ML model rollback, as described later.

[0428] (C6) model deactivation

[0429] When LCM control involves model deactivation, the LCM control entity sends control information instructing the deactivation of the AI / ML model used for transmit or receive processing as information about the AI / ML model to the communication device being controlled (e.g., at least one of a first and a second communication device). For example, suppose the LCM control entity is the first communication device. In this case, the first communication device sends control information to the second communication device, instructing the deactivation of the AI / ML model used for at least one of the first receive processing and the second transmit processing described above. The communication device being controlled (e.g., the second communication device) performs the deactivation of the AI / ML model associated with the instruction based on the received control information.

[0430] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40, instructing the deactivation of the AI / ML model. Terminal device 40 performs the deactivation of the AI / ML model associated with the received control information.

[0431] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20, which instructs the deactivation of the AI / ML model. Base station 20 performs the deactivation of the AI / ML model associated with the instruction based on the received control information.

[0432] For example, suppose the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), and the LCM control entity is the terminal device 40 acting as the transmitting device. In this case, the terminal device 40 acting as the transmitting device sends control information to the terminal device 40 acting as the receiving device, which instructs the deactivation of the AI / ML model. Based on the received control information, the terminal device 40 acting as the receiving device performs the deactivation of the AI / ML model associated with the instruction.

[0433] For example, suppose the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), and the LCM control entity is terminal device 40 acting as a receiving device. In this case, terminal device 40 acting as a receiving device sends control information to terminal device 40 acting as a transmitting device, which instructs the deactivation of the AI / ML model. Based on the received control information, terminal device 40 acting as a transmitting device performs the deactivation of the AI / ML model associated with the instruction.

[0434] Note that the control information instructing the deactivation of the AI / ML model may include information (e.g., a timer value) for the communication device being controlled, specifying the timing for AI / ML model deactivation. For example, when the communication device being controlled receives a timer value, it can start a timer with that timer value. The communication device being controlled can then perform AI / ML model deactivation at the timing specified by the information (e.g., when the timer expires). The information used to specify the timing for AI / ML model deactivation may be information indicating the time period (window, duration, or time interval) until AI / ML model deactivation is performed, or it may be information indicating the period (periodicity or cycle) for performing AI / ML model deactivation. Furthermore, deactivation here may be AI / ML model rollback, as described later.

[0435] (C7) Model Switching

[0436] When LCM control involves model switching, the LCM control entity sends control information, indicating the switching of the AI / ML model for sending or receiving processing, as information about the AI / ML model to the communication device being controlled (e.g., at least one of a first and a second communication device). Based on the received control information, the communication device being controlled switches the currently used AI / ML model to a different AI / ML model.

[0437] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40, instructing it to switch the AI / ML model. Based on the received control information, terminal device 40 switches the currently used AI / ML model to a different AI / ML model.

[0438] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20, instructing the switching of the AI / ML model. Based on the received control information, base station 20 switches the currently used AI / ML model to a different AI / ML model.

[0439] For example, suppose the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), and the LCM control entity is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, sends control information to the terminal device 40, which acts as the receiving device, instructing it to switch the AI / ML model. Based on the received control information, the terminal device 40, acting as the receiving device, switches the currently used AI / ML model to a different AI / ML model.

[0440] For example, suppose the wireless communication (or related processing) using the AI / ML model is a sidelink communication (or related processing), and the LCM control entity is the terminal device 40 acting as the receiving device. In this case, the terminal device 40 acting as the receiving device sends control information to the terminal device 40 acting as the transmitting device, instructing it to switch the AI / ML model. Based on the received control information, the terminal device 40 acting as the transmitting device switches the currently used AI / ML model to a different AI / ML model.

[0441] Note that model switching operations can be achieved through a combination of model activation and model deactivation operations.

[0442] Additionally, when a communication device becomes the controlled object, it can perform a model switching operation upon receiving model switching control information, provided that other active AI / ML models exist. Then, when the controlled communication device receives model switching control information again, it can choose not to perform a model switching operation if no other active AI / ML models exist. When the controlled communication device does not perform a model switching operation, it can notify the LCM control entity of this decision.

[0443] Additionally or alternatively, the control information indicating the switching of the AI / ML model for transmitting or receiving processing may include information indicating the timing for performing the AI / ML model switching (e.g., subframe, radio frame, SFN, time slot, symbol). The information indicating the timing for performing the AI / ML model switching may include information indicating the time period from receiving control information to the timing of performing the AI / ML model switching. For example, when the timing for receiving control information is n (e.g., the nth subframe, the nth radio frame, the nth SFN, the nth time slot, the nth symbol, where n is an integer), the timing for performing the AI / ML model switching may be indicated by n+k (e.g., the (n+k)th subframe, the (n+k)th radio frame, the (n+k)th SFN, the (n+k)th time slot, the (n+k)th symbol, where n and k are integers). The time period from receiving control information to the timing of performing the AI / ML model switching may be indicated by k.

[0444] (C8) Model rollback

[0445] When LCM control is a model rollback, the LCM control entity sends control information as information about the AI / ML model to the communication device that is the controlled object (e.g., at least one of a first and a second communication device). This control information instructs the rollback of the AI / ML model used for sending or receiving processing. For example, suppose the LCM control entity is the first communication device. In this case, the first communication device sends control information to the second communication device to cause at least one of the first receiving process and the second sending process to roll back from a process using the AI / ML model to a process not using the AI / ML model. Based on the received control information, the communication device that is the controlled object (e.g., the second communication device) rolls back the AI / ML model application process to a process not using the AI / ML model. For example, the second communication device rolls back at least one of the first receiving process and the second sending process from a process using the AI / ML model to a process not using the AI / ML model.

[0446] In the following description, processing without using an AI / ML model may be referred to as conventional processing or conventional signal processing. Alternatively, processing without using an AI / ML model may be referred to as non-AI / ML-based processing. Note that processing without using an AI / ML model is not limited to existing signal processing methods. The descriptions of "conventional processing" or "conventional signal processing" appearing in the following descriptions may be replaced with "processing without using an AI / ML model".

[0447] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40 instructing AI / ML model rollback. Terminal device 40 performs a rollback from AI / ML model application processing to regular processing based on the received control information.

[0448] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20 instructing AI / ML model rollback. Based on the received control information, base station 20 performs a rollback of the AI / ML model application processing to regular processing.

[0449] For example, suppose the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), and the LCM control entity is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, sends control information to the terminal device 40, which acts as the receiving device, instructing the AI / ML model to roll back. Based on the received control information, the terminal device 40, acting as the receiving device, performs a rollback of the AI / ML model application processing to regular processing.

[0450] For example, suppose the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), and the LCM control entity is the terminal device 40 acting as the receiving device. In this case, the terminal device 40 acting as the receiving device sends control information to the terminal device 40 acting as the transmitting device, instructing the AI / ML model to roll back. Based on the received control information, the terminal device 40 acting as the transmitting device executes the rollback of the AI / ML model application processing to regular processing.

[0451] Note that the LCM control entity can instruct rollback only for certain processes (e.g., functional processes) that apply the AI / ML model. For example, suppose a second communication device has multiple functionalities associated with at least one of the first receiving process and the second transmitting process described above. The first communication device can send information to the second communication device as control information (information about the AI / ML model) to cause the processing of some of the multiple functionalities to roll back from processing that uses the AI / ML model to processing that does not use the AI / ML model. The LCM control entity can instruct rollback for all processes.

[0452] When a communication device that becomes the controlled object receives control information from an LCM control entity, it may perform a rollback only for some processes (e.g., functional processes) of the applied AI / ML model, or it may perform a rollback for all processes.

[0453] Note that the control information instructing the rollback of the AI / ML model may include information (e.g., a timer value) for the communication device that is the controlled object, specifying the timing of the AI / ML model rollback. For example, when the communication device that is the controlled object receives a timer value, it can start a timer with that timer value. The communication device that is the controlled object can then perform the rollback of the AI / ML model at the timing specified from the information (e.g., when the timer expires).

[0454] Furthermore, AI / ML model rollback can involve performing a process to revert to the default AI / ML model. For example, a terminal device receiving an AI / ML model rollback notification can change the AI / ML model used for sending or receiving processing from the currently applied AI / ML model to a pre-configured default AI / ML model.

[0455] Additionally or alternatively, the control information indicating the backoff of the AI / ML model for transmitting or receiving processing may include information indicating the timing for performing the AI / ML model backoff (e.g., subframe, radio frame, SFN, time slot, symbol). The information indicating the timing for performing the AI / ML model backoff may include information indicating the time period from receiving the control information to the timing of performing the AI / ML model backoff. For example, when the timing for receiving the control information is n (e.g., the nth subframe, the nth radio frame, the nth SFN, the nth time slot, the nth symbol, where n is an integer), the timing for performing the AI / ML model backoff may be indicated by n + k (e.g., the (n + k)th subframe, the (n + k)th radio frame, the (n + k)th SFN, the (n + k)th time slot, the (n + k)th symbol, where n and k are integers). The time period from receiving the control information to the timing of performing the AI / ML model backoff may be indicated by k.

[0456] (C9) Model Monitoring

[0457] When LCM control is model monitoring, the LCM control entity sends control information, instructing the execution of AI / ML model monitoring, as information about the AI / ML model to the communication device that becomes the controlled object (e.g., at least one of a first communication device and a second communication device). When the communication device that becomes the controlled object receives the control information, it performs the AI / ML model monitoring operation.

[0458] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40 instructing it to perform AI / ML model monitoring. When terminal device 40 receives this control information, it performs the AI / ML model monitoring operation.

[0459] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20 instructing it to perform AI / ML model monitoring. When base station 20 receives this control information, it performs the AI / ML model monitoring operation.

[0460] For example, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), and the LCM control entity is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, sends control information to the terminal device 40, which acts as the receiving device, to instruct it to perform AI / ML model monitoring. When the terminal device 40, acting as the receiving device, receives the control information, it performs the AI / ML model monitoring operation.

[0461] For example, suppose the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), and the LCM control entity is the terminal device 40 acting as the receiving device. In this case, the terminal device 40 acting as the receiving device sends control information to the terminal device 40 acting as the transmitting device to instruct it to perform AI / ML model monitoring. When the terminal device 40 acting as the transmitting device receives the control information, it performs the AI / ML model monitoring operation.

[0462] Note that the communication device being controlled can perform performance measurements of the currently applied AI / ML model as an AI / ML model monitoring operation. AI / ML model performance measurements can be performed for a predetermined period of time or at predetermined intervals. That is, the control information instructing AI / ML model monitoring can be information indicating the period (window, duration, or time interval) for performing AI / ML model performance measurements, or it can be information indicating the cycle (periodicity or periodicity) for performing AI / ML model performance measurements. Furthermore, the control information instructing AI / ML model monitoring can be a combination of these.

[0463] (C10) Model Update

[0464] LCM control involves the LCM control entity sending control information, as information about the AI / ML model, to the communication device (e.g., at least one of a first and a second communication device) that is being controlled during model updates. When the controlled communication device receives the control information, it performs the AI / ML model update operation. For example, when the controlled communication device receives the control information, it performs retraining of the currently used AI / ML model.

[0465] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is base station 20. In this case, base station 20 sends control information to terminal device 40 instructing it to perform an AI / ML model update. When terminal device 40 receives this control information, it performs the AI / ML model update operation.

[0466] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), and the LCM control entity is terminal device 40. In this case, terminal device 40 sends control information to base station 20 instructing it to perform an AI / ML model update. When base station 20 receives this control information, it performs the AI / ML model update operation.

[0467] For example, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), and the LCM control entity is the terminal device 40, which acts as the transmitting device. In this case, the terminal device 40, acting as the transmitting device, sends control information to the terminal device 40, which acts as the receiving device, to instruct the execution of an AI / ML model update. When the terminal device 40, acting as the receiving device, receives the control information, it executes the AI / ML model update operation.

[0468] For example, assuming the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the LCM control entity is terminal device 40, which acts as a receiving device. In this case, terminal device 40, acting as a receiving device, sends control information to terminal device 40, which acts as a transmitting device, to instruct the execution of an AI / ML model update. When terminal device 40, acting as a transmitting device, receives the control information, it executes the AI / ML model update operation.

[0469] Note that a communication device that becomes the controlled object cannot use the AI / ML model during an AI / ML model update operation. Therefore, when a communication device that becomes the controlled object performs an AI / ML model update operation, it can roll back the currently executing processing to regular processing that does not apply the AI / ML model. Alternatively, when a communication device that becomes the controlled object performs an AI / ML model update operation, it can switch the AI / ML model associated with the update operation to a different AI / ML model.

[0470] <4-1-4. Case where both communication devices become control entities of the LCM>

[0471] As described above, both the first and second communication devices can serve as the control entity of the LCM. For example, when the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), both the base station 20 and the terminal device 40 can serve as the control entity of the LCM. When the wireless communication (or related processing) using the AI / ML model is sidelink communication (or related processing), both the terminal device 40, which serves as the transmitting device, and the terminal device 40, which serves as the receiving device, can serve as the control entity of the LCM.

[0472] When both the first and second communication devices become the control entity of the LCM, coordination can be performed in advance between the communication devices to determine which of the first and second communication devices will become the control entity. For example, either the first or second communication device can send information about the control entity as information about the AI / ML model to the other communication device. If the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), either the base station 20 or the terminal device 40 can send information about the control entity to the other communication device. Furthermore, if the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), either the terminal device 40 that is the transmitting device or the terminal device 40 that is the receiving device can send information about the control entity to the other communication device.

[0473] When both the first communication device and the second communication device become the control entity of the LCM, it can be determined which of the first and second communication devices becomes the control entity for each LCM (e.g., for each of (M1) to (M25)). The means for determining which becomes the control entity of each LCM can be either determination means 1 or determination means 2.

[0474] (Method 1)

[0475] The device determines the control entity of the LCM based on static information. For example, the first and second communication devices can determine the control entity of the LCM based on specifications (e.g., rules shared in advance between the communication devices).

[0476] (Method 2)

[0477] The device determines the control entity of the LCM based on semi-static information. For example, in advance, one of the first and second communication devices can determine which LCM it will control and which LCM the other communication device will control. Then, one communication device can notify the other communication device which LCM it will control.

[0478] Note that the determining device can be either the first communication device or the second communication device, or both. Furthermore, the determining device can be a communication device other than the first or second communication device.

[0479] When both the first and second communication devices become control entities of the LCM, some LCM controls included in the LCM can be executed by both the first and second communication devices. For example, both the first and second communication devices can perform model training. In this case, the first and second communication devices can each train an AI / ML model and use the trained model separately.

[0480] <4-1-5 Others>

[0481] LCM control (including the LCMs described above or below) can be performed by management device 10. For example, LCM control can be performed by network functions (LCM-related network functions) included in the core network. LCM-related network functions are network functions that provide LCM-related control functions. LCM-related network functions have functions such as controlling the above (M1) to (M25) or providing the information required to perform the above (M1) to (M25).

[0482] Figure 9 An example of a network configuration including LCM-related network functions is shown. LCM-related network functions can be included in a server (e.g., MEC server 552) connected to UPF 521. Furthermore, LCM-related network functions can be included in a server (e.g., MEC server 553) connected to base station 20 ((R) AN 530). Moreover, LCM-related network functions can be defined as one of the network functions. Then, this network function ( Figure 9The LCM-related network function 554 shown can be included in the core network as one of the new network functions. Furthermore, the LCM-related network function can be included in existing network functions. For example, the functionality of the LCM-related network function can be included in... Figure 9 As shown in MTLF 551, LCM-related network functions can exist in one or more locations.

[0483] Furthermore, LCM control can be performed by management device 10, base station 20, relay station 30, terminal device 40, or other communication devices (e.g., server devices on the Internet). Each of these communication devices can become the control entity of the LCM.

[0484] <4-2. Control methods for AI / ML model management>

[0485] The control entities of LCM have been described above. Next, the control devices of LCM (AI / ML model management) will be described.

[0486] <4-2-1. Wireless Communication Using AI / ML Models>

[0487] When using AI / ML models to implement wireless communication (hereinafter also referred to as AI communication), the control entity of the LCM implements control related to the application of the AI / ML model (hereinafter referred to as LCM control). LCM control can be a function-based LCM or a model ID-based LCM. The LCM control entity may apply only one of the function-based LCM and the model ID-based LCM, or a combination of both.

[0488] (D1) Functional LCM

[0489] Function-based LCM directives enable AI / ML model management based on functionality. For example, function-based LCM directives enable AI / ML model management for each use case and / or function.

[0490] For example, suppose a functionality indicates the communication requirements needed for AI communication in use cases and / or functions that apply an AI / ML model. In this case, the LCM's control entity notifies the communication device that becomes the controlled object of information indicating which functionality's AI communication should be activated. For example, the LCM's control entity determines which functionality's AI communication should be activated and notifies the communication device that becomes the controlled object of the ID of the functionality associated with that determination (hereinafter referred to as the functionality ID).

[0491] (Sending capability information)

[0492] At this point, the terminal device that becomes the controlled object can send capability information as information about the AI / ML model to the communication device that becomes the control entity of the LCM. For example, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. In this case, the first communication device can send information indicating which of a plurality of functionalities it corresponds to in relation to at least one of the first transmission process and the second reception process as capability information to the second communication device. Here, the capability information can be the ID of the supported functionality (functionality ID) or a list of functionality IDs.

[0493] For example, assuming the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, the terminal device 40, which becomes the controlled object, sends capability information indicating which functionality it corresponds to to base station 20.

[0494] For example, assuming the wireless communication (or related processing) using an AI / ML model is sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device. In this case, the terminal device 40 that acts as the control object will send capability information corresponding to which function to the terminal device 40 that acts as the control entity of the LCM.

[0495] (Sending training completion information)

[0496] Furthermore, the terminal device that becomes the controlled object can send training completion information as information about the AI / ML model to the communication device that becomes the control entity of the LCM. For example, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. Also, suppose the AI / ML model is linked to a functionality associated with at least one of the first transmission process and the second reception process. In this case, the first communication device can send information indicating which of the multiple functionalities associated with at least one of the first transmission process and the second reception process has completed training (is in a state where model inference is possible) as training completion information to the second communication device. Here, the training completion information can be the ID of the functionality whose AI / ML model training has been completed (functional ID), or it can be a list of functional IDs.

[0497] For example, suppose the wireless communication (or related processing) that applies the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is base station 20. In this case, the terminal device 40, which becomes the controlled object, sends training completion information to base station 20, indicating which functional AI / ML model has completed training.

[0498] For example, assuming the wireless communication (or related processing) of the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device. In this case, the terminal device 40 that acts as the control object will send training completion information indicating which functional AI / ML model has been trained to the terminal device 40 that acts as the control entity of the LCM.

[0499] (activation)

[0500] The control entity of an LCM determines which functional AI communication to activate based on information received from the controlled object. For example, suppose a second communication device is the control entity of an LCM, and a first communication device is the controlled object of an LCM. The second communication device determines which functional AI communication to activate based on information received from the first communication device (e.g., capability information and / or training completion information).

[0501] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is base station 20. In this case, base station 20 determines which AI communication function to activate based on information received from the terminal device 40, which becomes the controlled object (e.g., capability information and / or training completion information). Base station 20 then notifies the terminal device 40 of the determination result (e.g., the function ID).

[0502] For example, assuming the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 acting as a transmitting device and the terminal device 40 acting as a receiving device. The terminal device 40 acting as the control entity of the LCM determines which AI communication function to activate based on information received from the terminal device 40 acting as the controlled device (e.g., capability information and / or training completion information). Then, the terminal device 40 acting as the control entity of the LCM notifies the terminal device 40 acting as the controlled device of the determination of the result (e.g., the functionality ID).

[0503] In the case of a functional LCM, the communication device that becomes the control entity of the LCM can perform LCM control without receiving a list of AI / ML models, etc., maintained by the communication device that becomes the controlled object. That is, based on information notified by the communication device that becomes the control entity of the LCM, the communication device that becomes the controlled object determines which specific AI / ML model to use.

[0504] (D2) LCM based on model ID

[0505] Model ID-based LCM means AI / ML model management based on AI / ML models.

[0506] For example, suppose a logical ID is assigned to an AI / ML model that can be supported by a communication device (at least one of a first and a second communication device) performing AI communication. In this case, the LCM's control entity will notify the communication device that is the controlled object of information indicating which AI / ML model to activate for AI communication. For example, the LCM's control entity determines which AI / ML model to activate and notifies the controlled object of the ID of the AI / ML model associated with that determination (hereinafter referred to as the model ID). Note that the model ID is not limited to the logical ID. The LCM's control entity may notify the controlled object of the physical ID linked to the AI / ML model separately, rather than the logical ID.

[0507] (Sending capability information)

[0508] The terminal device that becomes the controlled object can send capability information as information about the AI / ML model to the communication device that becomes the control entity of the LCM. For example, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. In this case, the first communication device can send information as capability information to the second communication device indicating which of a plurality of AI / ML models it corresponds to with at least one of the first transmission process and the second reception process. Here, the capability information can be the ID of the supported AI / ML model (model ID) or a list of model IDs.

[0509] For example, assuming the wireless communication (or related processing) using an AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, the terminal device 40, which becomes the controlled entity, sends capability information indicating which AI / ML model it corresponds to to base station 20.

[0510] For example, assuming the wireless communication (or related processing) using an AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device. In this case, the terminal device 40 that acts as the control object will send capability information indicating which AI / ML model it corresponds to to the terminal device 40 that acts as the control entity of the LCM.

[0511] (Sending training completion information)

[0512] Furthermore, the terminal device that becomes the controlled object can send training completion information as information about the AI / ML model to the communication device that becomes the control entity of the LCM. For example, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. In this case, the first communication device can send information indicating which of the multiple AI / ML models associated with at least one of the first and second receiving processes has completed training (is in a state capable of model inference) as training completion information to the second communication device. Here, the training completion information can be the ID of the AI / ML model that has completed training (model ID), or it can be a list of model IDs.

[0513] For example, assuming the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, the terminal device 40, which becomes the controlled entity, sends training completion information to base station 20, indicating which AI / ML model has completed training.

[0514] For example, assuming the wireless communication (or related processing) of the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that becomes the transmitting device and the terminal device 40 that becomes the receiving device. In this case, the terminal device 40 that becomes the controlled object sends training completion information to the terminal device 40 that becomes the control entity of the LCM, indicating which AI / ML model has completed training.

[0515] (activation)

[0516] The control entity of an LCM determines which AI / ML model to activate based on information received from the controlled object. For example, suppose a second communication device is the control entity of an LCM, and a first communication device is the controlled object of an LCM. The second communication device determines which AI / ML model to activate based on information received from the first communication device (e.g., capability information and / or training completion information).

[0517] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), and the control entity for the LCM is base station 20. In this case, base station 20 determines which AI / ML model to activate based on information received from the terminal device 40, which becomes the controlled object (e.g., capability information and / or training completion information). Then, base station 20 notifies the terminal device 40 of the determination result (e.g., model ID).

[0518] For example, assuming the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 acting as a transmitting device and the terminal device 40 acting as a receiving device. The terminal device 40 acting as the control entity of the LCM determines which AI / ML model to activate based on information received from the terminal device 40 acting as the controlled device (e.g., capability information and / or training completion information). Then, the terminal device 40 acting as the control entity of the LCM notifies the terminal device 40 acting as the controlled device of the information indicating the determination result (e.g., model ID).

[0519] <4-2-2. Methods for Sending AI / ML Model Information>

[0520] The communication device can send information about the AI / ML model (e.g., control information for LCM control) as follows.

[0521] (E1) Unicast

[0522] An LCM control entity can control a control object. In this case, the LCM control entity can send information about the AI / ML model (e.g., control information used for LCM control) to a control object.

[0523] For example, assuming the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, base station 20 can send information (control information) about the AI / ML model to a terminal device 40.

[0524] Furthermore, assuming that the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is terminal device 40. In this case, terminal device 40 can send information (control information) about the AI / ML model to a base station 20.

[0525] For example, assuming that the wireless communication (or related processing) applying the AI / ML model is a sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device. The terminal device 40 that acts as the control entity of the LCM can send information (control information) about the AI / ML model to another terminal device 40.

[0526] Furthermore, a communication control device that controls at least one of the first and second communication devices can send information (control information) about the AI / ML model to one of the first and second communication devices. For example, a communication control device that controls at least one of a communication device that acts as a transmitting device and a communication device that acts as a receiving device can send information (control information) about the AI / ML model to one of the transmitting device and the receiving device.

[0527] (E2) group shared

[0528] The control entity of an LCM can control multiple specific control objects. In this case, the LCM control entity can send information about the AI / ML model (e.g., control information for LCM control) to the multiple specific control objects.

[0529] For example, assuming the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is base station 20. In this case, base station 20 can send information (control information) about the AI / ML model to a plurality of specific terminal devices 40.

[0530] Furthermore, assuming that the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing), the control entity for the LCM is the terminal device 40. In this case, the terminal device 40 can send information (control information) about the AI / ML model to multiple specific base stations 20.

[0531] For example, assuming that the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing), the control entity of the LCM is one of the terminal device 40 that acts as a transmitting device and the terminal device 40 that acts as a receiving device. The terminal device 40 that acts as the control entity of the LCM can send information (control information) about the AI / ML model to a plurality of specific terminal devices 40.

[0532] Furthermore, a communication control device that controls at least one of the first and second communication devices can send information (control information) about the AI / ML model to both the first and second communication devices. For example, a communication control device that controls at least one of a communication device that acts as a transmitting device and a communication device that acts as a receiving device can send information (control information) about the AI / ML model to both the transmitting and receiving devices.

[0533] (E3) Broadcast

[0534] The control entity of the LCM can control multiple unspecified control objects. In this case, the LCM control entity can send information about the AI / ML model (e.g., control information for LCM control) to the multiple unspecified control objects. For example, base station 20 can send information about the AI / ML model (control information) to multiple unspecified terminal devices 40, or terminal device 40 can send information about the AI / ML model (control information) to multiple unspecified base stations 20. Terminal device 40 can send information about the AI / ML model (control information) to multiple unspecified terminal devices 40. The communication control device can send information about the AI / ML model (control information) to multiple unspecified communication devices, including at least one of a first communication device and a second communication device.

[0535] (E4) Other

[0536] In (E1) to (E3) above, the communication device that becomes the control entity of the LCM sends information about the AI / ML model. However, the communication device that becomes the control object can also send information about the AI / ML model. In this case, the communication device that becomes the control object can send information about the AI / ML model to one communication device. Moreover, the communication device that becomes the control object can send information about the AI / ML model to multiple communication devices. At this time, the communication device that becomes the control object can send information related to the AI / ML model to multiple specific communication devices, or it can send information related to the AI / ML model to multiple unspecified communication devices.

[0537] <4-2-3. Information Sharing Regarding LCM Control>

[0538] When using AI / ML models to implement wireless communication, the first and second communication devices need to jointly understand which AI / ML model or which function should be implemented for the LCM.

[0539] The following sections describe several examples of information sharing regarding LCM control. Note that the examples shown below (F1) through (F4) can be implemented independently. Alternatively, at least a portion of the examples shown below (F1) through (F4) can be implemented in combination with other portions. Furthermore, at least a portion of the information sent from the first communication device to the second communication device (or from the second communication device to the first communication device) in the examples shown below (F1) through (F4) can be information about the AI / ML model described above or below, or can be included in the information about the AI / ML model described above or below.

[0540] Note that in the following description, it is assumed that the first communication device sends information to the second communication device, but the second communication device may send information to the first communication device. In this case, the first and second communication devices may be base station 20 and terminal device 40, or terminal device 40 and another terminal device 40. Furthermore, a third communication device (e.g., a communication control device), different from the first and second communication devices, may send information to the first and / or second communication devices. In this case, the third communication device may be management device 10 or master terminal device 40.

[0541] The description of "first communication device" in the following description may be replaced with "second communication device" or "third communication device". Furthermore, the description of "second communication device" in the following description may be replaced with "first communication device" or "first communication device and / or second communication device".

[0542] (F1) Information sharing via model ID

[0543] The first communication device notifies the second communication device of the AI / ML model's ID (model ID). For example, base station 20 notifies terminal device 40 of the model ID. Alternatively, one of the terminal device 40 acting as a transmitting device and the terminal device 40 acting as a receiving device notifies the other terminal device 40 of the model ID. The second communication device determines the AI / ML model to use based on the notified model ID.

[0544] The first communication device can notify the ID linked to the AI / ML model as the model ID, or it can notify a logically assigned ID. The first communication device can notify the model ID information as semi-static information, such as system information or RRC signaling. Alternatively, the first communication device can notify the model ID information as dynamic information, such as MAC CE (MAC Control Element), DCI (Downlink Control Information), UCI (Uplink Control Information), or SCI (Sidelink Control Information).

[0545] The second communication device can activate the AI / ML model specified by the notified model ID. Alternatively, the second communication device can save the notification information (e.g., model ID information) as information about available AI / ML models until it receives an activation notification for the AI / ML model specified by the notified model ID.

[0546] The information notified from the first communication device to the second communication device may be, for example, the following notification information (e.g., RRCIE (information element)). The following notification information (e.g., ModelIDConfig) may further include at least a portion of the information about the AI / ML model described above or below. Alternatively, at least a portion of the following notification information may be included in the information about the AI / ML model described above or below. The first communication device may notify the second communication device of at least a portion of the notification information shown below.

[0547] ModelIDConfig::= SEQUENCE { modelId CHOICE { modelID1 SEQUENCE {modelId, numberOfInputNode, numberOfOutputNode, numberOfLayer,informationOfWeight, …}, modelID2 SEQUENCE { modelId, numberOfInputNode,numberOfOutputNode, numberOfLayer, informationOfWeight, …}, …}}

[0548] Here, `modelId` indicates the physical or logical identifier assigned to the AI / ML model. `numberOfInputNode` represents the number of input nodes in the AI / ML model's network structure. `numberOfOutputNode` represents the number of output nodes in the AI / ML model's network structure. `numberOfLayer` represents the number of layers in the AI / ML model's network structure. `informationOfWeight` represents the weights assigned between each node in the AI / ML model's network structure.

[0549] (F2) Information sharing through functional IDs

[0550] The first communication device notifies the second communication device of the functional ID. For example, base station 20 notifies the terminal device 40 of the functional ID. Alternatively, one of the terminal device 40 acting as a transmitting device and the terminal device 40 acting as a receiving device notifies the other terminal device 40 of the functional ID. The second communication device determines the AI / ML model to be used based on the notified functional ID.

[0551] The first communication device can notify the ID linked to the functional model as a functional ID, or it can notify a logically assigned ID. The first communication device can notify the functional ID information as semi-static information, such as system information or RRC signaling. Furthermore, the first communication device can notify the functional ID information as dynamic information, such as MAC CE (MAC Control Element), DCI (Downlink Control Information), UCI (Uplink Control Information), or SCI (Sidelink Control Information).

[0552] The second communication device determines the functionality to which the AI / ML model is applied based on the notified functionality ID. The second communication device can activate the AI / ML model specified by the determined functionality. Note that the second communication device can save the determination information (e.g., model ID information) as information about the available AI / ML models until it receives an activation notification for the determined AI / ML model. Alternatively, the second communication device can retain the notification information (e.g., functionality ID information) until it receives an activation notification for the functionality specified by the notified functionality ID.

[0553] The information notified from the first communication device to the second communication device may be, for example, the following notification information (e.g., RRCIE (information element)). The following notification information (e.g., FunctionalityConfig) may also include at least a portion of the information about the AI / ML model described above or below. Alternatively or additionally, at least a portion of the following notification information may be included in the information about the AI / ML model described above or below. The first communication device may notify the second communication device of at least a portion of the notification information shown below.

[0554] FunctionalityConfig::= SEQUENCE { Functionality CHOICE {functionality1 SEQUENCE { functionalityId, feature, requirementList,capableModelList, …}, functionality2 SEQUENCE { functionalityId, feature,requirementList, capableModelList, …},}}

[0555] Here, `functionalityId` indicates the physical or logical identifier assigned to a functionality. `feature` represents the characteristics that contain that functionality. `requirementList` indicates the required values ​​for that functionality. `capableModelList` indicates information about the AI / ML models that can be applied to that functionality.

[0556] (F3) Information sharing through model configuration

[0557] The first communication device notifies the second communication device of the RRC signaling defined for each AI / ML model. For example, base station 20 notifies terminal device 40 of the RRC signaling defined for each AI / ML model. Alternatively, one of the terminal devices 40 acting as a transmitting device and the other as a receiving device notifies the other terminal device 40 of the RRC signaling defined for each AI / ML model. In this case, the first communication device can perform the notification using different messages, different information fields, or different information elements for each AI / ML model.

[0558] The second communication device distinguishes the notified AI / ML model based on RRC signaling defined for each AI / ML model. The second communication device can distinguish the notified AI / ML model based on which message, information field, or information element it receives. Note that the second communication device can retain distinguishing information (e.g., model ID information) as information about the available AI / ML model until it receives an activation notification for the distinguished AI / ML model.

[0559] The information notified from the first communication device to the second communication device may be, for example, the following notification information (e.g., RRCIE (information element)). The following notification information (e.g., ModelConfigA, ModelConfigB) may further include at least a portion of the information regarding the AI / ML model described above or below. Alternatively, at least a portion of the following notification information may be included in the information regarding the AI / ML model described above or below. The first communication device may notify the second communication device of at least a portion of the notification information shown below. In this case, when notifying ModelConfigA, the second communication device can distinguish that the AI / ML model associated with the notification is the AI / ML model of ModelConfigA. Furthermore, when notifying ModelConfigB, the second communication device can distinguish that the AI / ML model associated with the notification is the AI / ML model of ModelConfigB.

[0560] ModelConfigA::= SEQUENCE {

[0561] numberOfInputNode

[0562] numberOfOutputNode,

[0563] numberOfLayer

[0564] informationOfWeight

[0565] …

[0566] }

[0567] ModelConfigB::= SEQUENCE {

[0568] numberOfInputNode

[0569] numberOfOutputNode,

[0570] numberOfLayer

[0571] informationOfWeight

[0572] …

[0573] }

[0574] Here, `modelId` indicates the physical or logical identifier assigned to the AI / ML model. `numberOfInputNode` represents the number of input nodes in the AI / ML model's network structure. `numberOfOutputNode` represents the number of output nodes in the AI / ML model's network structure. `numberOfLayer` represents the number of layers in the AI / ML model's network structure. `informationOfWeight` represents the weights assigned between each node in the AI / ML model's network structure.

[0575] (F4) Information sharing through function configuration

[0576] The first communication device notifies the second communication device of the RRC signaling defined for each functional ID. For example, base station 20 notifies terminal device 40 of the RRC signaling defined for each functional ID. Alternatively, one of the terminal device 40 acting as a transmitting device and the terminal device 40 acting as a receiving device notifies the other terminal device 40 of the RRC signaling defined for each functional ID. In this case, the first communication device can perform the notification using different messages, different information fields, or different information elements for each functional ID.

[0577] The second communication device distinguishes the functionality to which the AI / ML model is applied based on the notified functionality ID. The second communication device may distinguish the notified functionality (or AI / ML model) based on which message, information field, or information element it receives. The second communication device may activate the AI / ML model specified by the distinguished functionality. Note that the second communication device may retain distinguishing information (e.g., model ID information) as information about the available AI / ML model until it receives an activation notification for the distinguished AI / ML model. Alternatively, the second communication device may retain distinguishing information (e.g., functionality ID information) until it receives an activation notification for the functionality specified by the notified functionality ID.

[0578] The information notified from the first communication device to the second communication device may be, for example, the following notification information (e.g., RRCIE (information element)). The following notification information (e.g., FunctionalityConfigA, FunctionalityConfigB) may further include at least a portion of the information regarding the AI / ML model described above or below. Alternatively or additionally, at least a portion of the following notification information may be included in the information regarding the AI / ML model described above or below. The first communication device may notify the second communication device of at least a portion of the notification information shown below. In this case, when FunctionalityConfigA is notified, the second communication device can distinguish that the functionality (or AI / ML model) associated with the notification is the functionality of FunctionalityConfigA (or the AI / ML model of FunctionalityConfigA). Furthermore, when FunctionalityConfigB is notified, the second communication device can distinguish that the functionality (or AI / ML model) associated with the notification is the functionality of FunctionalityConfigB (or the AI / ML model of FunctionalityConfigB).

[0579] FunctionalityConfigA::= SEQUENCE {

[0580] feature,

[0581] requirementList,

[0582] capableModelList,

[0583] …

[0584] }

[0585] FunctionalityConfigB::= SEQUENCE {

[0586] feature,

[0587] requirementList,

[0588] capableModelList,

[0589] …

[0590] }

[0591] Here, "features" represents features that include functionality. "requirementList" indicates the required values ​​for the functionality. "capableModelList" indicates information about AI / ML models that can be applied to this functionality.

[0592] <4-2-4. LCM Execution Timing>

[0593] In LCM (AI / ML model management) control, either event-triggered control or period-based control can be introduced.

[0594] Note that in the following description, it is assumed that the first communication device performs LCM control, but the second communication device can also perform LCM control. In this case, the first communication device can be base station 20 or terminal device 40. Furthermore, a third communication device (e.g., a communication control device), different from the first and second communication devices, can perform LCM control. In this case, the third communication device can be management device 10 or main terminal device 40 (main terminal device 40).

[0595] The description of "first communication device" in the following description may be replaced with "second communication device" or "third communication device". Furthermore, "first communication device" or "first communication device and / or second communication device" may be used instead of the description of "second communication device" or "third communication device" in the following description.

[0596] (G1) Example of event-triggered control

[0597] The first communication device can implement LCM control at a time when a predetermined event occurs. For example, the first communication device can send information about the AI / ML model to other communication devices (e.g., a second communication device and / or a third communication device) at the time when the predetermined event occurs. Additionally, the first communication device can execute one or more of the above (M1) to (M25) at the time when the predetermined event occurs. Note that the time when the predetermined event occurs can be the time when the first communication device receives a predetermined notification from another communication device. For example, the time when the predetermined event occurs can be the time when the first communication device receives a predetermined notification sent by another communication device at the time the predetermined event occurs.

[0598] The timing for the predetermined event to occur can be at least one of the timings shown below. Note that the timings shown below are merely examples. The timing for the predetermined event to occur is not limited to the following. Furthermore, information instructing the first communication device from other communication devices (a second communication device and / or a third communication device) to apply which of the following timings or events is appropriate.

[0599] ● Timing when AI / ML model training is complete

[0600] ● Timing when AI / ML model training begins

[0601] ● Timing when the AI / ML model's validity period expires (timer expires, etc.)

[0602] ● Timing for determining the validity of an AI / ML model as invalid

[0603] ● Timing when it is determined that the AI / ML model needs to be updated

[0604] ● Timing when the training data expires (timer expires, etc.)

[0605] ● Timing during AI / ML model download

[0606] ● Timing when AI / ML models are uploaded

[0607] ● Implement timing during AI / ML model transmission

[0608] ● Implement timing during AI / ML model inference

[0609] ● Implement timing for AI / ML model activation

[0610] ● Timing when deactivating AI / ML models

[0611] ● Timing when switching AI / ML models

[0612] ● Timing when the AI / ML model is rolled back and switched to regular processing without using the AI / ML model.

[0613] (G2) Example of periodic-based control

[0614] The first communication device can perform LCM control at a predetermined time. For example, the first communication device can send information about the AI / ML model to other communication devices (e.g., the second and / or the third communication device) at the predetermined time. In addition, the first communication device can execute one or more of the above (M1) to (M25) at the predetermined time.

[0615] Note that the timing for reaching the predetermined time can be a regular time interval (e.g., at least one of the number of symbols, the number of time slots, the number of subframes, and absolute time). Alternatively, the timing for reaching the predetermined time can be set as a time timing (e.g., at least one of the symbol index, time slot index, subframe index, and absolute time).

[0616] (G3) Notification Information

[0617] In event-triggered control and / or period-based control, the information notified from the first communication device to other communication devices (e.g., the second and / or third communication devices) may be, for example, the following notification information (e.g., RRC IE (Information Element)). This notification information (e.g., LCMConfig) may also include at least a portion of information about the AI / ML model described above or below. Alternatively or additionally, at least a portion of the following notification information may be included in the information about the AI / ML model described above or below. The first communication device may notify other communication devices of at least a portion of the notification information shown below.

[0618] LCMConfig::= SEQUENCE {

[0619] modelTrainingConfig SEQUENCE {

[0620] onlineTrainingEnabled

[0621] trainingEventTrigger,

[0622] trainingPeridoicTimer

[0623] …

[0624] },

[0625] modelIdentificationConfig SEQUENCE {

[0626] identificationModelList,

[0627] identificationEventTrigger,

[0628] identificationPeriodicTimer

[0629] …

[0630] },

[0631] functionalityIdentificationConfig SEQUENCE {

[0632] identificationFunctionalityList,

[0633] identificationEventTrigger,

[0634] identificationPeriodicTimer,

[0635] …

[0636] },

[0637] modelTransferConfig SEQUENCE {

[0638] modelId,

[0639] numberOfInputNode,

[0640] numberOfOutputNode,

[0641] numberOfLayer,

[0642] informationOfWeight,

[0643] …

[0644] },

[0645] modelInferenceConfig SEQUENCE {

[0646] modelId,

[0647] numberOfInputNode,

[0648] numberOfOutputNode,

[0649] numberOfLayer,

[0650] informationOfWeight,

[0651] …

[0652] },

[0653] modelActivationConfig SEQUENCE {

[0654] modelId,

[0655] functionalityId,

[0656] feature,

[0657] requirementList,

[0658] capableModelList,

[0659] modelValidationTimer

[0660] …

[0661] },

[0662] functionalityActivationConfig SEQUENCE {

[0663] functionalityId,

[0664] feature,

[0665] requirementList,

[0666] capableModelList,

[0667] functionalityValidationTimer

[0668] …

[0669] },

[0670] modelDeactivationConfig SEQUENCE {

[0671] modelId,

[0672] functionalityId,

[0673] feature,

[0674] requirementList,

[0675] capableModelList,

[0676] modelValidationTimer

[0677] …

[0678] },

[0679] functionalityDeactivationConfig SEQUENCE {

[0680] functionalityId,

[0681] feature,

[0682] requirementList,

[0683] capableModelList,

[0684] functionalityValidationTimer

[0685] …

[0686] },

[0687] modelSwitchingConfig SEQUENCE {

[0688] modelId,

[0689] functionalityId,

[0690] feature,

[0691] requirementList,

[0692] capableModelList,

[0693] modelValidationTimer

[0694] …

[0695] },

[0696] functionalitySwitchingConfig SEQUENCE {

[0697] functionalityId,

[0698] feature,

[0699] requirementList,

[0700] capableModelList,

[0701] functionalityValidationTimer

[0702] …

[0703] },

[0704] modelFallbackConfig SEQUENCE {

[0705] modelId,

[0706] functionalityId,

[0707] feature,

[0708] requirementList,

[0709] capableModelList,

[0710] modelValidationTimer,

[0711] fallbackConventionalConfiguration

[0712] …

[0713] },

[0714] modelMonitoringConfig SEQUENCE {

[0715] modelId,

[0716] functionalityId,

[0717] feature,

[0718] requirementList,

[0719] capableModelList,

[0720] monitoringEventTrigger,

[0721] monitoringPeriodicTimer,

[0722] …

[0723] },

[0724] modelUpdateConfig SEQUENCE {

[0725] modelId,

[0726] functionalityId,

[0727] feature,

[0728] requirementList,

[0729] capableModelList,

[0730] updateEventTrigger,

[0731] updatePeriodicTimer,

[0732] …

[0733] },

[0734] …

[0735] }

[0736] Here, `onlineTrainingEnable` indicates whether online training is enabled or disabled. Online training can refer to the means of training the model using training data acquired in real time. `trainingEventTrigger` indicates information about the event that triggers the start of model training. `trainingPeridoicTimer` indicates information about the periodic timer used to implement model training.

[0737] In addition, `identificationModelList` represents information about the list of model IDs used in model identification. `identificationFunctionalityList` represents information about the list of model IDs used in model identification. `identificationEventTrigger` indicates information about the event trigger that initiates model identification. `identificationPeriodicTimer` represents information about the periodic timer that implements model identification.

[0738] Furthermore, `modelId` indicates the physical or logical identifier assigned to the AI / ML model. `numberOfInputNode` represents the number of input nodes in the AI / ML model's network structure. `numberOfOutputNode` represents the number of output nodes in the AI / ML model's network structure. `numberOfLayer` represents the number of layers in the AI / ML model's network structure. `informationOfWeight` represents the weights assigned between each node in the AI / ML model's network structure.

[0739] Similarly, `functionalityId` indicates the physical or logical identifier assigned to a functionality. `feature` represents the feature containing that functionality. `requirementList` indicates information about the required values ​​for the functionality. `capableModelList` indicates information about the AI / ML models that can be applied to the functionality. `modelValidationTimer` indicates information about a timer representing the period during which AI / ML models can be used, or about a timer representing the period until they become unavailable. `functionalityValidationTimer` indicates information about a timer representing the period until the functionality information is updated.

[0740] Furthermore, when falling back to regular signal processing, `fallbackConventionalConfiguration` indicates semi-static information about the signal processing after the fallback. `monitoringEventTrigger` indicates information about the event triggering that begins model monitoring. `monitoringPeriodicTimer` indicates information about the periodic timer implementing model monitoring. `updateEventTrigger` indicates information about the event triggering that begins model updating. `updatePeriodicTimer` indicates information about the periodic timer implementing model updating.

[0741] <4-3. Fallback to Normal Signal Processing>

[0742] Next, we will describe the fallback to the conventional processing without using AI / ML models.

[0743] Figure 11 This is a diagram used to illustrate rollback processing. For example, a signal processing method using an AI / ML model (transmit signal processing method and / or receive signal processing method) is designated as the first signal processing method, and a signal processing method not using an AI / ML model (transmit signal processing method and / or receive signal processing method) is designated as the second signal processing method. Figure 11 In the example, the AI / ML-based encoder processing is the transmit signal processing using the first signal processing method, and the AI / ML-based decoder processing is the receive signal processing using the first signal processing method. Furthermore, in Figure 11 In the example, conventional signal processing is transmit / receive signal processing using the second signal processing method. Note that conventional signal processing can be referred to as non-AI / ML-based encoder processing or non-AI / ML-based decoder processing.

[0744] The communication devices (first communication device and / or second communication device) can perform a switching process (hereinafter also referred to as rollback processing) of the signal processing method from a first signal processing method using an AI / ML model to a second signal processing method not using an AI / ML model when predetermined conditions are met. Here, the first communication device can be a transmitting device and the second communication device can be a receiving device, or the first communication device can be a receiving device and the second communication device can be a transmitting device.

[0745] For example, suppose the wireless communication (or related processing) applying the AI / ML model is uplink / downlink communication (or related processing). In this case, base station 20 and terminal device 40 can perform fallback processing when predetermined conditions are met. Furthermore, suppose the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing). In this case, terminal device 40, which becomes a transmitting device, and terminal device 40, which becomes a receiving device, can perform fallback processing when predetermined conditions are met.

[0746] Here, conventional signal processing (signal processing using the second signal processing method) can be at least one of the signal processing methods (H1) to (H3) shown below. Note that the signal processing shown below is an example. Conventional signal processing is not limited to the following processing.

[0747] (H1)CSI Feedback

[0748] Conventional signal processing can be traditional CSI feedback processing within CSI feedback. Here, conventional CSI feedback can be at least one of type 1 single-panel codebook, type 1 multi-panel codebook, type 2 codebook, type 2 port selection codebook, enhanced type 2 codebook, enhanced type 2 port selection codebook, and further enhanced type 2 port selection codebook.

[0749] (H2) Beam Management

[0750] Conventional signal processing can refer to conventional beam management processing within beam management. Here, conventional beam management can be the operation of receiving an SSB (Synchronization Signal Block) and transmitting a PRACH preamble on the PRACH resource corresponding to the SSB. Alternatively, conventional beam management can be the operation of receiving a CSI-RS and reporting a CSI report.

[0751] (H3) Positioning

[0752] Conventional signal processing can be considered routine positioning processing in positioning. Here, conventional positioning might involve transmitting a PRS (Positioning Reference Signal) on the downlink and performing position estimation. Alternatively, conventional positioning might involve transmitting an SRS (Sound Reference Signal) on the uplink and performing position estimation.

[0753] As described above, when predetermined conditions are met, the communication devices (first communication device and / or second communication device) can perform rollback processing. In this case, the communication devices can perform rollback processing when a predetermined event occurs, when a predetermined trigger is activated, or when a predetermined time is reached.

[0754] More specifically, the communication device may perform rollback processing in at least one of the following cases (I1) to (I4).

[0755] (I1) When a rollback notification is received

[0756] When a communication device receives a rollback notification from another communication device instructing it to switch to normal signal processing, the communication device may perform rollback processing. For example, one of the first and second communication devices may perform rollback processing upon receiving a rollback notification from the other communication device. The first and / or second communication devices may also perform rollback processing upon receiving a rollback notification from a third communication device (a communication control device that controls the first and / or second communication devices).

[0757] (I2) When communication quality deteriorates

[0758] When communication quality deteriorates, the communication device can perform a rollback process. For example, when communication quality deteriorates and the initial access procedure becomes necessary, the communication device can perform a rollback process.

[0759] (I3) When the performance of the AI / ML model no longer meets the requirements

[0760] When the performance of an AI / ML model no longer meets requirements, the communication device can perform a rollback. For example, when the communication device determines, based on model monitoring results, that the performance of the AI / ML model no longer meets the functional requirements of the application, the communication device can perform a rollback.

[0761] (I4) When the validity period of the AI / ML model expires

[0762] When the validity period of an AI / ML model expires, the communication device can perform rollback processing. For example, the communication device can perform rollback processing when the validation timer of the AI / ML model expires and retraining or reacquiring the AI / ML model becomes necessary.

[0763] <4-4. Changes to LCM-Controlled Entities>

[0764] When the communication device used as the control entity of the LCM is mobile, it may move outside the communication area. Therefore, it is necessary to consider the possibility of changes in the control entity when performing LCM control. Therefore, the communication devices (first communication device and / or second communication device) can perform LCM control as follows.

[0765] For example, a valid time period can be set for LCM control. Then, when the LCM control entity executes LCM control, the communication device starts counting a timer. Here, the communication device counting the timer can be either the LCM control entity or a controlled object. The communication device determines that AI / ML model-related information (configuration information and / or control information) is valid information (available information) until the timer reaches zero. When the timer reaches zero, the communication device determines that the AI / ML model-related information is invalid.

[0766] In addition, communication devices can transmit an already activated AI / ML model to another communication device so that the other communication device can use the same AI / ML model.

[0767] For example, suppose the wireless communication (or related processing) using the AI / ML model is uplink / downlink communication (or related processing). In this case, base station 201 can transmit the AI / ML model used in communicating with terminal device 40 to base station 202. Thus, even if terminal device 40 moves outside the communication range of base station 201, base station 202 can still communicate smoothly with terminal device 40.

[0768] Furthermore, it is assumed that the wireless communication (or related processing) applying the AI / ML model is sidelink communication (or related processing). In this case, terminal device 402 can transmit the AI / ML model used for communication with terminal device 401 to terminal device 403. This allows terminal device 403 to communicate smoothly with terminal device 401, even if terminal device 401 moves outside the communication range of terminal device 402.

[0769] <4-5. Examples of Input / Output Information for AI / ML Models>

[0770] The communication devices (first and / or second communication devices) can perform signal processing based on input information given to the AI / ML model and / or output information from the AI / ML model. Furthermore, the information processing device can perform training of the AI / ML model based on the input information and / or output information shown below. Here, the information processing device can be one or both of the first and second communication devices. Additionally, the information processing device can be a device other than the first and second communication devices.

[0771] Here, the input information and / or output information may include, for example, at least one of the information shown in (J1) to (J15) below. Of course, the input information and / or output information is not limited to the following information.

[0772] (J1)CQI (Channel Quality Information)

[0773] (J2) PMI (Precoding Matrix Indicator)

[0774] (J3)CRI (CSI-RS Resource Indicator)

[0775] (J4)SSBRI (SS / PBCH resource block indicator)

[0776] (J5)LI (Layer Indicator)

[0777] (J6)RI (rank indicator)

[0778] (J7)L1-RSRP (Layer 1 Reference Signal Received Power)

[0779] (J8) Interference levels in other cells

[0780] (J9) Location estimation information of terminal device 40

[0781] Location information of (J10) base station 20

[0782] (J11) Channel Matrix

[0783] (J12) eigenvectors

[0784] (J13) Beam ID

[0785] (J14) Timing error information

[0786] (J15) LOS (Line of Sight) / NLOS (Non-Line of Sight) Information

[0787] <4-6. Coexistence with communication devices that do not support AI communication>

[0788] Next, we will explain the coexistence between communication devices that support AI communication and those that do not. As mentioned above, AI communication is wireless communication that uses AI / ML models.

[0789] In the following description, wireless communication that does not use an AI / ML model (wireless communication that is not AI communication) may be referred to as conventional communication or conventional wireless communication. Note that wireless communication that does not use an AI / ML model (wireless communication that is not AI communication) is not necessarily limited to existing wireless communication. The references to "conventional communication" or "conventional wireless communication" in the following description can be replaced with "wireless communication that does not use an AI / ML model" or "wireless communication that is not AI communication".

[0790] Note that the wireless communication described below can be uplink communication with the terminal device 40 as a transmitting device and the base station 20 as a receiving device, or it can be downlink communication with the base station 20 as a transmitting device and the terminal device 40 as a receiving device. Furthermore, the wireless communication described below can be sidelink communication with the terminal device 40 as a sidelink data transmitting device and another terminal device 40 as a sidelink data receiving device.

[0791] For AI communication to be performed, both the sending and receiving devices need to support AI communication. In this case, the coexistence of communication devices that support AI communication and those that do not needs to be considered.

[0792] On the other hand, if one of the transmitting and receiving devices is a communication device that supports AI communication and the other is a communication device that can perform conventional wireless communication without using AI / ML models, it is not necessarily necessary to consider the coexistence between communication devices that support AI communication and those that do not.

[0793] Here, in the processing related to AI / ML models of communication devices that support AI communication, information may need to be provided by communication devices that do not support AI communication. For example, in at least one of the processes shown above (M1) to (M25), such as model training processing or model inference processing, information may need to be provided by communication devices that do not support AI communication. Therefore, even when communication devices that support AI communication are capable of performing conventional wireless communication, it may be necessary to consider the coexistence between communication devices that support AI communication and those that do not.

[0794] Note that when one of the transmitting and receiving devices is a communication device that supports AI communication, and the other is a communication device that cannot perform traditional wireless communication without using AI / ML models, the coexistence between communication devices that support AI communication and those that do not still needs to be considered.

[0795] When considering the coexistence of communication devices that support AI communication and those that do not, the operation of the communication devices is as follows. The operation of the communication devices is explained below for each of broadcast communication, multicast communication, and unicast communication.

[0796] (L1) Broadcast Communication

[0797] For example, when a transmitting device uses an AI / ML model for signal processing and sends data, sometimes a communication device that supports AI communication can receive the data, but a communication device that does not support AI communication cannot receive the data.

[0798] (L2) multicast communication

[0799] In the case of multicast communication, the communication device determines whether the communication device performing wireless communication (the first communication device and / or the second communication device) belongs to a group that supports AI communication or a group that does not support AI communication. Here, the communication device performing the determination can be the first communication device, the second communication device, or a third communication device that is different from the first and second communication devices (e.g., a communication control device that controls the first and / or second communication devices).

[0800] Then, based on the determination result, the communication device determines whether to perform signal processing using the AI / ML model of the first communication device and / or the second communication device, or to perform signal processing without using the AI / ML model. The first communication device and / or the second communication device then performs signal processing using the AI / ML model or without using the AI / ML model based on the determination result.

[0801] (L3) Unicast Communication

[0802] In the case of unicast communication, the communication device determines whether the communication device performing wireless communication (the first communication device and / or the second communication device) can support AI communication. Here, the communication device performing the determination can be the first communication device, the second communication device, or a third communication device different from the first and second communication devices (e.g., a communication control device that controls the first and / or second communication devices).

[0803] At this point, the communication device performing wireless communication can send capability information, including information indicating whether it can support AI communication, to the communication device performing the determination. Then, the communication device performing the determination can determine whether the first communication device and / or the second communication device can support AI communication based on the capability information.

[0804] Then, based on the determination result, the communication device determines whether to make the signal processing of the first communication device and / or the signal processing of the second communication device use an AI / ML model or not. The first communication device and / or the second communication device then performs the signal processing using or not using an AI / ML model based on the determination result.

[0805] Note that one of the first and second communication devices can notify the other of control information regarding whether to use AI communication or conventional communication. This control information may include capability information, which includes indications of whether AI communication is supported. The other communication device then performs signal processing using or without an AI / ML model based on the received control information.

[0806] <5. Sequence Examples>

[0807] Based on the above, an example sequence of communication processing according to this embodiment will be described.

[0808] In the following description, examples of sequences of wireless communication performed between a transmitting device and a receiving device will be described. Note that the wireless communication shown in the following sequence examples can be uplink communication with a terminal device 40 acting as a transmitting device and a base station 20 acting as a receiving device, or it can be downlink communication with a base station 20 acting as a transmitting device and a terminal device 40 acting as a receiving device. Furthermore, the wireless communication shown in the following sequence examples can be sidelink communication with another terminal device 40 acting as a sidelink data transmitting device and a sidelink data receiving device.

[0809] Note that the communication processing in this embodiment is not limited to the communication processing shown in the following sequence examples. The communication processing in this embodiment can also be implemented using other sequences.

[0810] <5-1. Case where the transmitting device is an LCM control entity>

[0811] First, a sequence example will be described when the transmitting device is the LCM control entity and the receiving device is the controlled object. This sequence example will describe the operation of the transmitting and receiving devices in both-side models. The two-side model is a case where AI / ML models are used in both the transmitting and receiving devices.

[0812] <5-1-1. Basic Sequence Example>

[0813] First, a basic sequence example will be described. Figure 12A This is a diagram illustrating an example of the basic sequence of communication processing when the transmitting device is an LCM control entity. Note that, for ease of explanation, details of each step from S11 to S15 are described below, but not all steps are essential components for specifying the invention. For example, the invention may be specified by step S11 and its details (e.g., details regarding information about the AI / ML model). In other words, steps S12 to S15 may be optional. Reference will be made below. Figure 12A Describe the communication processing based on this sequence example.

[0814] The transmitting device sends information related to the AI / ML model (information related to the AI / ML model described above or below) to the receiving device (control object) (step S11). At this time, the transmitting device may send data of the trained AI / ML model (learned AI / ML model) as information about the AI / ML model, may send activation notifications of the AI / ML model, or may send feedback information (information for model monitoring). Furthermore, the transmitting device may send information about the aforementioned LCM as information about the AI / ML model.

[0815] The receiving device receives information about the AI / ML model (information about the AI / ML model described above or below). Then, the receiving device performs processing based on the information about the AI / ML model (step S12). For example, if data of a trained AI / ML model, which serves as information about the AI / ML model, is received, the receiving device sets the received AI / ML model (or an AI / ML model generated based on the received AI / ML model data) as the AI / ML model for receiving processing. Furthermore, if an activation notification of the AI / ML model is received as information about the AI / ML model, the receiving device activates the AI / ML model associated with that notification. If feedback information, which serves as information about the AI / ML model, is received, the receiving device performs model monitoring based on the received feedback information.

[0816] The transmitting device generates transmission data by performing transmission signal processing (step S13). Here, the transmitting device can generate transmission data by performing transmission signal processing (first transmission processing) using an AI / ML model. Note that when both sides of the model are required, the transmitting device can use the AI / ML model to perform transmission signal processing. Alternatively, when the transmitting device uses the AI / ML model in the case of using only one side of the model, the transmitting device can use the AI / ML model to perform transmission signal processing.

[0817] The transmitting device sends the generated transmission data to the receiving device (step S14).

[0818] The receiving device decodes the received data by performing receive signal processing (first receive processing) (step S15). Here, the receiving device can decode the received data by performing receive signal processing using an activated AI / ML model. Note that when both sides of the model are required, the receiving device can use the AI / ML model to perform receive signal processing. Alternatively, when the receiving device uses the AI / ML model with only one side of the model, the receiving device can use the AI / ML model to perform receive signal processing.

[0819] <5-1-2. Specific Sequence Examples>

[0820] Next, specific sequence examples will be described. Figure 12B This is a diagram illustrating a specific sequence example of communication processing when the transmitting device is an LCM control entity. The following will refer to... Figure 12B Describe the communication processing based on this sequence example.

[0821] The transmitting device performs training on the AI / ML model (step S101). As a result, the transmitting device obtains the trained AI / ML model (the learned AI / ML model).

[0822] The transmitting device sends the trained AI / ML model to the receiving device (step S102). The transmitting device may notify the receiver of the AI / ML model as semi-static information (e.g., RRC signaling).

[0823] The receiving device notifies the sending device of the success / failure (ACK / NACK) of the reception of the received AI / ML model (step S103).

[0824] The transmitting device sends an activation notification of the AI / ML model to the receiving device (step S104). Here, the transmitting device may notify, for example, the ID of the AI / ML model to be activated (model ID) and / or the ID of the functionality to be activated (functional ID). The transmitting device may notify the model ID and / or functional ID as semi-static information such as RRC signaling, or as dynamic information such as MAC CE, DCI, UCI, or SCI.

[0825] The receiving device activates the AI / ML model based on the activation notification (step S105).

[0826] The transmitting device generates transmission data by performing transmission signal processing (step S106). Here, the transmitting device can generate transmission data by performing transmission signal processing using an AI / ML model. Note that when both sides of the model are required, the transmitting device can use the AI / ML model to perform transmission signal processing. Alternatively, when the transmitting device uses the AI / ML model with only one side of the model, the transmitting device can use the AI / ML model to perform transmission signal processing.

[0827] The transmitting device sends the generated transmission data to the receiving device (step S107).

[0828] The receiving device decodes the received data by performing receive signal processing (step S108). Here, the receiving device can decode the received data by performing receive signal processing using the activated AI / ML model. Note that when both sides of the model are required, the receiving device can use the AI / ML model to perform receive signal processing. Alternatively, when the receiving device uses the AI / ML model with only one side of the model, the receiving device can use the AI / ML model to perform receive signal processing.

[0829] The receiving device can feed back information used for model monitoring to the transmitting device (step S109). For example, the receiving device can feed back decoded received data to the transmitting device.

[0830] The sending device performs model monitoring based on the received feedback information (step S110). Then, based on the model monitoring results, the sending device determines whether the characteristics of the AI / ML model meet the requirements (step S111). For example, the sending device may determine whether the characteristics of the AI / ML model meet the requirements based on the requirement information set in the notified functionality.

[0831] Here, it is assumed that the transmitting device determines that the characteristics of the AI / ML model do not meet the requirements. In this case, the transmitting device notifies the receiving device of a rollback to conventional signal processing (step S112).

[0832] When the receiving device receives a fallback notification, it stops using the AI / ML model for signal processing. Then, the receiving device falls back to regular signal processing without using the AI / ML model (step S113).

[0833] The transmitting device performs training on the AI / ML model (step S114). Then, the transmitting device transmits the trained AI / ML model to the receiving device (step S115).

[0834] The receiving device notifies the sending device of the success / failure (ACK / NACK) of the reception of the received AI / ML model (step S116).

[0835] The sending device notifies the receiving device of the activation of the AI / ML model (step S117). Here, the sending device may notify, for example, the ID of the AI / ML model to be activated and / or the ID of the functionality to be activated.

[0836] The receiving device activates the AI / ML model based on the activation notification (step S118).

[0837] <5-2. Case where the receiving device is an LCM control entity>

[0838] Next, a sequence example will be described when the receiving device is the LCM control entity and the transmitting device is the controlled object. This sequence example will describe the operation of the transmitting and receiving devices in both-side models. The two-side model is a case where AI / ML models are used in both the transmitting and receiving devices.

[0839] <5-2-1. Basic Sequence Example>

[0840] First, a basic sequence example will be described. Figure 13A This is a diagram illustrating an example of the basic sequence of communication processing when the receiving device is an LCM control entity. Note that, for ease of explanation, details of each step from S21 to S25 are described below, but not all steps are essential components for specifying the invention. For example, the invention may be specified by step S21 and its details (e.g., details regarding information about the AI / ML model). In other words, steps S22 to S25 may be optional. Reference will be made below. Figure 13A Describe the communication processing based on this sequence example.

[0841] The receiving device sends information about the AI / ML model (information about the AI / ML model described above or below) to the sending device (the controlled object) (step S21). At this time, the receiving device may send data of the trained AI / ML model (the learned AI / ML model) as information about the AI / ML model, may send an activation notification of the AI / ML model, or may send feedback information (information for model monitoring). Furthermore, the receiving device may send information about the aforementioned LCM as information about the AI / ML model.

[0842] The transmitting device receives information related to the AI / ML model (information related to the AI / ML model as described above or below). Then, the transmitting device performs processing based on the information related to the AI / ML model (step S22). For example, if data from a trained AI / ML model is received as information about the AI / ML model, the transmitting device sets the received AI / ML model (or the AI / ML model generated based on the received AI / ML model data) as the AI / ML model for transmission processing. Furthermore, if an activation notification of the AI / ML model is received as information about the AI / ML model, the transmitting device performs activation of the AI / ML model associated with that notification. If feedback information as information about the AI / ML model is received, the transmitting device performs model monitoring based on the received feedback information.

[0843] The transmitting device generates transmission data by performing transmission signal processing (step S23). Here, the transmitting device can generate transmission data by performing transmission signal processing (second transmission processing) using an AI / ML model. Note that when both sides of the model are required, the transmitting device can use the AI / ML model to perform transmission signal processing. Alternatively, when the transmitting device uses the AI / ML model in the case of using only one side of the model, the transmitting device can use the AI / ML model to perform transmission signal processing.

[0844] The transmitting device sends the generated transmission data to the receiving device (step S24).

[0845] The receiving device decodes the received data by performing receive signal processing (step S25). Here, the receiving device can decode the received data by performing receive signal processing (second receive processing) using the activated AI / ML model. Note that when both models are required, the receiving device can use the AI / ML model to perform receive signal processing. Alternatively, when the receiving device uses the AI / ML model in the case of using only one model, the receiving device can use the AI / ML model to perform receive signal processing.

[0846] <5-2-2. Specific Sequence Examples>

[0847] Next, specific sequence examples will be described. Figure 13B This is a diagram illustrating a specific sequence example of communication processing when the receiving device is an LCM control entity. The following will refer to... Figure 13B Describe the communication processing based on this sequence example.

[0848] The receiving device performs training on the AI / ML model (step S201). As a result, the receiving device acquires the trained AI / ML model (the learned AI / ML model).

[0849] The receiving device transmits the trained AI / ML model to the sending device (step S202). The receiving device can notify the user of the AI / ML model as semi-static information (e.g., RRC signaling).

[0850] The transmitting device notifies the receiving device of the success / failure (ACK / NACK) of the reception of the received AI / ML model (step S203).

[0851] The receiving device sends an activation notification for the AI / ML model to the sending device (step S204). Here, the receiving device may notify, for example, the ID of the AI / ML model to be activated (model ID) and / or the ID of the functionality to be activated (functional ID). The receiving device may notify the model ID and / or functional ID as semi-static information such as RRC signaling, or as dynamic information such as MAC CE, DCI, UCI, or SCI.

[0852] The sending device activates the AI / ML model based on the activation notification (step S205).

[0853] The transmitting device generates transmission data by performing transmission signal processing (step S206). Here, the transmitting device can generate transmission data by performing transmission signal processing using an AI / ML model. Note that when both sides of the model are required, the transmitting device can use the AI / ML model to perform transmission signal processing. Alternatively, when the transmitting device uses the AI / ML model with only one side of the model, the transmitting device can use the AI / ML model to perform transmission signal processing.

[0854] The transmitting device sends the generated transmission data to the receiving device (step S207).

[0855] The receiving device decodes the received data by performing receive signal processing (step S208). Here, the receiving device can decode the received data by performing receive signal processing using an activated AI / ML model. Note that when both sides of the model are required, the receiving device can use the AI / ML model to perform receive signal processing. Alternatively, when the receiving device uses the AI / ML model with only one side of the model, the receiving device can use the AI / ML model to perform receive signal processing.

[0856] The receiving device performs model monitoring based on the decoded received data (step S209). Then, the transmitting device determines whether the characteristics of the AI / ML model meet the requirements based on the model monitoring results (step S210). For example, the transmitting device may determine whether the characteristics of the AI / ML model meet the requirements based on the requirement information set in the notified functionality.

[0857] Here, it is assumed that the receiving device determines that the characteristics of the AI / ML model do not meet the requirements. In this case, the receiving device notifies the receiving device to fall back to normal signal processing (step S211).

[0858] When the receiving device receives the fallback notification, it stops using the AI / ML model for signal processing. Then, the transmitting device falls back to regular signal processing that does not use the AI / ML model (step S212).

[0859] The receiving device performs training on the AI / ML model (step S213). Then, the receiving device transmits the trained AI / ML model to the sending device (step S214).

[0860] The transmitting device notifies the receiving device of the success / failure (ACK / NACK) of the reception of the received AI / ML model (step S215).

[0861] The receiving device sends an activation notification for the AI / ML model to the sending device (step S216). Here, the receiving device may notify, for example, the ID of the AI / ML model to be activated and / or the ID of the functionality to be activated.

[0862] The receiving device activates the AI / ML model based on the activation notification (step S217).

[0863] <5-3. Other Sequence Examples>

[0864] The above sequence examples illustrate a communication processing example involving multiple LCM controls. However, even partial sequences from the above sequence examples can constitute the communication processing of this embodiment. Other sequence examples of the communication processing of this embodiment will be described below.

[0865] (N1) Model Transfer

[0866] Figure 14 This diagram illustrates an example sequence of communication processes related to model transmission. The communication device that acts as the control entity of the LCM performs training on the AI / ML model (step S301). Then, the communication device that acts as the control entity of the LCM transmits the trained AI / ML model to the communication device that acts as the control object (step S302). The communication device that acts as the control object performs received signal processing using the received AI / ML model (step S303).

[0867] (N2) model activation

[0868] Figure 15 This is a diagram illustrating an example sequence of communication processes related to model activation. The communication device that becomes the control entity of the LCM sends an activation notification for the AI / ML model to the communication device that becomes the control object (step S401). The communication device that becomes the control object activates the AI / ML model based on the activation notification (step S402).

[0869] (N3) Signal processing using AI / ML models

[0870] Figure 16 This diagram illustrates an example sequence of signal processing using an AI / ML model. The communication device that becomes the control entity of the LCM generates transmission data by performing transmission signal processing using the AI / ML model (step S501). The communication device that becomes the control entity of the LCM sends the generated transmission data to the communication device that becomes the control object (step S502). The communication device that becomes the control object decodes the received data by performing reception signal processing using the AI / ML model (step S503).

[0871] (N4) Model Monitoring

[0872] Figure 17 This is a diagram illustrating an example sequence of communication processes related to model monitoring. The communication device that becomes the controlled object decodes the received data by performing received signal processing (step S601). The communication device that becomes the controlled object sends the decoded received data as feedback information to the communication device that becomes the control entity of the LCM (step S602). The communication device that becomes the control entity of the LCM performs model monitoring based on the received feedback information (step S603).

[0873] (N5) Model rollback

[0874] Figure 18This is a diagram illustrating an example sequence of communication processes related to model rollback. The communication device acting as the control entity of the LCM determines whether the characteristics of the AI / ML model meet the requirements based on model monitoring results (step S701). When the characteristics of the AI / ML model do not meet the requirements, the communication device acting as the control entity of the LCM notifies the communication device acting as the controlled object to roll back to normal signal processing (step S702). Upon receiving the rollback notification, the communication device acting as the controlled object stops using signal processing for the AI / ML model and rolls back the received signal processing to normal signal processing without using the AI / ML model (step S703). The communication device acting as the controlled object performs training on the AI / ML model (step S704).

[0875] << 6. Revisions >>

[0876] The above embodiments illustrate examples, and various modifications and applications are possible.

[0877] In the above embodiments, wireless communication between base station 20 and terminal device 40, or wireless communication between terminal device 40 and another terminal device 40, is used as an example to illustrate the technology of this disclosure. However, the application scope of this embodiment is not limited thereto. For example, the technology of this disclosure is also applicable to wireless communication between multiple communication devices selected from management device 10, base station 20, relay station 30, and terminal device 40. Furthermore, the technology of this disclosure is also applicable to wireless communication between management devices 10, between base stations 20, between relay stations 30, or between terminal devices 40. Moreover, the technology of this disclosure is also applicable to processing independently performed by management device 10, base station 20, relay station 30, and terminal device 40 using AI / ML models.

[0878] Furthermore, the functions of each block (acquisition unit 231 to transmission unit 234) of the control unit 23 of the base station 20 can be the same as the functions of each block (acquisition unit 431 to transmission unit 434) of the control unit 43 of the terminal device 40. Additionally, the functions of each block (acquisition unit 331 to transmission unit 334) of the control unit 33 of the relay station 30 can be the same as the functions of each block of the control unit 23 of the base station 20, or the functions of each block of the control unit 23 of the terminal device 40. Furthermore, the functions of each block of the control unit 43 of the terminal device 40 can be the same as the functions of each block of the control unit 23 of the base station 20. Furthermore, the functions of the control unit 13 of the management device 10 can be the same as the functions of the control unit 23 of the base station 20, or the functions of the control unit 33 of the relay station 30, or the functions of the control unit 43 of the terminal device 40.

[0879] The control device that controls the management device 10, base station 20, relay station 30 or terminal device 40 in this embodiment can be implemented by a dedicated computer system or by a general-purpose computer system.

[0880] For example, the communication program used to perform the above operations is stored and distributed on a computer-readable recording medium such as an optical disc, semiconductor memory, magnetic tape, or floppy disk. Then, for example, the control device is configured by installing the program on a computer and performing the above processes. At this time, the control device can be a device external to the management device 10, base station 20, relay station 30, or terminal device 40 (e.g., a personal computer). Alternatively, the control device can be a device internal to the management device 10, base station 20, relay station 30, or terminal device 40 (e.g., control unit 13, control unit 23, control unit 33, or control unit 43).

[0881] Furthermore, the communication program can be stored on the disk device of a server on a network such as the Internet, so that it can be downloaded to the computer. Additionally, the above functionality can be achieved through cooperation between the OS (operating system) and application software. In this case, parts other than the OS can be stored on media and distributed, or parts other than the OS can be stored on a server so that they can be downloaded to the computer.

[0882] Furthermore, in the processes described in the above embodiments, all or part of the processes described as automatically executed can also be executed manually, or all or part of the processes described as manually executed can also be executed automatically by known methods. Additionally, unless otherwise stated, the processing procedures, specific names, and information including various data and parameters shown in the above documents and figures can be arbitrarily changed. For example, the various information shown in each figure is not limited to the information illustrated.

[0883] Furthermore, each component of each illustrated device is functionally conceptual and does not necessarily need to be physically configured as shown. That is, the specific form of distribution or integration of the devices is not limited to the illustrated form, and all or part of them can be functionally or physically distributed or integrated in any unit according to various loads, usage conditions, etc. Note that such a distributed or integrated configuration can be implemented dynamically.

[0884] Furthermore, the above embodiments can be appropriately combined in areas where the processing content does not conflict. Additionally, the order of each step shown in the flowcharts and sequence diagrams of the above embodiments can be appropriately changed.

[0885] Furthermore, this embodiment can be implemented as any configuration constituting a device or system, such as a processor such as a system LSI (Large-Scale Integration), a module using multiple processors, a unit using multiple modules, wherein a set of other functions is added to the unit (i.e., a partial configuration of the device).

[0886] A system LSI can be referred to as a SoC (System-on-a-Chip). In other words, each of the devices described above or below (e.g., management device 10, base station 20, relay station 30, and terminal device 40) can be interpreted as a processor (e.g., CPU) of a system LSI (e.g., SoC), or a module that uses or constitutes a processor. Furthermore, or alternatively, this embodiment can be implemented by any configuration constituting the device or system, such as a modem chip (baseband chip) or an RF (radio frequency) unit or a combination thereof. An RF unit includes at least one of RF circuitry or an RF front end. In other words, each of the devices described above or below can be interpreted as a modem chip (baseband chip) or an RF unit or a combination thereof. Furthermore, or alternatively, each of the devices described above or below can be interpreted as a module that uses or constitutes a modem chip or an RF unit.

[0887] The modem chip performs signal processing related to communication within the device (including the devices described above or below). The modem chip may have at least the functions of a modulator or demodulator. The RF unit may have at least one of the functions of an RF transceiver (RF up-converter, RF down-converter), a power amplifier, and a low-noise amplifier. The RF transceiver converts between baseband signals and RF frequencies. The power amplifier amplifies the signal transmitted from the antenna. The low-noise amplifier amplifies the weak signal received from the antenna. Additionally or alternatively, the RF unit (specifically, the RF front end) may include at least one of the aforementioned power amplifier, low-noise amplifier, envelope tracker, filter, duplexer, multiplexer, antenna switch, and antenna tuner.

[0888] A combination of a modem chip and an RF unit can be referred to as a modem-RF system. At least a portion of the modem chip or RF unit, or a combination thereof, may be included in a system LSI (e.g., a System-on-a-Chip). For example, at least a portion of the MAC layer processing / PHY layer processing handled by at least a portion of the modem chip or RF unit, or a combination thereof, may be implemented by the system LSI. Here, the MAC layer processing or PHY layer processing may be at least a portion of the processing performed by the means (e.g., management device 10, base station 20, relay station 30, and terminal device 40) in the above or following embodiments.

[0889] Note that in this embodiment, a system means a collection of multiple components (devices, modules (parts), etc.), and it is not important whether all components are housed in the same enclosure. Therefore, multiple devices housed in separate enclosures and connected via a network, as well as a single device in which multiple modules are housed in one enclosure, are both systems.

[0890] Furthermore, for example, this embodiment can employ a cloud computing configuration, in which a function is shared and processed jointly by multiple devices via a network.

[0891] << 7. Conclusion>>

[0892] As described above, the first communication device (e.g., base station 20 or terminal device 40) in this embodiment is configured to wirelessly communicate with the second communication device (e.g., base station 20 or terminal device 40). Furthermore, at least one of the first and second communication devices can perform wireless communication-related processing using an AI / ML model.

[0893] The first communication device sends information about the AI / ML model to the second communication device when implementing LCM (AI / ML model management). Alternatively, the first communication device sends information about the AI / ML model to the second communication device implementing LCM.

[0894] For example, a first communication device sends information about an AI / ML model to a second communication device, wherein the AI / ML model is applied to at least one of (P1) a first transmission process of the communication device regarding a wireless signal transmitted from the first communication device to the second communication device, (P2) a first reception process of another communication device regarding a wireless signal transmitted from the first communication device to another communication device, (P3) a second transmission process of another communication device regarding a wireless signal transmitted from another communication device to the first communication device, and (P4) a second reception process of the first communication device regarding a wireless signal transmitted from another communication device to the first communication device.

[0895] The second communication device uses or manages the AI / ML model based on information received from the first communication device regarding the AI / ML model. For example, suppose the first communication device is the control entity of the LCM, and the second communication device is the controlled object of the LCM. In this case, the second communication device can perform processing related to its signal processing (first receiving processing or second transmitting processing) based on the information about the AI / ML model. Furthermore, suppose the second communication device is the control entity of the LCM, and the first communication device is the controlled object of the LCM. In this case, the second communication device can perform LCM related to the first communication device's signal processing (first transmitting processing or second receiving processing) based on the information about the AI / ML model.

[0896] This enables the efficient use or management of AI / ML models, thus allowing the first and / or second communication devices to achieve high communication performance.

[0897] The various embodiments of the present invention have been described above, but the technical scope of the present invention is not limited to the above embodiments, and various modifications can be made without departing from the spirit of the present invention. Furthermore, components from different embodiments and modifications can be appropriately combined.

[0898] Furthermore, the effects of each embodiment described in this specification are merely illustrative and not restrictive, and other effects may exist.

[0899] Note that this technology can also be configured as follows.

[0900] (1) A communication device for performing wireless communication with other communication devices, comprising:

[0901] The processor controls a wireless transceiver for performing wireless communication.

[0902] The processor controls the wireless transceiver to send information about an AI / ML model to the other communication device regarding at least one of the following processes:

[0903] The first transmission process of a communication device regarding wireless signals transmitted from the communication device to the other communication device.

[0904] The first receiving process of the other communication device regarding the wireless signals transmitted from the communication device to the other communication device.

[0905] The second transmission process of the other communication device regarding the wireless signals sent from the other communication device to the communication device, or

[0906] The second receiving process of the communication device regarding wireless signals transmitted from the other communication device to the communication device.

[0907] (2) According to the communication equipment described in (1),

[0908] The processor is configured to determine the control entity for managing the AI / ML model based on setting information pre-stored by the communication device or control information received from the other communication device, and the determined control entity is one or both of the communication device and the other communication device.

[0909] (3) The communication equipment described in (2),

[0910] The management of AI / ML models includes at least two of the following:

[0911] Data collection,

[0912] Model training,

[0913] Function recognition,

[0914] Model recognition,

[0915] Model delivery,

[0916] Model testing,

[0917] Model activation,

[0918] Functional activation,

[0919] Model deactivation,

[0920] Functional deactivation

[0921] Model switching,

[0922] Functional switching,

[0923] Model rollback,

[0924] Model monitoring,

[0925] Model update

[0926] Model deployment,

[0927] Model configuration,

[0928] Model selection, or

[0929] Terminal capabilities

[0930] And when the processor identifies both the communication device and the other communication device as control entities, the processor determines the control entity for each of at least two AI / ML model managements.

[0931] (4) The communication device according to any one of (1) to (3),

[0932] Information regarding AI / ML models includes at least one of the following:

[0933] Data applied to an AI / ML model in at least one of the first receiving process or the second sending process.

[0934] Instructs the activation of information applied to the AI / ML model in at least one of the first receiving process or the second transmitting process.

[0935] Instructions to activate information applied to the AI / ML model in at least one of the first receiving process or the second transmitting process.

[0936] Indicates switching information of the AI / ML model applied to at least one of the first receiving process or the second transmitting process, or

[0937] Information used to cause at least one of the first receiving process or the second sending process that applies an AI / ML model to fall back to a process that does not use an AI / ML model.

[0938] (5) The communication device according to (4),

[0939] The information indicating the activation of the AI / ML model includes at least one of the following:

[0940] Information used to specify the activation timing for the other communication devices, or

[0941] Information used by the other communication devices to specify the effective period of the activation.

[0942] (6) Based on the communication equipment in (4) or (5),

[0943] The information indicating the deactivation of the AI / ML model includes information for the other communication devices to specify the timing of the deactivation.

[0944] (7) The communication device according to any one of (4) to (6),

[0945] The information indicating the switching of the AI / ML model includes information indicating the timing for performing the switching of the AI / ML model, and the information indicating the timing for performing the switching of the AI / ML model indicates the time period from receiving the information indicating the timing for performing the switching of the AI / ML model to the timing for performing the switching of the AI / ML model.

[0946] (8) The communication device according to any one of (4) to (7),

[0947] The information used to enable rollback includes information indicating the timing of rollback for the AI / ML model of the other communication device, and the information indicating the timing of the rollback indicates the time period from receiving the information indicating the timing of the rollback to executing the rollback.

[0948] (9) The communication device according to (8),

[0949] The other communication device has multiple functionalities associated with at least one of the first receiving process and the second sending process, and the information for causing rollback is information for causing the processing of some of the multiple functionalities to roll back from processing using an AI / ML model to processing without using an AI / ML model.

[0950] (10) The communication device according to any one of (1) to (8),

[0951] Information regarding AI / ML models includes at least one of the following:

[0952] ModelIDConfig is a configuration information for the identifier of an AI / ML model, or

[0953] FunctionalityConfig contains configuration information regarding the functionality of the applied AI / ML model.

[0954] ModelIDConfig includes at least one of the following:

[0955] ModelId is an identifier that indicates the AI / ML model.

[0956] numberOfInputNode represents the number of input nodes in the AI / ML model network structure that constitutes the AI / ML model.

[0957] numberOfOutputNode represents the number of output nodes in the AI / ML model network structure that constitutes the AI / ML model.

[0958] numberOfLayer represents the number of layers in the AI / ML model network structure that makes up the AI / ML model, or

[0959] informationOfWeight represents the weights assigned among each node in the AI / ML model network structure that makes up the AI / ML model.

[0960] And FunctionalityConfig includes at least one of the following:

[0961] The functionalityId represents a functional identifier.

[0962] This refers to a feature that includes functional characteristics.

[0963] A requirementsList representing the required values ​​for functionality, or

[0964] The `capableModelList` represents information about AI / ML models that can be applied to the aforementioned functionality.

[0965] (11) The communication device according to (8) or (9),

[0966] The processor controls the wireless transceiver to send information for causing rollback in at least one of the following situations:

[0967] When communication quality deteriorates

[0968] When the performance of the AI / ML model no longer meets the predetermined requirements, or

[0969] When the validity period of an AI / ML model has expired.

[0970] (12) The communication device according to any one of (1) to (11),

[0971] The processor controls the wireless transceiver to send information to the other communication devices indicating which of a plurality of functionalities the communication devices correspond to in relation to at least one of the first transmission process and the second reception process.

[0972] (13) The communication device according to any one of (1) to (12),

[0973] The AI / ML model is linked to a functionality associated with at least one of the first transmission process and the second reception process, and the processor controls the wireless transceiver to transmit information indicating which of the plurality of functionalities has completed AI / ML model training as information about the AI / ML model to the other communication device.

[0974] (14) The communication device according to any one of (1) to (13),

[0975] The processor controls the wireless transceiver to send information to the other communication device indicating which AI / ML model among a plurality of AI / ML models the communication device corresponds to, which is associated with at least one of the first transmission process and the second reception process.

[0976] (15) The communication device according to any one of (1) to (14),

[0977] The identification information is linked to the AI / ML model, and the processor sends information indicating which of the multiple AI / ML models has completed training as information about the AI / ML model to the other communication devices.

[0978] (16) The communication device according to any one of (1) to (15),

[0979] The processor controls the wireless transceiver to send information about the AI / ML model to multiple communication devices among the other communication devices.

[0980] (17) The communication device according to any one of (1) to (16),

[0981] The processor controls the wireless transceiver to send information about the AI / ML model to the other communication device at a predetermined time when a predetermined event occurs or at a predetermined time.

[0982] (18) The communication device according to any one of (1) to (17),

[0983] The communication device is one of the base station and the terminal device, and the other communication device is the other of the base station and the terminal device.

[0984] (19) The communication device according to any one of (1) to (17),

[0985] ...

Claims

1. A communication device for performing wireless communication with other communication devices, comprising: The processor controls a wireless transceiver for performing wireless communication. The processor controls the wireless transceiver to send information about an AI / ML model to the other communication device regarding at least one of the following processes: The first transmission process of a communication device regarding wireless signals transmitted from the communication device to the other communication device. The first receiving process of the other communication device regarding the wireless signals transmitted from the communication device to the other communication device. The second transmission process of the other communication device regarding the wireless signals sent from the other communication device to the communication device, or The second receiving process of the communication device regarding wireless signals transmitted from the other communication device to the communication device.

2. The communication device according to claim 1, in, The processor is configured to determine a control entity for managing the AI / ML model based on setting information pre-stored by the communication device or control information received from the other communication device, and the determined control entity is one or both of the communication device and the other communication device.

3. The communication device according to claim 2, The management of AI / ML models includes at least two of the following: Data collection, Model training, Function recognition, Model recognition, Model delivery, Model testing, Model activation, Functional activation, Model deactivation, Functional deactivation Model switching, Functional switching, Model rollback, Model monitoring, Model update Model deployment, Model configuration, Model selection, or Terminal capabilities And when the processor identifies both the communication device and the other communication device as control entities, the processor determines the control entity for each of at least two AI / ML model managements.

4. The communication device according to claim 1, in, Information about AI / ML models includes at least one of the following: Data applied to an AI / ML model in at least one of the first receiving process or the second sending process. Instructs the activation of information applied to the AI / ML model in at least one of the first receiving process or the second transmitting process. Instructions to activate information applied to the AI / ML model in at least one of the first receiving process or the second transmitting process. Indicates switching information of the AI / ML model applied to at least one of the first receiving process or the second transmitting process, or Information used to cause at least one of the first receiving process or the second sending process that applies an AI / ML model to fall back to a process that does not use an AI / ML model.

5. The communication device according to claim 4, in, Information indicating the activation of an AI / ML model includes at least one of the following: Information used to specify the activation timing for the other communication devices, or Information used by the other communication devices to specify the effective period of the activation.

6. The communication device according to claim 4, in, The information instructing the deactivation of the AI / ML model includes information for specifying the timing of the deactivation for the other communication devices.

7. The communication device according to claim 4, in, The information indicating the switching of the AI / ML model includes information indicating the timing for performing the switching of the AI / ML model, and the information indicating the timing for performing the switching of the AI / ML model indicates the time period from receiving the information indicating the timing for performing the switching of the AI / ML model to the timing for performing the switching of the AI / ML model.

8. The communication device according to claim 4, in, The information used to enable rollback includes information indicating the timing of rollback for the AI / ML model of the other communication device, and the information indicating the timing of the rollback indicates the time period from receiving the information indicating the timing of the rollback to the timing of executing the rollback.

9. The communication device according to claim 8, in, The other communication device has multiple functionalities associated with at least one of the first receiving process and the second transmitting process, and the information for causing rollback is information for causing the processing of some of the multiple functionalities to roll back from processing using an AI / ML model to processing without using an AI / ML model.

10. The communication device according to claim 8, in, Information about AI / ML models includes at least one of the following: ModelIDConfig is a configuration information for the identifier of an AI / ML model, or FunctionalityConfig contains configuration information regarding the functionality of the applied AI / ML model. ModelIDConfig includes at least one of the following: ModelId is an identifier that indicates the AI / ML model. `numberOfInputNode` represents the number of input nodes in the AI / ML model network structure that constitutes the AI / ML model. `numberOfOutputNode` represents the number of output nodes in the AI / ML model network structure that constitutes the AI / ML model. `numberOfLayer` represents the number of layers in the AI / ML model network structure that constitutes the AI / ML model, or informationOfWeight represents the weights assigned among each node in the AI / ML model network structure that constitutes the AI / ML model. And FunctionalityConfig includes at least one of the following: The functionalityId represents a functional identifier. This refers to a feature that includes functional characteristics. A requirementsList representing the required values ​​for functionality, or The `capableModelList` represents information about AI / ML models that can be applied to the aforementioned functionality.

11. The communication device according to claim 8, in, The processor controls the wireless transceiver to send information for enabling rollback in at least one of the following situations: When communication quality deteriorates When the performance of the AI / ML model no longer meets the predetermined requirements, or When the validity period of an AI / ML model has expired.

12. The communication device according to claim 1, in, The processor controls the wireless transceiver to send information to the other communication devices indicating which of a plurality of functionalities the communication devices correspond to in relation to at least one of the first transmission process and the second reception process.

13. The communication device according to claim 1, The AI / ML model is linked to a functionality associated with at least one of the first transmission process and the second reception process, and the processor controls the wireless transceiver to transmit information indicating which of the plurality of functionalities has completed AI / ML model training as information about the AI / ML model to the other communication device.

14. The communication device according to claim 1, The processor controls the wireless transceiver to send information to the other communication device indicating which AI / ML model among a plurality of AI / ML models the communication device corresponds to, which is associated with at least one of the first transmission process and the second reception process.

15. The communication device according to claim 1, in, The identification information is linked to the AI / ML model, and the processor sends information indicating which of the multiple AI / ML models has completed training as information about the AI / ML model to the other communication devices.

16. The communication device according to claim 1, The processor controls the wireless transceiver to send information about the AI / ML model to multiple other communication devices.

17. The communication device according to claim 1, The processor controls the wireless transceiver to send information about the AI / ML model to the other communication device at a predetermined time when a predetermined event occurs or at a predetermined time.

18. The communication device according to claim 1, in, The communication device is one of the base station and the terminal device, and the other communication device is the other of the base station and the terminal device.

19. The communication device according to claim 1, in, The communication device is a terminal device, and the other communication device is another terminal device.

20. A communication method, wherein a communication device performing wireless communication with other communication devices sends information to the other communication devices regarding an AI / ML model applied to at least one of the following processes: The first transmission process of a communication device regarding wireless signals transmitted from the communication device to the other communication device. The first receiving process of the other communication device regarding the wireless signals transmitted from the communication device to the other communication device. The second transmission process of the other communication device regarding the wireless signals sent from the other communication device to the communication device, or The second receiving process of the communication device regarding wireless signals transmitted from the other communication device to the communication device.

Images