Communication methods and related devices

Abstract

The present application provides a communication method and a related device for determining and distributing an AI model in a communication network through interaction between different communication devices, which enables the computational power of the communication device to be applied to AI model processing. In the method, a first communication device transmits first information and then receives second information to determine a first AI model group, wherein the second information includes model parameters of the first AI model group or model parameters of the first AI model within the first AI model group, and the first AI model group includes a first AI model distributed to the first communication device and a second AI model distributed to the second communication device.

Description

Technology Field

[0001] This application claims priority to Chinese patent application No. 202311600304.7, titled “Communication method and related device,” filed with the Chinese Intellectual Property Office on November 24, 2023, the entirety of which is incorporated by reference into this specification.

[0002] Technology field

[0003] This application relates to the field of communication technology, and in particular to communication methods and related devices. Background Technology

[0004] Wireless communication is transmission between two or more communication nodes without using conductors or cables. Communication nodes generally include network devices and terminal devices.

[0005] Currently, in wireless communication systems, communication nodes generally possess both signal transmission and reception capabilities and computational capabilities. Taking a network device with computational capabilities as an example, this capability primarily provides computational support for transmission and reception (e.g., signal processing), thereby enabling communication with other communication nodes.

[0006] However, in a communication network, in addition to the ability to provide computational power support for the aforementioned communication tasks, communication nodes may have additional surplus computational power. How to utilize this surplus computational power is an important technical challenge.

[0007] The present application provides a communication method and a related device for determining and distributing an artificial intelligence (AI) model in a communication network through interaction between different communication devices, which enables the computational capabilities of the communication devices to be applied to the processing of the AI ​​model.

[0008] A first aspect of the present application provides a communication method. The method is performed by a first communication device. The first communication device may be a communication device (e.g., a terminal device), the first communication device may be a component within the communication device (e.g., a processor, a chip, or a chip system), or the first communication device may be a logical module or software capable of implementing all or some functions of the communication device. In the method, the first communication device transmits first information, the first information is used to determine a first AI model group, the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to 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 second information from the second communication device, the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0009] Based on the aforementioned technical solution, the second communication device is used as a receiver of the first information, and the second communication device determines a first AI model group based on the first information from the first communication device and can distribute the first AI model to the first communication device through the second information. Accordingly, when a communication device within a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the first information from the first communication device may be used as one of the decision 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 to the maximum extent, and accordingly, the processing performance of the model processing subsequently performed by the first communication device based on the AI ​​model is improved.

[0010] In this application, terms such as AI model, neural network model, AI neural network model, machine learning model, and AI processing model may be substituted for each other.

[0011] It should be understood that the first AI model group including a first AI model and a second AI model can be understood as the function of the first AI model group being implemented through at least model processing of the first AI model and model processing of the second AI model. That is, after receiving the second information, the first communication device may distribute the first AI model to the first communication device through the model parameters of the first AI model group or the model parameters of the first AI model included in the second information and perform model processing on the first AI model. Correspondingly, the second communication device may perform model processing on the second AI model distributed to the second communication device. Optionally, model processing may include one or more of model update processing, model training processing, and model inference processing.

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

[0013] For example, the second communication device may be a terminal device. Correspondingly, the first communication device and the second communication device may communicate with each other over a sidelink (SL). In this case, the first AI model and the second AI model may be referred to as an end-to-end model, an end-to-end collaboration model, etc.

[0014] In another example, the second communication device may be a network device (e.g., an access network device). Correspondingly, the first communication device and the second communication device may communicate with each other over uplink and downlink communication links. In this case, the first AI model and the second AI model may be referred to as an edge-terminal model, an edge-terminal collaboration model, a terminal-edge model, a terminal-edge collaboration model, etc.

[0015] Optionally, when the first AI model group is considered as one AI model, the first AI model and the second AI model can be understood as two AI submodels in the AI ​​model.

[0016] In the present application, the fact that one AI model is distributed to one communication device (e.g., a first AI model is distributed to a first communication device and a second AI model is distributed to a second communication device) can be understood as the communication device acquiring the model parameters of the AI ​​model, acquiring / generating / configuring the AI ​​model based on the model parameters of the AI ​​model, and subsequently performing model processing on the AI ​​model.

[0017] Optionally, model parameters may include one or more of the model's hyperparameters, the model's dataset (including the model's input data and label data corresponding to the input data), and the model's structural parameters.

[0018] Optionally, one AI model group may include two or more AI models. For example, in addition to the first AI model and the second AI model, the first AI model group may additionally include another AI model. The other AI model may be deployed to a different communication device different from the first communication device and the second communication device. This is not limited in the present specification.

[0019] It must be understood that the transmission of wireless communication signals (e.g., the reception and transmission of configuration information of communication resources and the reception and transmission of reference signals) can be performed between different communication devices (e.g., a first communication device and a second communication device).

[0020] Optionally, the AI ​​models of the present application (e.g., the first AI model, the second AI model, and the subsequent third through sixth AI models) may be used for the management of wireless communication signals (including at least one of configuration, updating, and optimization). For example, the AI ​​models may include one or more of an AI model used for modulation and / or demodulation, an AI model used for channel prediction, an AI model used for beam management, an AI model used for auxiliary positioning, an AI model used for channel compression, an AI model used for resource scheduling, and an AI model used to replace one or more modules in a transmitter and / or receiver. Alternatively, the AI ​​models of the present application may be AI models used for other AI tasks, e.g., an AI model used for image recognition, an AI model used for natural language processing, or an AI model used for computer vision.

[0021] A second aspect of the present application provides a communication method. The method is performed by a second communication device. The second communication device may be a communication device (e.g., a network device or a terminal device), or the second communication device may be a component within the communication device (e.g., a processor, a chip, or a chip system), or a logical module or software capable of implementing all or some functions of the communication device. In the method, the second communication device receives first information. The second communication device determines a first AI model group based on the first information, and the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to the second communication device, and 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 transmits second information, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0022] Based on the aforementioned technical solution, the second communication device is used as a receiver of the first information, and the second communication device can determine a first AI model group based on the first information from the first communication device and distribute the first AI model to the first communication device through the second information. Accordingly, when a communication device of a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the first information from the first communication device may be used as one of the grounds 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 to the maximum extent, thereby improving the processing performance of the model processing subsequently performed by the first communication device based on the AI ​​model.

[0023] In a possible implementation of the first or second embodiment, the second communication device is a functional entity that determines a list of AI model groups based on the first information, and the list of AI model groups includes one or more AI model groups, and one or more AI model groups include the first AI model group.

[0024] Based on the aforementioned technical solution, the second communication device can communicate with one or more first communication devices, and the second communication device can receive information (e.g., one or more first information) from one or more first communication devices to create / acquire / determine one or more AI model groups. In other words, the second communication device can collect information and create a model based on the collected information. Subsequently, the second communication device can distribute the AI ​​model to one or more first communication devices.

[0025] Optionally, the list of AI model groups may include one or more AI model groups, and each AI model group may include two or more AI models. As described above, the relationship between AI model groups and AI models can also be understood as the relationship between an AI model and an AI submodel. Therefore, the list of AI model groups can alternatively be replaced by a list of AI models. That is, the list of AI models may include one or more AI models.

[0026] Optionally, list can be replaced with other terms such as set, dictionary, combination, space, etc.

[0027] In a possible implementation of the first or second embodiment, the second communication device is a functional entity that determines a list of AI model groups based on first information and selects some or all of the AI ​​model groups to be used by the first communication device from the list of AI model groups.

[0028] Based on the aforementioned technical solution, after the second communication device determines a list of AI model groups based on the first information, the function implemented by the second communication device may further include selecting some or all of the AI ​​model groups used by the first communication device from the list of AI model groups. In other words, in addition to collecting information and generating a model based on the collected information, the second communication device may additionally perform model selection, so that the second communication device may subsequently distribute an AI model adapted to one or more first communication devices to one or more first communication devices.

[0029] In a possible implementation of the first or second embodiment, the first information includes first dimensional information or second dimensional information. Where the input of the first AI model includes the output of the second AI model, the first dimensional information is used to determine the dimensional information of the input data of the first AI model or the dimensional information of the output data of the second AI model. Where the input of the second AI model includes the output of the first AI model, the second dimensional information is used to determine the dimensional information of the output data of the first AI model or the dimensional information of the input data of the second AI model.

[0030] Based on the aforementioned technical solution, the second communication device can determine the dimension information of the data transmitted over 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 over the communication link is related to the processing performance of the AI ​​model, the implementation process for the second communication device to determine the first AI model group can be simplified in this way, thereby reducing the complexity of the second communication device and improving the processing performance of the AI ​​model included in the first AI model group.

[0031] In a possible implementation of the first or second embodiment, the dimension information includes at least one of an upper limit of a dimension, a lower limit of a dimension, a dimension expected by the first communication device (or a dimension not expected by the first communication device), and a value range of the dimension expected by the first communication device (or a value range of the dimension not expected by the first communication device).

[0032] Optionally, the dimension information may include the value of any of the aforementioned items, the quantized value, the index of the value, and the index of the quantized value. Other implementations are also possible.

[0033] Based on the aforementioned technical solution, the dimensional information determined based on the first dimensional information or the second dimensional information may include at least one of the aforementioned items, thereby improving the flexibility of implementing the solution.

[0034] For example, if the dimension information includes an upper limit and / or a lower limit of the dimension, the second communication device may use the range indicated by the upper limit and / or lower limit as one of the grounds for determining the AI ​​model, thereby improving the flexibility of implementing the solution.

[0035] In another example, if the dimension information includes the dimension expected by the first communication device and / or the range of values ​​of the dimension expected by the first communication device, the AI ​​model determined by the second communication device based on the dimension information can satisfy the expectations of the first communication device.

[0036] In a possible implementation of the first or second embodiment, the first dimension information or the second dimension information is determined based on channel state information (CSI).

[0037] Based on the aforementioned technical solution, the first dimensional information or the second dimensional information included in the first information can be determined based on channel state information, so that the first dimensional information or the second dimensional information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to a certain extent. In this way, an AI model that can be subsequently acquired based on the first information is adapted to the channel characteristics of the wireless channel, thereby improving the transmission performance of the AI ​​data corresponding to the AI ​​model. Furthermore, when the AI ​​model acquired based on the first information can be adapted to the channel characteristics of the wireless channel, the data transmitted over the wireless link can also satisfy channel bandwidth requirements to the maximum extent, thereby improving the model performance of the AI ​​model included in the first AI model group.

[0038] Optionally, the channel state information may include information about a channel from a first communication device to a second communication device and / or information about a channel from a second communication device to a first communication device. When the first communication device is a terminal device and the second communication device is a network device, information about a channel from the first communication device to the second communication device may be understood as uplink channel information, and information about a channel from the second communication device to the first communication device may be understood as downlink channel information.

[0039] Optionally, channel status information can be obtained based on a reference signal.

[0040] For example, when a first communication device and a second communication device communicate with each other 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.

[0041] In another example, when the first communication device and the second communication device communicate with each other via an uplink and a downlink, the reference signal may include a channel state information reference signal (CSI-RS), a sounding reference signal (SRS), etc.

[0042] In a possible implementation of the first or second embodiment, the first information comprises at least one of the following: input data of the first AI model, label data of the input data of the first AI model, local computational capability status information of the first communication device, and channel status information.

[0043] Based on the aforementioned technical solution, the first information may include at least one of the aforementioned information. That is, the second communication device may determine a first AI model group based on at least one of the aforementioned information to improve the flexibility of implementing the solution.

[0044] In one embodiment, when the first information includes input data of a first AI model and label data of the input data of the first AI model, the input data can be used as an input to the first AI model, and the label data can be used as one of the grounds for determining the model processing performance of the first AI model; therefore, in the case of a second communication device, the second communication device can obtain an AI model with excellent performance based on the two pieces of information.

[0045] In addition, in the case of the second communication device, the second communication device implements a mutual information-based mathematical calculation based on two pieces of information received and transmitted by the second communication device over a wireless link and AI data (e.g., input data of the second AI model or output data of the second AI model), determines a first AI model group based on the result of the mathematical calculation, and improves the model performance of the AI ​​models included in the first AI model group under the premise that the data over the wireless link satisfies the bandwidth.

[0046] In another implementation example, if the first information includes local computational capability state information of the first communication device, the complexity requirement of the model processing of the AI ​​model may be related to the local computational capability state of the first communication device. Accordingly, in the case of the second communication device, the first AI model group determined by the second communication device based on the local computational capability state information can be adapted to the local computational capability state of the first communication device, thereby providing a first AI model that satisfies the local computational capability state, which improves the success rate of the first communication device performing model processing based on the first AI model.

[0047] In another implementation example, where the first information includes channel state information, the channel state information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device. Therefore, in the case of the second communication device, the second communication device determines a first AI model group to adapt to the channel characteristics based on the channel state information, thereby ensuring that the data transmitted over the wireless link fully satisfies the channel bandwidth requirements and improving the transmission performance of the AI ​​data corresponding to the AI ​​model.

[0048] If the first information includes two or more of the information described above, it may be understood that additional overlapping benefits may be obtained through the two or more pieces of information based on the technical benefits achieved from any of the information described above.

[0049] In a possible implementation of the first or second embodiment, the first information being used to determine a first AI model group comprises: the first information being used to obtain a first AI model group by updating a second AI model group; the second AI model group comprising a third AI model and a fourth AI model, wherein the third AI model is deployed to a first communication device and the fourth AI model is deployed to a second communication device, and the input of the third AI model comprises the output of the fourth AI model, or the input of the fourth AI model comprises the output of the third AI model.

[0050] Optionally, "update" may be replaced with other terms such as, for example, "modification," "iteration," "optimization," "processing," etc.

[0051] Optionally, when the second AI model group is considered as one AI model, the third AI model and the fourth AI model can be understood as two AI sub-models in the AI ​​model.

[0052] Based on the aforementioned technical solution, in the case of a second communication device, after the second communication device receives the first information, the second communication device can obtain the first AI model group by updating the second AI model group based on the first information. That is, since the first information transmitted by the first communication device can be used to update another AI model, the solution can be applied to an AI model update scenario.

[0053] Optionally, the third and fourth AI models included in the second AI model group may be general models or dedicated models, so that different types of models are updated.

[0054] It should be understood that general models may be referred to as basic models, large models, or L0 models. Dedicated models may be referred to as small models, L1 models, L2 models, etc.

[0055] Large models are used as an example. Large models can be machine learning models with a large number of parameters and complex structures, capable of processing large-scale data and completing various complex tasks such as natural language processing, computer vision, and speech recognition.

[0056] Optionally, large models are typically built based on deep neural networks and have billions or hundreds of billions of parameters.

[0057] Optionally, large models can be designed to improve the model's representational power and predictive performance, and to handle more complex tasks and data.

[0058] Optionally, large models can learn complex patterns and features by training on large-scale data, possess stronger generalization capabilities, and accurately predict unprocessed data.

[0059] In contrast, a small model may be a model with fewer parameters and fewer layers. Generally, compared to small models, large models usually have more parameters and deeper layers, and possess stronger representational power and higher accuracy. However, large models also require more computing resources and time for training and inference, and are applicable to scenarios with large amounts of data and sufficient computing resources, such as cloud computing, high-performance computing, or artificial intelligence.

[0060] Optionally, the compact model offers advantages such as lightweight design, high efficiency, and easy deployment, and is applicable to scenarios with small data volumes and limited computing resources, such as mobile applications, embedded devices, or the Internet of Things.

[0061] In a possible implementation of the first or second embodiment, the first information is information transmitted periodically and / or the second information is information transmitted periodically.

[0062] Based on the aforementioned technical solution, the first information may be one of the grounds for determining the AI ​​model, and the second information may be used to deploy the AI ​​model. The AI ​​model may be periodically determined and / or periodically deployed by periodically transmitting the first information and / or the second information between the first communication device and the second communication device, so that the AI ​​model is repeatedly updated multiple times through a periodic process.

[0063] In a possible implementation of the first or second embodiment, the AI ​​model of the first AI model group is a dedicated model.

[0064] Based on the aforementioned technical solution, the first communication device may be a terminal device. Accordingly, the AI ​​model deployed to the terminal device may be a dedicated model, and the first information transmitted by the terminal device may be used to determine the dedicated model. Different terminal devices may have different device-side characteristics (e.g., different local data, different local computational capabilities, and different channel characteristics). Accordingly, in this method where a dedicated model is deployed to the terminal device, the AI ​​model deployed to the terminal device can be adapted to the device-side characteristics of the terminal device, thereby improving the model processing performance of the AI ​​model.

[0065] A third aspect of the present application provides a communication method. The method is performed by a second communication device. The second communication device may be a communication device (e.g., a cloud server or a core network device), the second communication device may be a component within the communication device (e.g., a processor, a chip, or a chip system), or the second communication device may be a logical module or software capable of implementing all or some functions of the communication device. In the method, the second communication device transmits third information, the third information is used to determine a third AI model group, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to the first communication device, and the sixth AI model is distributed to the second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The second communication device receives fourth information from the third communication device, the fourth information includes model parameters of the third AI model group.

[0066] Based on the aforementioned technical solution, the third communication device is used as a receiver of the third information. The third communication device can determine a third AI model group based on the third information from the second communication device, and through the fourth information, the second communication device can subsequently distribute a fifth AI model to the first communication device and a sixth AI model to the second communication device. Therefore, when a communication device within a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the third information from the second communication device may be used as one of the decision bases for the third communication device to determine the AI ​​model, thereby maximally adapting the AI ​​model determined by the third communication device to the second communication device and improving the success rate of performing model processing on the AI ​​model by the second communication device.

[0067] It can be understood that the third AI model group includes the fifth AI model and the sixth AI model, and that the function of the third AI model group is implemented through at least the model processing of the fifth AI model and the model processing of the sixth AI model. That is, after receiving the fourth information, the second communication device can determine the model parameters of the fifth AI model and the model parameters of the sixth AI model. Additionally, the second communication device can distribute the fifth AI model to the first communication device and distribute the sixth AI model to the second communication device so that the model processing of the fifth AI model and the sixth AI model is implemented. Optionally, the model processing may include one or more of model update processing, model training processing, and model inference processing.

[0068] Optionally, if the third AI model group is considered as one AI model, the fifth AI model and the sixth AI model can be understood as two AI sub-models in the AI ​​model.

[0069] It should be noted that the second communication device and the third communication device may be implemented in multiple ways. The second communication device may be a terminal device or an access network device, and the third communication device may be a cloud server or a core network device. For example, if the third communication device is a cloud server, the second communication device may communicate with the cloud server through the core network device. As another example, if the third communication device is a core network device, the second communication device may be a terminal device, and the terminal device may communicate with the core network device through the access network device. As yet another example, if the third communication device is a core network device, the second communication device may be an access network device, and the access network device may communicate with the core network device through a communication interface between the access network device and the core network device.

[0070] A fourth aspect of the present application provides a communication method. The method is performed by a third communication device. The third communication device may be a network device (e.g., an access network device, a core network device, or a cloud server), or the third communication device may be a component within the network device (e.g., a processor, a chip, or a chip system), or a logical module or software capable of implementing all or some functions of the network device. In the method, the third communication device receives third information. The third communication device determines a third AI model group based on the third information, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to the first communication device, and the sixth AI model is distributed to the second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The third communication device transmits fourth information, and the fourth information includes model parameters of the third AI model group.

[0071] Based on the aforementioned technical solution, after receiving third information used to determine a third AI model group, the third communication device may transmit fourth information, and the third information includes model parameters of the third AI model group. That is, the third communication device is used as a receiver of the third information. The third communication device may determine a third AI model group based on the third information from the second communication device, and through the fourth information, the second communication device may subsequently distribute a fifth AI model to the first communication device and a sixth AI model to the second communication device. Therefore, when a communication device within a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the third information from the second communication device may also be used as one of the decision bases for the third communication device to determine an AI model, so that the AI ​​model determined by the third communication device can be adapted to the second communication device as much as possible, thereby improving the success rate of performing model processing on the AI ​​model by the second communication device thereafter.

[0072] In a possible implementation of the third or fourth embodiment, the third communication device is a functional entity that determines a third AI model group based on the third information.

[0073] Based on the aforementioned technical solution, the third communication device can communicate with one or more second communication devices, and the third communication device can receive information (e.g., one or more third information) from one or more second communication devices to create / acquire / determine a third AI model group. In other words, the second communication device can collect information and create a model based on the collected information. Subsequently, the second communication device can distribute the AI ​​model to one or more second communication devices (and corresponding first communication devices).

[0074] Optionally, the AI ​​model of the third AI model group is a general model. In this way, the third communication device can determine the general model through information from one or more second communication devices (e.g., one or more third information). Subsequently, the general model having high generalization and excellent universality can be distributed to each of the one or more second communication devices and to the first communication device connected to each second communication device.

[0075] In a possible implementation of the third or fourth embodiment, the third information includes third dimensional information or fourth dimensional information. Where the input of the fifth AI model includes the output of the sixth AI model, the third dimensional information is used to determine the dimensional information of the input data of the fifth AI model or the dimensional information of the output data of the sixth AI model. Where the input of the sixth AI model includes the output of the fifth AI model, the fourth dimensional information is used to determine the dimensional information of the output data of the fifth AI model or the dimensional information of the input data of the sixth AI model.

[0076] Based on the aforementioned technical solution, the third communication device can determine the dimension information of the data transmitted in the communication link between the first communication device and the second communication device through the third information. Since the dimension of the data transmitted in the communication link is related to the processing performance of the AI ​​model when the communication bandwidth between the first communication device and the second communication device is fixed, the implementation process of the third communication device determining the first AI model group can be simplified in this way, thereby reducing the complexity of the third communication device and improving the processing performance of the AI ​​model included in the first AI model group.

[0077] In a possible implementation of the third or fourth embodiment, the dimension information comprises at least one of the following: an upper limit of a dimension, a lower limit of a dimension, a dimension expected by the second communication device (or a dimension not expected by the second communication device), and a value range of the dimension expected by the second communication device (or a value range of the dimension not expected by the second communication device).

[0078] Optionally, the dimension information may include the value of any of the aforementioned items, the quantized value, the index of the value, and the index of the quantized value. Other implementations are also possible.

[0079] Based on the aforementioned technical solution, the dimensional information determined based on the third-dimensional information or the fourth-dimensional information may include at least one of the aforementioned items, thereby improving the flexibility of implementing the solution.

[0080] For example, if the dimension information includes an upper limit and / or a lower limit of the dimension, the third communication device may use the range indicated by the upper limit and / or lower limit as one of the grounds for determining the AI ​​model, thereby enhancing the flexibility to implement the solution.

[0081] As another example, if the dimension information includes the dimension expected by the second communication device and / or the range of values ​​of the dimension expected by the second communication device, the AI ​​model determined by the third communication device based on the dimension information can satisfy the expectations of the second communication device.

[0082] In a possible implementation of the third or fourth embodiment, the third-dimensional information or the fourth-dimensional information is determined based on the channel state information.

[0083] Based on the aforementioned technical solution, the third-dimensional information or the fourth-dimensional information included in the third information can be determined based on channel state information, so that the third-dimensional information or the fourth-dimensional information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to a certain extent. In this way, an AI model that can be subsequently acquired based on the third information can be adapted to the channel characteristics of the wireless channel, thereby improving the transmission performance of the AI ​​data corresponding to the AI ​​model. Furthermore, when the AI ​​model acquired based on the third information can be adapted to the channel characteristics of the wireless channel, the data transmitted over the wireless link can also fully satisfy the channel bandwidth requirements, thereby improving the model performance of the AI ​​model included in the third AI model group.

[0084] In a possible implementation of the third or fourth embodiment, the third information comprises at least one of the following: model parameters of an AI model in one or more groups of AI models—each group of one or more groups of AI models comprises a dedicated model distributed to a first communication device and a dedicated model distributed to a second communication device—; data from one or more first communication devices connected to a second communication device; and input data of an AI model distributed to a second communication device and label data of the input data of an AI model distributed to a second communication device.

[0085] Based on the aforementioned technical solution, the third information can be implemented through at least one of the aforementioned items, thereby improving the flexibility of implementing the solution.

[0086] In one embodiment, if the third information includes model parameters of an AI model of one or more AI model groups, the third communication device may obtain one or more dedicated models distributed to the first communication device and the second communication device based on the third information. In this way, the third communication device may obtain model characteristics of one or more dedicated models and, by reflecting the characteristics obtained in the third AI model group, improve the generalization (or universality) of the general models included in the third AI model group.

[0087] In another embodiment, where the third information includes data from one or more first communication devices (e.g., terminal devices) connected to the second communication device, different terminal devices may have different data characteristics (e.g., different data may be collected at different geographical locations, different data may be collected at different times, and different data may correspond to different wireless channels of the user). Accordingly, in this manner, the third communication device obtains a third AI model group based on these data characteristics and improves the generalization (or universality) of the general models included in the third AI model group.

[0088] In another implementation example, where the third information includes input data of an AI model distributed to the second communication device and label data of the input data of the AI ​​model distributed to the second communication device, since the input data can be used as input to the sixth AI model and the label data can be used as one of the grounds for determining the model processing performance of the sixth AI model, in the case of the third communication device, the third communication device can obtain an AI model with superior performance based on the two pieces of information.

[0089] In addition, in the case of the third communication device, the third communication device implements a mutual information-based mathematical calculation based on two pieces of information received and transmitted by the second communication device over a wireless link and AI data (e.g., input data of the sixth AI model or output data of the sixth AI model), and determines a third AI model group based on the result of the mathematical calculation, thereby improving the model performance of the AI ​​models included in the third AI model group under the premise that the data over the wireless link satisfies the bandwidth.

[0090] If the third information includes two or more of the information described above, it can be understood that additional overlapping benefits may be obtained through the two or more pieces of information based on the technical benefits achieved from any of the information described above.

[0091] Optionally, each piece of information included in the third piece of information may be part of the information obtained by the second communication device through filtering from a plurality of pieces of information.

[0092] In a possible implementation of the third or fourth embodiment, the third information being used to determine the third AI model group comprises: the third information being used to obtain the third AI model group by updating the fourth AI model group; the fourth AI model group comprising the seventh AI model and the eighth AI model, wherein the seventh AI model is deployed to the first communication device and the eighth AI model is deployed to the second communication device, and the input of the seventh AI model comprises the output of the eighth AI model, or the input of the eighth AI model comprises the output of the seventh AI model.

[0093] Optionally, when the 4th AI model group is considered as one AI model, the 7th AI model and the 8th AI model can be understood as two AI sub-models within the AI ​​model.

[0094] Based on the aforementioned technical solution, in the case of the third communication device, after receiving the third information, the third communication device can obtain the third AI model group by updating the fourth AI model group based on the third information. That is, since the third information transmitted by the second communication device can be used to update other AI models, the solution can be applied to an AI model update scenario.

[0095] In a possible implementation of the third or fourth embodiment, the third information is information transmitted periodically and / or the fourth information is information transmitted periodically.

[0096] Based on the aforementioned technical solution, the third information may be one of the grounds for determining the AI ​​model, and the fourth information may be used to deploy the AI ​​model. The AI ​​model may be periodically determined and / or periodically deployed by periodically transmitting the third information and / or the fourth information between the second communication device and the third communication device, so that the AI ​​model is updated repeatedly multiple times through a periodic process.

[0097] A fifth aspect of the present application provides a communication device. The device is a first communication device, and the device comprises a transceiver unit and a processing unit. The processing unit is configured to determine first information. The transceiver unit is configured to transmit the first information, and the first information is used to determine a first artificial intelligence (AI) model group, and the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to 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 further configured to receive second information from the second communication device, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0098] In a fifth aspect of the present application, a constituent unit of the communication device may be further configured to perform the steps performed in a possible embodiment of the first aspect and achieve a corresponding technical effect. For details, refer to the first aspect. Details are not described herein.

[0099] A sixth aspect of the present application provides a communication device. The device is a second communication device, and the device includes a transceiver unit and a processing unit. The transceiver unit receives first information. The processing unit is configured to determine a first AI model group based on the first information, and the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to the second communication device, and 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 further configured to transmit second information, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0100] In the sixth aspect of the present application, a constituent unit of the communication device may be further configured to perform the steps performed in a possible implementation of the second aspect and achieve a corresponding technical effect. For details, refer to the second aspect. Further details are not described herein.

[0101] A seventh aspect of the present application provides a communication device. The device is a second communication device, and the device includes a transceiver unit and a processing unit. The processing unit is configured to determine third information. The transceiver unit is configured to transmit the third information, and the third information is used to determine a third AI model group, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to the first communication device, and the sixth AI model is distributed to the second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The transceiver unit is further configured to receive fourth information from the third communication device, and the fourth information includes model parameters of the third AI model group.

[0102] In the seventh aspect of the present application, a constituent unit of the communication device may be further configured to perform the steps performed in a possible implementation of the third aspect and achieve a corresponding technical effect. For details, refer to the third aspect. Details are not described again herein.

[0103] An eighth aspect of the present application provides a communication device. The device is a third communication device, and the device comprises a transceiver unit and a processing unit. The transceiver unit is configured to receive third information. The processing unit is configured to determine a third AI model group based on the third information, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to a first communication device, and the sixth AI model is distributed to a second communication device; the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The transceiver unit is further configured to transmit fourth information, and the fourth information includes model parameters of the third AI model group.

[0104] In the eighth aspect of the present application, a constituent unit of the communication device may be further configured to perform the steps performed in a possible implementation of the fourth aspect and achieve a corresponding technical effect. For details, refer to the fourth aspect. Details are not described again herein.

[0105] A ninth aspect of the present application provides a communication device comprising at least one processor. At least one processor is coupled to a memory. The memory is configured to store a program or instruction. At least one processor is configured to execute the program or instruction so that the device implements a method according to any possible implementation of any one of the first through fourth aspects.

[0106] In possible implementations, the communication device includes additional memory. Optionally, the processor and memory are integrated together.

[0107] A tenth aspect of the present application provides a communication device comprising at least one logic circuit and an input / output interface. The logic circuit is configured to perform a method according to any possible implementation of any one of the first to fourth aspects.

[0108] A first aspect of the present application provides a communication system. The communication system includes the aforementioned first communication device and second communication device. Alternatively, the communication system includes the aforementioned second communication device and third communication device. Alternatively, the communication system includes the aforementioned first communication device, second communication device, and third communication device.

[0109] A twelfth aspect of the present application provides a computer-readable storage medium. The storage medium is configured to store one or more computer-executable instructions. When a computer-executable instruction is executed by a processor, the processor performs a method according to any possible implementation of any one of the first through fourth aspects.

[0110] A 13th aspect of the present application provides a computer program product (or also referred to as a computer program). When the computer program of the computer program product is executed by a processor, the processor performs a method according to any possible implementation of any one of the first through fourth aspects.

[0111] A fourth aspect of the present application provides a chip system. The chip system includes at least one processor configured to support a communication device in implementing a method according to any one of the possible implementations of the first through fourth aspects.

[0112] In a possible design, the chip system may additionally include memory. The memory is configured to store program instructions and data required by a communication device. The chip system may include a chip, or may include a chip and other separate devices. Optionally, the chip system may additionally include an interface circuit, and the interface circuit provides program instructions and / or data to at least one processor.

[0113] For technical effects achieved by any design method of the third through tenth embodiments, refer to the technical effects achieved by different design methods of the first through fourth embodiments. Details are not described further in this specification. Brief explanation of the drawing

[0114] FIGS. 1a and FIGS. 1b are drawings of a communication system according to the present application. FIGS. 2a to 2g are drawings of an AI processing process according to the present application. FIG. 3 is an interaction diagram of a communication method according to the present application. FIGS. 4 to 6 are interaction diagrams of a communication method according to the present application. FIGS. 7 to 11 are drawings of a communication device according to the present application. Specific details for implementing the invention

[0115] First, to aid the understanding of those skilled in the art, some terms are explained and described in the embodiments of this application.

[0116] (1) The terminal device may be a wireless terminal device capable of receiving scheduling and display information of a network device. The wireless terminal device may be a device that provides voice and / or data access to a user, a handheld device with wireless connection capabilities, or another processing device connected to a wireless modem.

[0117] A terminal device may communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device may be a mobile terminal device, such as a mobile phone (also called a “cellular” phone or mobile phone), a computer, and a data card. For example, the terminal device may be a portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile device that exchanges voice and / or data with a radio access network. For example, the terminal device may be a device such as a personal communications service (PCS) phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a pad, or a computer with wireless transceiver capabilities. Wireless terminal devices may also be referred to as systems, subscriber units, subscriber stations, mobile stations (MS), remote stations, access points (AP), remote terminals, access terminals, user terminals, user agents, subscriber stations (SS), customer premises equipment (CPE), terminals, user equipment (UE), mobile terminals (MT), etc.

[0118] As an example, not a limitation, in the embodiments of this application, the terminal device may alternatively be a wearable device. A wearable device may also be referred to as a wearable intelligent device, an intelligent wearable device, etc., and is a general term for wearable devices intelligently designed and developed to be worn daily using wearable technology, e.g., glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that can be worn directly on the body or integrated into a user's clothing or accessories. A wearable device is not only a hardware device but also enables powerful functions through software support, data exchange, and cloud interaction. In a broad sense, a wearable intelligent device includes large, fully functional devices capable of implementing all or part of functions without relying on a smartphone, e.g., smartwatches or smart glasses, and devices dedicated to only one type of application function that must operate in conjunction with other devices such as smartphones, e.g., various smart bands, smart helmets, or smart jewelry for monitoring physical signs.

[0119] Alternatively, the terminal may be an unmanned aerial vehicle, 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.

[0120] Furthermore, terminal devices may alternatively be terminal devices of communication systems that have evolved after the 5th generation (5G) communication system (e.g., 6th generation (6G) communication system), terminal devices of future evolved public land mobile networks (PLMN), etc. For example, 6G networks can further expand the forms and functions of 5G communication terminals, and 6G terminals include, but are not limited to, vehicles, cellular network terminals (integrating the functions of satellite terminals), unmanned aerial vehicles, and Internet of Things (IoT) devices.

[0121] In an embodiment of the present application, the terminal device may further acquire AI services provided by a network device. Optionally, the terminal device may further have AI processing capabilities.

[0122] (2) A network device may be a device within a wireless network. For example, a network device may be a RAN node (or device) that connects a terminal device to a wireless network, and may also be a base station. Currently, some examples of RAN devices include a base station, an evolved NodeB (eNodeB), a base station gNB (gNodeB) of a 5G communication system, a transmission reception point (TRP), an evolved NodeB (eNB), a radio network controller (RNC), a NodeB (NodeB, NB), a home base station (e.g., a home evolved NodeB or a home NodeB, HNB), a baseband unit (BBU), and a wireless fidelity (Wi-Fi) access point (AP). Furthermore, in a network structure, a network device may include a central unit (CU) node, a distributed unit (DU) node, or a RAN device including a CU node and a DU node.

[0123] Optionally, the RAN node may alternatively be a macro base station, micro base station, indoor base station, relay node, or donor node, or a radio controller in a cloud radio access network (CRAN) scenario. Alternatively, the RAN node may be a server, wearable device, vehicle, vehicle-mounted device, etc. For example, in V2X technology, the access network device may be a roadside unit (RSU).

[0124] In another possible scenario, multiple RAN nodes cooperate to help terminals implement radio access, and different RAN nodes individually implement parts of the base station's functions. For example, RAN nodes may be a central unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP), a radio unit (RU), etc. The CU and DU may be deployed individually or may be included in the same network element, for example, a baseband unit (BBU). The RU may be included in a radio frequency device or radio frequency unit, for example, a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0125] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meanings. For example, in an open access network (open RAN, O-RAN, or ORAN) system, CU may be referred to as O-CU (open CU), DU may be referred to as O-DU, CU-CP may be referred to as O-CU-CP, CU-UP may also be referred to as O-CU-UP, and RU may also be referred to as O-RU. For convenience of explanation, CU, CU-CP, CU-UP, DU, and RU are used as descriptive examples in this application. Any one of CU (or CU-CP or CU-UP), DU, and RU in this application may be implemented using a software module, a hardware module, or a combination of a software module and a hardware module.

[0126] Communication between an access network device and a terminal device follows a specific 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 a radio resource control (RRC) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, a media access control (MAC) layer, or a physical (PHY) layer. The user plane protocol layer may include at least one of a service data adaptation protocol (SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, a physical layer, etc.

[0127] Refer to Table 1 for the correspondence between the network elements of an ORAN system and the protocol layer functions that can be implemented by the network elements.

[0128] ORAN network elements 3GPP Protocol Layer Functions O-CU-CP RRC+PDCP-Control Plane (PDCP-C) O-CU-UP SDAP+PDCP-User Plane (PDCP-U) O-DU RLC+MAC+PHY-high O-RU PHY-low

[0129] The network device may be another device that provides wireless communication functions to the terminal device. The specific technology and specific device type used by the network device are not limited to the embodiments of this application. For ease of explanation, this is not limited to the embodiments of this application.

[0130] A network device may further include a core network device. For example, a core network device includes network elements such as a mobility management entity (MME), home subscriber server (HSS), serving gateway (S-GW), policy and charging rules function (PCRF), and public data network gateway (PDN gateway, P-GW) of a 4th generation (4G) network, and an access and mobility management function (AMF), user plane function (UPF), or session management function (SMF) of a 5G network. Furthermore, a core network device may further include other core network devices of a 5G network and a next-generation network of a 5G network.

[0131] In an embodiment of the present application, the network device may alternatively be a network node having AI capabilities and may provide AI services to a terminal or other network device, and may be, for example, an AI node, a computing capability node, a RAN node having AI capabilities, or a core network element having AI capabilities on the network side (access network or core network).

[0132] In the embodiments of the present application, a device configured to implement the function of a network device may be a network device or a device capable of supporting the implementation of the function of a network device (e.g., a chip system). The device may be mounted on a network device. In the technical solution provided in the embodiments of the present application, a network device is used as an example of a device configured to implement the function of a network device.

[0133] (3) Configuration and Pre-configuration: Both configuration and pre-configuration are used in this application. Configuration means that a network device / server transmits configuration information of some parameters or the value of a parameter to a terminal via a message or signaling, and the terminal determines the communication parameters or resources used for transmission based on the value or information. Pre-configuration is similar to configuration and may be parameter information or parameter values ​​pre-negotiated by the network device / server and the terminal device, parameter information or parameter values ​​used by the base station / network device or the terminal device and specified in a standard protocol, or parameter information or parameter values ​​pre-stored in the base station / server or the terminal device. This is not limited in this application.

[0134] In addition, these values ​​and parameters may be changed or updated.

[0135] (4) In embodiments of the present application, the terms “system” and “network” may be used interchangeably. “Plural” refers to two or more. “And / or” describes an association between associated objects and indicates that three relationships may exist. For example, A and / or B may represent the following three cases: A alone exists, both A and B exist, and B alone exists, where A and B may be singular or plural. The character ' / ' generally indicates an “or” relationship between associated objects. At least one of the following items (parts) or a similar expression thereof represents any combination of these items, including any combination of singular items (parts) or plural items (parts). For example, “at least one of A, B and C” includes A, B, C, AB, AC, BC, or ABC. Additionally, unless otherwise specified, in the embodiments of this application, ordinal numbers such as "first" and "second" are used to distinguish multiple objects and are not used to limit the order, chronological order, priority, or importance of the multiple objects.

[0136] (5) In the embodiments of the present application, “transmission” and “reception” indicate the direction of signal transmission. For example, “transmitting information to XX” may be understood as the destination of the information being XX, which may include direct transmission through an air interface or indirect transmission through an air interface by another unit or module. “Receiving information from YY” may be understood as the source of the information being YY, which may include direct reception from YY through an air interface or indirect reception from YY by another unit or module through an air interface. “Transmission” may also be understood as the “output” of the chip interface, and “reception” may also be understood as the “input” of the chip interface.

[0137] In other words, transmission and reception may be performed between devices, for example, between a network device and a terminal device, or within a device, for example, through a bus, cable, or interface, between components, modules, chips, software modules, or hardware modules within the device.

[0138] While necessary processing such as encoding and modulation may be performed on information between the source end and the destination end where information is transmitted, it can be understood that the destination end is capable of recognizing valid information from the source end. A similar description of the present application is likewise understood, and details are not described again.

[0139] (6) In embodiments of the present application, “representation” may include direct representation and indirect representation, or explicit representation and implicit representation. Information represented by one piece of information (e.g., the representation information below) is referred to as the representation information. In a particular implementation process, the representation information may be represented in a plurality of ways, such as, for example, by directly representing the representation information, or by representing the representation information or an index of the representation information, but is not limited thereto. Alternatively, the representation information may be represented indirectly by representing other information. There is an association between the other information and the representation information. Alternatively, only a portion of the representation information may be represented, and the remainder of the representation information may be known or pre-agreed upon. For example, the representation overhead may be reduced to some extent by representing the specific information using an alternatively pre-agreed sequence of information (e.g., pre-defined in a protocol). The specific representation method in the present application is not limited. It can be understood that in the case of a transmitter of display information, the display information can display target information, and in the case of a receiver of display information, the display information can be used to determine target information.

[0140] In this application, unless otherwise specified, cross-references may be made between embodiments regarding identical or similar parts of the embodiments. In the embodiments of this application and methods / designs / implements in the embodiments, terms and / or descriptions between different embodiments and between methods / designs / implements in the embodiments may be consistent and cross-referenced unless otherwise specified or a logical conflict arises, and technical features in different embodiments and methods / designs / implements in the embodiments may be combined based on their internal logical relationships to form new embodiments, methods, or implementations. The following implementations of this application do not limit the scope of protection of this application.

[0141] The present application may be applied to a long term evolution (LTE) system, a new radio (NR) system, or a communication system evolved after 5G (e.g., 6G). The communication system includes at least one network device and / or at least one terminal device.

[0142] FIG. 1a is a diagram of the architecture of a communication system (1000) to which an embodiment of the present application is applied. As illustrated in FIG. 1a, the communication system includes a RAN (100) and a core network (200). Optionally, the communication system (1000) may further include an internet (300). The RAN (100) may include at least one RAN node (e.g., 110a and 110b of FIG. 1a, collectively referred to as 110) and may further include at least one terminal (e.g., 120a to 120j of FIG. 1a, collectively referred to as 120). The RAN (100) may further include other RAN nodes, such as, for example, a wireless relay device and / or a wireless backhaul device (not illustrated in FIG. 1a). A terminal (120) is connected to a RAN node (110) wirelessly, and a RAN node (110) is connected to a core network (200) via a wired or wireless method. A core network device within the core network (200) and a RAN node (110) within the RAN (100) may be different, independent physical devices, or they may be the same physical device that integrates the logical functions of the core network device and the logical functions of the RAN node. Terminals may be connected to each other via a wired or wireless method, and RAN nodes may be connected to each other via a wired or wireless method.

[0143] The RAN (100) may be an evolved universal terrestrial radio access (E-UTRA) system, an NR system, and a future radio access system defined by the 3rd generation partnership project (3GPP). The RAN (100) may further include two or more of the aforementioned different radio access systems. Alternatively, the RAN (100) may be an open RAN (O-RAN).

[0144] For ease of explanation, the following description uses an example where a base station is used as a RAN node.

[0145] Base stations and terminals may be in a fixed location or may be movable. Base stations and terminals may be deployed on the ground, on water, or on airplanes, balloons, and satellites, including being deployed indoors or outdoors, or being deployed as handhelds or mounted on vehicles. Application scenarios for base stations and terminals are not limited to the embodiments of this application.

[0146] The roles of base stations and terminals can be relative. For example, the helicopter or unmanned aerial vehicle (120i) of FIG. 1a can be configured as a mobile base station. In the case of a terminal (120j) accessing the wireless access network (100) through 120i, the terminal (120i) is a base station. However, in the case of a base station (110a), 120i is a terminal. That is, communication between 110a and 120i is performed according to a wireless air interface protocol. Of course, communication between 110a and 120i may alternatively be performed according to an interface protocol between base stations. In this case, with respect to 110a, 120i is also a base station. Thus, both base stations and terminals can be collectively referred to as communication devices. 110a and 110b in FIG. 1a may be referred to as communication devices having the function of a base station, and 120a to 120j in FIG. 1a may be referred to as communication devices having the function of a terminal.

[0147] Communication between a base station and a terminal, between base stations, or between terminals may be performed in licensed spectrum, in unlicensed spectrum, or in both licensed and unlicensed spectrum. Communication may be performed in spectrum below 6 gigahertz (GHz), in spectrum above 6 GHz, or in both spectrum below 6 GHz and spectrum above 6 GHz. The spectrum resources used for wireless communication are not limited to the embodiments of this application.

[0148] In an embodiment of the present application, the function of a base station may be performed by a module (e.g., a chip) within the base station, or by a control subsystem including the function of a base station. Here, the control subsystem including the function of a base station may be a control center in the aforementioned application scenarios, such as a smart grid, industrial control, smart transportation, and a smart city. The function of a terminal may be performed by a module (e.g., a chip or a modem) within the terminal, or by a device including the function of a terminal.

[0149] FIG. 1b is another diagram of a communication system according to one embodiment of the present application. FIG. 1b describes an example in which the network device is a base station, and both Device 1 and Device 2 are terminal devices. As illustrated in FIG. 1b, the communication link between Device 1 and Device 2 may be referred to as a sidelink (SL), and the communication link between Device 1 (or Device 2) and the base station may be referred to as an uplink-downlink including an uplink and a downlink. It can be seen that a sidelink is a communication mechanism in which different terminal devices communicate directly with each other without using a network device.

[0150] Optionally, in a sidelink (SL), typically, the transmitting device and the receiving device may be terminal devices or network devices of the same type, or a roadside unit (RSU) and a terminal device. In terms of physical entities, the RSU is a roadside station or a roadside unit. In terms of function, the RSU may be a terminal device or a network device. This is not limited in the present application. That is, the transmitting device may be a terminal device and the receiving device may also be a terminal device; the transmitting device may be a roadside station and the receiving device may also be a terminal device; or the transmitting device may be a terminal device and the receiving device may also be a roadside station. Additionally, the sidelink may alternatively be a link between base station devices of the same type or different types. In this case, the function of the sidelink is similar to that of the trunk link, but the air interface technology used by the sidelink may be the same or different from that used by the trunk link.

[0151] For example, broadcast, unicast, and multicast are supported in sidelink.

[0152] When a terminal device (e.g., device 1) communicates directly with another terminal device (e.g., device 2) without a network device, the two terminal devices can communicate with each other through a proximity-based services communication 5 (PC5) interface.

[0153] A typical application of Sidelink is V2X communication. V2X communication utilizes and enhances current cellular network capabilities and elements to enable low-latency and high-reliability communication between various nodes within a vehicle network, including Vehicle-to-Vehicle (V2V), Vehicle-to-Pedestrian (V2P), Vehicle-to-Infrastructure (V2I), and Vehicle-to-Network (V2N) communication.

[0154] The technical solution provided in this application may be applied to a wireless communication system (e.g., the system illustrated in FIG. 1a or FIG. 1b). For example, an AI network element may be introduced into the communication system provided in this application to implement some or all of the AI-related operations. The AI ​​network element may also be referred to as an AI node, AI device, AI entity, AI module, AI model, AI unit, etc. The AI ​​network element may be embedded in a network element of the communication system. For example, the AI ​​network element may be an AI module that implements AI-related functions embedded in a terminal device, an access network device, a core network device, a cloud server, or an operations, administration and maintenance (OAM). The OAM may be used as a network management system of the core network device and / or a network management system of the access network device. Alternatively, the AI ​​network element may be a network element deployed independently in the communication system. Optionally, the terminal or a chip embedded in the terminal may include an AI entity configured to implement AI-related functions.

[0155] The following briefly describes the artificial intelligence (AI) that may be used in this application.

[0156] Artificial intelligence (AI) enables machines to possess human intelligence, allowing them to, for example, simulate some intelligent human behaviors using computer software and hardware. To implement artificial intelligence, machine learning methods can be used. In machine learning methods, a machine acquires a model through learning (or training) using training data. A model represents a mapping from input to output. The model acquired through learning can be used for inference (or prediction). Specifically, the model can be used to predict an output corresponding to a given input. The output can also be referred to as the result of inference (or prediction).

[0157] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Unsupervised learning can also be referred to as unsupervised learning.

[0158] From the perspective of supervised learning, based on collected sample values ​​and sample labels, the mapping relationship between sample values ​​and sample labels is learned using machine learning algorithms, and the learned mapping relationship is represented using an AI model. The process of training a machine learning model is the process of learning mapping relationships. During training, sample values ​​are input into the model to obtain predicted values, and model parameters are optimized by calculating the error between the model's predicted values ​​and the sample labels (ideal values). After the learning of mapping relationships is complete, new sample labels can be predicted using the learned mappings. Mapping relationships learned through supervised learning can include linear or non-linear mappings. The learning task can be classified into classification and regression tasks based on the type of label.

[0159] From the perspective of unsupervised learning, internal patterns of samples are autonomously explored using algorithms based on collected sample values. For specific types of unsupervised learning algorithms, samples are used as supervised signals. In other words, the model learns the mapping relationships between samples, which is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples. Self-supervised learning can be used for signal compression and decompression. Common algorithms include autoencoders and generative adversarial networks.

[0160] Reinforcement learning differs from supervised learning and is an algorithm that learns policies to solve problems by interacting with an environment. Unlike supervised and unsupervised learning, reinforcement learning does not have data with clear "correct" action labels. The algorithm must interact with the environment to acquire reward signals fed back by the environment and adjust decision actions to obtain larger reward signal values. For example, in downlink power control, a reinforcement learning model expects to achieve higher system throughput by adjusting the downlink transmission power for each user based on the total system throughput fed back by the wireless network. The goal of reinforcement learning is also to learn the mapping relationship between the environmental state and better (e.g., optimal) decision actions. However, the labels for "correct actions" cannot be acquired in advance. Therefore, the network cannot be optimized by calculating the error between an action and a "correct action." Reinforcement learning training is implemented through iterative interaction with the environment.

[0161] Neural networks (NNs) are specific models of machine learning technology. According to the Universal Approximation Theorem, neural networks can theoretically approximate any continuous function, and thus possess the ability to learn arbitrary mappings. In conventional communication systems, designing communication modules requires extensive expertise. However, in neural network-based deep learning communication systems, implicit pattern structures can be automatically discovered from large datasets and mapping relationships between data can be established, thereby achieving better performance than conventional modeling methods.

[0162] The idea behind neural networks originates from the neuron structure of brain tissue. For example, each neuron performs a weighted summation operation on its input values ​​and outputs the result using an activation function.

[0163] Figure 2a is a diagram of the neuron structure. The input to the neuron is and the corresponding weight Assuming that, where n is a positive integer, and and It can be various possible types such as decimal numbers, integers (e.g., 0, positive integers, or negative integers), or complex numbers. Is It is a weight for. The bias, for example, b, is used for weighted summation of the input values. Activation functions can take various forms. For example, the activation function of a neuron is If so, the output is is. In contrast, the activation function is If so, the output is b can be of various possible types, such as decimal numbers, integers (e.g., 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.

[0164] A neural network generally comprises multiple layers, and each layer may contain one or more neurons. Increasing the depth and / or width of a neural network increases its representational capabilities, providing powerful information extraction and abstraction for complex systems. Depth refers to the number of layers included in the neural network, and width refers to the number of neurons included in each layer. In one implementation, the neural network includes an input layer and an output layer. The input layer performs neuronal processing on received input information and transmits the processing result to the output layer, and the output layer outputs the final result. In another implementation, the neural network includes an input layer, a hidden layer, and an output layer. The input layer performs neuronal processing on received input information and transmits the processing result to the hidden layer. The hidden layer performs a computation on the received processing result and then transmits the computation result to the output layer or the next adjacent hidden layer. The output layer outputs the final result. A neural network may include a single hidden layer or multiple sequentially connected hidden layers. This is not limited to this.

[0165] A neural network is, for example, a deep neural network (DNN). Based on the network configuration method, a DNN may include a feedforward neural network (FNN), a convolutional neural network (CNN), and a recurrent neural network (RNN).

[0166] Figure 2b is a diagram of an FNN network. In an FNN network, neurons in adjacent layers are fully connected, requiring significant storage space and causing high computational complexity.

[0167] CNNs are neural networks dedicated to processing grid-based data, such as time-series data (discrete-time sampling) and image data (two-dimensional discrete sampling). Instead of performing operations on all input information simultaneously, CNNs significantly reduce the computational load of model parameters by capturing partial information through fixed-size windows and performing convolution operations. Furthermore, based on the different types of information captured through the windows (for example, people and objects within the same image are different types of information), different convolution kernel operations can be applied to the windows, enabling better feature extraction.

[0168] An RNN is a DNN network that uses feedback time-series information. The inputs of an RNN include the current input and the previous output. RNNs are suitable for capturing temporally correlated sequential features and are ideal for applications such as speech recognition and channel encoding and decoding.

[0169] In the aforementioned model training process of machine learning, a loss function may be defined to explain the gap or difference between the model output and the target value. The loss function may be expressed in multiple forms, and the specific form of the loss function is not limited. The model training process involves adjusting some or all of the model parameters to minimize the loss function below a threshold or to meet target requirements.

[0170] A model may also be referred to as an AI model, a rule, or by other names. An AI model can be regarded as a specific method for implementing AI functions. An AI model represents a mapping relationship or function between the inputs and outputs of a model. AI functions may include one or more of the following: data collection, model training (or model learning), model information distribution, model inference (also referred to as model inference, inference, prediction, etc.), model monitoring or model validation, distribution of inference results, etc. AI functions may also be referred to as AI (AI-related) behaviors or AI-related functions.

[0171] Below, an example of the neural network implementation process is explained with reference to the attached drawings.

[0172] 1. Fully connected neural network, also known as a multilayer perceptron (MLP).

[0173] As illustrated in FIG. 2c, an MLP comprises one input layer (left), one output layer (right), and multiple hidden layers (middle). Each layer of the MLP contains multiple nodes, which are called neurons. Neurons in two adjacent layers are connected to each other in pairs.

[0174] Optionally, when considering neurons in two adjacent layers, the output h of a neuron in the lower layer is the weighted sum of all neurons x in the upper layer connected to the neuron in the lower layer, and can be expressed using an activation function as follows:

[0175]

[0176] Here, w is a weight matrix, and b is the bias vector, and f is the activation function.

[0177] Additionally, optionally, the output of the neural network can be expressed recursively as follows:

[0178]

[0179] n is the index of a neural network layer, 1≤n≤N, and N is the total number of layers in the neural network.

[0180] In other words, a neural network can be understood as a mapping relationship from an input dataset to an output dataset. Neural networks are typically initialized randomly, and random based on existing data w and b The process of obtaining mapping relationships from is called neural network training.

[0181] Optionally, a specific training method evaluates the output of the neural network using a loss function.

[0182] As illustrated in FIG. 2d, errors can be backpropagated, and ( w and b Neural network parameters (including) can be iteratively optimized using gradient descent until the loss function reaches a minimum value, i.e., the "better point (e.g., optimal point)" in FIG. 2d. It can be understood that the neural network parameters corresponding to the "better point (e.g., optimal point)" in FIG. 2d can be used as neural network parameters in the trained AI model information.

[0183] Additionally, optionally, the gradient descent process can be expressed as follows:

[0184]

[0185] Is ( w and b The parameter to be optimized (including ), L is the loss function, and is the learning rate for controlling the gradient descent step, and represents an induction operation, and silver for Represents the derivative value of.

[0186] Additionally, optionally, the backpropagation process uses the chain rule to calculate partial derivatives.

[0187] As shown in FIG. 2e, the gradient of the parameter in the previous layer can be obtained through the recursive calculation of the gradient of the parameter in the next layer, and can be expressed as follows:

[0188]

[0189] is the weight of node i connected to node j, and is the weighted sum of the inputs of node i.

[0190] 2. Federated Learning (FL)

[0191] The concept of federated learning is proposed to effectively address the challenges currently faced by artificial intelligence development. While ensuring user data privacy and security, federated learning enables various edge devices and central servers to cooperate to efficiently complete the training tasks of models.

[0192] As illustrated in Figure 2f, the FL architecture is the current training architecture in the FL field. For example, the FedAvg algorithm is the fundamental algorithm of FL, and the algorithmic procedure of the FedAvg algorithm is approximately as follows:

[0193] (1) The central unit is a model to be trained Initialize, broadcast the model to all client devices, and send it.

[0194] (2) ( In the )th round, client is a local dataset Global model received based on About Local training results by performing epochs Acquires and reports local training results to the central node.

[0195] (3) The central node summarizes and collects local training results from all (or some) clients. The set of clients uploading local models in the nth round is It is assumed that the central unit obtains a new global model by performing a weighted average using the quantity of samples from corresponding clients as weights, and specific update rules are is. Next, the central unit installs the latest version of the global model on all client devices for a new round of training. Broadcasts and transmits.

[0196] (4) Steps (2) and (3) are repeated until the model finally converges or the number of training rounds reaches an upper limit.

[0197] Local model In addition to reporting, the local gradients obtained through training This can also be reported. The central node calculates the average of the local gradients and updates the global model based on the direction of the average gradient.

[0198] In the FL framework, it can be seen that datasets reside on distributed nodes. Specifically, distributed nodes collect local datasets, perform local training, and report the local results (models or gradients) obtained through training to the central node. The central node does not possess the datasets; it is solely responsible for fusing the training results from the distributed nodes to acquire a global model and transmitting the global model to the distributed nodes.

[0199] 3. Decentralized Learning

[0200] Unlike federated learning, another distributed learning architecture is decentralized learning.

[0201] As illustrated in Fig. 2g, a fully distributed system without a central node is considered. The design goal of a decentralized learning system is usually the target of all nodes The average value of and, here is the quantity of distributed nodes, and is a parameter to be optimized, and in machine learning is a parameter of a machine learning (e.g., neural network) model. Each node has local data and a local target Local gradient using After calculating, the local gradient is transmitted to adjacent nodes reachable in communication. After receiving the gradient information transmitted by an adjacent node of an arbitrary node, the node [calculates] the parameters of the local model according to the following formula You can update:

[0202]

[0203] is my i It represents the parameters of the local model obtained through the (k+1)th (k is a natural number) update at the node, and Eun Je i It represents the parameters of the local model obtained through the k-th update at the node (if k is 0, this now i Indicates that it is a parameter of the local model that exists before being updated in the node), represents the optimization coefficient, and is a node i It is a set of adjacent nodes of, and The node i The quantity of elements in the set of adjacent nodes, i.e., nodes i It represents the quantity of adjacent nodes. Through the exchange of information between nodes, the decentralized learning system will eventually learn an integrated model.

[0204] The technical solution provided in this application may be applied to a wireless communication system (e.g., a system illustrated in FIG. 1a or FIG. 1b). In a wireless communication system, a communication node typically has signal reception and transmission functions and computational functions. A network device having computational functions is used as an example. The computational function of the network device is primarily intended to provide computational capability support for signal reception and transmission functions (e.g., performing transmission processing and reception processing for a signal) to implement communication operations between the network device and other communication nodes.

[0205] In a communication network, in addition to computational functions to provide computational power support for the aforementioned communication tasks, communication nodes may have additional surplus computational power. How to utilize this surplus computational power is an important technical challenge.

[0206] In a possible implementation, communication nodes can be used as participating nodes in an AI learning system, and the computational power of the communication nodes is applied to the stages of the AI ​​learning system. With the advent of the era of large models, deep learning models with large parameters, such as bidirectional encoder representations from transformers (BERT) or generative pre-trained transformers (GPT), can complete increasingly complex tasks and achieve superior performance. However, for large models, even the model's inference process is limited by device capacity. Therefore, large models are typically stored on central cloud servers. Furthermore, each device on the network generates a large amount of raw data daily, and this data requires large models to be invoked multiple times for inference. Typically, devices (e.g., communication nodes) can transmit data to a central server, which performs inference based on the data and then returns the inference results to the devices. This process consumes a significant amount of communication resources for data transmission, and the privacy of device data is at risk.

[0207] One scholar proposes a distributed inference technique for deep neural networks to better reduce communication overhead and protect user data privacy. The model is distributed across devices, and the devices' local computational power is used to perform inference based on the model, thereby reducing communication overhead and preserving data privacy. However, in communication systems, there is currently no relevant literature providing a solution for how to determine the AI ​​model used by communication nodes (e.g., how to generate and update it).

[0208] FIG. 3 is a diagram illustrating an implementation of a communication method according to the present application. The method includes the following steps.

[0209] It should be noted that in FIG. 3, an example in which a first communication device and a second communication device (a second communication device and a third communication device in FIG. 6) are used as the executing entities in the interaction diagram is used to explain the method. However, the executing entities in the interaction diagram are not limited in this application. For example, in FIG. 3 and FIG. 6 below, the executing entities of the method may be replaced by a chip, chip system, processor, logic module, software, etc. within the communication device. In FIG. 3, the first communication device may be a network device, and the second communication device may be a terminal device. Alternatively, both the first communication device and the second communication device are terminal devices (for example, the method may be applied to a communication process between different terminal devices in a sidelink communication scenario).

[0210] S301: A first communication device transmits first information, and correspondingly, a 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 a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to 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.

[0211] S302: The second communication device transmits the second information, and in response, the first communication device receives the second information. The second information includes the model parameters of the first AI model group or the model parameters of the first AI model.

[0212] In this application, terms such as AI model, neural network model, AI neural network model, machine learning model, and AI processing model may be substituted for each other.

[0213] It should be understood that the first information transmitted by the first communication device in step S301 is used to determine the first AI model group. It can be understood that the first AI model group includes a first AI model and a second AI model, and that the functions of the first AI model group are implemented through at least the model processing of the first AI model and the model processing of the second AI model. That is, after receiving the second information, the first communication device may distribute the first AI model to the first communication device through the model parameters of the first AI model group or the model parameters of the first AI model included in the second information, and perform model processing on the first AI model. Correspondingly, the second communication device may perform model processing on the second AI model distributed to the second communication device. Optionally, model processing may include one or more of model update processing, model training processing, and model inference processing.

[0214] It must be understood that the transmission of wireless communication signals (e.g., the reception and transmission of configuration information of communication resources and the reception and transmission of reference signals) can be performed between different communication devices (e.g., a first communication device and a second communication device).

[0215] Optionally, the AI ​​models of the present application (e.g., the first AI model, the second AI model, and subsequent third through sixth AI models) may be used for the management of wireless communication signals (including at least one of configuration, updating, and optimization). For example, the AI ​​models may include one or more of an AI model used for modulation and / or demodulation, an AI model used for channel prediction, an AI model used for beam management, an AI model used for auxiliary positioning, an AI model used for channel compression, an AI model used for resource scheduling, and an AI model used to replace one or more modules in a transmitter and / or receiver. Alternatively, the AI ​​models of the present application may be AI models used for other AI tasks, for example, an AI model used for image recognition, an AI model used for natural language processing, or an AI model used for computer vision.

[0216] Optionally, when the first AI model group is considered as one AI model, the first AI model and the second AI model can be understood as two AI submodels in the AI ​​model.

[0217] In the present application, the fact that one AI model is distributed to one communication device (e.g., a first AI model is distributed to a first communication device and a second AI model is distributed to a second communication device) can be understood as the communication device acquiring the model parameters of the AI ​​model, acquiring / generating / configuring the AI ​​model based on the model parameters of the AI ​​model, and subsequently performing model processing on the AI ​​model.

[0218] Optionally, model parameters may include one or more of the model's hyperparameters, the model's dataset (including the model's input data and label data corresponding to the input data), and the model's structural parameters.

[0219] Optionally, one AI model group may include two or more AI models. For example, in addition to the first AI model and the second AI model, the first AI model group may additionally include another AI model. The other AI model may be deployed to a different communication device different from the first communication device and the second communication device. This is not limited in the present specification.

[0220] In a possible implementation, a second communication device is a functional entity that determines a list of AI model groups based on first information, and the list of AI model groups includes one or more AI model groups, and one or more AI model groups include a first AI model group. Specifically, the second communication device can communicate with one or more first communication devices, and the second communication device can create / acquire / determine one or more AI model groups by receiving information (e.g., one or more first information) from one or more first communication devices. In other words, the second communication device can collect information and create a model based on the collected information. Subsequently, the second communication device can distribute the AI ​​model to one or more first communication devices.

[0221] Optionally, the list of AI model groups may include one or more AI model groups, and each AI model group may include two or more AI models. As previously mentioned, the relationship between AI model groups and AI models can also be understood as the relationship between an AI model and an AI submodel. Therefore, the list of AI model groups can alternatively be replaced by a list of AI models. That is, the list of AI models may include one or more AI models.

[0222] Optionally, list can be replaced with other terms such as set, dictionary, combination, space, etc.

[0223] Optionally, the second communication device being a functional entity that determines an AI model group list based on the first information includes the second communication device determining an AI model group list based on the first information and selecting some or all of the AI ​​model groups to be used by the first communication device from the AI ​​model group list. Specifically, after the second communication device determines an AI model group list based on the first information, the function implemented by the second communication device may further include selecting some or all of the AI ​​model groups to be used by the first communication device from the AI ​​model group list. That is, in addition to collecting information and generating a model based on the collected information, the second communication device may additionally perform model selection, so that the second communication device may subsequently distribute an AI model adapted to one or more first communication devices to one or more first communication devices.

[0224] For the sake of understanding, examples of AI models deployed in the first communication device and the second communication device are described below using the examples shown in FIGS. 4 and 5.

[0225] In the example illustrated in FIG. 4, the first AI model is distributed to the first communication device, and the second AI model is distributed to the second communication device, and the input of the first AI model distributed to the first communication device includes the output of the second AI model distributed to the second communication device. In this example, an example is used where the input data of the second AI model is X. Through the processing of the second AI model, the second communication device can acquire and transmit data Z; and through transmission over a wireless channel, the data received by the first communication device It is expressed as (due to interference such as transmission path loss and noise in wireless channels may differ from Z, It can be understood that this can be understood as an estimated value of Z, a measured value of Z, etc.). Afterwards, the first communication device data It can be used as input to the first AI model, and through the processing of the first AI model, data You can obtain.

[0226] In the example illustrated in FIG. 5, the first AI model is deployed to the first communication device, and the second AI model is deployed to the second communication device, and the input of the second AI model includes the output of the first AI model. In this example, an example is used where the input data of the first AI model is X. Through the processing of the first AI model, the first communication device can acquire and transmit data Z; and through transmission over a wireless channel, the data received by the second communication device It is expressed as. Afterwards, the second communication device data It can be used as input to the second AI model, and through the processing of the second AI model, data You can obtain.

[0227] It can be understood that the first communication device and the second communication device can be implemented in a plurality of ways.

[0228] For example, the second communication device may be a terminal device. Correspondingly, the first communication device and the second communication device may communicate with each other over a sidelink (SL). In this case, the first AI model and the second AI model may be referred to as an end-to-end model, an end-to-end collaboration model, etc.

[0229] As another example, the second communication device may be a network device (e.g., an access network device). Correspondingly, the first communication device and the second communication device may communicate with each other over uplink and downlink communication links. In this case, the first AI model and the second AI model may be referred to as an edge-terminal model, an edge-terminal collaboration model, a terminal-edge model, a terminal-edge collaboration model, etc. For example, if the first communication device is a terminal device and the second communication device is an access network device, the scenario illustrated in FIG. 4 can be understood as terminal-edge collaboration implemented based on a downlink scenario, and the scenario illustrated in FIG. 5 can be understood as terminal-edge collaboration implemented based on an uplink scenario.

[0230] In FIGS. 4 and 5, data Y may be label data corresponding to data X, and the processing results of label data Y and the first AI model and the second AI model It should be noted that the correlation between them can be used to detect or determine the processing performance of the first AI model and the second AI model. For example, the correlation can be determined using gradient information, loss functions, etc.

[0231] From the implementation process illustrated in FIG. 3, it can be seen that in the case of the second communication device, after receiving the first information in step S301, the second communication device can obtain a first AI model group by performing a model generation process based on the first information. Below, an example of a process in which the second communication device determines a first AI model group based on the first information is described.

[0232] First, the theoretical basis for generating an AI model is explained through Method A.

[0233] Generally, in conventional networks for connections or sessions, the design goal between different communication devices (receivers and transmitters) is to ensure that the receiver accurately recovers all data transmitted by the transmitter, that is, lossless data transmission. However, in future intelligent networks, due to the presence of large amounts of data and various AI task objectives, full data transmission may no longer be necessary, but the transmission of data valuable to AI tasks is required. Therefore, in a network, it is possible to optimize AI model performance (e.g., maximize accuracy) while minimizing wireless data transmission. To achieve this, in the example illustrated in FIGS. 4 and 5, AI model performance can be expressed based on mutual information between different data.

[0234] In one embodiment, in FIGS. 4 and 5, input data X of an AI model, label data Y corresponding to input data X, and data received by a receiver (e.g., a first communication device in FIG. 4 or a second communication device in FIG. 5). satisfies method A:

[0235]

[0236] is a variable and It indicates the amount of mutual information between them. Specifically, wireless channel transmission data and label data It represents mutual information (accurate / expected results of AI tasks) between them. A large amount of mutual information is wireless channel transmitted data. Label data included in It means more information about, which can be understood as higher AI model accuracy and better model performance. is the original input data and wireless link data Indicates the amount of mutual information between. A small amount of mutual information indicates less data transmitted over the wireless link, i.e., lower wireless communication overhead. Configurable parameter ( The value range of ) can be [0, 1]) can be used to control the ratio between two pieces of mutual information. Therefore, the aforementioned mathematical formula Minimizing it maximizes AI processing accuracy while minimizing wireless communication overhead. The subscript "IB" indicates information bottleneck (IB) theory (or distributed information bottleneck, deterministic information bottleneck, other information theories, etc.). That is, can be replaced with other symbols. This is just an example of implementation.

[0237] Based on method A, the basis for model generation may include one or more of the following data A to data C. That is, an AI model group may be generated based on the following data A to data C (for example, a second communication device may generate a first AI model group). Data A to data C are described below.

[0238] Data A: Used as input data data-label pairs ( is the batch size of the batch data, and is the input data of the m-th pair of data, and is the label data of the m-th pair of data).

[0239] Data B: Dimensions of data transmitted over a wireless link (e.g., of FIG. 4 or FIG. 5) or For example, the dimension of the data can be expressed as the quantity of tokens / words / marks (hereinafter uniformly referred to as tokens), or the dimension of the data can be expressed as an embedding vector. A token can be understood as similar to a "word" and as a basic unit for dividing intermediate transmissions. An embedding is a vector representation of an intermediate transmission. Each token can be understood as a constituent unit within the embedding vector; that is, the quantity of constituent units within the embedding is the token dimension.

[0240] Data C: Channel status information.

[0241] Optionally, a neural network may be used to generate an AI model group based on one or more of data A through data C. That is, the neural network inputs one or more of data A through data C and processes them to obtain model parameters for the AI ​​model group.

[0242] In one implementation example, the loss function of the neural network can be expressed as method B:

[0243]

[0244] In another implementation example, To reduce the computational complexity of variational processing, the calculation of mutual information can be introduced. For example, the loss function of a neural network can be expressed as method C:

[0245]

[0246] The subscript "VIB" indicates a variational information bottleneck;

[0247] represents a neural network or model parameter placed on the user side (e.g., a first communication device);

[0248] represents a neural network or model parameter deployed on the base station side (e.g., a second communication device);

[0249] represents the Lagrange multiplier for balancing AI processing accuracy and wireless communication overhead;

[0250] P is given x Represents the conditional probability density of;

[0251] q is Representing the variational probability density of;

[0252] p and q actually represent different probability density functions.

[0253] is the probability density function Under the premise that it is known It indicates calculating the expected / average value of some parts in.

[0254] is the probability density function Under the premise that this is known It indicates calculating the expected / average value of some parts in.

[0255] represents a conditional probability distribution with parameter θ, and approximate conditional probability It is the form of the variational distribution.

[0256] is a parameter It represents a conditional probability distribution having .

[0257] is an approximate probability distribution It represents the shape of the variational distribution.

[0258] is a probability distribution and probability distribution It represents the Kullback-Leibler (KL) divergence between them.

[0259] In another implementation example, assuming the token / embedding dimension is constrained by wireless link bandwidth, the loss function of the neural network can be expressed as method D:

[0260] max

[0261] The constraints Indicates that, represents the maximum wireless link bandwidth threshold, and the constraint The physical meaning of is that the amount of data transmitted over the wireless link satisfies the air interface bandwidth constraint. That is, in method D, the accuracy of the AI ​​task can be maximized under the premise that the wireless channel transmission information satisfies the bandwidth constraint.

[0262] In another implementation example, assuming the token / embedding dimension is constrained by wireless link bandwidth, the loss function of the neural network can be expressed by method E:

[0263] min

[0264] The constraints Indicates that, represents the minimum threshold of accuracy for the AI ​​task, and constraints The physical meaning of is that the accuracy of the AI ​​task is greater than the minimum threshold. That is, in method E, when the accuracy of the AI ​​task is greater than the minimum threshold, the amount of information transmitted in the wireless channel can be minimized.

[0265] Based on the aforementioned implementation process, and under the premise that the token / embedding dimension is subject to bandwidth constraints, the inference accuracy of the acquired AI model is high. For example, based on different datasets, the simulation results obtained in Method C are as follows: Test on the Canadian Institute for Advanced Research (CIFAR) dataset: training epochs=319, intermediate output dimension=20, and accuracy=92.37%;

[0266] Test on the modified national institute of standards and technology (MNIST) dataset: epoch=400, intermediate dim=64, and accuracy=97.62%.

[0267] The intermediate output dimension is Z / It is the dimension of, and accuracy is the accuracy of the AI ​​task.

[0268] From the foregoing description, it can be seen that the first information transmitted by the first communication device in step S301 can be used to determine the first AI model group. Based on parameters in the process illustrated in methods A through E, the first information may include one or more of the following information A through E. That is, the second communication device can obtain the parameters required in methods A through D through one or more of the information A through E included in the following first information, and can generate the first AI model group using one of methods A through E.

[0269] Information A: First dimensional information. If the input of the first AI model includes the output of the second AI model, the first dimensional information is used to determine the dimensional information of the input data of the first AI model or the dimensional information of the output data of the second AI model.

[0270] Information B: Second dimensional information. If the input of the second AI model includes the output of the first AI model, the second dimensional information is used to determine the dimensional information of the output data of the first AI model or the dimensional information of the input data of the second AI model.

[0271] Information C: Input data of the first AI model and label data of the input data of the first AI model.

[0272] Information D: Local computing capability status information of the first communication device.

[0273] Info E: Channel status information.

[0274] For information A or information B, the second communication device can determine the dimension information of the data transmitted over the communication link through the first information. For example, the dimension information may be the quantity of tokens or the dimension of an embedding vector. When the communication bandwidth between the first communication device and the second communication device is fixed, since the dimension of the data transmitted over the communication link is associated with the processing performance of the AI ​​model, the implementation process in which the second communication device determines the first AI model group can be simplified in this way, thereby reducing the complexity of the second communication device and improving the processing performance of the AI ​​model included in the first AI model group.

[0275] Optionally, the dimension information includes at least one of the following: an upper limit of the dimension, a lower limit of the dimension, a dimension expected by the first communication device (or a dimension not expected by the first communication device), and a value range of the dimension expected by the first communication device (or a value range of the dimension not expected by the first communication device).

[0276] For example, if the dimension information includes an upper limit and / or a lower limit of the dimension, the second communication device can use the range indicated by the upper limit and / or lower limit as one of the grounds for determining the AI ​​model, thereby improving the flexibility to implement the solution.

[0277] As another example, if the dimension information includes the dimension expected by the first communication device and / or the range of values ​​of the dimension expected by the first communication device, the AI ​​model determined by the second communication device based on the dimension information can satisfy the expectations of the first communication device.

[0278] Optionally, in information A or information B, the first dimensional information or the second dimensional information is determined based on channel state information (CSI). Specifically, the first dimensional information or the second dimensional information included in the first information is determined based on the channel state information so that the first dimensional information or the second dimensional information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to some extent. In this way, an AI model that can be subsequently acquired based on the first information can be adapted to the channel characteristics of the wireless channel, thereby improving the transmission performance of the AI ​​data corresponding to the AI ​​model. Furthermore, when the AI ​​model acquired based on the first information can be adapted to the channel characteristics of the wireless channel, the data transmitted over the wireless link can also satisfy channel bandwidth requirements as much as possible, thus improving the model performance of the AI ​​model included in the first group of AI models.

[0279] Optionally, the channel state information may include information about a channel from the first communication device to the second communication device and / or information about a channel from the second communication device to the first communication device. Where the first communication device is a terminal device and the second communication device is a network device, information about a channel from the first communication device to the second communication device may be understood as uplink channel information, and information about a channel from the second communication device to the first communication device may be understood as downlink channel information.

[0280] Optionally, channel state information can be obtained based on a reference signal.

[0281] For example, when a first communication device and a second communication device communicate with each other 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.

[0282] In another example, when a first communication device and a second communication device communicate with each other via an uplink and a downlink, the reference signal may include a channel state information reference signal (CSI-RS), a sounding reference signal (SRS), etc.

[0283] In the case of information C, when the first information includes input data of the first AI model and label data of the input data of the first AI model, the input data can be used as an input to the first AI model, and the label data can be used as one of the grounds for determining the model processing performance of the first AI model; therefore, in the case of the second communication device, the second communication device can obtain an AI model with excellent performance based on the two pieces of information.

[0284] In addition, in the case of the second communication device, the second communication device implements a mutual information-based mathematical calculation based on two pieces of information and AI data (e.g., input data of the second AI model or output data of the second AI model) received and transmitted by the second communication device on a wireless link, and determines a first AI model group based on the result of the mathematical calculation, thereby improving the model performance of the AI ​​model included in the first AI model group under the premise that the data on the wireless link satisfies the bandwidth.

[0285] With respect to information D, if the first information includes local computational capability status information of the first communication device, the complexity requirements of the model processing of the AI ​​model may be related to the local computational capability status of the first communication device. Accordingly, in the case of the second communication device, a first AI model group determined by the second communication device based on the local computational capability status information can be adapted to the local computational capability status of the first communication device, and accordingly, a first AI model satisfying the local computational capability status is provided, thereby improving the success rate of the first communication device performing model processing based on the first AI model.

[0286] In the case of information E, if 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, in the case of the second communication device, the second communication device determines a first AI model group based on the channel state information and adapts it to the channel characteristics so that the data transmitted in the wireless link satisfies the channel bandwidth requirements to the maximum extent, thereby improving the transmission performance of AI data corresponding to the AI ​​model.

[0287] If the first information includes two or more pieces of information in the aforementioned information A through information E, it can be understood that additional overlapping benefits may be obtained through the two or more pieces of information based on the technical benefits achieved in any of the information in the foregoing description.

[0288] In a possible implementation, the first communication device transmits the first information in step S301. The first information is used to determine the first AI model group, which is used to update the second AI model group to obtain the first AI model group; the second AI model group includes a third AI model and a fourth AI model, the third AI model is distributed to the first communication device and the fourth AI model is distributed to the second communication device, and the input of the third AI model includes the output of the fourth AI model, or the input of the fourth AI model includes the output of the third AI model. Specifically, in the case of the second communication device, after the second communication device receives the first information, the second communication device can obtain the first AI model group by updating the second AI model group based on the first information. That is, the first information transmitted by the first communication device can be used to update other AI models, so the solution can be applied to an AI model update scenario.

[0289] Optionally, "update" may be replaced with other terms, e.g., "modification," "iteration," "optimization," "processing," etc.

[0290] Optionally, if the second AI model group is considered as one AI model, the third AI model and the fourth AI model can be understood as two AI sub-models in the AI ​​model.

[0291] Optionally, the third and fourth AI models included in the second AI model group may be general models or dedicated models, so that different types of models are updated.

[0292] It should be understood that general models may be referred to as basic models, large models, or L0 models. Dedicated models may be referred to as small models, L1 models, L2 models, etc.

[0293] Large models are used as examples. Large models can be machine learning models with a large number of parameters and complex structures, capable of processing large-scale data and completing various complex tasks such as natural language processing, computer vision, and speech recognition.

[0294] Optionally, large models are typically built based on deep neural networks and have billions or hundreds of billions of parameters.

[0295] Optionally, large models can be designed to handle more complex tasks and data by improving the model's representational power and predictive performance.

[0296] Optionally, large models can learn complex patterns and features by training on large-scale data, possess stronger generalization capabilities, and accurately predict unprocessed data.

[0297] In contrast, a small model may be a model with fewer parameters and fewer layers. Typically, compared to a small model, a large model usually has more parameters and deeper layers, and possesses stronger representational power and higher accuracy. However, large models also require more computing resources and time for training and inference, and are applicable to scenarios with large amounts of data and sufficient computing resources, such as cloud computing, high-performance computing, or artificial intelligence.

[0298] Optionally, the compact model offers advantages such as lightweight design, high efficiency, and easy deployment, and is applicable to scenarios with small data volumes and limited computing resources, such as mobile applications, embedded devices, or the Internet of Things.

[0299] In a possible implementation, the first information transmitted by the first communication device in step S301 is information transmitted periodically and / or the second information transmitted by the second communication device in step S302 is information transmitted periodically. Specifically, the first information may be one of the basis for determining an AI model, and the second information may be used to deploy an AI model. The AI ​​model may be periodically determined and / or periodically deployed by periodically transmitting the first information and / or the second information between the first communication device and the second communication device, so that the AI ​​model is updated repeatedly multiple times through a periodic process.

[0300] In a possible implementation, the first information transmitted by the first communication device in step S301 is used to determine the first AI model group, and the AI ​​model in the first AI model group is a dedicated model. Specifically, the first communication device may be a terminal device. Accordingly, the AI ​​model deployed to the terminal device may be a dedicated model, and the first information transmitted by the terminal device may be used to determine the dedicated model. Different terminal devices may have different device-side characteristics (e.g., different local data, different local computational capabilities, and different channel characteristics). Accordingly, in this manner in which a dedicated model is deployed to the terminal device, the AI ​​model deployed to the terminal device can be adapted to the device-side characteristics of the terminal device, thereby improving the model processing performance of the AI ​​model.

[0301] Based on the technical solution illustrated in FIG. 3, the first communication device transmits first information used to determine a first AI model group in step S301, and then the first communication device receives second information in step S302, wherein the second information includes model parameters of the first AI model group or model parameters of the first AI model within the first AI model group. That is, the second communication device is used as a receiver of the first information, and the second communication device determines the first AI model group based on the first information from the first communication device and can distribute the first AI model to the first communication device through the second information. Accordingly, when a communication device within a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the first information from the first communication device can also be used as one of the decision 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 to the maximum extent, thereby improving the processing performance of the model processing subsequently performed by the first communication device based on the AI ​​model.

[0302] FIG. 6 is a diagram illustrating an implementation of a communication method according to the present application. The method includes the following steps.

[0303] S601: A second communication device transmits third information, and in response, a third communication device receives third information. The third information is used to determine a third AI model group, and the third AI model group includes a fifth AI model and a sixth AI model; the fifth AI model is distributed to a first communication device and the sixth AI model is distributed to a second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model.

[0304] S602: The third communication device transmits the fourth information, and in response, the second communication device receives the fourth information. The fourth information includes the model parameters of the third AI model group.

[0305] It should be noted that the second communication device and the third communication device may be implemented in multiple ways. The second communication device may be a terminal device or an access network device, and the third communication device may be a cloud server or a core network device. For example, if the third communication device is a cloud server, the second communication device may communicate with the cloud server through the core network device. As another example, if the third communication device is a core network device, the second communication device may be a terminal device, and the terminal device may communicate with the core network device through the access network device. As yet another example, if the third communication device is a core network device, the second communication device may be an access network device, and the access network device may communicate with the core network device through a communication interface between the access network device and the core network device.

[0306] It can be understood that the third AI model group includes the fifth AI model and the sixth AI model, and that the function of the third AI model group is implemented through at least the model processing of the fifth AI model and the model processing of the sixth AI model. In other words, after receiving the fourth information, the second communication device can determine the model parameters of the fifth AI model and the model parameters of the sixth AI model. Additionally, the second communication device can distribute the fifth AI model to the first communication device and distribute the sixth AI model to the second communication device so that the model processing of the fifth AI model and the sixth AI model is implemented. Optionally, the model processing may include one or more of model update processing, model training processing, and model inference processing.

[0307] Optionally, when the third AI model group is considered as one AI model, the fifth AI model and the sixth AI model can be understood as two AI sub-models in the AI ​​model.

[0308] In a possible implementation, a third communication device is a functional entity that determines a third AI model group based on third information. Specifically, the third communication device may communicate with one or more second communication devices, and the third communication device may receive information (e.g., one or more third information) from one or more second communication devices to create / acquire / determine a third AI model group. In other words, the second communication device may collect information and create a model based on the collected information. Subsequently, the second communication device may distribute the AI ​​model to one or more second communication devices (and corresponding first communication devices).

[0309] Optionally, the AI ​​model of the third AI model group is a general model. In this way, the third communication device can determine the general model through information from one or more second communication devices (e.g., one or more third information). Subsequently, the general model having high generalization and excellent universality can be distributed to each of the plurality of second communication devices and to the first communication device connected to each of the second communication devices.

[0310] From the foregoing description, it can be seen that the third information transmitted by the second communication device in step S601 can be used to determine the third AI model group. From the process illustrated in methods A through E, it can be seen that the third information includes one or more of the following information 1 through information 5. That is, the third communication device can obtain the parameters required in methods A through E through one or more of the information 1 through information 5 included in the following third information, and can generate the third AI model group using one of methods A through E.

[0311] Information 1: Third dimensional information. If the input of the 5th AI model includes the output of the 6th AI model, the third dimensional information is used to determine the dimensional information of the input data of the 5th AI model or the dimensional information of the output data of the 6th AI model.

[0312] Information 2: Fourth dimension information. If the input of the sixth AI model includes the output of the fifth AI model, the fourth dimension information is used to determine the dimension information of the output data of the fifth AI model or the dimension information of the input data of the sixth AI model.

[0313] Information 3: Model parameters of an AI model in one or more AI model groups. Each group of one or more AI model groups includes a dedicated model deployed to a first communication device and a dedicated model deployed to a second communication device.

[0314] Information 4: Data from one or more first communication devices connected to the second communication device.

[0315] Information 5: Input data of an AI model deployed to a second communication device and label data of the input data of an AI model deployed to a second communication device.

[0316] Regarding Information 1 and Information 2, the third communication device can determine the dimension information of the data transmitted over the communication link between the first communication device and the second communication device through the third information. For example, the dimension information may be the quantity of tokens or the dimension of an embedding vector. When the communication bandwidth between the first communication device and the second communication device is fixed, since the dimension of the data transmitted over the communication link is associated with the processing performance of the AI ​​model, the implementation process of the third communication device determining the first AI model group can be simplified in this way, thereby reducing the complexity of the third communication device and improving the processing performance of the AI ​​model included in the first AI model group.

[0317] Optionally, the dimension information includes at least one of an upper limit of the dimension, a lower limit of the dimension, a dimension expected by the second communication device (or a dimension not expected by the second communication device), and a value range of the dimension expected by the second communication device (or a value range of the dimension not expected by the second communication device). Specifically, the dimension information determined based on the third dimension information or the fourth dimension information may include at least one of the aforementioned items to enhance flexibility in implementing the solution.

[0318] Optionally, the dimension information may include a value, a quantized value, an index of the value, and an index of the quantized value of any of the aforementioned items. Other implementations are also possible.

[0319] For example, if the dimension information includes an upper limit and / or a lower limit of the dimension, the third communication device can use the range indicated by the upper limit and / or lower limit as one of the grounds for determining the AI ​​model, thereby improving the flexibility of the solution implementation.

[0320] As another example, if the dimension information includes the dimension expected by the second communication device and / or the range of values ​​of the dimension expected by the second communication device, the AI ​​model determined by the third communication device based on the dimension information can satisfy the expectations of the second communication device.

[0321] Optionally, third-dimensional information or fourth-dimensional information is determined based on channel state information. Specifically, the third-dimensional information or fourth-dimensional information included in the third information can be determined based on channel state information, so that the third-dimensional information or fourth-dimensional information can reflect the channel characteristics of the wireless channel between the first communication device and the second communication device to some extent. In this way, an AI model that can be subsequently acquired based on the third information is adapted to the channel characteristics of the wireless channel, thereby improving the transmission performance of AI data corresponding to the AI ​​model. Furthermore, when the AI ​​model acquired based on the first information can be adapted to the channel characteristics of the wireless channel, the data transmitted over the wireless link can also fully satisfy channel bandwidth requirements, thereby improving the model performance of the AI ​​model included in the first group of AI models.

[0322] In the case of Information 3, when the third information includes model parameters of an AI model of one or more AI model groups, the third communication device may acquire one or more dedicated models distributed to the first communication device and the second communication device based on the third information. In this way, the third communication device may acquire model characteristics of one or more dedicated models and reflect the acquired characteristics in the third AI model group to improve the generalization (or universality) of the general models included in the third AI model group.

[0323] In the case of Information 4, if the third information includes data from one or more first communication devices (e.g., terminal devices) connected to the second communication device, different terminal devices may have different data characteristics (e.g., different data may be collected at different geographical locations, different data may be collected at different times, and different data may correspond to different wireless channels of the user). Accordingly, in this way, the third communication device acquires a third AI model group based on these data characteristics and improves the generalization (or universality) of the general model included in the third AI model group.

[0324] Regarding information 5, if the third information includes input data of an AI model distributed to the second communication device and label data of the input data of the AI ​​model distributed to the second communication device, the input data can be used as input to the sixth AI model, and the label data can be used as one of the basis for determining the model processing performance of the sixth AI model. Therefore, in the case of the third communication device, the third communication device can obtain an AI model with excellent performance based on the two pieces of information.

[0325] In addition, in the case of the second communication device, the second communication device implements a mutual information-based mathematical calculation based on two pieces of information and AI data (e.g., input data of the sixth AI model or output data of the sixth AI model) received and transmitted by the second communication device over a wireless link, and determines a third AI model group based on the result of the mathematical calculation, thereby improving the model performance of the AI ​​models included in the third AI model group under the premise that the data over the wireless link satisfies the bandwidth.

[0326] If the third information includes two or more pieces of information from the aforementioned information 1 to information 5, it can be understood that additional overlapping benefits may be obtained through the two or more pieces of information based on the technical benefits achieved from any of the information in the foregoing description.

[0327] Optionally, each piece of information included in the third information may be part of the information obtained by the second communication device filtering multiple pieces of information. Below, an example in which the third information includes information 3 is used for explanation.

[0328] In one embodiment, when the third information includes information 3, N (N is 2 or more) AI model groups may be distributed to the first communication device and the second communication device, and each AI model group includes a dedicated model distributed to the first communication device and a dedicated model distributed to the second communication device, and information 3 may include model parameters of k (k is a positive integer) AI model groups from the N AI model groups.

[0329] In the example of Information 3, the second communication device can determine k AI model groups from N AI model groups based on target information (the target information can be used to determine general AI model groups, general AI models, reference AI models, etc.; or the target information can be general AI model groups, general AI models, reference AI models, etc.). For example, the second communication device can determine a difference value (e.g., a cosine value corresponding to the logit) between the representation parameter of N AI model groups (e.g., Logit) and the representation parameter of an AI model / AI model group determined based on the target information, that is, the second communication device determines N difference values, determines k corresponding AI model groups based on k larger difference values ​​from the N difference values, and can include the model parameters of the k AI model groups in Information 3.

[0330] In another example of Information 3, N AI model groups may include 2N dedicated models. For example, N of the 2N dedicated models are deployed to a first communication device, and the remaining N of the 2N dedicated models are deployed to a second communication device. For ease of reference, the N dedicated models deployed to the first communication device are referred to as N1 dedicated models below, and the N dedicated models deployed to the second communication device are referred to as N2 dedicated models. Then, the second communication device can determine k AI model groups from the 2N AI models based on target information (the target information may be used to determine a general AI model, a reference AI model, etc.; or the target information may be a general AI model, a reference AI model, etc.).

[0331] Additionally, for any AI model in 2N dedicated models, the second communication device can determine a difference value between the representation parameter of the arbitrary AI model (e.g., logit) and the representation parameter of the AI ​​model determined based on target information (e.g., cosine value between logits); that is, the second communication device can determine 2N difference values ​​corresponding to each of the 2N dedicated models. Subsequently, based on the 2N difference values, the second communication device can determine k larger difference values ​​from the N difference values ​​corresponding to N1 dedicated models, determine k corresponding AI model groups, and include the model parameters of the k AI model groups in Information 3. Alternatively, based on the 2N difference values, the second communication device can determine k larger difference values ​​from the N difference values ​​corresponding to N2 dedicated models, determine k corresponding AI model groups, and include the model parameters of the k AI model groups in Information 3.

[0332] Optionally, target information may be pre-configured in the second communication device or configured for the second communication device by a third communication device (or other device). This is not limited to the above.

[0333] Likewise, if third information includes information 1, information 2, information 4, or information 5, refer to the implementation described above. Details are not described herein.

[0334] In a possible implementation, at step S601, the second communication device may transmit third information. That the third information is used to determine the third AI model group includes: the third information is used to obtain the third AI model group by updating the fourth AI model group; the fourth AI model group includes the seventh AI model and the eighth AI model, the seventh AI model is distributed to the first communication device and the eighth AI model is distributed to the second communication device, and the input of the seventh AI model includes the output of the eighth AI model, or the input of the eighth AI model includes the output of the seventh AI model. Specifically, in the case of the third communication device, after receiving the third information, the third communication device may obtain the third AI model group by updating the fourth AI model group based on the third information. That is, since the third information transmitted by the second communication device may be used to update other AI models, the solution is applicable to AI model update scenarios. Optionally, when the fourth AI model group is considered as one AI model, the seventh AI model and the eighth AI model may be understood as two AI submodels within the AI ​​model.

[0335] In a possible implementation, the third information transmitted by the second communication device in step S601 is information transmitted periodically and / or the fourth information transmitted by the third communication device in step S602 is information transmitted periodically. Specifically, the third information may be one of the grounds for determining an AI model, and the fourth information may be used to deploy an AI model. The AI ​​model may be periodically determined and / or periodically deployed by periodically transmitting the third information and / or the fourth information between the second communication device and the third communication device, so that the AI ​​model is updated repeatedly multiple times through a periodic process.

[0336] Based on the technical solution illustrated in FIG. 6, after the second communication device transmits third information used to determine a third AI model group in step S601, the second communication device may receive fourth information in step S602, wherein the third information includes model parameters of the third AI model group. That is, the third communication device is used as a receiver of the third information. The third communication device may determine a third AI model group based on the third information from the second communication device, and through the fourth information, the second communication device may subsequently distribute a fifth AI model to the first communication device and a sixth AI model to the second communication device. Therefore, when a communication device of a communication system is used as an AI participation node and its computational power can be applied to AI model processing, the third information from the second communication device may be used as one of the decision bases for the third communication device to determine an AI model, thereby enabling the AI ​​model determined by the third communication device to be adapted to the second communication device as much as possible, and thereby improving the success rate of the second communication device performing model processing on the AI ​​model thereafter.

[0337] Refer to FIG. 7. One embodiment of the present application provides a communication device (700). The communication device (700) can implement the functions of a second communication device or a first communication device in the above-described method embodiment, and thus can also implement the advantageous effects of the above-described method embodiment. In this embodiment of the present application, the communication device (700) may be a first communication device (or a second communication device), or an integrated circuit, element, etc. inside a first communication device (or a second communication device), and may be, for example, a chip.

[0338] It should be noted that the transceiver unit (702) may include a transmitting unit and a receiving unit configured to perform transmission and reception, respectively.

[0339] In a possible implementation, when the device (700) is configured to perform the method performed by the first communication device of the above-described embodiment, the device (700) includes a processing unit (701) and a transceiver unit (702). The processing unit (701) is configured to determine first information. The transceiver unit (702) is configured to transmit the first information, and the first information is used to determine a first artificial intelligence (AI) model group, and the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to 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 also configured to receive second information from the second communication device, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0340] In a possible implementation, when the device (700) is configured to perform the method performed by the second communication device in the above-described 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, and the first AI model group includes a first AI model and a second AI model. The first AI model is distributed to the first communication device, and the second AI model is distributed to the second communication device, and 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 further configured to transmit second information, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0341] In a possible implementation, when the device (700) is configured to perform the method performed by the first communication device of the above-described embodiment, the device (700) includes a processing unit (701) and a transceiver unit (702). The processing unit (701) is configured to determine third information. The transceiver unit (702) is configured to transmit the third information, and the third information is used to determine a third AI model group, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to the first communication device, and the sixth AI model is distributed to the second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The transceiver unit (702) is also configured to receive fourth information from the third communication device, and the fourth information includes model parameters of the third AI model group.

[0342] In a possible implementation, when the device (700) is configured to perform the method performed by the second communication device of the aforementioned embodiment, the device (700) includes a processing unit (701) and a transceiver unit (702). The transceiver unit (702) is configured to receive third information. The processing unit (701) is configured to determine a third AI model group based on the third information, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is distributed to the first communication device, and the sixth AI model is distributed to the second communication device; the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The transceiver unit (702) is further configured to transmit fourth information, and the fourth information includes model parameters of the third AI model group.

[0343] It should be noted that regarding content such as the information execution process of a unit in the communication device (700), reference should be made to the description in the method embodiment described above in this application. Details are not described again in this specification.

[0344] FIG. 8 is another diagram of the structure of a communication device (800) according to the present application. The communication device (800) includes a logic circuit (801) and an input / output interface (802). The communication device (800) may be a chip or an integrated circuit.

[0345] The transceiver unit (702) illustrated in FIG. 7 may be a communication interface. The communication interface may be an input / output interface (802) of FIG. 8. The input / output interface (802) may include an input interface and an output interface. Alternatively, the communication interface may be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit.

[0346] Optionally, the logic circuit (801) is configured to determine first information. The input / output interface (802) is configured to transmit the first information, the first information is used to determine a first artificial intelligence (AI) model group, the first AI model group includes a first AI model and a second AI model. The first AI model is deployed to a first communication device, and the second AI model is deployed to a 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 input / output interface (802) is also configured to receive second information from the second communication device, the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0347] Optionally, the input / output interface (802) receives first information. The logic circuit (801) is configured to determine a first AI model group based on the first information, and the first AI model group includes a first AI model and a second AI model. The first AI model is deployed to a first communication device and the second AI model is deployed to a second communication device, and 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 input / output interface (802) is further configured to transmit second information, and the second information includes model parameters of the first AI model group or model parameters of the first AI model.

[0348] Optionally, the logic circuit (801) is configured to determine third information. The input / output interface (802) is configured to transmit the third information, and the third information is used to determine a third AI model group, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is deployed to a first communication device, and the sixth AI model is deployed to a second communication device, and the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The input / output interface (802) is further configured to receive fourth information from the third communication device, and the fourth information includes model parameters of the third AI model group.

[0349] Optionally, the input / output interface (802) is configured to receive third information. The logic circuit (801) is configured to determine a third AI model group based on the third information, and the third AI model group includes a fifth AI model and a sixth AI model. The fifth AI model is deployed to a first communication device, and the sixth AI model is deployed to a second communication device; the input of the fifth AI model includes the output of the sixth AI model, or the input of the sixth AI model includes the output of the fifth AI model. The input / output interface (802) is further configured to transmit fourth information, and the fourth information includes model parameters of the third AI model group.

[0350] The logic circuit (801) and the input / output interface (802) may further perform other steps performed by the first communication device or the second communication device in any embodiment, and achieve corresponding advantageous effects. Details are not described again herein.

[0351] In a possible implementation, the processing unit (701) shown in FIG. 7 may be the logic circuit (801) of FIG. 8.

[0352] Optionally, the logic circuit (801) may be a processing unit, and some or all functions of the processing unit may be implemented using software. Some or all functions of the processing unit may be implemented using software.

[0353] Optionally, the processing unit may include memory and a processor. The memory is configured to store a computer program, and the processor reads and executes the computer program stored in the memory to perform the corresponding processing and / or steps in any method embodiment.

[0354] Optionally, the processing unit may include only a processor. Memory configured to store computer programs is located outside the processing unit, and the processor is connected to the memory via circuits / wires to read and execute computer programs stored in the memory. The memory and the processor may be integrated together or may be physically independent of each other.

[0355] Optionally, the processing unit may be one or more chips or one or more integrated circuits. For example, the processing unit may be one or more field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), system-on-chip (SoCs), central processing units (CPUs), network processors (NPs), digital signal processors (DSPs), microcontroller units (MCUs), programmable logic devices (PLDs), other integrated chips, or any combination of the aforementioned chips or processors.

[0356] FIG. 9 illustrates a communication device (900) in the aforementioned embodiment according to one embodiment of the present application. Specifically, the communication device (900) may be a communication device used as a terminal device in the aforementioned embodiment. In the example illustrated in FIG. 9, the terminal device is implemented using a terminal device (or a component within the terminal device).

[0357] In the drawing of a possible logical structure of a communication device (900), the communication device (900) may include at least one processor (901) and a communication port (902), but is not limited thereto.

[0358] The transceiver unit (702) illustrated in FIG. 7 may be a communication interface. The communication interface may be a communication port (902) of FIG. 9. The communication port (902) may include an input interface and an output interface. Alternatively, the communication port (902) may be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit.

[0359] Additionally, optionally, the device may further include at least one of a memory (903) and a bus (904). In this embodiment of the application, at least one processor (901) is configured to perform control processing for the behavior of the communication device (900).

[0360] Additionally, 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, a transistor logic device, a hardware component, or any combination thereof. The processing module may implement or execute various exemplary logical blocks, modules, and circuits described with reference to the contents disclosed in this application. Alternatively, the processor may be a combination of processors implementing computing functions, for example, a combination of one or more microprocessors, or a combination of a digital signal processor and a microprocessor. For convenience and brevity of description, those skilled in the art will clearly understand that for detailed operation processes of the aforementioned systems, devices, and units, reference is made to the corresponding processes of the aforementioned method embodiments. Details are not described again herein.

[0361] It should be noted that the communication device (900) illustrated in FIG. 9 may be specifically configured to implement the steps implemented by the terminal device in the method embodiment described above and to achieve a technical effect corresponding to the terminal device. For a specific implementation of the communication device illustrated in FIG. 9, refer to the description of the method embodiment described above. Details are not described one by one in this specification.

[0362] FIG. 10 is a diagram of the structure of a communication device (1000) in the aforementioned embodiment according to one embodiment of the present application. Specifically, the communication device (1000) may be a communication device used as a network device in the aforementioned embodiment. In the example illustrated in FIG. 10, the network device is implemented using a network device (or a component within a network device). For the structure of the communication device, reference is made to the structure illustrated in FIG. 10.

[0363] A communication device (1000) includes at least one processor (1011) and at least one network interface (1014). 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), memory (1012), transceiver (1013), and network interface (1014) are connected, for example, via a bus. In this embodiment of the application, the connection may include various interfaces, transmission lines, buses, etc. This is not limited to this embodiment. The antenna (1015) is connected to the transceiver (1013). The network interface (1014) is configured to enable the communication device to communicate with another communication device via a communication link. For example, the network interface (1014) may include a network interface between the communication device and a core network device, for example, an S1 interface. The network interface may include a network interface between a communication device and another communication device (e.g., another network device or a core network device), e.g., an X2 or Xn interface.

[0364] The transceiver unit (702) illustrated in FIG. 7 may be a communication interface. The communication interface may be the network interface (1014) of FIG. 10. The network interface (1014) may include an input interface and an output interface. Alternatively, the network interface (1014) may be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit.

[0365] The processor (1011) is configured to primarily process communication protocols and communication data, control the entire communication device, execute software programs, and process data of software programs, and is configured to support the communication device in performing operations described in the embodiments, for example. The communication device may include a baseband processor and a central processing unit. The baseband processor is configured primarily to process communication protocols and communication data. The central processing unit is configured primarily to control the entire terminal device, execute software programs, and process data of software programs. The processor (1011) of FIG. 10 may integrate the functions of the baseband processor and the central processing unit. Those skilled in the art will understand that the baseband processor and the central processing unit may alternatively be independent processors and may be interconnected using technology such as a bus. Those skilled in the art will understand that the terminal device may include multiple baseband processors to adapt to different network standards, the terminal device may include multiple central processing units to enhance the processing capabilities of the terminal device, and the components of the terminal device may be connected using various buses. The baseband processor may also be represented as a baseband processing circuit or a baseband processing chip. The central processing unit can also be represented as a central processing circuit or a central processing chip. Functions for processing communication protocols and communication data can be embedded in the processor or stored in memory in the form of software programs, and the processor implements baseband processing functions by executing these software programs.

[0366] The memory is configured to primarily store software programs and data. The memory (1012) may exist independently and is connected to the processor (1011). Optionally, the memory (1012) and the processor (1011) may be integrated together, for example, into a single chip. The memory (1012) may store program code for executing the technical solution in the embodiments of the present application, and the processor (1011) controls the execution. Various types of executed computer program code may also be considered as drivers for the processor (1011).

[0367] FIG. 10 illustrates only one memory and one processor. An actual terminal device may have multiple processors and multiple memories. The memory may also be referred to as a storage medium, a storage device, etc. The memory may be a storage element on the same chip as the processor, i.e., an on-chip storage element, or an independent storage element. This is not limited to the embodiments of the present application.

[0368] A transceiver (1013) may be configured to support the reception or transmission of radio frequency signals between a communication device and a terminal, and the transceiver (1013) may be connected to an antenna (1015). The transceiver (1013) includes a transmitter (Tx) and a receiver (Rx). Specifically, one or more antennas (1015) may receive radio frequency signals. The receiver (Rx) of the transceiver (1013) is configured to receive 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 a processor (1011) so that the processor (1011) further processes the digital baseband signals or digital intermediate frequency signals, for example, performs demodulation and decoding. Additionally, the transmitter (Tx) of the transceiver (1013) is further configured to receive a modulated digital baseband signal or a digital intermediate frequency signal from the processor (1011), convert the modulated digital baseband signal or the digital intermediate frequency signal into a radio frequency signal, and transmit the radio frequency signal through one or more antennas (1015). Specifically, the receiver (Rx) can acquire a digital baseband signal or a digital intermediate frequency signal by selectively performing one or more levels of downmixing processing and analog-to-digital conversion processing on the radio frequency signal, and the order of downmixing processing and analog-to-digital conversion processing is adjustable. The transmitter (Tx) can acquire a radio frequency signal by selectively performing one or more levels of upmixing processing and digital-to-analog conversion processing on the modulated digital baseband signal or the digital intermediate frequency signal. The order of upmixing processing and digital-to-analog conversion processing is adjustable. The digital baseband signal and the digital intermediate frequency signal may be collectively referred to as digital signals.

[0369] The transceiver (1013) may also be referred to as a transceiver unit, a transceiver machine, a transceiver device, etc. Optionally, a device in the transceiver unit configured to implement a receiving function may be considered a receiving unit. A device in the transceiver unit configured to implement a transmitting function may be considered a transmitting unit. That is, the transceiver unit includes a receiving unit and a transmitting unit. The receiving unit may also be referred to as a receiver machine, an input port, a receiving circuit, etc. The transmitting unit may be referred to as a transmitter machine, a transmitter, a transmitting circuit, etc.

[0370] It should be noted that the communication device (1000) illustrated in FIG. 10 may be configured to specifically implement the steps implemented by the network device in the method embodiment described above and to achieve a technical effect corresponding to the network device. For specific implementations of the communication device (1000) illustrated in FIG. 10, refer to the description of the method embodiment described above. Details are not described again in detail in this specification.

[0371] FIG. 11 is a diagram of the structure of a communication device in the aforementioned embodiment according to one embodiment of the present application.

[0372] The communication device (110) may be understood to include, for example, modules, units, elements, circuits, or interfaces appropriately configured together to perform the technical solution provided in this application. The communication device (110) may be the aforementioned terminal device or network device, or a component (e.g., a chip) within such devices, and is configured to implement the method described in the method examples below. The communication device (110) includes one or more processors (111). The processor (111) may be a general-purpose processor, a dedicated processor, etc. For example, the processor may be a baseband processor or a central processing unit. The baseband processor may be configured to process communication protocols and communication data. The central processing unit may be configured to control the communication device (e.g., a RAN node, a terminal, or a chip), execute a software program, and process data from the software program.

[0373] Optionally, in the design, the processor (111) may include a program (113) (which may sometimes be referred to as code or instructions). The program (113) may be executed on the processor (111) so that the communication device (110) performs the method described in the following embodiment. In other possible designs, the communication device (110) includes a circuit (not shown in FIG. 11).

[0374] Optionally, the communication device (110) may include one or more memories (112). The memory stores a program (114) (which may sometimes be referred to as code or instructions). The program (114) can be executed on a processor (111) so that the communication device (110) performs the method described in the above-described method embodiment.

[0375] Optionally, the processor (111) and / or memory (112) may include AI modules (117 and 118), and the AI ​​modules are configured to implement AI-related functions. The AI ​​modules may be implemented using software, hardware, or a combination of software and hardware. For example, the AI ​​modules may include radio intelligence control (RIC) modules. For example, the AI ​​modules may be quasi-real-time RICs or non-real-time RICs.

[0376] Optionally, the processor (111) and / or memory (112) may additionally store data. The processor and memory may be placed separately or integrated together.

[0377] Optionally, the communication device (110) may further include a transceiver (115) and / or an antenna (116). The processor (111) may sometimes be referred to as a processing unit and controls the communication device (e.g., a RAN node or terminal). The transceiver (115) may sometimes be referred to as a transceiver unit, transceiver machine, transceiver circuit, transceiver, etc., and is configured to implement the transmission and reception functions of the communication device through the antenna (116).

[0378] The processing unit (701) illustrated in FIG. 7 may be a processor (111). The transceiver unit (702) illustrated in FIG. 7 may be a communication interface. The communication interface may be the transceiver (115) of FIG. 11. The transceiver (115) may include an input interface and an output interface. Alternatively, the transceiver (115) may be a transceiver circuit, and the transceiver circuit may include an input interface circuit and an output interface circuit.

[0379] One embodiment of the present application further provides a computer-readable storage medium. The storage medium is configured to store one or more computer-executable instructions. When a computer-executable instruction is executed by a processor, the processor performs a method according to a possible implementation of the first communication device or the second communication device in the above-described embodiment.

[0380] One embodiment of the present application further provides a computer program product (or also referred to as a computer program). When the computer program product is executed by a processor, the processor performs a method according to a possible implementation of a first communication device or a second communication device.

[0381] One embodiment of the present application further provides a chip system. The chip system includes at least one processor configured to support a communication device in implementing a function in the aforementioned possible implementation of the communication device. Optionally, the chip system further includes an interface circuit, and the interface circuit provides program instructions and / or data to at least one processor. In a possible design, the chip system may further include memory. The memory is configured to store program instructions and data required by the communication device. The chip system may include a chip or may include a chip and other separate devices. The communication device may specifically be the first communication device or the second communication device in the aforementioned method embodiment.

[0382] One embodiment of the present application further provides a communication system. The network system architecture includes a first communication device and a second communication device in any one of the aforementioned embodiments.

[0383] It should be understood that in the various embodiments provided in this application, the disclosed systems, devices, and methods may be implemented in different ways. For example, the described device embodiments are merely examples. For example, a unit partition is merely a logical functional partition and may be a different partition in actual implementation. For example, multiple units or components may be combined or integrated into different systems, or some features may be ignored or not performed. Additionally, the mutual coupling, direct coupling, or communication connections displayed or discussed may be implemented through some interfaces. Indirect coupling or communication connections between devices or units may be implemented electronically, mechanically, or in other forms.

[0384] Units described as separate components may or may not be physically separated, and components indicated as units may or may not be physical units, that is, they may be located in a single location or distributed across multiple network units. Some or all of the units may be selected based on actual requirements to achieve the objectives of the solutions of the embodiments.

[0385] Additionally, in the embodiments of the present application, the functional units may be integrated into a single processing unit, each unit may exist physically independently, or two or more units may be integrated into a single unit. The integrated unit may be implemented in the form of hardware or in the form of a software functional unit. When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, the integrated unit may be stored on a computer-readable storage medium. Based on this understanding, an essential part, or a contributing part, of the technical solution of the present application, or all or part of the technical solution may be implemented in the form of a software product. A computer software product is stored on a storage medium and includes various instructions for instructing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. The storage medium includes any medium capable of storing program code, such as a USB flash drive, a removable hard disk, Read-Only Memory (ROM), Random Access Memory (RAM), a magnetic disk, or an optical disk.

Claims

Claim 1 A communication method applied to a first communication device, the method comprising: a step of transmitting first information - said first information is used to determine a first artificial intelligence (AI) model group, said first AI model group includes a first AI model and a second AI model, said first AI model is distributed to the first communication device and said second AI model is distributed to a 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 - and a step of receiving second information from the second communication device - said second information includes model parameters of the first AI model group or model parameters of the first AI model - Claim 2 A communication method according to claim 1, wherein the first information includes first dimensional information or second dimensional information, and when the input of the first AI model includes the output of the second AI model, the first dimensional information is used to determine the dimensional information of the input data of the first AI model or the dimensional information of the output data of the second AI model, and when the input of the second AI model includes the output of the first AI model, the second dimensional information is used to determine the dimensional information of the output data of the first AI model or the dimensional information of the input data of the second AI model. Claim 3 A communication method according to paragraph 2, wherein the dimension information comprises at least one of 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. Claim 4 A communication method according to claim 2 or 3, wherein the first dimension information or the second dimension information is determined based on channel state information. Claim 5 A communication method according to any one of claims 1 to 4, wherein the first information comprises at least one of the input data of the first AI model, label data of the input data of the first AI model, local computational capability status information of the first communication device, and channel status information. Claim 6 A communication method according to any one of claims 1 to 5, wherein the first information is used to determine the first AI model group, 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 a third AI model and a fourth AI model, the third AI model is distributed to the first communication device and the fourth AI model is distributed to the second communication device, and the input of the third AI model includes the output of the fourth AI model, or the input of the fourth AI includes the output of the third AI model. Claim 7 A communication method according to any one of claims 1 to 6, wherein the first information is information transmitted periodically and / or the second information is information transmitted periodically. Claim 8 A communication method according to any one of claims 1 to 7, wherein the AI ​​model of the first AI model group is a dedicated model. Claim 9 A communication method applied to a second communication device, comprising: a step of receiving first information; a step of determining a first AI model group based on the first information, wherein the first AI model group includes a first AI model and a second AI model, wherein the first AI model is distributed to the first communication device and the second AI model is distributed to the second communication device, and the input of the first AI model includes the output of the second AI model, or the input of the second AI includes the output of the first AI model; and a step of transmitting second information, wherein the second information includes model parameters of the first AI model group or model parameters of the first AI model. Claim 10 In claim 9, the second communication device is a functional entity that determines an AI model group list based on the first information, and the AI ​​model group list includes one or more AI model groups, and the one or more AI model groups include the first AI model group, a communication method. Claim 11 In paragraph 10, the communication method 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 to be used by the first communication device from the AI ​​model group list. Claim 12 A communication method according to any one of claims 9 to 11, wherein the first information includes first dimensional information or second dimensional information, and when the input of the first AI model includes the output of the second AI model, the first dimensional information is used to determine the dimensional information of the input data of the first AI model or the dimensional information of the output data of the second AI model, and when the input of the second AI model includes the output of the first AI model, the second dimensional information is used to determine the dimensional information of the output data of the first AI model or the dimensional information of the input data of the second AI model. Claim 13 In paragraph 12, the communication method comprises at least one of the dimension information, an upper limit of the dimension, a lower limit of the dimension, the dimension expected by the first communication device, and a value range of the dimension expected by the first communication device. Claim 14 A communication method according to claim 12 or 13, wherein the first dimension information or the second dimension information is determined based on channel state information. Claim 15 A communication method according to any one of claims 11 to 14, wherein the first information comprises at least one of the input data of the first AI model, label data of the input data of the first AI model, local computational capability status information of the first communication device, and channel status information. Claim 16 A communication method according to any one of claims 11 to 15, wherein determining the first AI model group based on the first information comprises updating the second AI model group based on the first information to obtain the first AI model group, wherein the second AI model group includes a third AI model and a fourth AI model, wherein the third AI model is distributed to the first communication device and the fourth AI model is distributed to the second communication device; and wherein the input of the third AI model includes the output of the fourth AI model, or the input of the fourth AI includes the output of the third AI model. Claim 17 A communication method according to any one of claims 11 to 16, wherein the first information is information transmitted periodically and / or the second information is information transmitted periodically. Claim 18 A communication method according to any one of claims 11 to 17, wherein the AI ​​model of the first AI model group is a dedicated model. Claim 19 A communication device comprising a module configured to perform a method according to any one of claims 1 through 18. Claim 20 A communication device comprising at least one processor, wherein the at least one processor is coupled to a memory; and wherein the at least one processor is configured to perform a method according to any one of claims 1 to 18. Claim 21 In paragraph 20, the communication device is a chip or chip system. Claim 22 A readable storage medium, wherein the storage medium stores a computer program or instructions, and when the computer program or instructions are executed by a communication device, a method according to any one of claims 1 to 18 is implemented.

Images