Communication method and related equipment

Abstract

A communication method and related equipment are used for realizing at least one of determination, updating and deployment of an AI model in a communication network through interaction between different communication devices. In the method, a second communication device, as a receiver of first information, may determine a first AI model group based on the first information from the first communication device, and transmit second information indicating a first AI model of the first AI model group to the first communication device. Therefore, under the condition that the communication device in the communication system is used as an AI participation node, the computing power of the communication device can be applied to the processing of the AI model, and the first information from the first communication device can also be used as one of the determination bases for determining the AI model by the second communication device; the AI model determined by the second communication device can be adapted to the first communication device as much as possible, so that the processing performance of the subsequent first communication device for model processing based on the AI model is improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method and related equipment. Background Art

[0002] Wireless communication can be the transmission communication between two or more communication nodes without propagation through conductors or cables. The communication nodes generally include network equipment and terminal equipment.

[0003] At present, in wireless communication systems, communication nodes generally have signal transceiving capabilities and computing capabilities. Taking network devices with computing capabilities as an example, the computing capabilities of network devices mainly provide computing power support for signal transceiving capabilities (for example: sending and receiving signals) to achieve communication between network devices and other communication nodes.

[0004] However, in a communication network, the computing power of communication nodes may have surplus computing power in addition to providing computing power support for the above communication tasks. Therefore, how to utilize this computing power is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present application provides a communication method and related equipment for realizing the determination and deployment of artificial intelligence (AI) models in a communication network through interaction between different communication devices, so that the computing power of the communication device can be applied to the processing of the AI ​​model.

[0006] The first aspect of the present application provides a communication method, which is performed by a first communication device, which may be a communication device (such as a terminal device), or the first communication device may be a partial component in the communication device (such as a processor, a chip or a chip system, etc.), or the first communication device may also be a logic module or software that can realize all or part of the functions of the communication device. In this method, the first communication device sends a first information, and the first information is used to determine a first AI model group, and the first AI model group includes K AI models and a second AI model, K is an integer greater than or equal to 1; wherein the input of any AI model in the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more AI models in the K AI models; the second AI model is deployed in the second communication device, and the K AI models are used to be deployed in M ​​communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; the first communication device receives a second information, and the second information is used to determine a first AI model, and the first AI model includes one or more models in the K AI models.

[0007] Based on the above technical solution, as the recipient of the first information, the second communication device can determine the first AI model group based on the first information from the first communication device, and send the second information for determining the first AI model in the first AI model group to the first communication device. Thus, when the communication device in the communication system is an AI participating node, while enabling the computing power of the communication device to be applied to the processing of the AI model, the first information from the first communication device can also be used as one of the bases for the second communication device to determine the AI model, so that the AI model determined by the second communication device can be as adaptable as possible to the first communication device, thereby improving the processing performance of the subsequent model processing by the first communication device based on the AI model.

[0008] In this application, terms such as AI model, neural network model, AI neural network model, machine learning model, and AI processing model can be replaced with each other.

[0009] It should be understood that the first AI model group includes K AI models and a second AI model. It can be understood that the functions of the first AI model group are at least realized through the model processing of the K AI models and the model processing of the second AI model. In other words, after the first communication device receives the second information, the first communication device can determine the first AI model through the second information and perform model processing on the first AI model; correspondingly, the second communication device can perform model processing on the second AI model deployed on the second communication device. Optionally, the model processing may include one or more of model update processing, model training processing, and model inference processing.

[0010] Optionally, the M communication devices used to deploy the K AI models can be referred to as the cooperation set of the second communication device, or the cooperation set of the first AI model group, or the cooperation set of the second AI model.

[0011] Optionally, when the first AI model group is regarded as one AI model, the K AI models and the second AI model can be understood as different AI sub-models in the one AI model.

[0012] It should be understood that the second communication device can be implemented in various ways.

[0013] For example, any one of the M communication devices is a terminal device and the second communication device can be a terminal device. Correspondingly, the first communication device and the second communication device can communicate on the sidelink (SL). In this case, the K AI models and the second AI model can be referred to as end-to-end models, or end-to-end cooperation models, etc. Exemplarily, taking the end-to-end cooperation model as an example, when K is equal to 1, it can be called a single-link end-to-end cooperation model; when K is greater than 1, it can be called a multi-link end-to-end cooperation model.

[0014] For another example, any one of the M communication devices is a terminal device and the second communication device can be a network device (such as an access network device). Correspondingly, the first communication device and the second communication device can communicate on the uplink and downlink communication links. In this case, the K AI models and the second AI model can be referred to as edge models, edge collaboration models, end-edge models, end-edge collaboration models, etc. Exemplarily, taking the end-edge collaboration model as an example, when K is equal to 1, it can be called a single-link end-edge collaboration model; when K is greater than 1, it can be called a multi-link end-edge collaboration model.

[0015] It should be understood that an AI model is deployed in a communication device (for example, the first AI model is deployed in the first communication device, the second AI model is deployed in the second communication device, etc.), which can be expressed as an AI model is used to be deployed in a communication device. In other words, after a communication device obtains the model parameters of an AI model, it can obtain / generate / construct the AI model based on the model parameters of the AI model, and subsequently, the communication device can perform model processing on the AI model.

[0016] It should be understood that K AI models are used to be deployed in M communication devices, where M is less than or equal to K.

[0017] For example, when M is equal to K, different AI models among the K AI models can be used to be deployed in different communication devices among the M communication devices, that is, the K AI models and the M communication devices can be in one-to-one correspondence. In other words, the i-th AI model among the K AI models is deployed in the i-th communication device among the M communication devices (the value range of i is from 1 to K or from 1 to M). Exemplarily, the first AI model used to be deployed in the first communication device can include one of the K AI models.

[0018] For another example, when M is less than K, at least two AI models among the K AI models can be used to be deployed in one of the M communication devices, that is, the K AI models and the M communication devices may not be in one-to-one correspondence. Exemplarily, the first AI model used to be deployed in the first communication device can include at least two AI models among the K AI models.

[0019] Optionally, the model parameters can include one or more of the hyperparameters of the model, the dataset of the model (including the input data of the model and the label data corresponding to the input data), and the structure parameters of the model.

[0020] It should be understood that wireless communication signals (such as the transceiver of configuration information of communication resources, the transceiver of reference signals, etc.) can be transmitted between different communication devices (such as a first communication device and a second communication device). Optionally, the AI models involved in this application (such as a first AI model, a second AI model, etc.) can be used to manage the wireless communication signals (including at least one of configuration, update, and optimization). For example, the AI model can include an AI model for modulation and / or demodulation, an AI model for channel prediction, an AI model for beam management, an AI model for assisted positioning, an AI model for channel compression, an AI model for resource scheduling, and one or more of an AI model for replacing one or more modules in a transmitter and / or a receiver. Alternatively, the AI models involved in this application can also be AI models for other AI tasks, such as an AI model for image recognition, an AI model for natural language processing, an AI model for computer vision, etc.

[0021] In a possible implementation, before the first communication device receives the second information, the method further includes: the first communication device receives third information, where the third information is used to configure the mapping relationship between P-dimensional information and Q indexes, and both P and Q are positive integers; the second information includes one or more of the Q indexes, and the one or more indexes are used to determine one of the P-dimensional information.

[0022] In this application, the index can be replaced by other terms, such as identification, number, etc.

[0023] Optionally, the mapping relationship configured by the third information can be implemented in ways such as tables, formulas, etc.

[0024] Based on the above technical solution, the first communication device can receive the third information for configuring the mapping relationship between P-dimensional information and Q indexes. Thereafter, the second information received by the first communication device can include one or more of the Q indexes, and the first communication device can determine one of the P-dimensional information based on the one or more indexes. Thus, through the implementation method of configuring the mapping relationship by the third information, the second information sent by the second communication device can indicate the dimensional information by carrying indexes, which can reduce the indication overhead.

[0025] Optionally, the third information is layer 3 (L3) signaling, and the second information is layer 1 (L1) and / or layer 2 (L2) signaling. For example, the L3 signaling may include non-access stratum (NAS) signaling and / or radio resource control (RRC) layer signaling. For another example, L1 signaling may be understood as physical layer signaling. For another example, L2 signaling may be understood as radio network layer signaling, including at least one of medium access control (MAC) layer signaling, radio link control (RLC) signaling, packet data convergence protocol (PDCP) signaling, and service data adaptation protocol (SDAP) layer signaling.

[0026] In a possible implementation, the second information includes at least one of the following: model parameters of the first AI model, model parameters of the first AI model group, dimension information of the input data of the first AI model, and dimension information of the output data of the first AI model.

[0027] It can be understood that when the second information includes the dimension information of the input data of the first AI model and / or the dimension information of the output data of the first AI model, the first communication device can update the initial model based on the dimension information of the input data and / or the output data to obtain the first AI model. Optionally, the initial model may be a model pre-configured in the first communication device or a model configured by the second communication device (such as the third AI model described later), which is not limited here.

[0028] Based on the above technical solutions, the second information for determining the first AI model may include at least one of the above. In other words, the second communication device can deploy the first AI model on the first communication device through at least one of the above to improve the flexibility of scheme implementation.

[0029] Optionally, the second information may be sent through one message / signaling or multiple messages / signaling, which is not limited here.

[0030] In a possible implementation, the first information includes first dimension information or second dimension information; when the input of the first AI model includes the output of the second AI model, the first dimension information is used to determine the dimension information of the input data of the first AI model or the dimension information of the output data of the second AI model; when the input of the second AI model includes the output of the first AI model, the second dimension information is used to determine the dimension information of the output data of the first AI model or the dimension information of the input data of the second AI model.

[0031] Based on the above technical solution, the second communication device can determine the dimension information of the data transmitted on the communication link through the first information. When the communication bandwidth between the first communication device and the second communication device is fixed, since the dimension of the data transmitted on the communication link is related to the processing performance of the AI model, in this way, the implementation process of the second communication device to determine the first AI model group can be simplified, and while reducing the complexity of the second communication device, the processing performance of the AI models included in the first AI model group can also be improved.

[0032] In a possible implementation, the dimension information includes at least one of the following: the upper limit value of the dimension, the lower limit value of the dimension, the dimension expected by the first communication device, and the value range of the dimension expected by the first communication device.

[0033] Optionally, the dimension information may include the value of at least one of the above, or the quantization value of the value of at least one of the above, or the index of the value of at least one of the above, the index of the quantization value of the value of at least one of the above, etc., or may be implemented in other ways, which is not limited herein.

[0034] Based on the above technical solution, the dimension information determined by the first dimension information or the second dimension information may include at least one of the above, so as to improve the flexibility of the solution implementation.

[0035] For example, when the above dimension information includes the upper limit value and / or the lower limit value of the dimension, the second communication device may use the range indicated by the upper limit value and / or the lower limit value as one of the bases for determining the AI model, which can improve the flexibility of the solution implementation.

[0036] For another example, when the above dimension information includes the dimension expected by the first communication device and / or the value range of the dimension expected by the first communication device, the AI model determined by the second communication device based on the dimension information can meet the expectations of the first communication device.

[0037] In a possible implementation, the first dimension information or the second dimension information is determined based on the channel state information.

[0038] Based on the above technical solution, the first dimension information or the second dimension information included in the first information may be determined based on the channel state information, such that the first dimension information or the second dimension information can, to a certain extent, reflect the channel characteristics of the wireless channel between the first communication device and the second communication device, so that the AI model obtained based on the first information can be adapted to the channel characteristics of the wireless channel, in order to improve the transmission performance of the AI data corresponding to the AI model. Moreover, when the AI model obtained based on the first information can be adapted to the channel characteristics of the wireless channel, it can also make the transmitted data of the wireless link satisfy the channel bandwidth requirements as much as possible, thereby improving the model performance of the AI models included in the first AI model group.

[0039] Optionally, the channel state information may include the channel information between the first communication device and the second communication device, and / or the channel information between the second communication device and the first communication device. Among them, when the first communication device is a terminal device and the second communication device is a network device, the channel information between the first communication device and the second communication device can be understood as uplink channel information, and the channel information between the second communication device and the first communication device can be understood as downlink channel information.

[0040] Optionally, the channel state information may be obtained based on reference signals.

[0041] For example, when the first communication device and the second communication device communicate via a sidelink, the reference signals may include sidelink synchronization signal / physical broadcast channel block (sidelink SSB, SL-SSB, or S-SS / PSBCH block), sidelink channel state information reference signal (SL-CSI-RS), etc.

[0042] For another example, when the first communication device and the second communication device communicate via uplink and downlink, the reference signals may include channel state information reference signal (CSI-RS), sounding reference signal (SRS), etc.

[0043] In one possible implementation, the first information includes at least one of the following: input data of the first AI model and label data of the input data of the first AI model, local computing power status information of the first communication device, and channel status information.

[0044] Based on the above technical solution, the first information may include at least one of the above information, so that the second communication device can determine the first AI model group adapted for the first communication device based on the above at least one of the information.

[0045] In an implementation example, when the first information includes input data of a first AI model and label data of the input data of the first AI model, since the input data can be used as input of the first AI model, the label data can be used as one of the bases for determining the model processing performance of the first AI model. Therefore, for the second communication device, the second communication device can obtain an AI model with better performance based on these two pieces of information.

[0046] In addition, for the second communication device, the second communication device can implement mathematical calculations based on mutual information based on these two information and the AI ​​data sent and received by the second communication device on the wireless link (for example, input data of the second AI model or output data of the second AI model, etc.), and determine the first AI model group based on the result of the mathematical calculation to improve the model performance of the AI ​​models included in the first AI model group under the premise that the wireless link data meets the bandwidth.

[0047] In another implementation example, when the first information includes the local computing power status information of the first communication device, the complexity requirement of the model processing of the AI ​​model may be related to the local computing power status of the first communication device. Therefore, for the second communication device, the first AI model group determined by the second communication device based on the local computing power status information can be adapted to the local computing power status of the first communication device to provide a first AI model that meets the local computing power status, thereby improving the success rate of the model processing performed by the first communication device based on the first AI model.

[0048] In another implementation example, when the first information includes channel state information, since the channel state information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device, for the second communication device, the second communication device determines a first AI model group based on the channel state information to adapt to the channel characteristics, so that the transmission data of the wireless link meets the channel bandwidth requirements as much as possible, in order to improve the transmission performance of the AI ​​data corresponding to the AI ​​model.

[0049] It can be understood that, when the first information includes two or more of the above-mentioned information, based on the technical gain brought by any one of the above-mentioned items, further superposition gain can be obtained through the two or more of the information.

[0050] Optionally, the first information may be sent via one message / signaling, or may be sent via multiple messages / signaling, which is not limited here.

[0051] In a possible implementation, the first information is used to determine a first AI model group, including: the first information is used to update the second AI model group to obtain the first AI model group; the second AI model group includes X AI models and a fourth AI model, where X is an integer greater than or equal to 1; wherein the input of any AI model among the X AI models includes the output of the fourth AI model, or the input of the fourth AI model includes the output of any AI model among the X AI models; a third AI model among the X AI models is deployed on a first communication device, and the third AI model includes one or more AI models among the X AI models; the fourth AI model is deployed on a second communication device, and the X AI models are deployed on Y communication devices, where the Y communication devices include the first communication device, and Y is less than or equal to X.

[0052] Optionally, update can be replaced by other terms such as modification, iteration, optimization, processing, etc.

[0053] Optionally, when the second AI model group is regarded as one AI model, the X AI models and the fourth AI model may be understood as multiple AI sub-models in the one AI model.

[0054] Based on the above technical solution, for the second communication device, after receiving the first information, the second communication device can update the second AI model group based on the first information to obtain the first AI model group. In other words, the first information sent by the first communication device can be used to update other AI models, so that the solution can be applicable to AI model update scenarios.

[0055] Optionally, the third AI model and the second AI model included in the second AI model group may be general models or special models to implement updates of different types of models.

[0056] It should be understood that the general model can be called a basic model, a large model or an L0 model, and the dedicated model can be called a small model, an L1 model, an L2 model, etc.

[0057] Taking the big model as an example, the big model can refer to a machine learning model with a large number of parameters and complex structure, which can process massive data and complete various complex tasks, such as natural language processing, computer vision, speech recognition, etc.

[0058] Alternatively, large models are often built from deep neural networks with billions or even hundreds of billions of parameters.

[0059] Optionally, the large model can be designed to improve the model's expressiveness and predictive performance, and to be able to handle more complex tasks and data.

[0060] Alternatively, large models can learn complex patterns and features by training on massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unprocessed data.

[0061] In contrast, a small model can refer to a model with fewer parameters and shallower layers. Generally speaking, compared with a small model, a large model usually has more parameters and deeper layers, has stronger expressive power and higher accuracy, but also requires more computing resources and time for training and reasoning. It is suitable for scenarios with large data volumes and sufficient computing resources, such as cloud computing, high-performance computing, artificial intelligence, etc.

[0062] Alternatively, the small model has the advantages of being lightweight, efficient, and easy to deploy, and is suitable for scenarios with small data volumes and limited computing resources, such as mobile applications, embedded devices, and the Internet of Things.

[0063] In a possible implementation, after the first communication device receives the second information, the method further includes: the first communication device receives fourth information, and the fourth information is used to instruct to stop processing of the first AI model.

[0064] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the first communication device can also receive fourth information for instructing to stop processing the first AI model, so that the first communication device stops participating in the processing of the first AI model group, which can release the resources of the first communication device and save the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0065] In addition, when K is greater than 1, that is, the communication devices participating in the first AI model group may include other K-1 communication devices in addition to the first communication device. The first communication device may stop processing the first AI model through the instruction of the fourth information, and prune some AI models (such as the first AI model) to prune inefficient devices / inefficient links and achieve the selection of a better (or optimal) collaboration set.

[0066] In one possible implementation, before the first communication device receives the fourth information, the method also includes: the first communication device sends fifth information, and the fifth information is used to determine the fourth information; wherein the fifth information includes at least one of the following: input data or output data of the first AI model, dimensional information of the input data of the first AI model or dimensional information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device.

[0067] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the first communication device can send fifth information so that the second communication device can determine fourth information based on at least one of the above information indicated by the fifth information. Thus, the fourth information indicating to the first communication device to stop model processing can be determined based on the status information of the first communication device indicated by the fifth information, so that the second communication device can implement pruning processing based on the status information of one or more communication devices in the collaborative set.

[0068] In a possible implementation, the method further includes: the first communication device sends sixth information, where the sixth information is used to request to stop processing of the first AI model.

[0069] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the first communication device can send sixth information for requesting to stop the processing of the first AI model, so that the second communication device can determine the fourth information based on the sixth information. Thus, the fourth information indicating to the first communication device to stop the model processing can be determined based on the request of the first communication device, so that the second communication device can implement pruning processing based on the request of one or more communication devices in the collaborative set.

[0070] Optionally, the first communication device may trigger determination (or sending) of the sixth information based on a variety of trigger conditions. For example, the trigger condition may include: the first communication device determines that the channel state indicated by the channel state information is lower than or equal to a threshold, the first communication device determines that the computing power indicated by its own computing power state information is lower than or equal to a threshold, and the first communication device determines that the dimension information of the input data of the first AI model or the dimension information of the output data indicates that the dimension is lower than a threshold. One or more of the following.

[0071] The second aspect of the present application provides a communication method, which is performed by a second communication device, which may be a communication device (such as a network device or a terminal device), or the second communication device may be a partial component in the communication device (such as a processor, a chip or a chip system, etc.), or the second communication device may also be a logic module or software that can realize all or part of the functions of the communication device. In this method, the second communication device receives first information; the second communication device determines a first AI model group based on the first information, the first AI model group includes K AI models and a second AI model, K is an integer greater than or equal to 1; wherein the input of any AI model in the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more AI models in the K AI models; the second AI model is deployed in the second communication device, the K AI models are deployed in M ​​communication devices, the M communication devices include the first communication device, M is less than or equal to K; the second communication device sends second information, the second information is used to determine the first AI model, the first AI model is deployed in the first communication device, and the first AI model includes one or more models in the K AI models.

[0072] Based on the above technical solution, the second communication device acts as a receiver of the first information. The second communication device can determine the first AI model group based on the first information from the first communication device, and send the second information used to determine the first AI model in the first AI model group to the first communication device. Thus, when the communication device in the communication system acts as an AI participating node, the computing power of the communication device can be applied to the processing of the AI ​​model. At the same time, the first information from the first communication device can also be used as one of the bases for the second communication device to determine the AI ​​model, so that the AI ​​model determined by the second communication device can be adapted to the first communication device as much as possible, so as to improve the processing performance of the subsequent model processing based on the AI ​​model by the first communication device.

[0073] In one possible implementation, before the second communication device sends the second information, the method also includes: the second communication device sends third information, the third information is used to configure a mapping relationship between P dimensional information and Q indexes, P and Q are both positive integers; the second information includes one or more indexes of the Q indexes, and the one or more indexes are used to determine one of the dimensional information in the P dimensional information.

[0074] Based on the above technical solution, the second communication device can send third information for configuring the mapping relationship between P dimensional information and Q indexes. After that, the second information sent by the second communication device can include one or more indexes of the Q indexes, and the first communication device can determine one of the dimensional information in the P dimensional information based on the one or more indexes. Thus, through the implementation method of configuring the mapping relationship by the third information, the second information sent by the second communication device can indicate the dimensional information by carrying the index, which can reduce the overhead of the indication.

[0075] Optionally, the third information is layer 3 signaling, and the second information is layer 1 and / or layer 2 signaling.

[0076] In one possible implementation, the second information includes at least one of the following: model parameters of the first AI model, model parameters of the first AI model group, dimensional information of input data of the first AI model, and dimensional information of output data of the first AI model.

[0077] It is understandable that, when the second information includes the dimensional information of the input data of the first AI model and / or the dimensional information of the output data of the first AI model, the first communication device can update the initial model based on the dimensional information of the input data and / or the dimensional information of the output data to obtain the first AI model. Optionally, the initial model can be a model preconfigured in the first communication device, or a model configured by the second communication device (such as the third AI model described later), which is not limited here.

[0078] Based on the above technical solution, the second information used to determine the first AI model may include at least one of the above items. In other words, the second communication device can deploy the first AI model on the first communication device through at least one of the above items to improve the flexibility of the implementation of the solution.

[0079] Optionally, the second information may be sent via one message / signaling, or may be sent via multiple messages / signaling, which is not limited here.

[0080] In one possible implementation, the first information includes first dimension information or second dimension information; when the input of the first AI model includes the output of the second AI model, the first dimension information is used to determine the dimension information of the input data of the first AI model or the dimension information of the output data of the second AI model; when the input of the second AI model includes the output of the first AI model, the second dimension information is used to determine the dimension information of the output data of the first AI model or the dimension information of the input data of the second AI model.

[0081] Based on the above technical solution, the second communication device can determine the dimension information of the data transmitted on the communication link through the first information. When the communication bandwidth between the first communication device and the second communication device is constant, since the dimension of the data transmitted on the communication link is related to the processing performance of the AI ​​model, in this way, the implementation process of the second communication device determining the first AI model group can be simplified, reducing the complexity of the second communication device while also improving the processing performance of the AI ​​models included in the first AI model group.

[0082] In a possible implementation, the dimension information includes at least one of the following: an upper limit value of the dimension, a lower limit value of the dimension, the dimension expected by the first communication device, and a value range of the dimension expected by the first communication device.

[0083] Optionally, the dimensional information may include the value of at least one of the above items, or the quantized value of the value of at least one of the above items, or the index of the value of at least one of the above items, the index of the quantized value of the value of at least one of the above items, etc., or may be implemented in other ways, which are not limited here.

[0084] Based on the above technical solution, the dimensional information determined by the first dimensional information or the second dimensional information may include at least one of the above items to enhance the flexibility of the solution implementation.

[0085] For example, when the above-mentioned dimensional information includes an upper limit value and / or a lower limit value of the dimension, the second communication device can use the range indicated by the upper limit value and / or the lower limit value as one of the bases for determining the AI ​​model, thereby improving the flexibility of the implementation of the solution.

[0086] For example, when the above-mentioned dimensional information includes the dimension expected by the first communication device and / or the value range of the dimension expected by the first communication device, the AI ​​model determined by the second communication device based on the dimensional information can meet the expectations of the first communication device.

[0087] In a possible implementation manner, the first dimension information or the second dimension information is determined based on channel state information.

[0088] Based on the above technical solution, the first dimension information or the second dimension information contained in the first information can be determined based on the channel state information, so that the first dimension information or the second dimension information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to a certain extent, so that the AI ​​model that can be obtained based on the first information can be adapted to the channel characteristics of the wireless channel, in order to improve the transmission performance of the AI ​​data corresponding to the AI ​​model. Moreover, when the AI ​​model obtained based on the first information can be adapted to the channel characteristics of the wireless channel, the transmission data of the wireless link can also meet the channel bandwidth requirements as much as possible, thereby improving the model performance of the AI ​​model included in the first AI model group.

[0089] In one possible implementation, the first information includes at least one of the following: input data of the first AI model and label data of the input data of the first AI model, local computing power status information of the first communication device, and channel status information.

[0090] Based on the above technical solution, the first information may include at least one of the above information, so that the second communication device can determine the first AI model group adapted to the first communication device based on the above at least one information. The specific possible implementation of the first information can refer to the relevant content in the first aspect, which will not be repeated here.

[0091] In a possible implementation, after the second communication device sends the second information, the method further includes: the second communication device sends fourth information, and the fourth information is used to instruct to stop processing the first AI model.

[0092] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the second communication device can also send fourth information for instructing to stop processing the first AI model, so that the first communication device stops participating in the processing of the first AI model group, which can release the resources of the first communication device and save the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0093] In addition, when K is greater than 1, that is, the communication devices participating in the first AI model group may include other K-1 communication devices in addition to the first communication device. The first communication device may stop processing the first AI model through the instruction of the fourth information, and prune some AI models (such as the first AI model) to prune inefficient devices / inefficient links and achieve the selection of a better (or optimal) collaboration set.

[0094] In one possible implementation, before the second communication device sends the fourth information, the method also includes: the second communication device receives fifth information, and the fifth information is used to determine the fourth information; wherein the fifth information includes at least one of the following: input data or output data of the first AI model, dimensional information of the input data of the first AI model or dimensional information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device.

[0095] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the second communication device can receive the fifth information so that the second communication device can determine the fourth information based on the at least one item of information indicated by the fifth information. Thus, the fourth information indicating to the first communication device to stop model processing can be determined based on the status information of the first communication device indicated by the fifth information, so that the second communication device can implement pruning processing based on the status information of one or more communication devices in the collaborative set.

[0096] In a possible implementation, the method further includes: the second communication device receives sixth information, where the sixth information is used to request to stop processing of the first AI model.

[0097] Based on the above technical solution, after the first communication device receives the second information for determining the first AI model, the first communication device can perform model processing based on the determined first AI model. Thereafter, the second communication device can receive the sixth information for requesting to stop the processing of the first AI model, so that the second communication device determines the fourth information based on the sixth information. Thus, the fourth information indicating to the first communication device to stop the model processing can be determined based on the request of the first communication device, so that the second communication device can implement pruning processing based on the request of one or more communication devices in the collaborative set.

[0098] The third aspect of the present application provides a communication method, which is performed by a first communication device, which may be a communication device (such as a terminal device), or the first communication device may be a partial component in the communication device (such as a processor, a chip or a chip system, etc.), or the first communication device may also be a logic module or software that can realize all or part of the functions of the communication device. In this method, the first communication device performs model processing based on a first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the first communication device receives fourth information, and the fourth information is used to indicate to stop the processing of the first AI model.

[0099] Based on the above technical solution, during the process of the first communication device performing model processing based on the first AI model, the first communication device can receive fourth information for instructing to stop processing the first AI model, so that the first communication device stops participating in the processing of the first AI model group, thereby releasing the resources of the first communication device and saving the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0100] In one possible implementation, the first AI model group also includes Z AI models, where Z is a positive integer, and the input of any AI model among the Z AI models includes the output of the second AI model, or the input of the second AI model includes the output of any one or more AI models among the Z AI models.

[0101] Based on the above technical solution, in addition to the first AI model deployed on the first communication device and the second AI model deployed on the second communication device, the first AI model group may also include Z AI models deployed on other communication devices. In other words, the communication devices participating in the first AI model group may include other communication devices in addition to the first communication device. The first communication device may stop processing the first AI model through the instruction of the fourth information, and prune some AI models (such as the first AI model) to prune inefficient devices / inefficient links and select a better (or optimal) collaboration set.

[0102] In one possible implementation, before the first communication device receives the fourth information, the method also includes: the first communication device sends fifth information, and the fifth information is used to determine the fourth information; wherein the fifth information includes at least one of the following: input data or output data of the first AI model, dimensional information of the input data of the first AI model or dimensional information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device.

[0103] Based on the above technical solution, the first communication device can send the fifth information so that the second communication device can determine the fourth information based on the at least one information indicated by the fifth information. Thus, the fourth information indicating to the first communication device to stop the model processing can be determined based on the state information of the first communication device indicated by the fifth information, so that the second communication device can implement pruning processing based on the state information of one or more communication devices in the cooperative set.

[0104] In a possible implementation, the method further includes: the first communication device sends sixth information, where the sixth information is used to request to stop processing of the first AI model.

[0105] Based on the above technical solution, the first communication device may send sixth information for requesting to stop processing of the first AI model, so that the second communication device determines the fourth information based on the sixth information. Thus, the fourth information indicating to the first communication device to stop model processing may be determined based on the request of the first communication device, so that the second communication device can implement pruning processing based on the request of one or more communication devices in the collaborative set.

[0106] Optionally, the first communication device may trigger determination (or sending) of the sixth information based on a variety of trigger conditions. For example, the trigger condition may include: the first communication device determines that the channel state indicated by the channel state information is lower than or equal to a threshold, the first communication device determines that the computing power indicated by its own computing power state information is lower than or equal to a threshold, and the first communication device determines that the dimension information of the input data of the first AI model or the dimension information of the output data indicates that the dimension is lower than a threshold. One or more of the following.

[0107] The fourth aspect of the present application provides a communication method, which is performed by a second communication device, which may be a communication device (such as a network device or a terminal device), or the second communication device may be a partial component in the communication device (such as a processor, a chip or a chip system, etc.), or the second communication device may also be a logic module or software that can implement all or part of the functions of the communication device. In this method, the second communication device determines fourth information, and the fourth information is used to indicate to stop processing the first AI model, and the first AI model and the second AI model are included in the first AI model group; wherein the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the second communication device sends the fourth information.

[0108] Based on the above technical solution, when the first communication device is performing model processing based on the first AI model, the second communication device can send fourth information for indicating to stop processing the first AI model, so that the first communication device stops participating in the processing of the first AI model group, thereby releasing the resources of the first communication device and saving the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0109] In a possible implementation of the fourth aspect, the first AI model group also includes Z AI models, the input of any AI model among the Z AI models includes the output of the second AI model, or the input of the second AI model includes the output of any one or more AI models among the Z AI models.

[0110] Based on the above technical solution, in addition to the first AI model deployed on the first communication device and the second AI model deployed on the second communication device, the first AI model group may also include Z AI models deployed on other communication devices. In other words, the communication devices participating in the first AI model group may include other communication devices in addition to the first communication device. The first communication device may stop processing the first AI model through the instruction of the fourth information, and prune some AI models (such as the first AI model) to prune inefficient devices / inefficient links and select a better (or optimal) collaboration set.

[0111] In a possible implementation manner of the fourth aspect, before the second communication device sends the fourth information, the method also includes: the second communication device receives fifth information, and the fifth information is used to determine the fourth information; wherein the fifth information includes at least one of the following: input data or output data of the first AI model, dimensional information of the input data of the first AI model or dimensional information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device.

[0112] Based on the above technical solution, the second communication device can receive the fifth information, so that the second communication device determines the fourth information based on the at least one information indicated by the fifth information. Thus, the fourth information indicating to the first communication device to stop the model processing can be determined based on the state information of the first communication device indicated by the fifth information, so that the second communication device can implement pruning processing based on the state information of one or more communication devices in the cooperative set.

[0113] In a possible implementation manner of the fourth aspect, the method further includes: the second communication device receives sixth information, where the sixth information is used to request to stop processing of the first AI model.

[0114] Based on the above technical solution, the second communication device may receive the sixth information for requesting to stop the processing of the first AI model, so that the second communication device determines the fourth information based on the sixth information. Thus, the fourth information indicating to the first communication device to stop the model processing may be determined based on the request of the first communication device, so that the second communication device can implement pruning processing based on the request of one or more communication devices in the collaborative set.

[0115] In a fifth aspect, the present application provides a communication device, which is a first communication device, and includes a transceiver unit and a processing unit; the processing unit is used to determine first information; the transceiver unit is used to send first information, and the first information is used to determine a first AI model group, and the first AI model group includes K AI models and a second AI model, K is an integer greater than or equal to 1; wherein the input of any AI model among the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more AI models among the K AI models; the second AI model is deployed in a second communication device, and the K AI models are used to be deployed in M ​​communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; the transceiver unit is also used to receive second information, and the second information is used to determine the first AI model, and the first AI model includes one or more models among the K AI models.

[0116] In the fifth aspect of the present application, the constituent modules of the communication device can also be used to execute the steps performed in each possible implementation method of the first aspect and achieve corresponding technical effects. For details, please refer to the first aspect and will not be repeated here.

[0117] In a sixth aspect of the present application, a communication device is provided, which is a second communication device, and includes a transceiver unit and a processing unit. The transceiver unit receives first information; the processing unit is used to determine a first AI model group based on the first information, and the first AI model group includes K AI models and a second AI model, K is an integer greater than or equal to 1; wherein the input of any AI model among the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more AI models among the K AI models; the second AI model is deployed in the second communication device, and the K AI models are deployed in M ​​communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; the transceiver unit is also used to send second information, and the second information is used to determine the first AI model, and the first AI model is deployed in the first communication device, and the first AI model includes one or more models among the K AI models.

[0118] In the sixth aspect of the present application, the constituent modules of the communication device can also be used to execute the steps performed in each possible implementation method of the second aspect and achieve corresponding technical effects. For details, please refer to the second aspect and will not be repeated here.

[0119] In the seventh aspect of the present application, a communication device is provided, which is a first communication device, and the device includes a transceiver unit and a processing unit; the processing unit is used to perform model processing based on a first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein the first AI model is deployed in the first communication device, and the second AI model is deployed in the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the transceiver unit is used to receive fourth information, and the fourth information is used to instruct to stop the processing of the first AI model.

[0120] The constituent modules of the communication device can also be used to execute the steps performed in various possible implementation methods of the third aspect and achieve corresponding technical effects. For details, please refer to the third aspect and will not be repeated here.

[0121] In an eighth aspect of the present application, a communication device is provided, which is a second communication device, and includes a transceiver unit and a processing unit. The processing unit is used to determine fourth information, and the fourth information is used to instruct to stop processing of the first AI model, and the first AI model and the second AI model are included in the first AI model group; wherein the first AI model is deployed in the first communication device, and the second AI model is deployed in the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the transceiver unit is used to send the fourth information.

[0122] The constituent modules of the communication device can also be used to execute the steps performed in each possible implementation method of the fourth aspect and achieve corresponding technical effects. For details, please refer to the fourth aspect and will not be repeated here.

[0123] In a ninth aspect of the present application, a communication device is provided, comprising at least one processor, wherein the at least one processor is coupled to a memory; the memory is used to store programs or instructions; the at least one processor is used to execute the program or instructions so that the device implements the method described in any possible implementation method of any one of the first to fourth aspects.

[0124] In a possible implementation, the communication device further includes a memory. Optionally, the processor and the memory are integrated together.

[0125] In a tenth aspect, the present application provides a communication device, comprising at least one logic circuit and an input / output interface; the logic circuit is used to execute the method described in any possible implementation method of any one of the first to fourth aspects.

[0126] In an eleventh aspect, the present application provides a communication system, which includes the above-mentioned first communication device and second communication device.

[0127] A twelfth aspect of the present application provides a computer-readable storage medium, which is used to store one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes a method as described in any possible implementation of any one of the first to fourth aspects above.

[0128] The thirteenth aspect of the present application provides a computer program product (or computer program). When the computer program in the computer program product is executed by the processor, the processor executes the method described in any possible implementation of any one of the first to fourth aspects above.

[0129] A fourteenth aspect of the present application provides a chip system, which includes at least one processor for supporting a communication device to implement the method described in any possible implementation of any one of the first to fourth aspects above.

[0130] In a possible design, the chip system may also include a memory for storing program instructions and data necessary for the communication device. The chip system may be composed of a chip, or may include a chip and other discrete devices. Optionally, the chip system also includes an interface circuit, which provides program instructions and / or data for the at least one processor.

[0131] Among them, the technical effects brought about by any design method in the fifth to fourteenth aspects can refer to the technical effects brought about by the different design methods in the first to fourth aspects mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0132] Figure 1a , Figure 1b A schematic diagram of a communication system provided for this application;

[0133] Figures 2a to 2g A schematic diagram of the AI ​​processing process involved in this application;

[0134] Figures 3 to 6 An interactive schematic diagram of the communication method provided by this application;

[0135] Figures 7 to 11 A schematic diagram of a communication device provided in this application. DETAILED DESCRIPTION

[0136] First, some terms in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.

[0137] (1) Terminal device: It can be a wireless terminal device that can receive network device scheduling and instruction information. The wireless terminal device can be a device that provides voice and / or data connectivity to users, or a handheld device with wireless connection function, or other processing devices connected to a wireless modem.

[0138] The terminal device can communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device can be a mobile terminal device, such as a mobile phone (or "cellular" phone, mobile phone), a computer and a data card, for example, a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device, which exchanges voice and / or data with the radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablet computers (Pads), computers with wireless transceiver functions, and other devices. The wireless terminal device can also be called a system, a subscriber unit, a subscriber station, a mobile station, a mobile station (MS), a remote station, an access point (AP), a remote terminal device (remote terminal), an access terminal device (access terminal), a user terminal device (user terminal), a user agent (user agent), a subscriber station (SS), a customer premises equipment (CPE), a terminal (terminal), a user equipment (UE), a mobile terminal (MT), etc.

[0139] As an example but not limitation, in the embodiments of the present application, the terminal device may also be a wearable device. Wearable devices may also be referred to as wearable smart devices or smart wearable devices, etc., which are a general term for the application of wearable technology to intelligently design and develop wearable devices for daily wear, such as glasses, gloves, watches, clothing and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also powerful functions achieved through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include full-featured, large-size, and independent of smartphones to achieve complete or partial functions, such as smart watches or smart glasses, etc., as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various types of smart bracelets, smart helmets, and smart jewelry for vital sign monitoring.

[0140] The terminal may also be a drone, a robot, a terminal in device-to-device (D2D) communication, a terminal in vehicle to everything (V2X), a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self driving, a wireless terminal in remote medical, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc.

[0141] In addition, the terminal device may also be a terminal device in a communication system that evolves after the fifth generation (5th generation, 5G) communication system (e.g., a sixth generation (6th generation, 6G) communication system, etc.) or a terminal device in a public land mobile network (PLMN) that evolves in the future, etc. Exemplarily, the 6G network can further expand the form and function of the 5G communication terminal, and the 6G terminal includes but is not limited to a car, a cellular network terminal (with integrated satellite terminal function), a drone, and an Internet of Things (IoT) device.

[0142] In an embodiment of the present application, the terminal device may also obtain AI services provided by the network device. Optionally, the terminal device may also have AI processing capabilities.

[0143] (2) Network equipment: It can be equipment in a wireless network, for example, the network equipment can be a RAN node (or device) that connects a terminal device to a wireless network, which can also be called a base station. At present, some examples of RAN equipment are: base station, evolved NodeB (eNodeB), gNB (gNodeB) in a 5G communication system, transmission reception point (TRP), evolved Node B (eNB), radio network controller (RNC), Node B (NB), home base station (e.g., home evolved Node B, or home Node B, HNB), base band unit (BBU), or wireless fidelity (Wi-Fi) access point AP, etc. In addition, in a network structure, the network equipment may include a centralized unit (CU) node, a distributed unit (DU) node, or a RAN device including a CU node and a DU node.

[0144] Optionally, the RAN node may also be a macro base station, a micro base station or an indoor station, a relay node or a donor node, or a wireless controller in a cloud radio access network (CRAN) scenario. The RAN node may also be a server, a wearable device, a vehicle or an onboard device, etc. For example, the access network device in the V2X technology may be a road side unit (RSU).

[0145] In another possible scenario, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, and different RAN nodes respectively implement part of the functions of the base station. For example, the RAN node can be a centralized unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU). The CU and DU can be set separately, or can also be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU) or a remote radio head (RRH).

[0146] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in an open radio access network (O-RAN or ORAN) system, CU may also be called O-CU (open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. For the convenience of description, this application takes CU, CU-CP, CU-UP, DU and RU as examples for description. Any unit of CU (or CU-CP, CU-UP), DU and RU in this application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0147] The communication between the access network device and the terminal device follows a certain protocol layer structure. The protocol layer may include a control plane protocol layer and a user plane protocol layer. The control plane protocol layer may include at least one of the following: RRC layer, PDCP layer, RLC layer, MAC layer, or physical layer. The user plane protocol layer may include at least one of the following: SDAP layer, PDCP layer, RLC layer, MAC layer, or physical layer.

[0148] For the correspondence between network elements in the ORAN system and their achievable protocol layer functions, refer to Table 1 below.

[0149] Table 1

[0150] ORAN network element Protocol layer functions of 3GPP O-CU-CP RRC + PCDP - Control Plane (PDCP-C) O-CU-UP SDAP + PCDP - User Plane (PDCP-U) O-DU RLC + MAC + PHY-high O-RU PHY-low

[0151] The network device may be any other device that provides wireless communication functions for the terminal device. The embodiments of the present application do not limit the specific technology and specific device form used by the network device. For the convenience of description, the embodiments of the present application do not limit.

[0152] The network equipment may also include core network equipment, such as mobility management entity (MME), home subscriber server (HSS), serving gateway (S-GW), policy and charging rules function (PCRF), public data network gateway (PDN gateway, P-GW) in the fourth generation (4G) network; access and mobility management function (AMF), user plane function (UPF) or session management function (SMF) and other network elements in the 5G network. In addition, the core network equipment may also include other core network equipment in the 5G network and the next generation network of the 5G network.

[0153] In an embodiment of the present application, the above-mentioned network device may also have a network node with AI capabilities, which can provide AI services for terminals or other network devices. For example, it may be an AI node on the network side (access network or core network), a computing node, a RAN node with AI capabilities, a core network element with AI capabilities, etc.

[0154] In the embodiment of the present application, the device for realizing the function of the network device may be a network device, or may be a device capable of supporting the network device to realize the function, such as a chip system, which may be installed in the network device. In the technical solution provided in the embodiment of the present application, the technical solution provided in the embodiment of the present application is described by taking the device for realizing the function of the network device as an example that the network device is used as the device.

[0155] (3) Configuration and pre-configuration: In this application, configuration and pre-configuration are used at the same time. Configuration refers to the network device / server sending some parameter configuration information or parameter values ​​to the terminal through messages or signaling, so that the terminal can determine the communication parameters or resources during transmission based on these values ​​or information. Pre-configuration is similar to configuration, and can be parameter information or parameter values ​​pre-negotiated between the network device / server and the terminal device, or parameter information or parameter values ​​used by the base station / network device or terminal device specified by the standard protocol, or parameter information or parameter values ​​pre-stored in the base station / server or terminal device. This application does not limit this.

[0156] Furthermore, these values ​​and parameters can be changed or updated.

[0157] (4) The terms "system" and "network" in the embodiments of the present application can be used interchangeably. "Multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the objects associated with each other are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in the embodiments of the present application are used to distinguish multiple objects, and are not used to limit the order, timing, priority or importance of multiple objects.

[0158] (5) "Send" and "receive" in the embodiments of the present application indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information is XX, which can include direct sending through the air interface, and also include indirect sending through the air interface by other units or modules. "Receive information from YY" can be understood as the source of the information is YY, which can include direct receiving from YY through the air interface, and also include indirect receiving from YY through the air interface from other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface.

[0159] In other words, sending and receiving can be performed between devices, for example, between a network device and a terminal device, or can be performed within a device, for example, sending or receiving between components, modules, chips, software modules, or hardware modules within the device through a bus, wiring, or interface.

[0160] It is understandable that information may be processed between the source and destination of information transmission, such as coding, modulation, etc., but the destination can understand the valid information from the source. Similar expressions in this application can be understood similarly and will not be repeated.

[0161] (6) In the embodiments of the present application, "indication" may include direct indication and indirect indication, and may also include explicit indication and implicit indication. The information indicated by a certain information (such as the indication information described below) is called information to be indicated. In the specific implementation process, there are many ways to indicate the information to be indicated, such as but not limited to, directly indicating the information to be indicated, such as the information to be indicated itself or the index of the information to be indicated. The information to be indicated may also be indirectly indicated by indicating other information, wherein the other information is associated with the information to be indicated; or only a part of the information to be indicated may be indicated, while the other part of the information to be indicated is known or agreed in advance. For example, the indication of specific information may be realized by means of the arrangement order of each information agreed in advance (such as predefined by the protocol), thereby reducing the indication overhead to a certain extent. The present application does not limit the specific method of indication. It is understandable that, for the sender of the indication information, the indication information may be used to indicate the information to be indicated, and for the receiver of the indication information, the indication information may be used to determine the information to be indicated.

[0162] In this application, unless otherwise specified, the same or similar parts between the various embodiments can refer to each other. In the various embodiments in this application, and the various methods / designs / implementations in each embodiment, if there is no special description and logical conflict, the terms and / or descriptions between different embodiments and the various methods / designs / implementations in each embodiment are consistent and can be referenced to each other. The technical features in different embodiments and the various methods / designs / implementations in each embodiment can be combined to form new embodiments, methods, or implementations according to their inherent logical relationships. The implementation methods of this application described below do not constitute a limitation on the scope of protection of this application.

[0163] The present application can be applied to a long term evolution (LTE) system, a new radio (NR) system, or a communication system evolved after 5G (such as 6G, etc.). The communication system includes at least one network device and / or at least one terminal device.

[0164] See also Figure 1a , is a schematic diagram of the architecture of a communication system 1000 used in an embodiment of the present application. Figure 1aAs shown, the communication system includes RAN 100 and core network 200. Optionally, the communication system 1000 may also include Internet 300. RAN 100 includes at least one RAN node (such as Figure 1a 110a and 110b in the figure, collectively referred to as 110), and may also include at least one terminal (such as Figure 1a RAN 100 may also include other RAN nodes, such as wireless relay equipment and / or wireless backhaul equipment ( Figure 1a (not shown in the figure). The terminal 120 is connected to the RAN node 110 in a wireless manner, and the RAN node 110 is connected to the core network 200 in a wireless or wired manner. The core network device in the core network 200 and the RAN node 110 in the RAN 100 can be independent and different physical devices, or can be the same physical device that integrates the logical functions of the core network device and the logical functions of the RAN node. Terminals and terminals and RAN nodes can be connected to each other in a wired or wireless manner.

[0165] RAN 100 may be an evolved universal terrestrial radio access (E-UTRA) system, an NR system, and a future radio access system defined in the 3rd generation partnership project (3GPP). RAN 100 may also include two or more of the above-mentioned different radio access systems. RAN 100 may also be an open RAN (open RAN, O-RAN).

[0166] For the convenience of description, a base station is taken as an example of a RAN node for description below.

[0167] Base stations and terminals can be fixed or movable. Base stations and terminals can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on the water surface; they can also be deployed on airplanes, balloons, and artificial satellites. The embodiments of this application do not limit the application scenarios of base stations and terminals.

[0168] The roles of the base station and the terminal can be relative, for example, Figure 1aThe helicopter or drone 120i in the figure can be configured as a mobile base station. For the terminal 120j that accesses the wireless access network 100 through 120i, the terminal 120i is a base station; but for the base station 110a, 120i is a terminal, that is, 110a and 120i communicate through the wireless air interface protocol. Of course, 110a and 120i can also communicate through the interface protocol between base stations. In this case, relative to 110a, 120i is also a base station. Therefore, base stations and terminals can be collectively referred to as communication devices. Figure 1a 110a and 110b in the figure may be referred to as communication devices having base station functions. Figure 1a 120a-120j in the figure can be called communication devices with terminal functions.

[0169] Base stations and terminals, base stations and base stations, and terminals and terminals can communicate through authorized spectrum, unauthorized spectrum, or both; they can communicate through spectrum below 6 gigahertz (GHz), spectrum above 6 GHz, or spectrum below 6 GHz and spectrum above 6 GHz. The embodiments of the present application do not limit the spectrum resources used for wireless communication.

[0170] In the embodiments of the present application, the functions of the base station may also be performed by a module (such as a chip) in the base station, or by a control subsystem including the base station function. The control subsystem including the base station function here may be a control center in the above-mentioned application scenarios such as smart grid, industrial control, smart transportation, and smart city. The functions of the terminal may also be performed by a module (such as a chip or a modem) in the terminal, or by a device including the terminal function.

[0171] Figure 1b Another schematic diagram of a communication system provided in an embodiment of the present application, in Figure 1b In the example, the network device is a base station, and both device 1 and device 2 are terminal devices. Figure 1b As shown, the communication link between device 1 and device 2 can be called a sidelink (SL), and the communication link between device 1 (or device 2) and the base station can be called an uplink and a downlink, including an uplink and a downlink. It can be seen that the sidelink is a communication mechanism in which different terminal devices communicate directly without going through network equipment.

[0172] Optionally, in SL, generally speaking, the transmitting device and the receiving device can be terminal devices or network devices of the same type, or they can be RSU and terminal devices, wherein RSU is a roadside station or roadside unit from a physical entity point of view, and can be a terminal device or a network device from a functional point of view, and this application does not impose any restrictions on this. That is, the transmitting device is a terminal device, and the receiving device is also a terminal device; or, the transmitting device is a roadside station, and the receiving device is also a terminal device; or, the transmitting device is a terminal device, and the receiving device is also a roadside station. In addition, the sidelink can also be a base station device of the same type or different types. At this time, the function of the sidelink is similar to that of the relay link, but the air interface technology used can be the same or different.

[0173] Exemplarily, broadcast, unicast, and multicast are supported on the sidelink.

[0174] When a terminal device (eg, device 1) communicates directly with another terminal device (eg, device 2) without going through a network device, the two terminal devices may communicate based on a proximity-based services communication 5 (PC5) port.

[0175] A typical application of the sidelink is V2X communication, which utilizes and enhances current cellular network functions and elements to achieve low-latency and high-reliability communication between various nodes in the vehicle network, including vehicle-to-vehicle communication (V2V), vehicle-to-pedestrian communication (V2P), vehicle-to-infrastructure communication (V2I), and vehicle-to-network communication (V2N).

[0176] The technical solution provided by this application can be applied to wireless communication systems (such as Figure 1a or Figure 1bThe system shown in the figure), for example, the communication system provided in the present application can introduce an AI network element to implement some or all AI-related operations. The AI ​​network element may also be referred to as an AI node, an AI device, an AI entity, an AI module, an AI model, or an AI unit, etc. The AI ​​network element may be a network element built into a communication system. For example, the AI ​​network element may be an AI module built into: a terminal device, an access network device, a core network device, a cloud server, or a network management (operation, administration and maintenance, OAM) to implement AI-related functions. The OAM may be a network management for a core network device and / or a network management for an access network device. Alternatively, the AI ​​network element may also be a network element independently set up in the communication system. Optionally, the terminal or the chip built into the terminal may also include an AI entity to implement AI-related functions.

[0177] The following is a brief introduction to artificial intelligence (AI) that may be involved in this application.

[0178] AI can give machines human intelligence, for example, it can allow machines to use computer hardware and software to simulate certain intelligent behaviors of humans. In order to realize artificial intelligence, machine learning methods can be used. In machine learning methods, machines use training data to learn (or train) a model. The model represents the mapping from input to output. The learned model can be used for reasoning (or prediction), that is, the model can be used to predict the output corresponding to a given input. Among them, the output can also be called the reasoning result (or prediction result).

[0179] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Among them, unsupervised learning can also be called unsupervised learning.

[0180] Supervised learning uses machine learning algorithms to learn the mapping relationship from sample values ​​to sample labels based on the collected sample values ​​and sample labels, and uses AI models to express the learned mapping relationship. The process of training a machine learning model is the process of learning this mapping relationship. During the training process, the sample values ​​are input into the model to obtain the model's predicted values, and the model parameters are optimized by calculating the error between the model's predicted values ​​and the sample labels (ideal values). After the mapping relationship is learned, the learned mapping can be used to predict new sample labels. The mapping relationship learned by supervised learning can include linear mapping or nonlinear mapping. According to the type of label, the learning task can be divided into classification task and regression task.

[0181] Unsupervised learning discovers the intrinsic patterns of samples by itself using algorithms based on the collected sample values. In unsupervised learning, there is a type of algorithm that uses the samples themselves as the supervision signal, that is, the model learns the mapping relationship from samples to samples, which is called self-supervised learning. During training, the model parameters are optimized by calculating the error between the predicted value of the model and the samples themselves. Self-supervised learning can be used in applications such as signal compression and decompression recovery. Common algorithms include autoencoders and generative adversarial networks, etc.

[0182] Reinforcement learning is different from supervised learning. It is a type of algorithm that learns strategies to solve problems by interacting with the environment. Different from supervised and unsupervised learning, there are no clear "correct" action label data in reinforcement learning problems. The algorithm needs to interact with the environment to obtain the reward signal feedback from the environment, and then adjust the decision-making actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the total system throughput rate feedback by the wireless network, and then expects to obtain a higher system throughput rate. The goal of reinforcement learning is also to learn the mapping relationship between the environmental state and the optimal (such as the best) decision-making actions. However, because the "correct actions" labels cannot be obtained in advance, the network cannot be optimized by calculating the error between the actions and the "correct actions". The training of reinforcement learning is achieved through iterative interactions with the environment.

[0183] A neural network (NN) is a specific model in machine learning technology. According to the universal approximation theorem, a neural network can theoretically approximate any continuous function, enabling the neural network to have the ability to learn any mapping. Traditional communication systems need to rely on rich expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover implicit pattern structures from large datasets, establish the mapping relationship between data, and obtain better performance than traditional modeling methods.

[0184] The idea of a neural network comes from the neuron structure of the brain tissue. For example, each neuron performs a weighted sum operation on its input values and outputs the operation result through an activation function.

[0185] As Figure 2a shown, it is a schematic diagram of the neuron structure. Assume that the input of the neuron is x = [x0, x1, …, x n , and the weights corresponding to each input are w = [w, w1, …, w n , where n is a positive integer, w i and x i can be various possible types such as decimals, integers (such as 0, positive integers, or negative integers, etc.), or complex numbers, etc. w i as the weight of x i is used to weight xi Weighting is performed. The bias for weighted summation of the input values according to the weights is, for example, b. There can be various forms of activation functions. Assuming the activation function of a neuron is: y = f(z) = max(0, z), then the output of this neuron is: For another example, the activation function of a neuron is: y = f(z) = z, then the output of this neuron is: Among them, b can be various possible types such as decimals, integers (such as 0, positive integers or negative integers), or complex numbers. The activation functions of different neurons in a neural network can be the same or different.

[0186] In addition, a neural network generally includes multiple layers, and each layer can include one or more neurons. By increasing the depth and / or width of the neural network, the expression ability of the neural network can be improved, providing a more powerful information extraction and abstract modeling ability for complex systems. Among them, the depth of the neural network can refer to the number of layers included in the neural network, and the number of neurons included in each layer can be called the width of that layer. In one implementation, the neural network includes an input layer and an output layer. The input layer of the neural network processes the received input information through neurons and passes the processing result to the output layer, and the output layer obtains the output result of the neural network. In another implementation, the neural network includes an input layer, a hidden layer, and an output layer. The input layer of the neural network processes the received input information through neurons and passes the processing result to the intermediate hidden layer. The hidden layer calculates the received processing result to obtain a calculation result. The hidden layer passes the calculation result to the output layer or the next adjacent hidden layer, and finally the output layer obtains the output result of the neural network. Among them, a neural network can include one hidden layer or multiple sequentially connected hidden layers, without limitation.

[0187] The neural network is, for example, a deep neural network (DNN). According to the construction method of the network, DNN can include a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0188] Figure 2b It is a schematic diagram of an FNN network. The characteristic of the FNN network is that neurons between adjacent layers are completely connected pairwise. This characteristic makes FNN usually require a large amount of storage space and leads to a high computational complexity.

[0189] A CNN is a neural network specifically designed to process data with a similar grid structure. For example, time series data (discretely sampled along the time axis) and image data (two-dimensionally discretely sampled) can both be considered data with a similar grid structure. Instead of using all the input information for computation at once, a CNN uses a window of a fixed size to intercept partial information for convolution operations, which greatly reduces the computational amount of model parameters. Additionally, depending on the type of information intercepted by the window (such as people and objects in the same picture being different types of information), different convolution kernels can be used for each window, enabling the CNN to better extract the features of the input data.

[0190] An RNN is a type of DNN network that utilizes feedback time series information. Its input includes the new input value at the current moment and its own output value at the previous moment. RNNs are suitable for obtaining sequential features that are temporally correlated and are particularly applicable to applications such as speech recognition and channel coding and decoding.

[0191] During the model training process of the above-mentioned machine learning, a loss function can be defined. The loss function describes the gap or difference between the output value of the model and the ideal target value. The loss function can be embodied in various forms, and no specific form of the loss function is restricted. The model training process can be regarded as the following process: by adjusting some or all of the parameters of the model, the value of the loss function is made less than the threshold value or meets the target requirements.

[0192] The model can also be called an AI model, a rule, or other names, etc. An AI model can be considered a specific method for implementing AI functions. An AI model characterizes the mapping relationship or function between the input and output of the model. AI functions can include one or more of the following: data collection, model training (or model learning), model information publishing, model inference (or also called model reasoning, reasoning, or prediction, etc.), model monitoring or model verification, or the release of inference results, etc. AI functions can also be called AI (related) operations, or AI-related functions.

[0193] Next, the implementation process of the neural network will be described exemplarily in conjunction with the accompanying drawings.

[0194] 1. Fully connected neural network, also known as a multilayer perceptron (MLP).

[0195] As Figure 2c shown, an MLP contains an input layer (on the left), an output layer (on the right), and multiple hidden layers (in the middle). Among them, each layer of the MLP contains several nodes, called neurons. Among them, the neurons in adjacent layers are pairwise connected.

[0196] Optionally, considering the neurons in adjacent two layers, the output h of the neurons in the next layer is the weighted sum of all the neurons x in the previous layer connected to it and passes through an activation function, which can be expressed as:

[0197] h = f(wx + b).

[0198] Among them, w is the weight matrix, b is the bias vector, and f is the activation function.

[0199] Further optionally, the output of the neural network can be recursively expressed as:

[0200] y = f n (w n f n-1 (…)+b n ).

[0201] Among them, n is the index of the neural network layer, 1 <= n <= N, where N is the total number of layers of the neural network.

[0202] In other words, the neural network can be understood as a mapping relationship from the input data set to the output data set. Usually, the neural network is randomly initialized, and the process of obtaining this mapping relationship from the existing data with random w and b is called the training of the neural network.

[0203] Optionally, the specific training method is to evaluate the output result of the neural network by using a loss function.

[0204] As Figure 2d shown, the error can be backpropagated, and the neural network parameters (including w and b) can be iteratively optimized by the method of gradient descent until the loss function reaches the minimum value, that is, the "optimal point (such as the optimal point)" in Figure 2d . It can be understood that the neural network parameters corresponding to the "optimal point (such as the optimal point)" in Figure 2d can be used as the neural network parameters in the trained AI model information.

[0205] Further optionally, the process of gradient descent can be expressed as:

[0206]

[0207] Among them, θ is the parameter to be optimized (including w and b), L is the loss function, η is the learning rate, which controls the step size of gradient descent, represents the derivative operation, represents the derivative of L with respect to θ.

[0208] Further optionally, the process of backpropagation utilizes the chain rule of partial derivatives.

[0209] AsFigure 2e As shown, the gradient of the parameters of the previous layer can be recursively calculated from the gradient of the parameters of the subsequent layer, which can be expressed as:

[0210]

[0211] where w ij is the weight connecting node j to node i, and s i is the weighted sum of inputs on node i.

[0212] 2. Federated Learning (FL).

[0213] The concept of federated learning effectively solves the dilemmas faced by the current development of artificial intelligence. On the premise of fully ensuring the privacy and security of user data, it enables each edge device and the central server to cooperate efficiently to complete the model learning task.

[0214] As Figure 2f shown, the FL architecture is a training architecture in the current FL field. Exemplarily, the FedAvg algorithm is the basic algorithm of FL, and its algorithm process is roughly as follows:

[0215] (1) The central server initializes the model to be trained and broadcasts it to all client devices.

[0216] (2) In the t-th round where t ∈ [1, T], client k ∈ [1, K] trains the received global model on the local dataset for E epochs to obtain the local training result and reports it to the central node.

[0217] (3) The central node aggregates the local training results from all (or part of) the clients. Assuming the set of clients uploading local models in the t-th round is the central server will perform weighted averaging with the number of samples of the corresponding clients as weights to obtain a new global model. The specific update rule is After that, the central server will broadcast the latest version of the global model to all client devices for a new round of training.

[0218] (4) Repeat steps (2) and (3) until the model finally converges or the number of training rounds reaches the upper limit.

[0219] In addition to reporting the local model it is also possible to report the local gradient of the training The central node will average the local gradients and update the global model according to the direction of this average gradient.

[0220] As can be seen, in the FL framework, the dataset exists at distributed nodes. That is, the distributed nodes collect the local datasets, perform local training, and report the local results (models or gradients) obtained from the training to the central node. The central node itself does not have a dataset and is only responsible for fusing the training results of the distributed nodes to obtain the global model and distributing it to the distributed nodes.

[0221] 3. Decentralized learning. Different from federated learning, another distributed learning architecture is decentralized learning.

[0222] As Figure 2g shown, consider a fully distributed system without a central node. The design objective f(x) of the decentralized learning system is generally the mean of the objective functions f i (x) of each node, that is where n is the number of distributed nodes, x is the parameter to be optimized, and in machine learning, x is the parameter of the machine learning (such as neural network) model. Each node uses the local data and the local objective f i (x) to calculate the local gradient and then sends it to the neighboring nodes that are communicable. After any node receives the gradient information sent by its neighboring nodes, it can update the parameter x of the local model according to the following formula:

[0223]

[0224] where represents the parameter of the local model after the (k + 1)-th (k is a natural number) update in the i-th node, represents the parameter of the local model after the k-th update in the i-th node (if k is 0, it means is the parameter of the local model of the i-th node that has not participated in the update), α k represents the tuning coefficient, N i is the set of neighboring nodes of node i, and |N i | represents the number of elements in the set of neighboring nodes of node i, that is, the number of neighboring nodes of node i. Through the information interaction between nodes, the decentralized learning system will finally learn a unified model.

[0225] The technical solution provided by this application can be applied to a wireless communication system (such as Figure 1a or Figure 1b the system shown). In a wireless communication system, communication nodes generally have signal transceiver capabilities and computing capabilities. Taking a network device with computing capabilities as an example, the computing capabilities of the network device mainly provide computing power support for the signal transceiver capabilities (for example: performing transmission processing and reception processing on signals) to achieve the communication tasks between the network device and other communication nodes.

[0226] In a communication network, the computing power of communication nodes may have surplus computing power in addition to providing computing power support for the above communication tasks. Therefore, how to utilize this computing power is a technical problem that needs to be solved urgently.

[0227] In one possible implementation, the communication node can be used as a participating node of the AI ​​learning system, and the computing power of the communication node is applied to a certain link of the AI ​​learning system. With the advent of the era of large models, deep learning models with massive parameters, such as bidirectional encoder representations from transformers (BERT) and generative pre-trained transformer (GPT), can complete more and more complex tasks and achieve better performance. However, for large models, even the reasoning process of the model will be limited by the device capacity, so generally large models are stored on cloud central servers. At the same time, each device in the network generates a huge amount of raw data every day, which requires multiple calls to the large model for reasoning. Generally speaking, the device (such as a communication node) can send data to the central server, the central server uses the data for reasoning, and then the central server returns the reasoning result to the device. This process will consume a lot of communication resources for data transmission, and the privacy of device data will also be at risk.

[0228] In order to better save communication overhead and protect the privacy of user data, one possible implementation is to use distributed reasoning technology of deep neural networks, which includes distributing the model to devices and using the local computing power of the devices to infer the model, thereby reducing communication overhead and obtaining data privacy protection. However, in the communication system, there is currently no relevant solution for how to determine the AI ​​model used by the communication node (for example, how to generate and how to update it).

[0229] See also Figure 3 , is a schematic diagram of an implementation of the communication method provided in this application, and the method includes the following steps.

[0230] It should be noted that in Figure 3 (and / or later Figure 6 ) takes the first communication device and the second communication device as the execution subjects of the interactive indication as an example to illustrate the method, but the present application does not limit the execution subjects of the interactive indication. Figure 3 (and / or later Figure 6 ), the execution subject of the method can be replaced by a chip, a chip system, a processor, a logic module or software in a communication device. Figure 3 (and / or laterFigure 6 ) Among them, the first communication device may be a network device and the second communication device may be a terminal device, or both the first communication device and the second communication device are terminal devices (for example, this method can be applied to the communication process of different terminal devices in a sidelink communication scenario).

[0231] S301. The first communication device sends first information. Correspondingly, the second communication device receives the first information. The first information is used to determine a first AI model group, and the first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; among the K AI models, the input of any one AI model includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed on the second communication device, and the K AI models are used to be deployed on M communication devices, and the M communication devices include the first communication device, and M is less than or equal to K.

[0232] S302. The second communication device sends second information. Correspondingly, the first communication device receives the second information. The second information is used to determine a first AI model, and the first AI model includes one or more of the K AI models.

[0233] In this application, terms such as AI model, neural network model, AI neural network model, machine learning model, and AI processing model can be replaced with each other.

[0234] It should be understood that the first information sent by the first communication device in step S301 is used to determine the first AI model group. The first AI model group includes K AI models and a second AI model. It can be understood that the function of the first AI model group is at least implemented through the model processing of the K AI models and the model processing of the second AI model. In other words, after the first communication device receives the second information, the first communication device can determine the first AI model through the second information and perform model processing on the first AI model; correspondingly, the second communication device can perform model processing on the second AI model deployed on the second communication device. Optionally, the model processing may include one or more of model update processing, model training processing, and model inference processing.

[0235] It should be understood that wireless communication signals (such as the transceiver of configuration information of communication resources, the transceiver of reference signals, etc.) can be transmitted between different communication devices (such as the first communication device and the second communication device).

[0236] Optionally, the AI models involved in this application (such as the first AI model, the second AI model, etc.) can be used to manage the wireless communication signal (including at least one of configuration, update, and optimization). For example, the AI model can include one or more of an AI model for modulation and / or demodulation, an AI model for channel prediction, an AI model for beam management, an AI model for assisting positioning, an AI model for channel compression, an AI model for resource scheduling, and an AI model for replacing one or more modules in a transmitter and / or a receiver. Alternatively, the AI models involved in this application can also be AI models for other AI tasks, such as an AI model for image recognition, an AI model for natural language processing, an AI model for computer vision, etc.

[0237] Optionally, in the case where the first AI model group is regarded as one AI model, the K AI models and the second AI model can be understood as different AI sub-models in the one AI model.

[0238] It should be understood that when an AI model is deployed in a communication device (for example, the first AI model is deployed in the first communication device, the second AI model is deployed in the second communication device, etc.), it can be stated that an AI model is used to be deployed in a communication device. In other words, after a communication device obtains the model parameters of an AI model, it can obtain / generate / construct the AI model based on the model parameters of the AI model. Subsequently, the communication device can perform model processing on the AI model. Optionally, the model parameters can include one or more of the hyperparameters of the model, the data set of the model (including the input data of the model and the label data corresponding to the input data), and the structure parameters of the model.

[0239] In a possible implementation, the second communication device is a functional entity that determines a list of AI model groups based on the first information. The list of AI model groups includes one or more AI model groups, and the one or more AI model groups include the first AI model group. Specifically, the second communication device can communicate with one or more first communication devices, and the second communication device can receive information (such as one or more first information) from the one or more first communication devices to generate / obtain / determine one or more AI model groups. In other words, the second communication device can perform information collection and perform model generation based on the collected information. Subsequently, the second communication device can deploy AI models in one or more first communication devices.

[0240] Optionally, the AI ​​model group list may include one or more AI model groups, and each AI model group may include two or more AI models. As mentioned above, the relationship between the AI ​​model group and the AI ​​model can also be understood as the relationship between the AI ​​model and the AI ​​sub-model. For this reason, the AI ​​model group list can also be replaced by an AI model list, that is, the AI ​​model list can include one or more AI models.

[0241] Optionally, list can be replaced with other terms such as set, dictionary, combination, space, etc.

[0242] Optionally, the second communication device is a functional entity that determines the AI ​​model group list based on the first information, including: the second communication device is a functional entity that determines the AI ​​model group list based on the first information, and selects some or all of the AI ​​model groups used by the first communication device from the AI ​​model group list. Specifically, after the second communication device determines the AI ​​model group list based on the first information, the functions implemented by the second communication device may also include selecting some or all of the AI ​​model groups used by the first communication device from the AI ​​model group list. In other words, in addition to performing information collection and performing model generation based on the collected information, the second communication device may also perform model selection so that the second communication device can subsequently deploy an AI model that is compatible with the one or more first communication devices in one or more first communication devices.

[0243] It should be understood that K AI models are used to be deployed in M ​​communication devices, where M is less than or equal to K. When the values ​​of M and K are different, the K AI models may have different implementations, which will be introduced below.

[0244] For example, when M is equal to K, different AI models among the K AI models can be used to be deployed in different communication devices among the M communication devices, that is, the K AI models and the M communication devices can be one-to-one corresponding. In other words, the i-th AI model among the K AI models can be used to be deployed in the i-th communication device among the M communication devices (the value of i is 1 to K or 1 to M). Exemplarily, the first AI model deployed in the first communication device may include one of the K AI models.

[0245] For another example, when M is less than K, at least two of the K AI models may be deployed in one of the M communication devices, that is, the K AI models may not correspond one-to-one to the M communication devices. Exemplarily, the first AI model deployed in the first communication device may include at least two of the K AI models.

[0246] Optionally, the M communication devices for deploying K AI models may be referred to as a collaboration set of the second communication device, or a collaboration set of the first AI model group, or a collaboration set of the second AI model.

[0247] For ease of understanding, the following will take the case where M is equal to K and K is greater than 1 (i.e., the M communication devices can be described as K communication devices) as an example to give an example description of the K communication devices and the AI models deployed by the second communication device.

[0248] As Figure 4 shown in the example, the K communication devices may be communication device 1,..., communication device K in the figure, and the K AI models may be AI model 1,..., AI model K in the figure. It should be understood that the first communication device may be one of the K communication devices, and the first AI model deployed on the first communication device may be one of the K AI models. In Figure 4 , the inputs of the K AI models deployed on the K communication devices include the output of the second AI model. Exemplarily, taking the input data of the second AI model as X as an example, after being processed by the second AI model, the second communication device may obtain and send data Z1,..., Z K to the K communication devices respectively; after transmission through the wireless channel, the data received by the K communication devices are respectively represented as (it can be understood that due to transmission path loss and interference such as noise on the wireless channel, may not be the same as Z1,..., may not be the same as Z K , but, generally has the same data dimension as Z1,..., generally has the same data dimension as Z K ; among them, can be understood as an estimate of Z1 or a measured value of Z1,..., can be understood as an estimate of Z K or a measured value of Z K etc.). Thereafter, the K communication devices may use the data they each receive as the input of the AI model, and obtain data after being processed by the AI model, that is, communication device 1 can obtain ,..., communication device K can obtain

[0249] Optionally, in Figure 4 the example shown, the data Z1,..., Z K sent by the second AI model to the K communication devices respectively may be partially the same, or all the same, or all different.

[0250] As Figure 5 shown in the example, the K communication devices can be communication device 1,..., communication device K in the figure, and the K AI models can be AI model 1,..., AI model K in the figure. It should be understood that the first communication device can be one of the K communication devices, and the first AI model to be deployed on the first communication device can be one of the K AI models. In Figure 5 , the input of the second AI model includes the outputs of the K AI models deployed on the K communication devices. Exemplarily, the inputs of the K AI models can be respectively represented as X1,..., X K , and after being processed by the K AI models respectively, Z1,..., Z K can be obtained; after transmission through the wireless channel, the data received by the second communication device is represented as (it can be understood that due to interference such as transmission path loss and noise on the wireless channel, may not be the same as Z1,..., and Z K may not be the same; however, and Z1 generally have the same data dimension,..., and Z K generally have the same data dimension; where, can be understood as an estimate of Z1 or a measured value of Z1,..., can be understood as an estimate of Z K or a measured value of Z K etc.). Thereafter, the second communication device can perform one or more AI model processes based on the received K data (i.e., ), and each AI model process can include one or more of the K data. In addition, the processing result after the one or more AI model processes can be represented as

[0251] It can be understood that the first communication device and the second communication device can have multiple implementations.

[0252] For example, any one of the M communication devices is a terminal device and the second communication device can be a terminal device. Correspondingly, the first communication device and the second communication device can communicate on the sidelink (SL). In this case, the K AI models and the second AI model can be called end-to-end models, or end-to-end collaborative models, etc. Exemplarily, taking the end-to-end collaborative model as an example, when K is equal to 1, it can be called a single-link end-to-end collaborative model; when K is greater than 1, it can be called a multi-link end-to-end collaborative model.

[0253] For another example, any one of the M communication devices is a terminal device and the second communication device can be a network device (such as an access network device). Correspondingly, the first communication device and the second communication device can communicate on the uplink and downlink communication links. In this case, the K AI models and the second AI model can be referred to as edge models, edge collaboration models, end-edge models, end-edge collaboration models, etc. Exemplarily, taking the end-edge collaboration model as an example, when K is equal to 1, it can be called a single-link end-edge collaboration model; when K is greater than 1, it can be called a multi-link end-edge collaboration model.

[0254] It should be noted that in Figure 4 and Figure 5 , the data Y, Y1,..., Y K can be the input data X, X1,..., X K of the model respectively, and the corresponding labeled data. The association relationship between the labeled data and the processing result of the AI model (such as ) can be used to detect or determine the processing performance of the K AI models and the second AI model. For example, the association relationship can be determined by means of gradient information, loss function, etc.

[0255] Optionally, for the K AI models, the labeled data Y1,..., Y K of the K AI models can be the same or different, which is not limited here.

[0256] From Figure 3 the implementation process shown, for the second communication device, after receiving the first information in step S301, the second communication device can generate (or determine) a model based on the first information to obtain a first group of AI models. The process of the second communication device determining the first group of AI models based on the first information will be described exemplarily below.

[0257] First, the theoretical basis for the generation of the AI model is described through the implementation process of the following method A.

[0258] Generally, in a traditional connection or session-oriented network, the design goal between different communication devices (i.e., transceivers) is: to enable the receiver to accurately reply all the data sent by the transmitter, that is, to pursue lossless data transmission. However, when facing future intelligent networks, due to the existence of massive data and different purposes of different AI tasks, it may no longer be necessary to transmit all data, but to transmit data valuable for AI tasks. Therefore, the network may transmit the minimum amount of wireless data to optimize the performance of the AI model (such as maximizing the accuracy of the AI model).

[0259] To achieve this goal, in Figure 4 and Figure 5In the illustrated example, the performance of the AI model can be characterized based on the mutual information between different data. As an implementation example, the following will use Figure 5 the scenario shown as an example to illustrate the implementation of mutual information. It should be understood that in Figure 4 the scenario shown, the reverse transmission can be implemented with reference to the following example.

[0260] In Figure 5 , the input data of the k-th (where k takes any value from 1 to K) AI model among the K AI models can be represented as X k , and the input data X k corresponds to the label data Y k and the data received by the second communication device satisfies Method A:

[0261]

[0262] where I(a; b) represents the mutual information content between variable a and variable b.

[0263] represents the set of.

[0264] represents the mutual information between the data transmitted over the wireless channel and the label data Y (the label data Y can be understood as the correct result / desired result of the AI task). The larger the value of this mutual information, the more information of the label data Y contained in the data transmitted over the wireless channel, which can be understood as the higher the accuracy of the AI model, that is, the better the model performance of the AI model.

[0265] represents the mutual information content between the original input data X k and the wireless link data . The smaller this mutual information content, the less data is transmitted in the wireless link, that is, the smaller the wireless communication overhead. Correspondingly, represents the sum of the data transmitted over K wireless links, the smaller it is, the smaller the total wireless communication overhead of the K links.

[0266] β k (for example, β k can take values in the range [0, 1]) can control the ratio between the two mutual informations, and is used to balance the accuracy of the AI processing result of the k-th link and the wireless communication overhead. Among them, the β k corresponding to different communication devices can be the same or different.

[0267] Therefore, minimizing the above formula That is to say: on the premise of ensuring that the wireless communication overhead is as small as possible, maximize the correctness of the processing results of the AI model. The subscript "IB" of represents the information bottleneck (IB) theory (or theories such as distributed information bottleneck, deterministic information bottleneck, other information theories, etc.). In other words, It can be replaced with other symbols. This is just an implementation example here.

[0268] Exemplarily, based on the implementation of Method A, during the generation of the AI model, the generation basis of the model can include one or more of the following Data A to Data C, that is, an AI model group can be generated based on the following Data A to Data C (for example, the second communication device can generate the first AI model group). The following will give an exemplary description of the Data A to Data C. Data A. M pairs of data and labels As input data (where M is the batch size of the batch data, x m，k represents the input data of the mth pair of data, and y m,k represents the label data of the mth pair of data).

[0269] Optionally, different label data can be the same, that is, y m,1 ,...y m,K can all be represented as y m . In other words, the input of multi-link data (that is, x m,1 ,...x m,K ) can be used to collaboratively infer the same result y m , that is, the above Data A can be represented as

[0270] Optionally, the quantity "M" of the M pairs of data and labels as input data has a different definition from the quantity "M" of the M communication devices mentioned above. Here, only the same letter is used to represent the quantity, and their values can be the same or different.

[0271] Data B. The dimension of the data transmitted in the wireless link (for example Figure 5 in or Z k(in terms of dimension). Exemplarily, the dimension of this data can be expressed as the number of tokens / words / markers (hereinafter uniformly referred to as tokens), or the dimension of this data can be expressed as the dimension of an embedding vector. It can be understood that Token: a concept similar to "word", which can be regarded as the basic unit for splitting the intermediate transmission quantity; Embedding: the vector representation of the intermediate transmission quantity. Tokens can be understood as each constituent unit in the Embedding vector, that is, the number of constituent units in the Embedding can be the dimension of the tokens.

[0272] Data C. Channel state information (such as: Gaussian white noise modeling, etc.).

[0273] Optionally, the process of generating an AI model group based on one or more of Data A to Data C can be implemented through a neural network. That is, the input data of this neural network includes one or more of Data A to Data C, and after being processed by this neural network, the model parameters of an AI model group (this AI model group includes K AI models and a second AI model) are obtained.

[0274] As an implementation example, the loss function of this neural network can be expressed as Method B:

[0275]

[0276] As another implementation example, in order to reduce the solution complexity, the calculation of variational mutual information can be introduced. For example, the loss function of this neural network can be expressed as Method C:

[0277]

[0278] The subscript "VDIB" of represents variational distributed information bottleneck (VDIB);

[0279] K: the number of user equipments / links in a multi-link scenario;

[0280] θ k : the neural network or model parameters arranged by user k (such as the first communication device);

[0281] ψ: the neural network or model parameters arranged on the base station side (such as the second communication device);

[0282] β k : Lagrange multiplier, used to balance the accuracy of the AI processing result of the kth link and the wireless communication overhead;

[0283] On user k, according to the neural network with parameter θ k given the input data x k in the case of the conditional probability density;

[0284] On the base station side, according to the neural network with parameter φ, given the received vector to obtain the cross-entropy between the inference variable and the target variable y;

[0285] Using the variational distribution to approximate the distribution of z k ;

[0286] Denote taking the expectation / mean of the part in {·} on the premise of knowing the probability density function p(x k , y);

[0287] Denote taking the expectation / mean of the part in {·} on the premise of knowing the probability density function ;

[0288] Denote the conditional probability distribution with parameter φ;

[0289] Denote the variational distribution form of the approximate probability distribution ;

[0290] Denote the Kullback-Leibler (KL) divergence between the probability distribution and the probability distribution .

[0291] As another implementation example, on the premise of constraining the tokens / embedding dimension according to the radio link bandwidth, the loss function of this neural network can be expressed in way D:

[0292]

[0293] where denotes that the constraint condition is Th1 k denotes the maximum threshold value of the k-th radio link bandwidth, and the physical meaning of the constraint condition is that the data volume transmitted by the radio link satisfies the air interface bandwidth constraint. In other words, through way D, the accuracy of the AI task can be maximized on the premise that the information transmitted through the radio channel satisfies the bandwidth constraint.

[0294] As another implementation example, on the premise of constraining the tokens / embedding dimension according to the wireless link bandwidth, the loss function of the neural network can be expressed in Equation E:

[0295]

[0296] where represents the constraint condition as Th2 represents the minimum threshold of the AI task accuracy, and the physical meaning of the constraint condition is that the accuracy of the AI task is greater than the minimum threshold. In other words, through Equation E, the amount of information transmitted over the wireless channel can be minimized on the premise that the accuracy of the AI task is greater than the minimum threshold.

[0297] Based on the above implementation process, on the premise of constraining the tokens / embedding dimension according to the bandwidth, the inference accuracy of the obtained AI model is relatively high. Exemplarily, based on different datasets, the simulation results using the above Equation C are as follows:

[0298] Testing on the Canadian Institute for Advanced Research (CIFAR) dataset: the number of K communication devices = 4, the number of training epochs = 320, the intermediate output dimension = 64, and the accuracy = 87.44%.

[0299] Testing on the Mixed National Institute of Standards and Technology Database (MNIST) dataset: the number of devices = 4, the number of epochs = 320, the intermediate dim = 64, and the accuracy = 98.44%.

[0300] where the intermediate output dimension is the dimension of the above or Z k and the accuracy is the precision of the AI task.

[0301] Optionally, taking K communication devices as K terminals and the second communication device as the base station as an example, the process of generating the model involved in any one of the above Equation A to Equation E can be summarized into the following steps.

[0302] Step 1. Based on the reparameterization trick and Monte Carlo sampling, an unbiased estimate of the VDIB gradient can be obtained and the objective function can be optimized using stochastic gradient descent. Given a mini-batch of data and randomly sampling for each group of data, the empirical estimate of VDIB can be obtained to satisfy manner F:

[0303]

[0304] Compared with manner C, the meaning of the newly added parameters in manner F is as follows:

[0305] M: The size of the batch size, that is, the size / dimension of the input data during each training.

[0306] L: The number of sampling times for the wireless channel environment during each training.

[0307] y m : The target vector of the m-th group of data.

[0308] The received data of the m-th group of data at the base station side under the l-th channel sampling.

[0309] Step 2. Based on manner F (i.e., the Lagrangian expression of VDIB), the base station side updates the model parameter ψ and backpropagates the gradient parameter to the terminal side to support the update of the terminal side model;

[0310] Step 3. The terminal updates the local model parameter based on the gradient parameter backpropagated by the base station

[0311] After multiple rounds of iteration of the above 3 steps, the optimal deep neural network parameters Device DNN of K terminals can be obtained The optimal deep neural network parameter BSDNNψ of the base station side * ; and the optimal tokens dimension of the wireless link corresponding to the highest task accuracy value

[0312] Next, it will be introduced in detail how the backpropagation of the gradient is realized during the model training process. During the model training process, by calculating the gradient of the objective function , the base station side updates ψ, and the terminal side updates θ k , k = 1, …, K, and the specific process is as follows:

[0313] First, the base station side calculates the gradient of the local model parameter ψ according to the following manner G, updates the local neural network parameter, and manner G satisfies:

[0314]

[0315] Secondly, calculate the gradient parameter φ provided to the terminal side at the base station side k , k = 1, …, K, satisfying method H:

[0316]

[0317] And transmit φ k , k = 1, …, K to the k-th terminal respectively. Therefore, the backpropagation parameters are

[0318] Finally, the k-th terminal obtains the gradient of the local model according to the received gradient parameter φ k and updates the local neural network parameter θ in the following way I K . Method I satisfies:

[0319]

[0320] In the above implementation, the parameters involved are as follows.

[0321] 1. The input information includes:

[0322] The number K of links (or: terminals) participating in cooperation;

[0323] The number of iterations T (or: number of epoch);

[0324] The channel state information of each link: the channel noise power The number of channel samplings L;

[0325] The number of samples for model training: batch size, which represents the specific number of samples in a group of samples input to the model during a single training process.

[0326] 2. The output information includes:

[0327] The optimal deep neural network parameters Device DNN of K terminals

[0328] The optimal deep neural network parameters BSDNN ψ of the base station side * ;

[0329] The task completion accuracy (Accuracy) corresponding to each tokens dimension N;

[0330] By traversing N, the accuracy (Accuracy) corresponding to each N value can be obtained. Therefore, the optimal tokens dimension of the wireless link corresponding to the maximum task accuracy can be output

[0331] 3. The key interaction quantities during the model training process:

[0332] For any given initial model, after one round of iteration (including T epochs), the final model parameters on the base station side and the terminal side are obtained. The main intermediate interaction information is: the backpropagation gradient parameters That is, the backpropagation parameters for each interaction can satisfy manner H.

[0333] In a possible implementation, the second information sent by the second communication device in step S302 includes at least one of the following: the model parameters of the first AI model, the model parameters of the first AI model group, the dimension information of the input data of the first AI model, and the dimension information of the output data of the first AI model. In other words, the second communication device can deploy the first AI model on the first communication device through at least one of the above, so as to improve the flexibility of the solution implementation.

[0334] It can be understood that when the second information includes the model parameters of the first AI model and / or the model parameters of the first AI model group, the first communication device can construct / generate the first AI model locally based on the model parameters included in the second information, and obtain the first AI model in this way.

[0335] It can be understood that when the second information includes the dimension information of the input data of the first AI model and / or the dimension information of the output data of the first AI model, the first communication device can update the initial model based on the dimension information of the input data and / or the dimension information of the output data to obtain the first AI model. Optionally, the initial model can be a model pre-configured in the first communication device or a model configured by the second communication device (such as the third AI model described later), which is not limited here.

[0336] In addition, when the second information includes the dimension information of the input data of the first AI model and / or the dimension information of the output data of the first AI model, the second information can be implemented in various ways. For example, the second information can include the specific values of the dimension information, and the second information can also include the index values of the dimension information.

[0337] For ease of understanding, the implementation process in which the second information includes the index values of the dimension information will be described below as an example.

[0338] In a possible implementation, before the first communication device receives the second information, the method further includes: the first communication device receives third information, where the third information is used to configure the mapping relationship between P dimension information and Q indexes, and both P and Q are positive integers; the second information includes one or more of the Q indexes, and the one or more indexes are used to determine one of the P dimension information. In other words, the first communication device can receive the third information for configuring the mapping relationship between P dimension information and Q indexes. Thereafter, the second information received by the first communication device in step S302 may include one or more of the Q indexes, and the first communication device can determine one of the P dimension information based on the one or more indexes. Thus, through the implementation manner of configuring the mapping relationship by the third information, the second information sent by the second communication device can indicate the dimension information by carrying indexes, which can reduce the indication overhead.

[0339] In this application, the index can be replaced by other terms, such as representation, number, etc.

[0340] Optionally, the third information is layer 3 (L3) signaling, and the second information is layer 1 (L1) and / or layer 2 (L2) signaling. For example, the L3 signaling may include NAS signaling and / or RRC layer signaling. Also, the L1 signaling can be understood as physical layer signaling. Also, the L2 signaling can be understood as radio network layer signaling, including at least one of MAC layer signaling, RLC signaling, PDCP signaling, and SDAP layer signaling.

[0341] Optionally, the mapping relationship configured by the third information can be implemented in ways such as tables, formulas, etc. The following will give an exemplary description of the implementation process in which the third information configures the mapping relationship through a table.

[0342] As shown in Table 2, it is an implementation example of the third information configuring the mapping relationship through a table.

[0343] Table 2

[0344] Index Dimension value 1 2 2 4 3 8 4 16 5 32 ... ...

[0345] In Table 2, the relationship between the P dimension information and the Q indexes can be a one-to-one correspondence, that is, the values of P and Q can be equal.

[0346] Exemplarily, in the example shown in Table 2, when the dimension value indicated by the second communication device is "8", the second information sent by the second communication device in step S302 may include the index value "3". When the dimension value indicated by the second communication device is "32", the second information sent by the second communication device in step S302 may include the index value "5". It can be seen that when the dimension value is relatively large, by including the index value in the second information, the indication overhead can be reduced.

[0347] Optionally, in the example shown in Table 2, when the dimension value indicated by the second communication device does not appear in the table, the second communication device can achieve this by rounding up, rounding down, or sending multiple index values. For example, when the dimension value indicated by the second communication device is "6", by rounding up, the second information sent by the second communication device in step S302 may include the index value "3", that is, indicating a dimension value of "8"; by rounding down, the second information sent by the second communication device in step S302 may include the index value "2", that is, indicating a dimension value of "8"; by sending multiple index values, the second information sent by the second communication device in step S302 may include the index values "1" and "2", that is, indicating a dimension value of "2 + 4 = 6", or, the second information sent by the second communication device in step S302 may include the index values "3" and "1", that is, indicating a dimension value of "8 - 1 = 6".

[0348] Optionally, in the example shown in Table 2, different dimension values can be represented by a formula. For example, the dimension value N satisfies:

[0349] N = 2 I ;

[0350] where I represents the index.

[0351] As shown in Table 3, it is an implementation example of configuring the mapping relationship through a table for the third information.

[0352] Table 3

[0353]

[0354] In Table 3, the relationship between P dimension information and Q indexes can be one-to-many, that is, the value of P is less than the value of Q.

[0355] Exemplarily, in the example shown in Table 3, there are some indexes that can indicate the values of two or more dimensions through one index value. For example, when the second communication device determines that the performance with dimension values of "2" and "4" is the same or similar based on any one of the foregoing methods A to E, the second information sent by the second communication device in step S302 may include the index value "1", so that the first communication device can locally decide to use one of the dimension values of "2" or "4" to determine the first AI model.

[0356] As shown in Tables 4 to 6, it is another implementation example of configuring the mapping relationship through a table for the third information.

[0357] Table 4

[0358]

[0359] Table 5

[0360]

[0361] Table 6

[0362]

[0363] In the implementation methods shown in Tables 4 to 6, the relationship between the P dimension information and the Q indexes can be a one-to-many relationship, that is, the value of P is less than the value of Q.

[0364] Exemplarily, in the example shown in Tables 4 to 6, the Q indexes included in the second information sent by the second communication device in step S302 may be two indexes, one index is used to indicate the category of the AI model (denoted as index 1), and the other index is used to indicate the dimension value in the category of the AI model (denoted as index 2).

[0365] For example, when the second communication device indicates that the category of the AI model is X and the dimension value is "16", the second information sent by the second communication device in step S302 may include the index value (1, 4), that is, the value of index 1 is 1 and the value of index 2 is 4. In this way, the first communication device can determine Table 5 based on the value of index 1 being 1 and determine "the dimension value corresponding to the AI model category X in Table 5 is 16" based on the value of index 2 being 4.

[0366] For another example, when the second communication device indicates that the dimension value of the AI model of category Y is "50", the second information sent by the second communication device in step S302 may include the index values (2, 5), that is, the value of index 1 is 2 and the value of index 2 is 5. In this way, the first communication device can determine Table 6 based on the value of index 1 being 2, and determine "the dimension value corresponding to the AI model of category X is 50" in Table 6 based on the value of index 2 being 5.

[0367] Optionally, in any of the above implementation manners, the one or more tables configured with the third information may be sent through one message / signaling, or may be sent through multiple messages / signaling, which is not limited herein.

[0368] Optionally, in the examples shown in Table 4 to Table 6, "the category of the AI model" may be replaced with other implementations, such as the identifier of the AI model, the function identifier of the AI model, the task category corresponding to the AI model, the data category corresponding to the AI model, the size category of the AI model, the function category of the AI model, the form category of the AI model, and one or more of the categories of AI model deployment.

[0369] As can be seen from the foregoing description, the first information sent by the first communication device in step S301 can be used to determine the first AI model group. Based on the parameters involved in the processes shown in the foregoing Manner A to Manner E, the first information may include one or more of the following Information A to Information E. In other words, the second communication device can obtain the parameters required for Manner A to Manner D through one or more of the Information A to Information E included in the following first information, and generate the first AI model group according to one of Manner A to Manner E.

[0370] Information A. First dimension information. Wherein, when the input of the first AI model includes the output of the second AI model, the first dimension information is used to determine the dimension information of the input data of the first AI model or the dimension information of the output data of the second AI model.

[0371] Information B. Second dimension information. When the input of the second AI model includes the output of the first AI model, the second dimension information is used to determine the dimension information of the output data of the first AI model or the dimension information of the input data of the second AI model.

[0372] Information C. The input data of the first AI model and the label data of the input data of the first AI model.

[0373] Information D. Local computing power status information of the first communication device.

[0374] Information E. Channel status information.

[0375] For information A or information B, the second communication device can determine the dimension information of the data transmitted on the communication link through the first information, for example, the dimension information can be the number of tokens, or the dimension of the embedding vector. When the communication bandwidth between the first communication device and the second communication device is constant, since the dimension of the data transmitted on the communication link is related to the processing performance of the AI ​​model, in this way, the implementation process of the second communication device determining the first AI model group can be simplified, reducing the complexity of the second communication device while also improving the processing performance of the AI ​​models included in the first AI model group.

[0376] Optionally, the dimensional information includes at least one of the following items: an upper limit value of the dimension, a lower limit value of the dimension, the dimension expected by the first communication device (or the dimension not expected by the first communication device), and a value range of the dimension expected by the first communication device (or the value range of the dimension not expected by the first communication device).

[0377] For example, when the above-mentioned dimensional information includes an upper limit value and / or a lower limit value of the dimension, the second communication device can use the range indicated by the upper limit value and / or the lower limit value as one of the bases for determining the AI ​​model, thereby improving the flexibility of the implementation of the solution.

[0378] For example, when the above-mentioned dimensional information includes the dimension expected by the first communication device and / or the value range of the dimension expected by the first communication device, the AI ​​model determined by the second communication device based on the dimensional information can meet the expectations of the first communication device.

[0379] Optionally, in information A or information B, the first dimension information or the second dimension information is determined based on channel state information (CSI). Specifically, the first dimension information or the second dimension information contained in the first information can be determined based on the channel state information, so that the first dimension information or the second dimension information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to a certain extent, so that the AI ​​model that can be subsequently obtained based on the first information can be adapted to the channel characteristics of the wireless channel, in order to improve the transmission performance of the AI ​​data corresponding to the AI ​​model. Moreover, in the case where the AI ​​model obtained based on the first information can be adapted to the channel characteristics of the wireless channel, the transmission data of the wireless link can also meet the channel bandwidth requirements as much as possible, thereby improving the model performance of the AI ​​model included in the first AI model group.

[0380] Optionally, the channel state information may include the channel information between the first communication device and the second communication device, and / or the channel information between the second communication device and the first communication device. Wherein, when the first communication device is a terminal device and the second communication device is a network device, the channel information between the first communication device and the second communication device may be understood as uplink channel information, and the channel information between the second communication device and the first communication device may be understood as downlink channel information.

[0381] Optionally, the channel state information may be obtained based on reference signals.

[0382] For example, when the first communication device and the second communication device communicate through a sidelink, the reference signal may include a sidelink-synchronization signal / physical broadcast channel block (sidelink SSB, SL-SSB, or S-SS / PSBCH block), a sidelink-channel state information reference signal (SL-CSI-RS), etc.

[0383] For another example, when the first communication device and the second communication device communicate through uplink and downlink, the reference signal may include a channel state information reference signal (CSI-RS), a sounding reference signal (SRS), etc.

[0384] For information C, when the first information includes the input data of the first AI model and the label data of the input data of the first AI model, since the input data can be used as the input of the first AI model and the label data can be used as one of the bases for determining the model processing performance of the first AI model, therefore, for the second communication device, the second communication device may obtain an AI model with better performance based on these two pieces of information.

[0385] In addition, for the second communication device, the second communication device may perform mathematical calculations based on mutual information on these two pieces of information and the AI data (such as the input data of the second AI model or the output data of the second AI model, etc.) transmitted and received by the second communication device on the wireless link, and determine a first group of AI models based on the results of the mathematical calculations, so as to improve the model performance of the AI models included in the first group of AI models on the premise that the wireless link data meets the bandwidth.

[0386] For information D, when the first information includes the local computing power status information of the first communication device, since the complexity requirements of the model processing of the AI model may be related to the local computing power status of the first communication device. Therefore, for the second communication device, the first AI model group determined by the second communication device based on the local computing power status information can be adapted to the local computing power status of the first communication device to provide a first AI model that meets the local computing power status, thereby improving the success rate of the first communication device in performing model processing based on the first AI model.

[0387] For information E, when the first information includes channel status information, since the channel status information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device. Therefore, for the second communication device, the first AI model group determined by the second communication device based on the channel status information is adapted to the channel characteristics, so that the transmitted data of the wireless link can meet the channel bandwidth requirements as much as possible, in order to improve the transmission performance of the AI data corresponding to the AI model.

[0388] It can be understood that when the first information includes two or more of the above information A to information E, based on the technical gains brought by any one of the above descriptions, further superimposed gains can be obtained through the two or more pieces of information.

[0389] Optionally, the first information can be sent through one message / signaling or multiple messages / signaling, which is not limited here.

[0390] In a possible implementation, the first information sent by the first communication device in step S301 is used to determine the first AI model group, including: the first information is used to update the second AI model group to obtain the first AI model group; the second AI model group includes X AI models and a fourth AI model, where X is an integer greater than or equal to 1; among them, the input of any one of the X AI models includes the output of the fourth AI model, or the input of the fourth AI model includes the output of any one of the X AI models; the third AI model among the X AI models is deployed on the first communication device, and the third AI model includes one or more of the X AI models; the fourth AI model is deployed on the second communication device, and the X AI models are deployed on Y communication devices, where the Y communication devices include the first communication device and Y is less than or equal to X. In other words, for the second communication device, after receiving the first information, the second communication device can update the second AI model group based on the first information to obtain the first AI model group. In other words, the first information sent by the first communication device can be used to update other AI models, so that the solution can be applied to the AI model update scenario.

[0391] Optionally, "update" can be replaced with other terms, such as "modify", "iterate", "optimize", "process", etc.

[0392] Optionally, when the second AI model group is regarded as one AI model, the X AI models and the fourth AI model can be understood as multiple AI sub-models in the one AI model.

[0393] Optionally, the third AI model and the fourth AI model included in the second AI model group can be general models or special models to achieve the update of different types of models.

[0394] It should be understood that the general model can be called the base model, the large model or the L0 model. The special model can be called the small model, the L1 model, the L2 model, etc.

[0395] Taking the large model as an example, the large model can refer to a machine learning model with a large number of parameters and a complex structure, which can process massive data and complete various complex tasks, such as natural language processing, computer vision, speech recognition, etc.

[0396] Optionally, the large model is usually constructed by a deep neural network and has billions or even hundreds of billions of parameters.

[0397] Optionally, the design purpose of the large model can be to improve the expression ability and prediction performance of the model and be able to process more complex tasks and data.

[0398] Optionally, large models can learn complex patterns and features by training on massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unprocessed data.

[0399] In contrast, small models can refer to models with fewer parameters and shallower layers. Generally, compared with small models, large models usually have more parameters and deeper layers, with stronger expressive power and higher accuracy, but also require more computing resources and time for training and inference, and are suitable for scenarios with large amounts of data and sufficient computing resources, such as cloud computing, high-performance computing, artificial intelligence, etc.

[0400] Optionally, small models have the advantages of being lightweight, highly efficient, and easy to deploy, and are suitable for scenarios with small amounts of data and limited computing resources, such as mobile applications, embedded devices, the Internet of Things, etc.

[0401] In a possible implementation manner, the first information sent by the first communication device in step S301 is information sent periodically, and / or the second information sent by the second communication device in step S302 is information sent periodically. Specifically, the first information can be one of the determination bases of the AI model, and the second information can deploy the AI model. Among them, by periodically sending the first information and / or the second information between the first communication device and the second communication device, the periodic determination and / or periodic deployment of the AI model can be realized, so as to achieve multiple iterative updates of the AI model through a periodic process.

[0402] In a possible implementation manner, the first information sent by the first communication device in step S301 is used to determine the first group of AI models, and the AI models in the first group of AI models are dedicated models. Specifically, the first communication device can be a terminal device. For this reason, the AI model deployed on this terminal device can be a dedicated model, and the first information sent by this terminal device can be used to determine this dedicated model. Since different terminal devices may have different end-side characteristics (such as different local data, different local computing powers, different channel characteristics, etc.), by deploying dedicated models in terminal devices, in this way, the AI models deployed in terminal devices can be adapted to the end-side characteristics of terminal devices, with a view to improving the model processing performance of AI models.

[0403] Based on Figure 3In the technical solution shown, the second communication device is the recipient of the first information. The second communication device can determine a first AI model group based on the first information from the first communication device and send second information for determining the first AI model in the first AI model group to the first communication device. Thus, when the communication devices in the communication system are AI participating nodes, while enabling the computing power of the communication devices to be applied to the processing of AI models, the first information from the first communication device can also be used as one of the bases for the second communication device to determine the AI model, so that the AI model determined by the second communication device can be adapted to the first communication device as much as possible, thereby improving the processing performance of the subsequent model processing by the first communication device based on the AI model.

[0404] In Figure 3 In a possible implementation manner of the method shown, after the first communication device receives the second information in step S302, the method further includes: the first communication device receives fourth information, and the fourth information is used to indicate to stop the processing of the first AI model. Specifically, the first communication device may also receive fourth information for indicating to stop the processing of the first AI model, so that the first communication device stops participating in the processing of the first AI model group, can release the resources of the first communication device, and save the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0405] In addition, when K > 1, that is, the communication devices participating in the first AI model group may include other K - 1 communication devices in addition to the first communication device, the indication by the fourth information can enable the first communication device to stop the processing of the first AI model, and can implement pruning processing on some AI models (such as the first AI model) to prune inefficient devices / inefficient links and select a better (or optimal) collaboration set.

[0406] In a possible implementation manner, the second communication device can send (or determine / generate, etc.) the fourth information based on various methods, which will be described below in combination with some implementation examples.

[0407] As an implementation example, the second communication device can obtain the proportion of the effective information volume of the M links between the second communication device and M communication devices and send the fourth information to the communication device on a certain link where the proportion of the effective information volume is lower than the threshold. For example, the threshold for the proportion of the effective information volume of a single link is predefined or preconfigured as Th ratio ; when the proportion of the effective information volume ratio k of the kth link among the M links satisfies:

[0408]

[0409] Among them, the definitions of the respective parameters can refer to the previous description.

[0410] In the above manner, the second communication device can send the fourth information to the communication device on a certain link whose proportion of effective information volume is lower than the threshold Th ratio . Through the indication of the fourth information, some communication devices (such as the first communication device) can be made to stop the processing of the first AI model, enabling pruning processing of some AI models (such as the first AI model) to prune inefficient devices / inefficient links and achieve the selection of a better (or optimal) collaboration set.

[0411] Optionally, in addition to including the proportion of effective information volume of the M links and the effective information volume ratio threshold, the determination basis of the above fourth information can also be achieved through other parameters. For example, the effective information volume and the effective information volume threshold of some or all of the M links, the mutual information volume and the mutual information volume threshold of some or all of the M links, the dimension of the input data and the input data dimension threshold of some or all of the M links, the dimension of the output data and the output dimension threshold of some or all of the M links, etc.

[0412] Optionally, any of the above thresholds can be determined by the second communication device based on various methods. For example, the local computing power status of the second communication device, the number of AI tasks processed by the second communication device, etc. Alternatively, any of the thresholds is pre-configured.

[0413] In a possible implementation, the second communication device may obtain some of the parameters in the determination basis of the above fourth information based on the information sent by M communication devices (such as the effective information volume of M links, the mutual information volume of some or all of the M links, the dimension of the input data of some or all of the M links, the dimension of the output data of some or all of the M links, etc.). For example, before the first communication device receives the fourth information, the method further includes: the first communication device sends fifth information to the second communication device, and the fifth information is used to determine the fourth information; wherein, the fifth information includes at least one of the following: the input data or output data of the first AI model, the dimension information of the input data of the first AI model or the dimension information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device. Specifically, after the first communication device receives the second information for determining the first AI model in step S302, the first communication device may perform model processing based on the determined first AI model. Thereafter, the first communication device may send the fifth information so that the second communication device can determine the fourth information based on at least one of the information indicated by the fifth information. Thus, the fourth information indicating to the first communication device to stop model processing may be determined based on the state information of the first communication device indicated by the fifth information, so that the second communication device can implement pruning processing based on the state information of one or more communication devices in the cooperation set.

[0414] As another implementation example, the second communication device may also trigger sending the fourth information to the first communication device based on other methods. For example, the method further includes: the first communication device sends sixth information to the second communication device, and the sixth information is used to request to stop the processing of the first AI model. Specifically, after the first communication device receives the second information for determining the first AI model in step S302, the first communication device may perform model processing based on the determined first AI model. Thereafter, the first communication device may send the sixth information for requesting to stop the processing of the first AI model so that the second communication device can determine the fourth information based on the sixth information. Thus, the fourth information indicating to the first communication device to stop model processing may be determined based on the request of the first communication device, so that the second communication device can implement pruning processing based on the requests of one or more communication devices in the cooperation set.

[0415] Optionally, the first communication device may trigger the determination (or transmission) of the sixth information based on multiple trigger conditions. For example, the trigger conditions may include: the first communication device determines that the channel state indicated by the channel state information is lower than or equal to a threshold, the first communication device determines that the computing power indicated by its own computing power state information is lower than or equal to a threshold, and the first communication device determines that the dimension indicated by the dimension information of the input data or the output data of the first AI model is lower than a threshold, one or more of which.

[0416] Please refer to Figure 6 , which is a schematic diagram of an implementation of the communication method provided in this application. The method includes the following steps.

[0417] S601. The first communication device performs model processing based on the first AI model. The first AI model and the second AI model are included in the first AI model group. Among them, the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device. The input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model.

[0418] S602. The second communication device sends the fourth information. Correspondingly, the first communication device receives the fourth information. The fourth information is used to indicate to stop the processing of the first AI model.

[0419] In a possible implementation manner, the first AI model group further includes Z AI models. The input of any one of the Z AI models includes the output of the second AI model, or the input of the second AI model includes the output of any one or more of the Z AI models. Specifically, in addition to the first AI model deployed on the first communication device and the second AI model deployed on the second communication device, the first AI model group may further include Z AI models deployed on other communication devices. In other words, the communication devices participating in the first AI model group may include other communication devices in addition to the first communication device. By instructing with the fourth information, the first communication device can be made to stop the processing of the first AI model, enabling pruning processing of some AI models (such as the first AI model) to prune inefficient devices / inefficient links and select a better (or optimal) collaboration set.

[0420] In a possible implementation, before the first communication device receives the fourth information, the method further includes: the first communication device sends fifth information, where the fifth information is used to determine the fourth information; where the fifth information includes at least one of the following: input data or output data of the first AI model, dimension information of the input data of the first AI model or dimension information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device. Specifically, the first communication device may send the fifth information so that the second communication device determines the fourth information based on at least one of the information indicated by the fifth information. Thus, the fourth information indicating to stop model processing for the first communication device can be determined based on the state information of the first communication device indicated by the fifth information, enabling the second communication device to perform pruning processing based on the state information of one or more communication devices in the cooperation set.

[0421] In a possible implementation, the method further includes: the first communication device sends sixth information, where the sixth information is used to request to stop the processing of the first AI model. Specifically, the first communication device may send the sixth information for requesting to stop the processing of the first AI model so that the second communication device determines the fourth information based on the sixth information. Thus, the fourth information indicating to stop model processing for the first communication device can be determined based on the request of the first communication device, enabling the second communication device to perform pruning processing based on the requests of one or more communication devices in the cooperation set.

[0422] Optionally, the first communication device may trigger the determination (or sending) of the sixth information based on multiple triggering conditions. For example, the triggering conditions may include: the first communication device determines that the channel state indicated by the channel state information is lower than or equal to a threshold, the first communication device determines that the computing power indicated by its own computing power state information is lower than or equal to a threshold, and the first communication device determines that the dimension indicated by the dimension information of the input data or output data of the first AI model is lower than a threshold, one or more of these.

[0423] It should be noted that in Figure 6 the shown implementation, the fourth information / fifth information / sixth information may also refer to the previous description and achieve corresponding technical effects, which will not be elaborated here.

[0424] Based on Figure 6 the shown technical solution, during the process of the first communication device performing model processing based on the first AI model, the first communication device may receive the fourth information for indicating to stop the processing of the first AI model, enabling the first communication device to stop participating in the processing of the first AI model group, releasing the resources of the first communication device, and saving the transmission overhead of the input data and / or output data corresponding to the first AI model.

[0425] Please refer to Figure 7 , an embodiment of the present application provides a communication device 700. The communication device 700 can implement the functions of the second communication device or the first communication device in the foregoing method embodiment, and thus can also achieve the beneficial effects possessed by the foregoing method embodiment. In the embodiment of the present application, the communication device 700 may be the first communication device (or the second communication device), or may be an integrated circuit or component inside the first communication device (or the second communication device), such as a chip.

[0426] It should be noted that the transceiver unit 702 may include a sending unit and a receiving unit, which are respectively used to perform sending and receiving.

[0427] In a possible implementation manner, when the device 700 is used to execute the method performed by the first communication device in the foregoing embodiment, the device 700 includes a processing unit 701 and a transceiver unit 702; the processing unit 701 is used to determine the first information; the transceiver unit 702 is used to send the first information, and the first information is used to determine the first AI model group. The first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; wherein, the input of any one of the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed on the second communication device, and the K AI models are used to be deployed on M communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; the transceiver unit 702 is further used to receive the second information, and the second information is used to determine the first AI model, and the first AI model includes one or more of the K AI models.

[0428] In a possible implementation, when the device 700 is used to execute the method performed by the second communication device in the foregoing embodiment, the device 700 includes a processing unit 701 and a transceiver unit 702; the transceiver unit 702 receives first information; the processing unit 701 is configured to determine a first AI model group based on the first information, the first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; wherein, the input of any one of the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed on the second communication device, and the K AI models are deployed on M communication devices, the M communication devices include the first communication device, and M is less than or equal to K; the transceiver unit 702 is further configured to send second information, and the second information is used to determine a first AI model, the first AI model is deployed on the first communication device, and the first AI model includes one or more of the K AI models.

[0429] In a possible implementation, when the device 700 is used to execute the method performed by the first communication device in the foregoing embodiment, the device 700 includes a processing unit 701 and a transceiver unit 702; the processing unit 701 is configured to perform model processing based on a first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein, the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the transceiver unit 702 is configured to receive fourth information, and the fourth information is used to instruct to stop the processing of the first AI model.

[0430] In a possible implementation, when the device 700 is used to execute the method performed by the second communication device in the foregoing embodiment, the device 700 includes a processing unit 701 and a transceiver unit 702; the processing unit 701 is configured to determine fourth information, and the fourth information is used to instruct to stop the processing of a first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein, the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; the input of the first AI model includes the output of the second AI model, or the input of the second AI model includes the output of the first AI model; the transceiver unit 702 is configured to send the fourth information.

[0431] It should be noted that for the content such as the information execution process of the units of the communication device 700 above, please refer to the description in the method embodiment shown in the foregoing of this application, and details are not described herein again.

[0432] Please refer toFigure 8 FIG. Figure 8 is another schematic structural diagram of the communication device 800 provided by this application. The communication device 800 includes a logic circuit 801 and an input / output interface 802. Among them, the communication device 800 can be a chip or an integrated circuit.

[0433] Among them, Figure 7 the shown transceiver unit 702 can be a communication interface, and this communication interface can be Figure 8 the input / output interface 802 in FIG. Figure 8 . This input / output interface 802 can include an input interface and an output interface. Alternatively, this communication interface can also be a transceiver circuit, and this transceiver circuit can include an input interface circuit and an output interface circuit.

[0434] Optionally, the logic circuit 801 is used to determine first information; the input / output interface 802 is used to send the first information, and the first information is used to determine a first AI model group. The first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; among them, the input of any one of the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed in a second communication device, and the K AI models are used to be deployed in M communication devices. The M communication devices include the first communication device, and M is less than or equal to K; the input / output interface 802 is further used to receive second information, and the second information is used to determine a first AI model. The first AI model includes one or more of the K AI models.

[0435] Optionally, the input / output interface 802 receives the first information; the logic circuit 801 is used to determine a first AI model group based on the first information. The first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; among them, the input of any one of the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed in a second communication device, and the K AI models are deployed in M communication devices. The M communication devices include the first communication device, and M is less than or equal to K; the input / output interface 802 is further used to send second information, and the second information is used to determine a first AI model. The first AI model is deployed in the first communication device, and the first AI model includes one or more of the K AI models.

[0436] Optionally, the logic circuit 801 is configured to perform model processing based on a first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein, the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; an input of the first AI model includes an output of the second AI model, or an input of the second AI model includes an output of the first AI model; the input / output interface 802 is configured to receive fourth information, and the fourth information is used to indicate to stop the processing of the first AI model.

[0437] Optionally, the logic circuit 801 is configured to determine fourth information, and the fourth information is used to indicate to stop the processing of the first AI model, and the first AI model and the second AI model are included in a first AI model group; wherein, the first AI model is deployed on the first communication device, and the second AI model is deployed on the second communication device; an input of the first AI model includes an output of the second AI model, or an input of the second AI model includes an output of the first AI model; the input / output interface 802 is configured to send the fourth information.

[0438] Wherein, the logic circuit 801 and the input / output interface 802 may also perform other steps performed by the first communication device or the second communication device in any of the embodiments and achieve corresponding beneficial effects, which will not be elaborated herein.

[0439] In a possible implementation manner, Figure 7 the processing unit 701 shown may be Figure 8 the logic circuit 801 in.

[0440] Optionally, the logic circuit 801 may be a processing device, and the functions of the processing device may be implemented partially or entirely by software. Among them, the functions of the processing device may be implemented partially or entirely by software.

[0441] Optionally, the processing device may include a memory and a processor. Among them, the memory is used to store a computer program, and the processor reads and executes the computer program stored in the memory to perform corresponding processing and / or steps in any method embodiment.

[0442] Optionally, the processing device may only include a processor. The memory for storing the computer program is located outside the processing device, and the processor is connected to the memory through a circuit / wire to read and execute the computer program stored in the memory. Among them, the memory and the processor may be integrated together, or may also be physically independent of each other.

[0443] Optionally, the processing device may be one or more chips or one or more integrated circuits. For example, the processing device may be one or more field-programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), system on chips (SoCs), central processor units (CPUs), network processors (NPs), digital signal processing circuits (DSPs), microcontroller units (MCUs), programmable logic devices (PLDs), or other integrated chips, or any combination of the above chips or processors, etc.

[0444] Please refer to Figure 9 , the communication device 900 involved in the above embodiments provided by the embodiments of the present application. The communication device 900 may specifically be the communication device acting as a terminal device in the above embodiments. Figure 9 The example shown is implemented by the terminal device (or components in the terminal device).

[0445] Among them, a possible schematic logical structure diagram of the communication device 900. The communication device 900 may include but is not limited to at least one processor 901 and a communication port 902.

[0446] Among them, Figure 7 The shown transceiver unit 702 may be a communication interface. The communication interface may be Figure 9 the communication port 902 in

[0447]

[0448] the communication port 902 may include an input interface and an output interface. Alternatively, the communication port 902 may also be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit. Further optionally, the device may further include at least one of a memory 903 and a bus 904. In the embodiments of the present application, the at least one processor 901 is used to control and process the actions of the communication device 900.

[0448] In addition, the processor 901 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in connection with the disclosure of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0449] It should be noted that Figure 9 the communication device 900 shown can specifically be used to implement the steps implemented by the terminal device in the foregoing method embodiments, and achieve the corresponding technical effects of the terminal device. Figure 9 For the specific implementation manners of the communication device shown, reference can be made to the descriptions in the foregoing method embodiments, and details will not be repeated here.

[0450] Please refer to Figure 10 FIG. 1000 is a schematic structural diagram of the communication device involved in the foregoing embodiments provided in the embodiments of the present application. The communication device 1000 may specifically be the communication device acting as a network device in the foregoing embodiments. Figure 10 The example shown is implemented by a network device (or a component in the network device). Among them, the structure of the communication device can refer to Figure 10 the structure shown.

[0451] The communication device 1000 includes at least one processor 1011 and at least one network interface 1014. Further optionally, the communication device further includes at least one memory 1012, at least one transceiver 1013, and one or more antennas 1015. The processor 1011, the memory 1012, the transceiver 1013, and the network interface 1014 are connected, for example, through a bus. In the embodiments of the present application, this connection may include various interfaces, transmission lines, or buses, etc., and this embodiment does not limit this. The antenna 1015 is connected to the transceiver 1013. The network interface 1014 is used to enable the communication device to communicate with other communication devices through a communication link. For example, the network interface 1014 may include a network interface between the communication device and a core network device, such as an S1 interface. The network interface may include a network interface between the communication device and other communication devices (such as other network devices or core network devices), such as an X2 or Xn interface.

[0452] Among them, Figure 7 the transceiver unit 702 shown may be a communication interface, and this communication interface may beFigure 10 The network interface 1014 therein, and the network interface 1014 may include an input interface and an output interface. Alternatively, the network interface 1014 may also be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit.

[0453] The processor 1011 is mainly used to process communication protocols and communication data, control the entire communication device, execute software programs, and process the data of software programs, for example, to support the communication device to perform the actions described in the embodiments. The communication device may include a baseband processor and a central processor. The baseband processor is mainly used to process communication protocols and communication data, and the central processor is mainly used to control the entire terminal device, execute software programs, and process the data of software programs. Figure 10 The processor 1011 therein may integrate the functions of the baseband processor and the central processor. Those skilled in the art can understand that the baseband processor and the central processor may also be independent processors interconnected through technologies such as a bus. Those skilled in the art can understand that the terminal device may include multiple baseband processors to adapt to different network systems, the terminal device may include multiple central processors to enhance its processing capabilities, and various components of the terminal device may be connected through various buses. The baseband processor may also be referred to as a baseband processing circuit or a baseband processing chip. The central processor may also be referred to as a central processing circuit or a central processing chip. The function of processing communication protocols and communication data may be built into the processor or stored in the memory in the form of a software program, and the processor executes the software program to implement the baseband processing function.

[0454] The memory is mainly used to store software programs and data. The memory 1012 may exist independently and be connected to the processor 1011. Optionally, the memory 1012 may be integrated with the processor 1011, for example, integrated within a single chip. Among them, the memory 1012 can store the program code for implementing the technical solution of the embodiments of the present application and be controlled by the processor 1011 to execute. Various types of computer program codes being executed may also be regarded as the driver programs of the processor 1011.

[0455] Figure 10 Only one memory and one processor are shown. In an actual terminal device, there may be multiple processors and multiple memories. The memory may also be referred to as a storage medium or a storage device, etc. The memory may be a storage element on the same chip as the processor, that is, an on-chip storage element, or an independent storage element, and the embodiments of the present application do not make any limitations in this regard.

[0456] The transceiver 1013 can be used to support the reception or transmission of radio frequency signals between a communication device and a terminal. The transceiver 1013 can be connected to the antenna 1015. The transceiver 1013 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 1015 can receive radio frequency signals. The receiver Rx of the transceiver 1013 is used to receive the radio frequency signals from the antenna, convert the radio frequency signals into digital baseband signals or digital intermediate frequency signals, and provide the digital baseband signals or digital intermediate frequency signals to the processor 1011 so that the processor 1011 can perform further processing on the digital baseband signals or digital intermediate frequency signals, such as demodulation processing and decoding processing. In addition, the transmitter Tx in the transceiver 1013 is also used to receive the modulated digital baseband signals or digital intermediate frequency signals from the processor 1011, convert the modulated digital baseband signals or digital intermediate frequency signals into radio frequency signals, and transmit the radio frequency signals through one or more antennas 1015. Specifically, the receiver Rx can selectively perform one-stage or multi-stage down-conversion processing and analog-to-digital conversion processing on the radio frequency signals to obtain digital baseband signals or digital intermediate frequency signals, and the sequence of the down-conversion processing and the analog-to-digital conversion processing can be adjusted. The transmitter Tx can selectively perform one-stage or multi-stage up-conversion processing and digital-to-analog conversion processing on the modulated digital baseband signals or digital intermediate frequency signals to obtain radio frequency signals, and the sequence of the up-conversion processing and the digital-to-analog conversion processing can be adjusted. The digital baseband signals and the digital intermediate frequency signals can be collectively referred to as digital signals.

[0457] The transceiver 1013 can also be referred to as a transceiver unit, a transceiver, a transceiver device, etc. Optionally, the devices used to implement the receiving function in the transceiver unit can be regarded as a receiving unit, and the devices used to implement the sending function in the transceiver unit can be regarded as a sending unit. That is, the transceiver unit includes a receiving unit and a sending unit. The receiving unit can also be referred to as a receiver, an input port, a receiving circuit, etc., and the sending unit can be referred to as a transmitter, a transmitter, or a transmitting circuit, etc.

[0458] It should be noted that Figure 10 The illustrated communication device 1000 can specifically be used to implement the steps implemented by the network device in the foregoing method embodiments and achieve the corresponding technical effects of the network device. Figure 10 For the specific implementation manners of the illustrated communication device 1000, reference can be made to the descriptions in the foregoing method embodiments, and details are not described herein again.

[0459] Please refer to Figure 11 , which is a schematic structural diagram of the communication device involved in the foregoing embodiments provided by the embodiments of the present application.

[0460] It can be understood that the communication device 110 includes, for example, modules, units, components, circuits, or interfaces, etc., which are appropriately configured together to execute the technical solutions provided in this application. The communication device 110 may be the terminal device or network device described above, or a component (such as a chip) in these devices, for implementing the methods described in the following method embodiments. The communication device 110 includes one or more processors 111. The processor 111 may be a general-purpose processor or a dedicated processor, etc. For example, it may be a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication device (such as a RAN node, a terminal, or a chip, etc.), execute software programs, and process the data of software programs.

[0461] Optionally, in one design, the processor 111 may include a program 113 (sometimes also referred to as code or instructions), and the program 113 can be run on the processor 111, so that the communication device 110 executes the methods described in the following embodiments. In another possible design, the communication device 110 includes a circuit ( Figure 11 not shown).

[0462] Optionally, the communication device 110 may include one or more memories 112, on which there is a program 114 (sometimes also referred to as code or instructions), and the program 114 can be run on the processor 111, so that the communication device 110 executes the methods described in the above method embodiments.

[0463] Optionally, the processor 111 and / or the memory 112 may include AI modules 117, 118, and the AI modules are used to implement AI-related functions. The AI modules can be implemented in a software, hardware, or a combination of software and hardware manner. For example, the AI modules may include a radio intelligence control (RIC) module. For example, the AI module may be a near-real-time RIC or a non-real-time RIC.

[0464] Optionally, data may also be stored in the processor 111 and / or the memory 112. The processor and the memory may be provided separately or integrated together.

[0465] Optionally, the communication device 110 may further include a transceiver 115 and / or an antenna 116. The processor 111 is sometimes also referred to as a processing unit, which controls the communication device (such as a RAN node or a terminal). The transceiver 115 is sometimes also referred to as a transceiver unit, a transceiver, a transceiver circuit, or a transceiver, etc., and is used to implement the transceiver function of the communication device through the antenna 116.

[0466] Among them, Figure 7 the processing unit 701 shown may be the processor 111. Figure 7 The transceiver unit 702 shown may be a communication interface, and this communication interface may be Figure 11 the transceiver 115 in, and this transceiver 115 may include an input interface and an output interface. Alternatively, this transceiver 115 may also be a transceiver circuit, and this transceiver circuit may include an input interface circuit and an output interface circuit.

[0467] An embodiment of the present application further provides a computer-readable storage medium, which is used to store one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes the method described in the possible implementation manners of the first communication device or the second communication device in the foregoing embodiments.

[0468] An embodiment of the present application further provides a computer program product (or computer program) containing programs or instructions. When the computer program product is executed by the processor, the processor executes the method of the possible implementation manners of the foregoing first communication device or second communication device.

[0469] An embodiment of the present application further provides a chip system, which includes at least one processor for supporting a communication device to implement the functions involved in the possible implementation manners of the foregoing communication device. Optionally, the chip system further includes an interface circuit, and the interface circuit provides program instructions and / or data for the at least one processor. In a possible design, the chip system may further include a memory for storing the necessary program instructions and data of the communication device. The chip system may be composed of chips or may include chips and other discrete devices, where the communication device may specifically be the first communication device or the second communication device in the foregoing method embodiments.

[0470] An embodiment of the present application further provides a communication system, and the network system architecture includes the first communication device and the second communication device in any of the foregoing embodiments.

[0471] In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other may be indirect couplings or communication connections through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0472] The unit described as a separation component may or may not be physically separated. The component displayed as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0473] In addition, each functional unit in various embodiments of the present application may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that makes a contribution, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

Claims

1. A communication method, characterized in that, The method is applied to a first communication device, and the method includes: Sending first information, where the first information is used to determine a first artificial intelligence (AI) model group, and the first AI model group includes K AI models and a second AI model, where K is an integer greater than or equal to 1; wherein, the input of any one of the K AI models includes the output of the second AI model, or, the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed on a second communication device, and the K AI models are used to be deployed on M communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; Receiving second information, where the second information is used to determine a first AI model, and the first AI model includes one or more of the K AI models.

2. The method according to claim 1, characterized in that, Before receiving the second information, the method further includes: Receiving third information, where the third information is used to configure the mapping relationship between P dimension information and Q indexes, and both P and Q are positive integers; the second information includes one or more of the Q indexes, and the one or more indexes are used to determine one of the P dimension information.

3. The method according to claim 2, characterized in that, The third information is a layer 3 signaling, and the second information is a layer 1 and / or layer 2 signaling.

4. The method according to any one of claims 1 to 3, characterized in that, The second information includes at least one of the following: The model parameters of the first AI model, the model parameters of the first AI model group, the dimension information of the input data of the first AI model, the dimension information of the output data of the first AI model.

5. The method according to any one of claims 1 to 4, characterized in that, The first information includes first dimension information or second dimension information; When the input of the first AI model includes the output of the second AI model, the first dimension information is used to determine the dimension information of the input data of the first AI model or the dimension information of the output data of the second AI model; When the input of the second AI model includes the output of the first AI model, the second dimension information is used to determine the dimension information of the output data of the first AI model or the dimension information of the input data of the second AI model.

6. The method according to claim 5, characterized in that The dimension information includes at least one of the following: The upper limit value of the dimension, the lower limit value of the dimension, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device.

7. The method according to claim 5 or 6, characterized in that, The first dimension information or the second dimension information is determined based on the channel state information.

8. The method according to any one of claims 1 to 7, characterized in that The first information includes at least one of the following: The input data of the first AI model and the label data of the input data of the first AI model, the local computing power state information of the first communication device, the channel state information.

9. The method according to any one of claims 1 to 8, characterized in that, The first information is used to determine the first AI model group, including: The first information is used to update the second AI model group to obtain the first AI model group; the second AI model group includes X AI models and a fourth AI model, where X is an integer greater than or equal to 1; among them, the input of any one of the X AI models includes the output of the fourth AI model, or the input of the fourth AI model includes the output of any one of the X AI models; the third AI model among the X AI models is deployed on the first communication device, and the third AI model includes one or more of the X AI models; the fourth AI model is deployed on the second communication device, and the X AI models are deployed on Y communication devices, and the Y communication devices include the first communication device, and Y is less than or equal to X.

10. The method according to any one of claims 1 to 9, characterized in that, After receiving the second information, the method further includes: Receiving fourth information, where the fourth information is used to indicate to stop the processing of the first AI model.

11. The method according to claim 10, characterized in that, Before receiving the fourth information, the method further includes: Sending fifth information, where the fifth information is used to determine the fourth information; where the fifth information includes at least one of the following: The input data or output data of the first AI model, the dimension information of the input data of the first AI model or the dimension information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, the channel state information of the first communication device.

12. The method according to claim 10 or 11, characterized in that, The method further includes: Sending sixth information, where the sixth information is used to request to stop the processing of the first AI model.

13. A communication method, characterized in that, Including: Receiving the first information; Determining a first artificial intelligence AI model group based on the first information, where the first AI model group includes K AI models and a second AI model, and K is an integer greater than or equal to 1; among them, the input of any one of the K AI models includes the output of the second AI model, or the input of the second AI model includes the output of one or more of the K AI models; the second AI model is deployed on the second communication device, and the K AI models are deployed on M communication devices, and the M communication devices include the first communication device, and M is less than or equal to K; Sending second information, where the second information is used to determine a first AI model, and the first AI model is deployed on the first communication device, and the first AI model includes one or more of the K AI models.

14. The method according to claim 13, wherein Before sending the second information, the method further includes: Sending third information, where the third information is used to configure the mapping relationship between P dimension information and Q indexes, and both P and Q are positive integers; the second information includes one or more of the Q indexes, and the one or more indexes are used to determine one of the P dimension information.

15. The method according to claim 14, wherein The third information is a layer 3 signaling, and the second information is a layer 1 and / or layer 2 signaling.

16. The method according to any one of claims 13 to 15, characterized in that, The second information includes at least one of the following: The model parameters of the first AI model, the model parameters of the first AI model group, the dimension information of the input data of the first AI model, and the dimension information of the output data of the first AI model.

17. The method according to any one of claims 13 to 16, characterized in that, The first information includes first dimension information or second dimension information; When the input of the first AI model includes the output of the second AI model, the first dimension information is used to determine the dimension information of the input data of the first AI model or the dimension information of the output data of the second AI model; When the input of the second AI model includes the output of the first AI model, the second dimension information is used to determine the dimension information of the output data of the first AI model or the dimension information of the input data of the second AI model.

18. The method according to claim 17, wherein The dimension information includes at least one of the following: The upper limit value of the dimension, the lower limit value of the dimension, the dimension expected by the first communication device, and the value range of the dimension expected by the first communication device.

19. The method according to claim 17 or 18, characterized in that, The first dimension information or the second dimension information is determined based on the channel state information.

20. The method according to any one of claims 13 to 19, characterized in that The first information includes at least one of the following: The input data of the first AI model and the label data of the input data of the first AI model, the local computing power state information of the first communication device, and the channel state information.

21. The method according to any one of claims 13 to 20, characterized in that The first information is used to determine the first AI model group, including: The first information is used to update the second AI model group to obtain the first AI model group; the second AI model group includes X AI models and a fourth AI model, where X is an integer greater than or equal to 1; among them, the input of any one of the X AI models includes the output of the fourth AI model, or the input of the fourth AI model includes the output of any one of the X AI models; the third AI model among the X AI models is deployed on the first communication device, and the third AI model includes one or more AI models among the X AI models; the fourth AI model is deployed on a second communication device, and the X AI models are deployed on Y communication devices, and the Y communication devices include the first communication device, and Y is less than or equal to X.

22. The method according to any one of claims 13 to 21, characterized in that, After sending the second information, the method further includes: Sending fourth information, where the fourth information is used to indicate to stop the processing of the first AI model.

23. The method according to claim 22, characterized in that, Before sending the fourth information, the method further includes: Receiving fifth information, where the fifth information is used to determine the fourth information; among them, the fifth information includes at least one of the following: The input data or output data of the first AI model, the dimension information of the input data of the first AI model or the dimension information of the output data of the first AI model, the dimension expected by the first communication device, the value range of the dimension expected by the first communication device, and the channel state information of the first communication device.

24. The method according to claim 22 or 23, characterized in that, The method further includes: Receiving sixth information, where the sixth information is used to request to stop the processing of the first AI model.

25. A communication device, characterized in that, Including a module for executing the method according to any one of claims 1 to 24.

26. A communication device, characterized in that, Comprising at least one processor, the at least one processor being coupled to a memory; the at least one processor is configured to execute a program or instructions stored in the memory, such that the communication device executes the method according to any one of claims 1 to 24.

27. The communication device according to claim 26, wherein The communication device is a chip or a chip system.

28. A readable storage medium, characterized in that, A computer program or instructions are stored in the storage medium, and when the computer program or instructions are executed by a communication device, the method according to any one of claims 1 to 24 is implemented.

29. A computer program product, characterized in that, Comprising computer instructions, when the instructions are run on a processor, the method according to any one of claims 1 to 24 is caused to be executed.

Images