Communication device and communication method

Abstract

This communication device comprises: a control circuit for performing at least one of an artificial intelligence (AI) / machine learning (ML) model configuration, a parameter configuration for the AI / ML model, and a data-related configuration used in the AI / ML model, in accordance with a type related to the sharing, among communication devices, of information related to the AI / ML model; and a transmission circuit for transmitting information corresponding to the at least one setting to the other communication device.

Description

Communication device and communication method

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

[0002] Artificial intelligence (AI) or machine learning (ML) technologies for wireless local area networks (WLANs) are being studied by industry and academia (for example, see Non-Patent Document 1).

[0003] The IEEE (Institute of Electrical and Electronics Engineers) 802.11 Working Group launched a Topic Interest Group (TIG) on AI / ML (hereinafter also referred to as AIML) in July 2022 and began discussions (e.g., Non-Patent Document 2).

[0004] As one use case for applying AI / ML to wireless LANs, a method of sharing AI / ML models, parameters, etc. between communication devices is being considered (for example, Non-Patent Documents 4 and 5).

[0005] S. Szott, K. Kosek-Szott, P. Gawlowicz, JT Gomez, B. Bellalta, A. Zubow, F. Dressler, “WiFi Meets ML: A Survey on Improving IEEE 802.11 Performance with Machine Learning” IEEE Communications Surveys & Tutorials, vol.24, no.3, pp.1843-1893IEEE 802.11-22 / 847r3, AIML TIG July 2022 AgendaIEEE 802.11-22 / 950r2, Discussion on Connection between AI / ML & Wireless LANIEEE 802.11-23 / 905r0, Proposed technical report text for AIML model sharing use caseIEEE 802.11-23 / 750r0, Discussions on Neural Network Model Sharing for WLAN

[0006] There has been insufficient consideration of how to appropriately share information about AI / ML, such as information about AI / ML models and parameters, between communication devices.

[0007] Non-limiting examples of the present disclosure contribute to providing a communication device and a communication method that can appropriately share information related to AI / ML between communication devices.

[0008] A communication device according to one embodiment of the present disclosure includes a control circuit that performs at least one of settings for an AI (Artificial Intelligence) / ML (Machine Learning) model, settings for parameters of the AI / ML model, and settings related to data used in the AI / ML model, in accordance with a type related to sharing of information related to the AI / ML model between communication devices, and a transmission circuit that transmits information corresponding to the at least one setting to another communication device.

[0009] These comprehensive or specific aspects may be realized as a system, an apparatus, a method, an integrated circuit, a computer program, or a recording medium, or may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

[0010] According to an embodiment of the present disclosure, information related to AI / ML can be appropriately shared between communication devices.

[0011] Further advantages and benefits of one embodiment of the present disclosure will become apparent from the specification and drawings. Such advantages and / or benefits may be provided by some embodiments and features described in the specification and drawings, respectively, but not necessarily all of them may be provided to obtain one or more identical features.

[0012] Diagram showing examples of types related to the transfer of AI / ML models to each terminal and the passing of AI / ML parametersDiagram showing an example of Type 1 control operation where there is no inter-terminal transfer of AI / ML models and AI / ML parametersDiagram showing an example of Type 2 control operation where there is inter-terminal transfer of AI / ML parametersDiagram showing an example of Type 3 control operation where there is inter-terminal transfer of AI / ML models and AI / ML parametersDiagram showing an example of an AIML measurement type definitionDiagram showing an example configuration where the AI / ML model is a neural network (NN) modelDiagram showing an example of the format of the Measurement Request elementDiagram showing an example of an AIML Measurement Type definitionDiagram showing an example of a method for converting measurement elementsDiagram showing an example of the format of the Data collection elementDiagram showing an example of the format of an NDP Announcement frameDiagram showing an example of the format of the Sounding Dialogue Token field in an NDP Announcement frameDiagram showing an example of an NDP Announcement frame variant definitionDiagram showing an example of the format of the STA Info field for an EHT NDP Announcement frameDiagram showing an example of the format of the STA Info field for an AIML NDP Announcement frameAID11 in an NDP Announcement frame Diagram showing an example of a subfield definitionDiagram showing an example of a data collection frame definition for each application / use caseDiagram showing an example of a control operation when data collection is performed in the application layerDiagram showing an example of an Action frame formatDiagram showing an example of an Action frame formatDiagram showing an example of an AI / ML algorithm ID (AIML Algorithm ID) definitionDiagram showing an example of an AIML model element formatDiagram showing an example of an AIML parameter element formatDiagram showing an example of an AIML model parameter element formatDiagram showing an example of an AI / ML algorithm ID (AIML Algorithm ID) definitionDiagramDiagram showing an example of the setup frame formatDiagram showing an example of the AIML parameter setup frame formatDiagram showing an example of the AIML model parameter setup frame formatDiagram showing an example of the definition of an AI / ML algorithm ID (AIML Algorithm ID)Diagram showing an example of an output parameter report for each AIML measurement typeDiagram showing an example of operation after an output parameter report for each AIML measurement typeDiagram showing an example of the configuration of a Controller NodeDiagram showing an example of the configuration of a Controlled Node

[0013] Hereinafter, each embodiment of the present disclosure will be described in detail with reference to the drawings.

[0014] (First Embodiment) As described above, application of AI / ML to wireless LANs is being considered. For example, in an AI / ML model such as a neural network, training (or learning) is performed, parameters of the AI / ML model (hereinafter referred to as AI / ML parameters) are set, and parameters suitable for wireless LAN communications are determined through inference from the AI / ML model. As one use case of applying AI / ML to such wireless LANs, a method of sharing information related to AI / ML, such as the AI / ML model and the AI / ML parameters, between communication devices is being considered.

[0015] In the first embodiment of the present disclosure, a data collection request / response, an AIML model setting, and an AIML parameter setting are selected and shared between terminals in accordance with the type of transfer of an AI / ML model to each terminal and the type of transfer of AI / ML parameters. Note that an AIML model may also be referred to as an AI / ML model or an AI model.

[0016] <Examples of Three Types> Figure 1 is a diagram showing examples of types related to the transfer of AI / ML models to each terminal and the exchange of AI / ML parameters. Note that a Controller Node (e.g., an Access Point (hereinafter, AP)) and a Controlled Node (e.g., a non-AP Station (hereinafter, non-AP STA)) are each examples of a terminal. Furthermore, a Controller Node (e.g., an AP) and a Controlled Node (e.g., a non-AP STA) may each be referred to as a communication device.

[0017] As shown in the example in Figure 1, types are classified according to whether or not information is transferred between devices, the source of the AI / ML model transfer, and the location where AI / ML training and inference are performed. Type 1 in Figure 1 is a type in which there is no transfer of AI / ML models and AI / ML parameters between devices, Type 2 is a type in which there is transfer of AI / ML parameters between devices, and Type 3 is a type in which there is transfer of AI / ML models and AI / ML parameters between devices. Below, examples of control operations for each type are shown in Figures 2 to 4.

[0018] Figure 2 shows an example of Type 1 control operation in which there is no inter-terminal transfer of AI / ML models and AI / ML parameters. In Type 1, it is assumed that the AI / ML model is stored in the terminal or transferred from outside the WLAN network. As shown in Figure 2, Type 1 can be divided into Type 1-1, in which AI / ML training and inference are performed only at the Controller Node (e.g., AP), and Type 1-2, in which AI / ML training and inference are performed at both the Controller Node (e.g., AP) and a Controlled Node (e.g., non-AP STA).

[0019] Before a series of AI / ML operations, the Controller Node and Controlled Node exchange capability information in management frames. Capability information includes the following: - Whether AI / ML operations are supported (for example, whether AI / ML operations can be supported) - Types of supported AI / ML models / algorithms - Types and ranges of supported parameters

[0020] The above-mentioned capability information may be included in the capability information field, or may be defined as a capability element for AI / ML (e.g., AIML capability element) and included in the management frame.

[0021] As shown in Figure 2, a series of operations related to AI / ML includes operations (or control operations) in each phase: a data collection phase, a model training phase, a model inference phase, and phases such as "AI model execution, monitoring, update, and deactivation." Examples of operations in each phase are described below.

[0022] In the data collection phase, information required for AI / ML operations is collected between the Controller Node and the Controlled Node. Figure 2 shows an example of information exchange using a data collection request and a data collection response. In Type 1-1, the Controller Node collects information required for AI / ML operations. In Type 1-2, the Controller Node and the Controlled Node collect information required for AI / ML operations. Information required for AI / ML operations may be, for example, information required for training, inference, etc. using an AI / ML model.

[0023] After the data collection phase, the model training phase involves model training and learning to determine the configuration of the AI ​​model and / or AI / ML parameters.

[0024] Determining the configuration of an AI model corresponds to, for example, determining the type of AI / ML model and the configuration information corresponding to that type. Examples of AI / ML model types include neural networks (NN), k-means, k-nearest neighbors, CNN, and RNN. If the AI / ML model type is NN, examples of configuration information include the number of NN input nodes, the number of NN hidden layers, the number of NN hidden layer nodes, and the number of NN output nodes. If the AI / ML model type is k-means, examples of configuration information include the number of clusters and the upper limit of iterations.

[0025] AI / ML parameters are determined according to the configuration of the AI ​​model described above. For example, if the AI / ML model is a neural network, parameter information may include a weight matrix, bias vector, activation function (Sigmoid, Tanh, ReLU, ELU, etc.), etc. If the AI / ML model is a k-means algorithm, parameter information may include an initialization method, number of executions, relative tolerance, random number seed, etc.

[0026] Then, in the model inference phase, inference is performed. After the model training phase and model inference phase, output parameters, which are the results of the training, learning, and inference, are notified to other terminals. Note that in this disclosure, parameters determined according to the configuration of the AI ​​model described above are called AI / ML parameters, and the results of the training, learning, and inference are called output parameters. However, the AI / ML parameters may be included in the configuration information of the AI / ML model, and the results of the training, learning, and inference may also be referred to as AI / ML parameters.

[0027] After the output parameters are notified to other terminals, operations such as execution, monitoring, update, and deactivation of the AI ​​model are performed. Here, data collection may be performed by a data collection request / response in order to monitor and update the model.

[0028] As shown in Figure 2, in Type 1-1, where AI / ML training and inference are performed only on the Controller Node, the Controller Node performs the series of operations described above, from data collection to AI model execution, monitoring, updating, and stopping. Also, as shown in Figure 2, in Type 1-2, where AI / ML training and inference are performed on the Controller Node and Controlled Node, both the Controller Node and Controlled Node perform the series of operations described above, from data collection to AI model execution, monitoring, updating, and stopping.

[0029] Figure 3 shows an example of Type 2 control operation in which AI / ML parameters are transferred between terminals. Type 2 control operation adds a step (or phase) to the Type 1 control operation shown in Figure 2, in which AI / ML parameter setup is exchanged between terminals after the AI ​​model configuration is established. As shown in Figure 3, Type 2 can be divided into Type 2-1, in which AI / ML training and inference are performed only at the Controller Node, as in Type 1, and Type 2-2, in which AI / ML training and inference are performed at both the Controller Node and the Controlled Node.

[0030] In Type 2-1, data collection for model training and model training / learning operations are performed only by the Controller Node, not by the Controlled Node. In this case, the AI ​​model is supplied from within the terminal or from outside the WLAN network, and AI / ML parameters are set. In Type 2-1, information corresponding to the AI / ML parameter setting is notified from the Controller Node to the Controlled Node, and the Controlled Node sets the parameters of the AI / ML model based on the notification. The information corresponding to the AI / ML parameter setting may be parameter information for the AI / ML parameters described above. Then, the control operations after model inference are performed by both the Controller Node and the Controlled Node.

[0031] In Type 2-2, the Controller Node and Controlled Node collect data for model training and perform model training and learning operations. In Type 2-2, information corresponding to AI / ML parameter settings is notified from the Controller Node to the Controlled Node, and from the Controlled Node to the Controller Node. The Controller Node and Controlled Node each set the parameters of the AI / ML model based on the notification. Then, the Controller Node and Controlled Node both perform control operations after model inference.

[0032] Figure 4 shows an example of Type 3 control operation in which an AI / ML model and parameters are transferred between terminals. Type 3 control operation adds a step to the Type 2 control operation shown in Figure 3 in which AI / ML parameter setup and AI / ML model setup are exchanged between terminals. As shown in Figure 4, Type 3 is assumed to be either Type 3-1, in which AI / ML training and inference are performed only at the Controller Node, as in Type 1, or Type 3-2, in which AI / ML training and inference are performed at both the Controller Node and the Controlled Node.

[0033] In Type 3-1, data collection for model training and model training / learning operations are performed only by the Controller Node, not by the Controlled Node. Note that in Type 3-1, information corresponding to the AI / ML model settings and information corresponding to the AI / ML parameter settings are notified from the Controller Node to the Controlled Node. The information corresponding to the AI / ML model settings may be the type of AI / ML model and / or configuration information corresponding to the model. Upon receiving notification of the information corresponding to the AI / ML model settings and AI / ML parameter settings, the Controlled Node establishes the configuration and parameters of the AI ​​model and performs control operations after model inference.

[0034] In Type 3-2, the Controller Node and Controlled Node collect data for model training and perform model training and learning operations. In Type 3-2, information corresponding to the AI / ML model settings and information corresponding to the AI / ML parameter settings are notified from the Controller Node to the Controlled Node, and from the Controlled Node to the Controller Node. The Controller Node and Controlled Node receive notifications of information corresponding to the AI / ML model settings and AI / ML parameter settings, and establish the AI ​​model configuration and parameters. Then, the Controller Node and Controlled Node together perform control operations after model inference.

[0035] The type of operation of the Controller Node and the Controlled Node may be specified in advance, or either the Controller Node or the Controlled Node may determine the type of operation. For example, either the Controller Node or the Controlled Node may determine the type of operation based on capability information.

[0036] According to the first embodiment described above, information on AI / ML models and / or AI / ML parameters can be efficiently shared between terminals in accordance with the transfer type of the AI / ML models and / or AI / ML parameters between terminals. This also allows information on AI / ML models and AI / ML parameters to be appropriately shared between communication devices.

[0037] The types of AI / ML model transfer to each terminal and AI / ML parameter transfer are not limited to the above types. For example, although three types are shown in FIG. 1, other types may be added.

[0038] (Embodiment 2) In embodiment 2, as a data collection request, a correspondence between a use case, a measurement value (or a measurement element) that is an input required by an AI / ML model, and a method for converting the value into an input format to the AI / ML model (or an output parameter format of AI / ML) is specified, and the correspondence is defined as an AIML measurement type and shared between terminals.

[0039] Figure 5 shows an example of an AIML measurement type definition. For each AIML measurement type, use cases, measurement element types, and a conversion method for converting the data into an input format for an AI / ML model are defined.

[0040] As shown in Fig. 5, the types of use cases include CSI Compression, Distributed Channel Access, Multi-AP (MAP), Link Adaptation, Beam Management, etc. However, use cases other than the examples shown in Fig. 5 may also be defined as the types of use cases.

[0041] As the measurement element, a measurement value is defined corresponding to each use case, as shown in Fig. 5. The number of measurement values ​​defined is not limited to one, and multiple values ​​may be defined.

[0042] The conversion method may include, for example, a method for quantizing measurements (e.g., changing the number of bits), an input index, and output content corresponding to a specified use case. Alternatively, instead of or in addition to specifying the format of input data to an AI / ML model (e.g., a neural network (NN)), the format (e.g., the number of quantization bits, the numerical representation format such as floating-point format or fixed-point format) of output parameters (e.g., learning / training results and / or inference results) of the AI / ML model may be specified as an indication of the conversion method.

[0043] The number of AIML measurement types that can be specified is not limited to one, and multiple AIML measurement types may be specified.

[0044] The Controller Node may include information indicating an AIML measurement type (e.g., a number or an index) or information indicating a use case (e.g., a number or an index) in a frame for a data collection request and transmit the frame. Alternatively, the Controller Node may transmit a frame corresponding to the AIML measurement type (or use case) for the data collection request. For example, if the use case is "CSI Compression," the Controller Node transmits an NDPA / NDP (Null Data PPDU (Physical layer Protocol Data Unit) Announcement / Null Data PPDU) frame. Furthermore, if the use case is "Distributed Channel Access," the Controller Node transmits a frame requesting statistical information on frame reception and / or frame transmission (e.g., throughput (data rate), frame error rate (e.g., CRC error rate) or number of frame errors (e.g., number of CRC errors), average remaining buffer space, delay time until transmission, channel busy rate, signal quality / strength). Also, for example, if the use case is "Multi-AP", the Controller Node may send a frame requesting AP (or BSS) statistical information.

[0045] Next, the conversion method shown in FIG. 5 will be explained using an example in which the AI / ML model is a neural network (NN) model.

[0046] FIG. 6 is a diagram showing a configuration example in which the AI / ML model is a neural network (NN) model.

[0047] As shown in Fig. 6, measurement values ​​corresponding to each use case are input to a neural network. As pre-processing before input to the neural network, the measurement values ​​are converted into a format suitable for input to the neural network. Furthermore, as post-processing after output from the neural network, the measurement values ​​are associated with the output of each use case. The conversion method shown in Fig. 5 includes the contents of these pre-processing and post-processing.

[0048] According to the second embodiment described above, by specifying the correspondence between the use case and the measurement values ​​and conversion methods that are the inputs required for the AI / ML model, and defining them as AIML measurement types, information related to AI / ML can be appropriately shared between terminals.

[0049] In the present disclosure, the method of data collection request and response is not particularly limited. Hereinafter, examples of the data collection request and response methods will be described in Embodiments 2-1 to 2-5.

[0050] (Embodiment 2-1) In embodiment 2-1, a data collection request is additionally defined as one of the measurement types of a measurement request element. In embodiment 2-1, a terminal making a data collection request notifies the terminal by including the measurement request element defined as a data collection request in management frames (beacon, association request / response, probe request / response, action frames, etc.). Note that this operation notified by the terminal is an example of an operation equivalent to the data collection request operation in embodiment 1 described above.

[0051] Fig. 7 is a diagram showing an example of the format of a Measurement Request element. Fig. 7 shows the relationship between the value and content of the Measurement Type field in the format of the Measurement Request element. In this embodiment 2-1, as shown in Fig. 7, an undefined value of 17 in the Measurement Type field is associated with AIML Measurement. When the value of the Measurement Type field is the undefined value of 17, the Measurement Request field is defined to include an AIML measurement type and an index for AI model input.

[0052] 7 shows an example in which "AIML Measurement" is associated with the value 17 in the Measurement Type field, but AIML Measurement may also be associated with other undefined values ​​other than 17. Also, while FIG. 7 shows a configuration in which one AIML measurement type and one AI model input index are included, multiple of each may also be included.

[0053] Fig. 8 is a diagram showing an example of an AIML Measurement Type definition. Fig. 8 defines the relationship between the value set in the AIML Measurement Type field in Fig. 7 and its contents. As shown in Fig. 8, for each AIML Measurement Type, a use case, a type of measurement element, and output parameters corresponding to the use case are defined.

[0054] Figure 9 shows an example of a conversion method for measurement elements. This defines how to convert each measurement element into an input format for an AI / ML model. For example, it defines the number of input and output bits and their respective formats. These AIML measurement types and measurement elements may be defined as vendor-specific, meaning their specifications are vendor-dependent.

[0055] According to the embodiment 2-1 described above, by additionally defining a data collection request as one of the Measurement types of the Measurement Request element and notifying it by including it in Management frames (Beacon, Association Request / Response, Probe Request / Response, Action frames), etc., it is possible to share between terminals what measurement type of data (e.g., measurement element) and what conversion method to use, thereby enabling information related to AI / ML to be shared appropriately between terminals.

[0056] (Embodiment 2-2) In embodiment 2-2, a data collection request element is newly defined. Note that in this embodiment, the newly defined data collection request element is called a Data Collection Element, but the present disclosure is not limited to this name. In embodiment 2-2, a terminal making a data collection request notifies the Data Collection Element by including it in Management frames (Beacon, Association Request / Response, Probe Request / Response, Action frames), etc. Note that the operation notified by this terminal is an example of an operation equivalent to the Data Collection Request operation in embodiment 1 described above.

[0057] Fig. 10 is a diagram showing an example of the format of a Data Collection Element. Here, the fact that a certain element is a Data Collection Element may be recognized by defining the Element ID or Element ID Extension to a specific, undefined value. In this case, as shown in Fig. 10, the Data Collection Element may be configured to include an AIML Measurement Type and an Index for AIML model input. The contents of these fields may be the same as those in embodiment 2-1.

[0058] Similar content including the AIML Measurement Type and Index for AIML model input may be defined as a new frame. For example, a new frame including the AIML Measurement Type and Index for AIML model input may be defined by setting the Subtype value or Control Frame Extension value of the Frame Control field in the frame to a specific value (undefined value).

[0059] According to the embodiment 2-2 described above, a new data collection request element (Data collection element) is defined and then included in management frames (Beacon, Association Request / Response, Probe Request / Response, Action frames), etc., and notified. This allows terminals to share information about the type of measurement data (e.g., measurement element) to be converted and the conversion method used, thereby enabling appropriate sharing of information related to AI / ML between terminals.

[0060] (Embodiment 2-3) In this embodiment 2-3, a data collection request is additionally defined in an NDP (Null Data Packet) Announcement frame. In this embodiment 2-3, a terminal making a data collection request notifies an NDP Announcement frame defined as a data collection request. Note that this operation of the terminal notifying is an example of an operation equivalent to the operation of the Data Collection Request in the above-mentioned embodiment 1.

[0061] Fig. 11 is a diagram showing an example of the format of an NDP Announcement frame. Fig. 12 is a diagram showing an example of the format of a Sounding Dialog Token field in the NDP Announcement frame. The Sounding Dialog Token field includes an NDP Announcement variant.

[0062] Fig. 13 is a diagram showing an example of the definition of an NDP Announcement frame variant. In Fig. 13, when the NDP Announcement Variant subfield is "3", the definition was conventionally an EHT NDP Announcement frame. In this embodiment, when the NDP Announcement Variant subfield is "3", it is defined as an EHT / AIML NDP Announcement frame, also supporting AIML.

[0063] Although it has been described herein that the NDP Announcement Variant subfield is "3" and is therefore compatible with AIML, the present disclosure is not limited to this. For example, the NDP Announcement Variant subfield may be configured to be compatible with AIML if it has a value other than "3."

[0064] An STA Info List is included in the NDP Announcement frame shown in Fig. 11. The STA Info List contains a plurality of STA Info fields, for example, as shown in Fig. 14 and / or Fig. 15.

[0065] Fig. 14 is a diagram showing an example of the format of the STA Info field in the case of an EHT NDP Announcement frame. Fig. 15 is a diagram showing an example of the format of the STA Info field in the case of an AIML NDP Announcement frame.

[0066] As shown in Fig. 15, the STA Info field in the case of an AIML NDP Announcement frame may include an AIML Measurement Type and an Index for AIML model input. Note that the contents of the AIML Measurement Type and Index for AIML model input fields in Fig. 15 may be the same as those in the above-described embodiment 2-1.

[0067] Note that whether the STA Info field is an AIML NDP Announcement frame may be determined based on whether the value of the AID11 subfield is a specific value (for example, 2008).

[0068] Fig. 16 is a diagram showing an example of the definition of the AID11 subfield in an NDP Announcement frame. As shown in Fig. 16, by defining the STA Info field as an AIML NDP Announcement frame to one of "2008" to "2042," which is a value not defined in any of the conventionally defined NDP Announcement frame types, VHT / HE / EHT / Ranging / Sensing, a terminal (e.g., a node) receiving the frame can recognize that the received frame is an AIML NDP Announcement frame.

[0069] According to the above-described embodiments 2-3, by additionally defining a data collection request in the NDP Announcement frame, it is possible to share between terminals what measurement type of data (e.g., measurement elements) to convert and by what conversion method, thereby enabling information related to AI / ML to be appropriately shared between terminals.

[0070] (Embodiment 2-4) This embodiment 2-4 defines data collection frames corresponding to measurement elements specific to an application / use case. In this embodiment 2-4, the terminal that issues a data collection request notifies the data collection frames defined corresponding to measurement elements specific to the application / use case. Note that this notification operation by the terminal is an example of an operation equivalent to the Data Collection Request operation in the above-mentioned embodiment 1.

[0071] Fig. 17 is a diagram showing an example of a definition of a data collection frame for each application / use case. Fig. 17 shows measurement elements required for each application / use case and frames required to collect the measurement elements. For example, in the case of CSI (Channel State Information) Compression, the measurement elements are CSI (SNR (Signal to Noise Ratio), Φ, Ψ (Φ, Ψ are quantized angles)), and the triggers or requests required to acquire CSI are an NDP Announcement frame, a sounding NDP, and a BFRP (Beamforming Report Poll) Trigger frame, and the report is Compressed Beamforming / CQI.

[0072] A frame indicated by a right-pointing arrow (i.e., "→") in the data collection frames column of Fig. 17 may be transmitted by a terminal making a data collection request (e.g., Controller Node) and received by a terminal making a data collection response (e.g., Controlled Node). Also, a frame indicated by a left-pointing arrow (i.e., "←") may be transmitted by a terminal making a data collection response (e.g., Controlled Node) and received by a terminal making a data collection request (e.g., Controller Node). The Controller Node and Controlled Node may transmit and receive frames in the order shown in Fig. 17, but this is not a limitation.

[0073] Taking the case where the application / use case is CSI Compression or Beam Management as an example, in this case, at least one of an NDP Announcement frame, a sounding NDP, and a BFRP Trigger is transmitted as a data collection trigger or request by a terminal (e.g., a Controller Node) making a data collection request, and is received by a terminal (e.g., a Controlled Node) making a data collection response. Then, as a data collection response or report, Compressed Beamforming or CQI is transmitted by a terminal (e.g., a Controlled Node) making a data collection response, and is received by a terminal (e.g., a Controller Node) making a data collection request. Note that the NDP Announcement frame, the sounding NDP, the BFRP Trigger, and Compressed Beamforming or CQI may be transmitted and received in the order shown in FIG. 17, or in an order different from the order shown in FIG. 17.

[0074] According to the above-described embodiments 2 to 4, by defining data collection frames corresponding to measurement elements specific to an application / use case, information related to AI / ML can be shared between terminals.

[0075] (Embodiment 2-5) This embodiment 2-5 takes the form of performing data collection in the application layer. In this embodiment 2-5, a terminal that makes a data collection request requests data collection in the application layer. Note that the operation of this terminal is an example of an operation equivalent to the operation of the data collection request in the above-mentioned embodiment 1.

[0076] 18 is a diagram showing an example of a control operation when data collection is performed in the application layer. In FIG. 18, data collection in the control operation of type 1-2 shown in FIG. 2 is performed in the application layer, for example.

[0077] In Fig. 18, in the phases of operations such as data collection, model execution, monitoring, update, and deactivation, information required for AI / ML operations is exchanged between the SME (station management entity) / MLME (MAC sublayer management entity) and the application layer via data collection request / response. Note that this may also be used for general applications other than WLAN.

[0078] According to the above-described embodiments 2 to 5, data collection is performed at the application layer, making it possible to share information related to AI / ML between applications and terminals.

[0079] In addition, in the second to fifth embodiments, an example in which data collection in the control operation of type 1-2 shown in Fig. 2 is performed in the application layer is shown, but the present disclosure is not limited to this. For example, data collection in the control operation of each type shown in the first embodiment (for example, each type shown in Figs. 2 to 4) may be performed in the application layer.

[0080] In the above-described embodiments 2-1 to 2-5, the terminal making the data collection request may be a Controller Node or a Controlled Node. For example, in the cases of types 1-1, 2-1, and 3-1 shown in embodiment 1, the terminal making the data collection request may be a Controller Node. For example, in the cases of types 1-2, 2-2, and 3-2 shown in embodiment 1, the terminal making the data collection request may be a Controller Node or a Controlled Node.

[0081] (Embodiment 3) In the above-mentioned embodiments 2-1 to 2-5, a data collection request / response method for exchanging information corresponding to a data collection request / response between terminals has been described. In this embodiment 3, a data collection request / response method for exchanging information corresponding to a data collection request / response, information corresponding to an AIML model setting, and information corresponding to an AIML parameter setting between terminals will be described.

[0082] For example, in the third embodiment, when the Category of the Action frame is a specific value, the frame is defined as including a data collection request / response, an AIML model setting, and an AIML parameter setting. For example, the specific value may be an undefined value such as 39 or 40.

[0083] 19A and 19B are diagrams showing examples of the format of an Action frame. The Category in FIGS. 19A and 19B defines the type of Action frame. Here, the types of data collection request / response, AIML model setting, and AIML parameter setting are defined in currently undefined values ​​such as 39 and 40. In the examples of FIGS. 19A and 19B, Data Collection request / response is defined in 39, and AIML model setting and AIML parameter setting (AIML Setup) are defined in 40. Note that although AIML model setting and AIML parameter setting are defined in the same value 40, the AIML model setting and AIML parameter setting may be defined as different values.

[0084] Fig. 19A shows a case where Category is set to Data Collection. A case where Category is set to Data Collection corresponds to a case where Category contains a value of 39, for example. In the example of Fig. 19A, a Data Collection Action field is included after Category.

[0085] The Data Collection Action field indicates, for example, whether the Action frame is a Data Collection Request or a Data Collection Report. Fig. 19A shows an example in which the Data Collection Request is defined as 0 and the Data Collection Report is defined as 1. Note that defining the Data Collection Request as 0 may be equivalent to the Data Collection Request being associated with 0. The content of the Data Collection Elements following the Data Collection Action field changes depending on the value of this Data Collection Action field.

[0086] When the value of the Data Collection Action field indicates a data collection request (for example, when the value of the Data Collection Action field is 0), for example, the Data Collection Action field is followed by AIML Measurement Type and Index for AIML model input. Note that the contents of the AIML Measurement Type and Index for AIML model input fields may be the same as those in embodiment 2-1.

[0087] If the value of the Data Collection Action field indicates a data collection response (Data Collection Report) (for example, if the value of the Data Collection Action field is 1), the result of the data collection request (Data Collection Report Info) is included after the Data Collection Action field.

[0088] 19B shows a case where Category is set to AIML model settings and AIML parameter settings (AIML Setup). A case where Category is set to AIML model settings and AIML parameter settings (AIML Setup) corresponds to a case where Category includes a value of 40, for example. In the example of FIG. 19B, an AIML Setup Action field is included after the Category field.

[0089] The AIML Setup Action field indicates, for example, any one of an AIML model setup request, an AIML model setup response, an AIML parameter setup request, and an AIML parameter setup response. In Fig. 19B, the AIML model setup request is defined as 0, the AIML model setup response is defined as 1, the AIML parameter setup request is defined as 2, and the AIML parameter setup response is defined as 3. The content of the AIML Setup Elements following the AIML Setup Action field changes depending on the value of this AIML Setup Action field.

[0090] When the AIML Setup Action field indicates an AIML Model Setup Request (for example, when the value of the AIML Setup Action field is 0), for example, an AI / ML algorithm ID and configuration information are included after the AIML Setup Action field.

[0091] When the AIML Setup Action field indicates an AIML model setup report (for example, when the value of the AIML Setup Action field is 1), the AIML model setup result is included after the AIML Setup Action field.

[0092] When the AIML Setup Action field indicates an AIML Parameter Setup Request (for example, when the value of the AIML Setup Action field is 2), for example, an AI / ML algorithm ID and parameter information are included after the AIML Setup Action field.

[0093] When the AIML Setup Action field indicates an AIML Parameter Setup Report (for example, when the value of the AIML Setup Action field is 3), the AIML parameter setup result is included after the AIML Setup Action field.

[0094] FIG. 19C is a diagram illustrating an example of the definition of an AI / ML algorithm ID. As shown in FIG. 19C , for the AI / ML algorithm ID, correspondence with the type of AI / ML model is defined, and correspondence with the respective configuration information (configuration information (Configuration info) in FIG. 19B ) and parameter information (parameter information (Parameter info) in FIG. 19B ) is defined. For example, types of AI / ML models include neural networks (NNs), k-means, k-nearest neighbors, convolutional neural networks (CNNs), and recurrent neural networks (RNNs). When the type of AI / ML model is NN, examples of configuration information include the number of NN input nodes, the number of NN hidden layers, the number of NN hidden layer nodes, and the number of NN output nodes. When the type of AI / ML model is k-means, examples of parameter information include the number of clusters and the upper limit of iterations. When the type of AI / ML model is NN, examples of parameter information include a weight matrix, a bias vector, an activation function (Sigmoid, Tanh, ReLU, ELU, etc.), etc. If the AI / ML model type is k-means, examples include initialization method, number of runs, relative tolerance, random number seed, etc.

[0095] Note that the above definition defines a specific AI / ML algorithm, but it is also possible to define an AI / ML algorithm as vendor-specific, whose specifications are vendor-dependent.

[0096] According to the third embodiment described above, when the Category of the Action frame is a specific value, it is defined as a frame that includes at least one of a data collection request / response, an AIML model setting, and an AIML parameter setting, thereby enabling information related to AI / ML to be shared between terminals.

[0097] In the above-described third embodiment, an example has been shown in which the Action frame is defined as a frame including at least one of a data collection request / response, an AIML model setting, and an AIML parameter setting, but the present disclosure is not limited to this. For example, any of the above-described embodiments 2-1 to 2-5 may be applied to the data collection request / response, and an Action frame may be used for the AIML model setting and the AIML parameter setting, as in the above-described third embodiment.

[0098] (Embodiment 4) In this embodiment 4, information elements (AIML model element, AIML parameter element, AIML model parameter element) containing the contents of the AIML model setting and the AIML parameter setting are included in a management frame (Beacon, Association Request / Response, Probe Request / Response, Action frames, etc.) and notified.

[0099] Fig. 20A is a diagram showing an example of the format of an AIML model element. Fig. 20B is a diagram showing an example of the format of an AIML parameter element. Fig. 20C is a diagram showing an example of the format of an AIML model parameter element. Whether a certain element is an AIML model element, an AIML parameter element, or an AIML model parameter element may be recognized by defining the Element ID or Element ID Extension of the element to a specific value that is currently undefined.

[0100] 20A , in the case of an AIML model element, AIML Model info is included. The contents of the AIML Model info include, for example, an AIML Algorithm ID and Configuration info. Note that the AIML Algorithm ID and Configuration info may be the same as those in the third embodiment.

[0101] 20B , in the case of an AIML parameter element, AIML Parameter info is included. The contents of the AIML Parameter info include, for example, an AIML Algorithm ID and Parameter info. Note that the AIML Algorithm ID and Parameter info may be the same as those in the third embodiment.

[0102] As shown in Fig. 20C , in the case of an AIML model parameter element, AIML Model Parameter info is included. The contents of AIML Model Parameter info include, for example, AIML Algorithm ID, Configuration info, and Parameter info. The AIML Algorithm ID, Configuration info, and Parameter info may be the same as those in the third embodiment. Note that, although the example in Fig. 20C includes both Configuration info and Parameter info, it may also be possible to specify whether or not each of them is included.

[0103] Fig. 20D is a diagram showing an example of the definition of an AI / ML algorithm ID. As shown in Fig. 20D, for the AI / ML algorithm ID, correspondence with the type of AI / ML model is defined, and correspondence with each configuration information and parameter information is defined. Note that Fig. 20D is similar to Fig. 19C, and therefore description thereof will be omitted.

[0104] According to the fourth embodiment described above, information elements including the contents of AIML model settings and AIML parameter settings are included in management frames and notified, thereby enabling information related to AI / ML to be shared between terminals.

[0105] Fifth Embodiment In a fifth embodiment, frames including AIML model settings and AIML parameter settings (AIML model setup frame, AIML parameter setup frame, AIML model parameter setup frame) are newly defined.

[0106] Fig. 21A is a diagram showing an example of the format of an AIML model setup frame. Fig. 21B is a diagram showing an example of the format of an AIML parameter setup frame. Fig. 21C is a diagram showing an example of the format of an AIML model parameter setup frame. Whether a certain frame is an AIML model setup frame, an AIML parameter setup frame, or an AIML model parameter setup frame may be recognized by setting the Subtype value or Control Frame Extension value of the Frame Control field to a specific value (for example, an undefined value).

[0107] 21A , in the case of an AIML model setup frame, AIML Model info is included. The contents of the AIML Model info include, for example, an AIML Algorithm ID and Configuration info. The AIML Algorithm ID and Configuration info may be the same as those in the third embodiment.

[0108] 21B , in the case of an AIML parameter setup frame, AIML Parameter info is included. The contents of the AIML Parameter info include, for example, an AIML Algorithm ID and Parameter info. The AIML Algorithm ID and Parameter info may be the same as those in the third embodiment.

[0109] As shown in Fig. 21C , in the case of an AIML model parameter setup frame, AIML Model Parameter info is included. The contents of the AIML Model Parameter info include, for example, an AIML Algorithm ID, Configuration info, and Parameter info. The AIML Algorithm ID, Configuration info, and Parameter info may be the same as those in the third embodiment. Note that, although the example in Fig. 21C is configured to include both Configuration info and Parameter info, it may also be possible to specify whether or not each of them is included.

[0110] Fig. 21D is a diagram showing an example of the definition of an AI / ML algorithm ID. As shown in Fig. 21D, for the AI / ML algorithm ID, correspondence with the type of AI / ML model is defined, and correspondence with each configuration information and parameter information is defined. Note that Fig. 21D is similar to Fig. 19C, and therefore description thereof will be omitted.

[0111] According to the fifth embodiment described above, by newly defining a frame including an AIML model setting and an AIML parameter setting, information related to AI / ML can be shared between terminals.

[0112] Sixth Embodiment In a sixth embodiment, an output parameter report is sent as a result of training or inference by AI / ML.

[0113] FIG. 22 shows an example of an output parameter report for each AIML measurement type. For each AIML measurement type, it shows a use case, an output example, and the element / field / frame that forms the basis of the output report. For each AIML measurement type / use case, the output report is configured by adding or replacing the output, which is the result of AI / ML training or inference, to the element / field / frame that forms the basis of the output report. A portion of an already defined element / field / frame may be used, or new elements / fields / frames may be added. The output may also be included in the AI / ML parameter setup and reported. The output parameter may be specified by notifying the AIML measurement type, which is the use case number, or the output parameter specified by the AIML measurement type and index may be notified.

[0114] According to the sixth embodiment described above, information related to AI / ML can be shared between terminals by reporting output parameters (output parameter report) as training or inference results by AI / ML.

[0115] (Seventh embodiment) In the seventh embodiment, after an output parameter report of the AI / ML training or inference results is received, the terminal that has received the parameter report operates with the updated parameters.

[0116] Fig. 23 shows an example of operation after an output parameter report for each AIML measurement type. Fig. 23 shows use cases and their operation examples for each AIML measurement type. For each AIML measurement type / use case, upon receiving an output parameter report, the operation shown in the example of Fig. 23 is performed using the updated parameters.

[0117] According to the seventh embodiment described above, after an output parameter report of the AI / ML training or inference results is received, the terminal that has received the parameter report operates with the updated parameters, thereby enabling operation optimized by AI / ML.

[0118] Note that, as in Type 1-1 in Fig. 2 shown in the first embodiment, a terminal that performs AI / ML training or inference while not receiving output parameter notifications from other terminals (for example, the Controller Node of Type 1-1 in Fig. 2) may update parameters based on its own AI / ML training or inference results and perform operations. Also, a terminal that receives output parameter notifications from other terminals may perform updated operations based on the notified output parameters and its own AI / ML training or inference results.

[0119] The above-described embodiments may be used in combination as appropriate, or may be used by switching between them.

[0120] Furthermore, in the above-described embodiments, the sharing of information related to AI / ML between two terminals (a controlled node and a controller node) has been described as an example, but the present disclosure is not limited to this. Information related to AI / ML may be shared between three or more terminals. In this case, at least one controller node may be included among the three or more terminals.

[0121] <Example of Device Configuration> An example of the configuration of a Controller Node (for example, an AP) and a Controlled Node (for example, a non-AP STA) according to each of the above-described embodiments will be described.

[0122] FIG. 24 is a diagram illustrating an example of the configuration of the Controller Node 100. The Controller Node 100 may also be referred to as a communication device 100. In the communication device 100 illustrated in FIG. 24 , a control unit 101 (e.g., corresponding to a control circuit) performs at least one of the following settings: an AI (Artificial Intelligence) / ML (Machine Learning) model setting (e.g., the AIML model setting described above), an AI / ML model parameter setting (e.g., the AIML parameter setting described above), and a setting related to data used in the AI / ML model (e.g., the data collection setting described above), corresponding to a type (e.g., the type illustrated in FIG. 1 ) related to sharing of information related to an AI / ML model between communication devices. A wireless transceiver 102 (e.g., corresponding to a transmission circuit) transmits information corresponding to at least one setting to another communication device (e.g., the Controlled Node 200).

[0123] The control unit 101 sets the measurement type, measurement elements, output parameters, conversion method, etc. as settings related to data used in the AI / ML model, based on the correspondence relationships exemplified in Figures 5, 8, and 9, corresponding to, for example, Type 1 in Figure 1. The wireless transceiver unit 102 transmits information corresponding to this setting (for example, the above-mentioned AIML Measurement Type and Index for AIML model input) to the controlled node 200.

[0124] The control unit 101 performs settings related to data used in the AI / ML model and settings of parameters of the AI / ML model, for example, corresponding to Type 2 in Fig. 1 . The control unit 101 sets parameter information such as that exemplified in Fig. 19C as the setting of the parameters of the AI / ML model. The wireless transceiver unit 102 transmits information corresponding to the setting of data used in the AI / ML model (for example, the above-mentioned AIML Measurement Type and Index for AIML model input) and information corresponding to the setting of the parameters of the AI / ML model (for example, the above-mentioned Parameter info, etc.) to the controlled node 200.

[0125] The control unit 101 performs settings related to data used in the AI / ML model, settings of parameters for the AI / ML model, and settings of the AI / ML model, for example, corresponding to Type 3 in Fig. 1 . The control unit 101 sets an AIML algorithm and configuration information as illustrated in Fig. 19C as the settings of the AI / ML model. The wireless transceiver unit 102 transmits, to the controlled node 200, information corresponding to the settings related to data used in the AI / ML model (e.g., the above-mentioned AIML Measurement Type and Index for AIML model input), information corresponding to the settings of parameters for the AI / ML model (e.g., the above-mentioned Parameter info, etc.), and information corresponding to the settings of the AI / ML model (e.g., the above-mentioned AIML Algorithm ID and Configuration info, etc.).

[0126] FIG. 25 is a diagram illustrating an example of the configuration of a controlled node 200. The controlled node 200 may also be referred to as a communication device 200. In the communication device 200 illustrated in FIG. 25 , a control unit 201 (e.g., corresponding to a control circuit) performs at least one of the following settings: an artificial intelligence (AI) / machine learning (ML) model setting, an AI / ML model parameter setting, and a setting related to data used in the AI / ML model, in accordance with a type related to sharing of information related to an AI / ML model between communication devices. A wireless transceiver 202 (e.g., corresponding to a receiving circuit) receives information corresponding to at least one setting from another communication device.

[0127] Although the embodiments have been described above with reference to the drawings, the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications or alterations within the scope of the claims. It is understood that such modifications or alterations also fall within the technical scope of the present disclosure. Furthermore, the components in the embodiments may be combined in any manner without departing from the spirit of the present disclosure.

[0128] In the above-described embodiments, the notation "... part" used for each component may be replaced with other notations such as "... circuit," "... assembly," "... device," "... unit," or "... module."

[0129] The interface names (frame names), field names, or subfield names described in the above-described embodiments may be other names.

[0130] In addition, in each of the above-described embodiments, the field (or subfield) used for notifying control information is an example, and other fields or subfields may be used. Furthermore, the number of bits used for notifying control information in each field or subfield is an example, and other numbers of bits may be used.

[0131] Furthermore, the signal formats described in each of the above-mentioned embodiments are merely examples, and other configurations may be used in which at least one of other fields is added and some fields is deleted, and other configurations may be used in which at least one of other subfields is added and some subfields are deleted in each of the above-mentioned fields.

[0132] Furthermore, in the above embodiment, as an example, a case based on the format defined in IEEE 802.11 has been described, but the format to which an embodiment of the present disclosure is applied is not limited to the IEEE 802.11 format.

[0133] The present disclosure can be realized by software, hardware, or software in conjunction with hardware. Each functional block used in the description of the above embodiments may be partially or entirely realized as an LSI, which is an integrated circuit, and each process described in the above embodiments may be partially or entirely controlled by a single LSI or a combination of LSIs. The LSI may be composed of individual chips, or may be composed of a single chip that includes some or all of the functional blocks. The LSI may have data input and output. Depending on the degree of integration, the LSI may also be called an IC, system LSI, super LSI, or ultra LSI.

[0134] The integrated circuit method is not limited to LSI, and may be realized by a dedicated circuit, a general-purpose processor, or a dedicated processor. Also, a field programmable gate array (FPGA) that can be programmed after LSI manufacturing, or a reconfigurable processor that can reconfigure the connections and settings of circuit cells within the LSI, may be used. The present disclosure may be realized as digital processing or analog processing.

[0135] Furthermore, if an integrated circuit technology that can replace LSI emerges due to advances in semiconductor technology or other derivative technologies, it is natural that such technology may be used to integrate functional blocks. The application of biotechnology, etc. is also a possibility.

[0136] The present disclosure may be implemented in any type of apparatus, device, or system (collectively referred to as a communications apparatus) that has a communications function. The communications apparatus may include a radio transceiver and processing / control circuitry. The radio transceiver may include a receiver and a transmitter, or both functions. The radio transceiver (transmitter and receiver) may include a radio frequency (RF) module and one or more antennas. The RF module may include an amplifier, an RF modulator / demodulator, or the like. Non-limiting examples of communication devices include telephones (e.g., cell phones, smartphones), tablets, personal computers (PCs) (e.g., laptops, desktops, notebooks), cameras (e.g., digital still / video cameras), digital players (e.g., digital audio / video players), wearable devices (e.g., wearable cameras, smartwatches, tracking devices), game consoles, digital book readers, telehealth / telemedicine devices, communication-enabled vehicles or mobile transportation (e.g., cars, airplanes, ships), and combinations of the above devices.

[0137] The communication devices are not limited to portable or mobile devices, but also include any kind of non-portable or fixed equipment, devices, and systems, such as smart home devices (such as home appliances, lighting equipment, smart meters or measuring devices, control panels, etc.), vending machines, and any other "things" that may exist on an IoT (Internet of Things) network.

[0138] Communications include data communications via cellular systems, wireless LAN systems, communication satellite systems, and the like, as well as data communications via combinations of these.

[0139] A communications apparatus also includes devices such as controllers and sensors connected or coupled to a communications device that performs the communications functions described in this disclosure, such as controllers and sensors that generate control and data signals used by the communications device to perform the communications functions of the communications apparatus.

[0140] The communication apparatus also includes infrastructure facilities, such as base stations, access points, and any other apparatus, device, or system that communicates with or controls the various apparatuses listed above, but are not limited to these.

[0141] Furthermore, in recent years, in the field of IoT (Internet of Things) technology, CPS (Cyber ​​Physical Systems) has been attracting attention as a new concept that creates new added value by linking information between physical space and cyberspace. This CPS concept can also be adopted in the above-described embodiment.

[0142] That is, as a basic configuration of a CPS, for example, an edge server located in physical space and a cloud server located in cyberspace can be connected via a network, and processing can be distributed and performed by processors installed on both servers. Here, it is preferable that each piece of processing data generated in the edge server or cloud server is generated on a standardized platform, and the use of such a standardized platform can improve the efficiency of building a system that includes a variety of sensor groups and IoT application software.

[0143] A communication device according to one embodiment of the present disclosure includes a control circuit that performs at least one of settings for an AI (Artificial Intelligence) / ML (Machine Learning) model, settings for parameters of the AI / ML model, and settings related to data used in the AI / ML model, in accordance with a type related to sharing of information related to the AI / ML model between communication devices, and a transmission circuit that transmits information corresponding to the at least one setting to another communication device.

[0144] In one embodiment of the present disclosure, when the control circuit performs settings related to data used in the AI / ML model, the settings are performed based on the correspondence between multiple use cases of the AI / ML model, measurement elements, and conversion methods for the measurement elements.

[0145] In one embodiment of the present disclosure, the transmitting circuitry transmits information corresponding to the at least one setting using an action frame, where a category field of the action frame is set to a specific value.

[0146] In one embodiment of the present disclosure, when the control circuit sets the AI / ML model and the parameters of the AI / ML model, the transmitting circuit transmits a management frame having information corresponding to the setting of the AI / ML model and the parameters of the AI / ML model.

[0147] In one embodiment of the present disclosure, when the control circuit sets the AI / ML model and the parameters of the AI / ML model, the transmitting circuit transmits a specific frame having information corresponding to the setting of the AI / ML model and the parameters of the AI / ML model, and the subtype value or control frame extension value of the frame control field of the specific frame is set to a specific value.

[0148] In one embodiment of the present disclosure, the type is set from a plurality of type candidates that differ from each other in at least one of whether or not information corresponding to the settings of the AI / ML model is transmitted, whether or not information corresponding to the settings of the parameters of the AI / ML model is transmitted, and whether or not information corresponding to the settings regarding the data used in the AI / ML model is transmitted.

[0149] In one embodiment of the present disclosure, the control circuit determines an output corresponding to a use case as a result of training or inference by the AI / ML model, and the transmission circuit transmits the determined output to the other communication device.

[0150] In an embodiment of the present disclosure, when the control circuit acquires update information from the other communication device, the control circuit updates an operation related to communication based on the update information.

[0151] In a communication method according to one embodiment of the present disclosure, a communication device performs at least one of settings for an AI (Artificial Intelligence) / ML (Machine Learning) model, settings for parameters of the AI / ML model, and settings related to data used in the AI / ML model, in accordance with a type related to sharing of information related to the AI / ML model between communication devices, and transmits information corresponding to the at least one setting.

[0152] The disclosures of the specification, drawings and abstract contained in Japanese Patent Application No. 2023-127890, filed on August 4, 2023, are incorporated herein by reference in their entirety.

[0153] One embodiment of the present disclosure is useful in wireless communication systems.

[0154] 100 Controller Node (communication device) 101, 201 Control unit 102, 202 Wireless transmission / reception unit 200 Controlled Node (communication device)

Claims

1. A communication device comprising: a control circuit that performs at least one of settings for an AI (Artificial Intelligence) / ML (Machine Learning) model, settings for parameters of the AI / ML model, and settings related to data used in the AI / ML model, in response to a type related to sharing of information related to an AI / ML model between communication devices; and a transmission circuit that transmits information corresponding to the at least one setting to another communication device.

2. The communication device according to claim 1, wherein, when the control circuit performs settings related to data used in the AI / ML model, the settings are performed based on a correspondence relationship between a plurality of use cases of the AI / ML model, measurement elements, and conversion methods of the measurement elements.

3. The communication device according to claim 1, wherein the transmission circuit transmits information corresponding to the at least one setting using an action frame, the action frame having a category field set to a specific value.

4. The communication device according to claim 1, wherein, when the control circuit sets the AI / ML model and the parameters of the AI / ML model, the transmission circuit transmits a management frame having information corresponding to the setting of the AI / ML model and the parameters of the AI / ML model.

5. The communication device according to claim 1, wherein, when the control circuit sets the AI / ML model and the parameters of the AI / ML model, the transmission circuit transmits a specific frame having information corresponding to the setting of the AI / ML model and the parameters of the AI / ML model, and the specific frame has a subtype value or a control frame extension value of a frame control field set to a specific value.

6. The communication device of claim 1, wherein the type is set from a plurality of type candidates differing from each other in at least one of the following: whether or not to transmit information corresponding to settings of the AI / ML model, whether or not to transmit information corresponding to settings of parameters of the AI / ML model, and whether or not to transmit information corresponding to settings related to data used in the AI / ML model.

7. The communication device of claim 1, wherein the control circuit determines an output corresponding to a use case as a result of training or inference by the AI / ML model, and the transmission circuit transmits the determined output to the other communication device.

8. The communication device according to claim 1, wherein, when said control circuit receives update information from said other communication device, said control circuit updates an operation relating to communication based on said update information.

9. A communication method, in which a communication device performs at least one of settings for an AI (Artificial Intelligence) / ML (Machine Learning) model, settings for parameters of the AI / ML model, and settings related to data used in the AI / ML model, in response to a type related to sharing of information related to an AI / ML model between communication devices, and transmits information corresponding to the at least one setting.

Images