Communication method and related device

Abstract

The invention discloses a communication method and a related device, in the method, a first communication device receives first information and then performs first model processing and / or second model processing on a first model based on the first information, the first model processing comprises model training of the first communication device, and the second model processing comprises model training of the first communication device; the second model processing comprises model training of the at least two communication devices, and the at least two communication devices comprise the first communication device, so that the first communication device is prevented from training the first model according to a fixed training mode, and the first communication device is facilitated to train the first model according to a more suitable training mode; and the model training efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of communications, and more particularly to a communication method and related apparatus. Background Technology

[0002] With the development of communication technology, communication equipment in communication systems can now perform not only traditional communication services but also other new types of services, such as artificial intelligence (AI) services. Generally, a communication system capable of handling AI services can also be called an AI system.

[0003] Currently, communication devices can serve as participating nodes in AI systems, providing their own computing power and data. For example, a communication device can perform one or more model processing operations (such as model training, model updating, or model fine-tuning) on ​​a model based on local data to obtain another model. One possible approach is to increase the complexity of the model (such as increasing the number of model parameters or the number of neural network layers in the model), which can effectively improve the model's performance.

[0004] However, as the complexity of the models increases, the processing time may become excessively long, and limitations in communication equipment resources (such as computing power and storage resources) may even prevent the processing from executing. Therefore, optimizing the model processing is a pressing technical problem that needs to be solved. Summary of the Invention

[0005] This application provides a communication method and related apparatus for improving model processing performance.

[0006] The first aspect of this application provides a communication method executed by a first communication device. The first communication device can be a communication equipment (such as a terminal device or network device), or it can be a component of the communication equipment (e.g., a circuit or chip responsible for communication functions (such as a modem chip, a system-on-chip (SoC) chip, such as an SoC chip containing a modem core, or a system-in-package (SIP) chip)). Alternatively, the first communication device can also be a logic module or software capable of implementing all or part of the functions of the communication equipment. In this method, the first communication device receives first information, which instructs the first model to undergo first model processing and / or second model processing. Then, the first communication device can perform the first model processing and / or second model processing on the first model based on the first information. The first model processing includes model training by the first communication device, and the second model processing includes model training by at least two communication devices, where the first communication device is among the at least two communication devices.

[0007] After receiving the first information, the first communication device can perform model training on the first model using the first information and / or model training on at least two other communication devices. In other words, the sender of the first information (e.g., a third communication device) can control the model processing method of the first communication device by sending the first information. For example, the first information can control the first communication device to perform model training on the first model including the first communication device itself (or control the first communication device to train the first model independently or independently), or it can control the first communication device to perform model training on the first model including at least two other communication devices (or control the first communication device to train the first model in conjunction with other communication devices or jointly train the first model), or it can control the first communication device to perform model training on the first model including both the first communication device and at least two other communication devices (or control the first communication device to perform independent model training on the first model in a time-sharing manner and in conjunction with other communication devices). This helps avoid the first communication device training the first model in a fixed training method, thus allowing it to train the first model using a more suitable method and improving model training efficiency.

[0008] In this application, training the first model independently can also be referred to as training the first model independently or training the first model independently.

[0009] In this application, the model (e.g., the first model) may include an AI model, a neural network model, an AI neural network model, a machine learning model, or an AI processing model, etc. In this application, training may also refer to one or more of the following: fine-tuning, adjustment, updating, iteration, or optimization.

[0010] In this application, "training the first model in conjunction with other communication devices" can refer to the first communication device training the first model based on data from other communication devices, rather than limiting the training of the first model to other communication devices. Taking the second communication device as an example, the data from the second communication device can include at least one of the following: the output of the model deployed by the second communication device, the model's parameters, or intermediate results. Therefore, "training the first model in conjunction with other communication devices" can also be described as "training in conjunction with other communication devices." In this process, each of the at least two communication devices trains its own model. For example, the first communication device trains the first model, and the other communication devices train their own models. Furthermore, each individual communication device needs to train its own model based on data from the other communication devices.

[0011] In one possible implementation of the first aspect, the first information includes information that satisfies or at least one of the following: the first information is used to indicate a switch from processing the first model to processing the second model; the first information is used to indicate a switch from processing the second model to processing the first model; or, the first information is used to indicate time information for processing the first model and / or the second model. The sender of the first information (e.g., a third communication device) can control the model processing method of the first communication device on the first model in various ways, which is beneficial to improving the flexibility of the solution implementation. This enables the first communication device to perform model training based on the model processing method desired by the sender of the first information. Optionally, the first communication device can perform model training using the corresponding model processing method based on the instructions of other communication devices, which can reduce the implementation complexity of the first communication device. Furthermore, the number of first communication devices can be one or more, and these one or more first communication devices can be distributed nodes. In the above solution, the first communication device, as a distributed node, can perform model training based on the model processing method indicated by other nodes (e.g., the sender of the first information), which can improve the overall model training efficiency of one or more distributed nodes.

[0012] It should be noted that in this application, the first information may carry an identifier of the first model, or the first information may not carry an identifier of the first model. When the first information does not carry an identifier of the first model, the first model indicated by the first information may be any model deployed on the first communication device or a model being trained on the first communication device.

[0013] The first information can be used to instruct the first communication device to switch the model processing mode. The first communication device can switch the model processing mode according to the first information, which helps to reduce the implementation complexity of the first communication device. It also helps the sender of the first information (e.g., a third communication device) to control the first communication device to switch the model processing mode with less latency.

[0014] Based on the fact that the first information can be used to indicate the timing of the first model processing and / or the second model processing, the first communication device can perform corresponding model processing according to the timing of each model processing method indicated by the first information. This helps to reduce the number of interactions between the sender of the first information (e.g., the third communication device) and the first communication device, thereby saving costs.

[0015] In one possible implementation of the first aspect, the method further includes sending second information indicating a first performance and / or a second performance of the first model. The first performance is determined based on first data from the first communication device, and the second performance is determined based on second data from a second communication device, which is different from the first communication device. This allows the recipient of the second information to determine the performance of the first model processing and / or the performance of the second model processing based on the first and / or second performance.

[0016] Optionally, the aforementioned at least two communication devices and / or other communication devices may include a second communication device. Optionally, the second performance may be determined based on data from the at least two communication devices.

[0017] In one possible implementation of the first aspect, the first information is determined based on first performance and / or second performance. That is, after the first communication device sends the second information, the sender of the first information (e.g., a third communication device) can receive the second information and determine the first information based on the first performance and / or second performance indicated by the second information. This facilitates more accurate control of the model processing method of the first communication device and improves model processing efficiency. For example, if the first performance is superior, the first communication device can perform first model processing (or independent training) based on the first information, eliminating the need to train the first model in conjunction with other communication devices, thereby reducing transmission overhead. If the first performance is inferior, the first communication device can perform second model processing (or joint training) in conjunction with other communication devices (e.g., including a second communication device) based on the first information, improving the performance of the first communication device in training the first model.

[0018] In one possible implementation of the first aspect, the second information is carried on a first resource, which is determined based on at least one of the following: the time of receiving the model processing information of the first model, the start time corresponding to the first model processing, or the end time corresponding to the first model processing. This allows the first communication device to determine the first resource for sending the second information based on at least one of the time information among the time of receiving the model processing information of the first model, the start time corresponding to the first model processing, and the end time corresponding to the first model processing, thereby improving the efficiency of the first communication device in sending the second information. Optionally, it also helps to avoid or reduce the need for the first communication device and the recipient of the second information to negotiate the first resource through additional messages or signaling, thus reducing transmission overhead.

[0019] In this application, "moment" can be replaced with other terms, such as "time node," "starting time domain position," or "ending time domain position." In some examples, "abortion" as mentioned in this application can mean interruption, which may be restarted later, and "termination" can mean end. In some examples, "abortion" as mentioned in this application can be replaced with "termination" or "end." In some examples, "termination" as mentioned in this application can be replaced with "abortion" or "end." In some examples, "end" as mentioned in this application can be replaced with "abortion" or "termination."

[0020] In one possible implementation of the first aspect, the first communication device may further receive third information, which is used to instruct model processing information. This enables the first communication device to process the first model using the model processing information indicated by the third information, thereby improving model processing efficiency.

[0021] Optionally, the receiving time of the model processing information in the first model can be the time when the first communication device receives the third information.

[0022] Optionally, the model processing information indicates at least one of the following: dataset, resources for collecting the dataset, model structure, model parameters, model hyperparameters, or the format of the model parameters.

[0023] Optionally, the first communication device sends its own status information; this status information is used to determine the third information. In this way, the sender of the third information (e.g., the third communication device) can determine a first sub-model that matches the status information of the first communication device, enabling the first communication device to perform model processing based on a model that matches its own capabilities. This avoids model processing failures due to model-capability mismatch, thereby improving the success rate of model processing.

[0024] Optionally, the status information may include computing power information of the first communication device (e.g., the computing power information may indicate one or more of total computing power, used computing power, and idle computing power), storage information (e.g., the storage information may indicate one or more of total storage space, used storage space, and idle storage space), AI performance information (e.g., the AI ​​performance information may indicate one or more of AI service latency and AI service accuracy), communication information (e.g., the communication information may indicate one or more of the antenna information of the first communication device, communication chip information, channel information between the first communication device and other communication devices, latency, throughput, packet loss rate, and load), or one or more of other information.

[0025] In one possible implementation of the first aspect, the first resource includes one or more of the following: a start time-domain position, a number of time-domain units, an end time-domain position, a start frequency-domain position, a number of frequency-domain units, and an end frequency-domain position. This allows the first communication device to more accurately determine the first resource, thereby improving the success rate of receiving the second information.

[0026] As described above, the first information received by the first communication device can be used to indicate the timing information of the first model processing and / or the second model processing, or in other words, the first information is used to indicate the timing information of the first model processing and / or the second model processing. The timing information of the model processing can include the start and / or end and / or pause and / or restart timing information of the model processing. The timing information can be a time node or a condition, and the model processing can be the first model processing and / or the second model processing. In one possible implementation of the first aspect, the timing information of the first model processing includes at least one of the following: the start time node of the first model processing, the end time node of the first model processing, a first condition for pausing the first model processing, or a third condition for restarting the first model processing. The timing information of the second model processing can include at least one of the following: the start time node of the second model processing, the end time node of the second model processing, a second condition for pausing the second model processing, or a fourth condition for restarting the second model processing. This allows the sender of the first information (e.g., a third communication device) to more precisely control the process of the first communication device training the first model, thereby improving the performance of the first communication device in training the first model.

[0027] In one possible implementation of the first aspect, the timing information of the first model processing and / or the second model processing satisfies at least one of the following: the start time node of the first model processing is multiple time nodes; the end time node of the first model processing is multiple time nodes; the start time node of the second model processing is multiple time nodes; the end time node of the first model processing is the same as the start time node of the second model processing; the end time node of the first model processing is different from the start time node of the second model processing; the end time node of the second model processing is the same as the start time node of the first model processing; or, the end time node of the second model processing is different from the start time node of the first model processing. Alternatively, the timing information of the first model processing and / or the second model processing satisfies at least one of the following: the start time node of the first model processing is multiple time nodes; the end time node of the first model processing is multiple time nodes; the start time node of the second model processing is multiple time nodes; or, the end time node of the second model processing is multiple time nodes. And / or, the timing information of the first model processing and / or the second model processing satisfies that the end time node of the first model processing is the same as or different from the start time node of the second model processing. And / or, the timing information for the first model processing and / or the second model processing satisfies that the end time node of the second model processing is the same as or different from the start time node of the first model processing. This facilitates the first communication device in performing multiple switching operations of model processing modes based on the first information, saving transmission overhead.

[0028] In one possible implementation of the first aspect, the first information can be transmitted or carried via a message / signaling. For example, the message / signaling can be downlink control information (DCI), sidelink control information (SCI), radio resource control (RRC) messages, or medium access control control element (MAC-CE), etc. The sender of the first information (e.g., a third communication device) can transmit the first information in multiple ways, which improves the flexibility of the solution implementation. In this application, RRC signaling can replace RRC messages, and DCI can also be called DCI messages.

[0029] A second aspect of this application provides a communication method executed by a third communication device. The third communication device can be a communication equipment (such as a terminal device or network device), or it can be a component of the communication equipment (e.g., a circuit or chip responsible for communication functions, such as a modem chip (also known as a baseband chip), a SoC chip, such as an SoC chip containing a modem core, or a SIP chip, etc.), or it can be a logic module or software capable of implementing all or part of the functions of the communication equipment. In this method, the third communication device sends first information, which instructs a first model to undergo first model processing and / or second model processing. The first model processing includes model training by the first communication device, and the second model processing includes model training by at least two communication devices, including the first communication device. Optionally, the third communication device can send the first information to the first communication device.

[0030] Optionally, the third communication device may determine the first information before sending it. Here, "determine" can mean "generate" or "acquire," etc.

[0031] After the third communication device sends the first information, it facilitates the first communication device to perform model training on the first model based on the first information, and / or model training on at least two other communication devices. In other words, the third communication device can control the model processing method of the first communication device by sending the first information. For example, it can control the first communication device to perform model training on the first model including the first communication device (or control the first communication device to train the first model independently or independently), or it can control the first communication device to perform model training on the first model including at least two other communication devices (or control the first communication device to train the first model in conjunction with other communication devices or control the first communication device to perform joint training on the first model), or it can control the first communication device to perform model training on the first model including the first communication device and model training on at least two other communication devices (or control the first communication device to perform independent model training on the first model in a time-sharing manner and model training in conjunction with other communication devices). This helps avoid the first communication device training the first model in a fixed training method, thus allowing the first communication device to train the first model in a more suitable way and improving model training efficiency.

[0032] In one possible implementation of the second aspect, the first information includes at least one of the following: the first information is used to indicate a switch from processing the first model to processing the second model; the first information is used to indicate a switch from processing the first model to processing the second model; or, the first information is used to indicate time information for processing the first model and / or the second model. The third communication device can control the model processing method of the first communication device on the first model in various ways, which is beneficial to improving the flexibility of the solution implementation.

[0033] In one possible implementation of the second aspect, the third communication device further receives second information indicating a first performance and / or a second performance of the first model, wherein the first performance is determined based on first data from the first communication device, and the second performance is determined based on second data from a second communication device, which is different from the first communication device. This allows the third communication device to determine the performance of the first model processing and / or the performance of the second model processing based on the first and / or second performance.

[0034] Optionally, the aforementioned at least two communication devices and / or other communication devices may include a second communication device. Optionally, the second performance may be determined based on data from the at least two communication devices.

[0035] In one possible implementation of the second aspect, the first information is determined based on first performance and / or second performance. That is, the third communication device can receive the second information and determine the first information based on the first and / or second performance indicated by the second information. This facilitates more accurate control of the model processing method of the first communication device and improves model processing efficiency. For example, if the first performance is superior, the third communication device can control the first communication device to perform first model processing (or independent training) by sending the first information, eliminating the need for it to train the first model in conjunction with other communication devices, thus reducing transmission overhead. If the first performance is inferior, the third communication device can control the first communication device to perform second model processing (or joint training) in conjunction with other communication devices (e.g., including the second communication device) by sending the first information, thereby improving the performance of the first communication device in training the first model.

[0036] In one possible implementation of the second aspect, the second information is carried on a first resource, which is determined based on at least one of the following: the time of receiving the model processing information of the first model, the start time corresponding to the first model processing, or the end time corresponding to the first model processing. This allows the first and third communication devices to determine the first resource for sending the second information based on at least one of the time information among the time of receiving the model processing information of the first model, the start time corresponding to the first model processing, and the end time corresponding to the first model processing. This improves the efficiency of the third communication device in receiving the second information and helps avoid or reduce the need for additional signaling between the first and third communication devices to negotiate the first resource, thus reducing transmission overhead.

[0037] In one possible implementation of the second aspect, the third communication device further transmits third information, which instructs the model processing information. This allows the third communication device to control the first communication device to process the first model according to the model processing information by transmitting the third information, thereby improving model processing efficiency.

[0038] Optionally, the receiving time of the model processing information in the first model can be the time when the first communication device receives the third information.

[0039] Optionally, the model processing information indicates at least one of the following: dataset, resources for collecting the dataset, model structure, model parameters, model hyperparameters, or the format of the model parameters.

[0040] Optionally, the third communication device receives status information from the first communication device; this status information is used to determine the third information. In this way, the third communication device can determine a first sub-model that matches the status information of the first communication device, enabling the first communication device to perform model processing based on a model that matches its own capabilities. This avoids model processing failures due to model-capability mismatch, thereby improving the success rate of model processing.

[0041] Optionally, the status information may include computing power information of the first communication device (e.g., the computing power information may indicate one or more of total computing power, used computing power, and idle computing power), storage information (e.g., the storage information may indicate one or more of total storage space, used storage space, and idle storage space), AI performance information (e.g., the AI ​​performance information may indicate one or more of AI service latency and AI service accuracy), communication information (e.g., the communication information may indicate one or more of the antenna information of the first communication device, communication chip information, channel information between the first communication device and other communication devices, latency, throughput, packet loss rate, and load), or one or more of other information.

[0042] In one possible implementation of the second aspect, the first resource includes one or more of the following: a starting time-domain position, a number of time-domain units, an ending time-domain position, a starting frequency-domain position, a number of frequency-domain units, and an ending frequency-domain position. This allows the first communication device to more accurately determine the first resource, thereby improving the success rate of receiving the second information.

[0043] As described above, the first information sent by the third communication device can be used to indicate the timing information of the first model processing and / or the second model processing, or in other words, the first information is used to indicate the timing information of the first model processing and / or the second model processing. The timing information of the model processing can include the start and / or end and / or pause and / or restart timing information of the model processing. The timing information can be a time node or a condition, and the model processing can be the first model processing and / or the second model processing. In one possible implementation of the second aspect, the timing information of the first model processing includes at least one of the following: the start time node of the first model processing, the end time node of the first model processing, a first condition for pausing the first model processing, or a third condition for restarting the first model processing. The timing information of the second model processing can include at least one of the following: the start time node of the second model processing, the end time node of the second model processing, a second condition for pausing the second model processing, or a fourth condition for restarting the second model processing. This allows the third communication device to more precisely control the process of the first communication device training the first model, improving the performance of the first communication device in training the first model.

[0044] In one possible implementation of the second aspect, the timing information of the first model processing and / or the second model processing satisfies at least one of the following: the start time node of the first model processing is multiple time nodes, the end time node of the first model processing is multiple time nodes, the start time node of the second model processing is multiple time nodes, and the end time node of the second model processing is multiple time nodes. And / or, the timing information of the first model processing and / or the second model processing satisfies that the end time node of the first model processing is the same as or different from the start time node of the second model processing. And / or, the timing information of the first model processing and / or the second model processing satisfies that the end time node of the second model processing is the same as or different from the start time node of the first model processing. This facilitates the third communication device in controlling the first communication device to perform multiple model processing mode switching operations by sending first information, thus saving transmission overhead.

[0045] In one possible implementation of the second aspect, the first information can be transmitted or carried via messages / signaling. For example, the message / signaling can be downlink control information (DCI), sidelink control information (SCI), radio resource control (RRC) messages, or medium access control control element (MAC-CE), etc. The sender of the first information (e.g., a third communication device) can transmit the first information in multiple ways, which helps to improve the flexibility of the solution implementation.

[0046] A third aspect of this application provides a communication device, which is a first communication device, comprising a transceiver unit and a processing unit; the transceiver unit is used to receive first information; the processing unit is used to perform first model processing and / or second model processing on a first model based on the first information, the first model processing including model training of the first communication device, and the second model processing including model training of at least two communication devices, wherein the at least two communication devices include the first communication device.

[0047] In the third aspect of this application, the constituent modules of the communication device can also be used to execute the steps performed in various possible implementations of the first aspect and achieve the corresponding technical effects. For details, please refer to the first aspect, which will not be repeated here.

[0048] A fourth aspect of this application provides a communication device, which is a third communication device. The device includes a transceiver unit and a processing unit. The processing unit is used to determine first information, and the transceiver unit is used to transmit the first information. The first information is used to instruct a first model to undergo first model processing and / or second model processing. The first model processing includes model training of a first communication device, and the second model processing includes model training of at least two communication devices, wherein the at least two communication devices include the first communication device.

[0049] In the fourth aspect of this application, the constituent modules of the communication device can also be used to perform the steps executed in various possible implementations of the second aspect and achieve the corresponding technical effects. For details, please refer to the second aspect, which will not be repeated here.

[0050] A fifth aspect of this application provides a communication device including at least one processor coupled to a memory; the memory is used to store a program or instructions; the at least one processor is used to execute the program or instructions to cause the device to implement the method described in any possible implementation of any of the first to second aspects. Optionally, the communication device may include the memory.

[0051] The sixth aspect of this application provides a communication device including at least one logic circuit and an input / output interface; the logic circuit is used to perform the method as described in any one of the possible implementations of the first to second aspects described above.

[0052] The seventh aspect of this application provides a communication system comprising the aforementioned at least two communication devices and a third communication device, wherein the at least two communication devices include a first communication device, and optionally, the at least two communication devices further include a second communication device.

[0053] An eighth aspect of this application provides a computer-readable storage medium for storing one or more computer-executable instructions, which, when executed by a processor, perform the method as described in any possible implementation of any of the first to second aspects described above.

[0054] The ninth aspect of this application provides a computer program product (or computer program) that, when executed by a processor, performs the method described in any possible implementation of any of the first to second aspects described above.

[0055] The tenth aspect of this application provides a chip or chip system including at least one processor for supporting a communication device in implementing the methods described in any possible implementation of any of the first to second aspects. For example, the chip may be a baseband chip, a modem chip, a SoC chip (such as an SoC chip containing a modem core), a SIP chip, or a communication module, etc.

[0056] In one possible design, the chip or chip system may further include a memory for storing program instructions and data necessary for the communication device. The chip system may be composed of chips or may include chips and other discrete devices. Optionally, the chip system may also include interface circuitry that provides program instructions and / or data to the at least one processor.

[0057] The technical effects of any of the design methods in aspects three through ten can be found in the technical effects of the different design methods in aspects one through two above, and will not be repeated here. Attached Figure Description

[0058] Figures 1a to 1c A schematic diagram of the communication system provided in this application;

[0059] Figures 2a to 2g This is a schematic diagram of the AI ​​processing involved in this application;

[0060] Figure 3-1 An interactive schematic diagram of the communication method provided in this application;

[0061] Figure 3-2 for Figure 2f The diagram shows a distributed node performing training.

[0062] Figure 3-3 schematically shown Figure 2f The process of joint training between node k and node m is shown.

[0063] Figure 4 This schematically illustrates an example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on the timing information of the first model processing and / or the second model processing;

[0064] Figure 5 This schematically illustrates another example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on the timing information of the first model processing and / or the second model processing;

[0065] Figure 6 This schematically illustrates another example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on the timing information of the first model processing and / or the second model processing;

[0066] Figure 7 This illustration shows the process of each distributed node sending the second message.

[0067] Figures 8-10 Examples of communication device 1-1 determining the supervision time nodes of the sub-model are illustrated separately;

[0068] Figure 11 The illustration shows an example of how communication device 1-1 determines the supervision time point of the sub-model and the supervision time point of the complete model;

[0069] Figure 12 This schematically illustrates an example of how communication device 1-1 determines the supervision time points of the complete model;

[0070] Figures 13 to 17 A schematic diagram of the communication device provided in this application. Detailed Implementation

[0071] First, some terms used in the embodiments of this application will be explained to facilitate understanding by those skilled in the art.

[0072] (1) Terminal device: can be a wireless terminal device that can receive network device scheduling and instruction information. The wireless terminal device can be a device that provides voice and / or data connectivity to the user, or a handheld device with wireless connection function, or other processing device connected to a wireless modem.

[0073] Terminal devices can communicate with one or more core networks or the Internet via a radio access network (RAN). Terminal devices can be mobile terminal devices, such as mobile phones (or "cellular" phones), computers, and data cards. For example, they can be portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile devices that exchange voice and / or data with the RAN. Examples include personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablets, and computers with wireless transceiver capabilities. Wireless terminal equipment can also be referred to as a system, subscriber unit, subscriber station, mobile station (MS), remote station, access point (AP), remote terminal, access terminal, user terminal, user agent, subscriber station (SS), customer premises equipment (CPE), terminal, user equipment (UE), mobile terminal (MT), etc.

[0074] By way of example and not limitation, in this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices or smart wearable devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those that focus on a specific type of application function and require the use of other devices such as smartphones, such as various smart bracelets, smart helmets, and smart jewelry for vital sign monitoring.

[0075] Terminals can also be drones, robots, devices in device-to-device (D2D) communication, vehicles to everything (V2X) communication, virtual reality (VR) terminal devices, augmented reality (AR) terminal devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in telemedicine or telehealth services, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc.

[0076] Furthermore, the terminal device can also be a terminal device for a communication system evolved from the fifth generation (5G) communication system (such as 5G Advanced or future communication systems). For example, the form and function of the communication terminal can be further expanded, including but not limited to vehicles, cellular network terminals (integrating satellite terminal functions), drones, and Internet of Things (IoT) devices.

[0077] In this embodiment, the terminal device can also obtain artificial intelligence (AI) services provided by the network device. Optionally, the terminal device can also have AI processing capabilities.

[0078] (2) Network equipment: This can be equipment within a wireless network. For example, network equipment can be a RAN node (or device) that connects terminal devices to the wireless network, and can also be called a base station. Currently, some examples of RAN equipment include: base station, evolved NodeB (eNodeB), gNB (gNodeB) in 5G communication systems, transmission reception point (TRP), evolved Node B (eNB), radio network controller (RNC), Node B (NB), home base station (e.g., home-evolved Node B, or home Node B, HNB), base band unit (BBU), or wireless fidelity (Wi-Fi) access point (AP), etc. In addition, in a network architecture, network equipment can include central unit (CU) nodes, distributed unit (DU) nodes, or RAN equipment including both CU and DU nodes.

[0079] Optionally, RAN nodes can also be macro base stations, micro base stations, indoor stations, relay nodes, donor nodes, or radio controllers in cloud radio access network (CRAN) scenarios. RAN nodes can also be servers, wearable devices, vehicles, or in-vehicle equipment. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU).

[0080] In another possible scenario, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, with different RAN nodes each implementing some of the base station's functions. For example, RAN nodes can be CUs, DUs, CUs (control plane, CP), CUs (user plane, UP), or radio units (RUs). CUs and DUs can be set up separately or included in the same network element, such as a baseband unit (BBU). RUs can be included in radio equipment or radio units, such as remote radio units (RRUs), active antenna units (AAUs), radio heads (RHs), or remote radio heads (RRHs).

[0081] 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 meaning. For example, in an open access network (open RAN, O-RAN, or ORAN) system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. For ease of description, this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0082] Communication between access network devices and terminal devices follows a specific protocol layer structure. This protocol layer may include a control plane protocol layer and a user plane protocol layer. The control plane protocol layer may include at least one of the following: radio resource control (RRC) layer, packet data convergence protocol (PDCP) layer, radio link control (RLC) layer, media access control (MAC) layer, or physical (PHY) layer, etc. The user plane protocol layer may include at least one of the following: service data adaptation protocol (SDAP) layer, PDCP layer, RLC layer, MAC layer, or physical layer, etc.

[0083] The correspondence between network elements and their achievable protocol layer functions in the ORAN system can be found in Table 1 below.

[0084] Table 1

[0085] 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

[0086] Network devices can be other devices that provide wireless communication functions for terminal devices. The embodiments of this application do not limit the specific technology or form of the network device. For ease of description, the embodiments of this application are not limited.

[0087] Network equipment may also include core network equipment, such as the 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 or P-GW) in 4th generation (4G) networks; and access and mobility management function (AMF), user plane function (UPF), or session management function (SMF) in 5G networks. Furthermore, this core network equipment may also include other core network equipment in 5G networks and next-generation networks of 5G networks.

[0088] In this embodiment of the application, the network device may also have network nodes with AI capabilities, which can provide AI services to terminals or other network devices. For example, it may be an AI node, computing node, RAN node with AI capabilities, or core network element with AI capabilities on the network side (access network or core network).

[0089] In this application embodiment, the device for implementing the function of the network device can be the network device itself, or it can be a device capable of supporting the network device in implementing the function, such as a chip system. This device can be disposed within the network device. In the technical solutions provided in this application embodiment, the example of a network device being used to implement the function of the network device is used to describe the technical solutions provided in this application embodiment.

[0090] (3) Configuration and Pre-configuration: In this application, both configuration and pre-configuration are used. Configuration refers to the network device / server sending configuration information or parameter values ​​to the terminal via messages or signaling, so that the terminal can determine communication parameters or resources for transmission based on these values ​​or information. Pre-configuration is similar to configuration; it can be parameter information or parameter values ​​pre-negotiated between the network device / server and the terminal device, parameter information or parameter values ​​specified by standard protocols for use by the base station / network device or terminal device, or parameter information or parameter values ​​pre-stored in the base station / server or terminal device. This application does not limit this.

[0091] Furthermore, these values ​​and parameters can be changed or updated.

[0092] (4) The terms "system" and "network" in the embodiments of this application can be used interchangeably. "Multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, or B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the related objects before and after are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in the embodiments of this application are used to distinguish multiple objects and are not used to limit the order, sequence, priority or importance of multiple objects.

[0093] (5) In the embodiments of this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, which may include sending directly through the air interface or sending indirectly through the air interface by other units or modules. "Receive information from YY" can be understood as the source of the information being YY, which may include receiving directly from YY through the air interface or receiving indirectly from YY through the air interface by other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface.

[0094] In other words, sending and receiving can occur between devices, such as between network devices and terminal devices, or within a device, such as between components, modules, chips, software modules, or hardware modules within the device via buses, wiring, or interfaces.

[0095] It is understandable that information may undergo necessary processing, such as encoding and modulation, between the source and destination, but the destination can understand the valid information from the source. Similar statements in this application can be interpreted in a similar way and will not be elaborated further.

[0096] (6) In the embodiments of this application, "instruction" may include direct instruction and indirect instruction, as well as explicit instruction and implicit instruction. The information indicated by a certain piece of information (as described below, the instruction information) is called the information to be instructed. In the specific implementation process, there are many ways to indicate the information to be instructed, such as, but not limited to, directly indicating the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly indicate the information to be instructed by indicating other information, where there is an association between the other information and the information to be instructed; or it can only indicate a part of the information to be instructed, while the other parts of the information to be instructed are known or pre-agreed upon. For example, the instruction can be implemented by using a pre-agreed (e.g., protocol pre-defined or pre-configured) arrangement order of various information, thereby reducing the instruction overhead to a certain extent. This application does not limit the specific method of instruction. It is understood that for the sender of the instruction information, the instruction information can be used to indicate the information to be instructed; for the receiver of the instruction information, the instruction information can be used to determine the information to be instructed.

[0097] In this application, unless otherwise specified, the same or similar parts between the various embodiments can be referred to each other. In the various embodiments of this application, and the various methods / designs / implementations within each embodiment, unless otherwise specified or logically conflicting, the terminology and / or descriptions between different embodiments and between the various methods / designs / implementations within each embodiment are consistent and can be mutually referenced. The technical features in different embodiments and the various methods / designs / implementations within each embodiment can be combined to form new embodiments, methods, or implementations based on their inherent logical relationships. The following descriptions of the embodiments of this application do not constitute a limitation on the scope of protection of this application.

[0098] This application can be applied to long-term evolution (LTE) systems, new radio (NR) systems, or future communication systems beyond 5G. These communication systems include at least one network device and / or at least one terminal device.

[0099] Please see Figure 1a This is a schematic diagram of the architecture of the communication system 1000 used in an embodiment of this application. Figure 1a As shown, the communication system may include a radio access network (RAN) 100. Optionally, the communication system 1000 may also include a core network 200 and an Internet 300. The RAN 100 includes at least one RAN node (e.g., Figure 1a 110a and 110b, collectively referred to as 110, may also include at least one terminal (such as...). Figure 1aRAN100, denoted as RAN100, comprises RAN nodes 120a-120j, collectively referred to as RAN120. RAN100 may also include other RAN nodes, such as wireless relay equipment and / or wireless backhaul equipment. Figure 1a (Not shown in the image). Terminal 120 connects wirelessly to RAN node 110, and RAN node 110 connects wirelessly or via a wired connection to core network 200. The core network equipment in core network 200 and RAN node 110 in RAN 100 can be independent physical devices, or they can be the same physical device integrating the logical functions of core network equipment and RAN nodes. Terminals can connect to each other, and RAN nodes can connect to each other, via wired or wireless connections.

[0100] by Figure 1a Taking the communication system shown as an example, in addition to performing communication-related services, different devices (including network devices and terminal devices, and / or terminal devices and terminal devices) may also perform AI-related services.

[0101] like Figure 1b As shown, taking a network device as a base station as an example, a base station can perform communication-related services and AI-related services with one or more terminal devices, and different terminal devices can also perform communication-related services and AI-related services.

[0102] like Figure 1c As shown, taking terminal devices including TVs and mobile phones as an example, TVs and mobile phones can also perform communication-related services and AI-related services.

[0103] The technical solution provided in this application can be applied to wireless communication systems (e.g.) Figure 1a , Figure 1b or Figure 1c The system shown, for example, the communication system provided in this application, can incorporate AI network elements to implement some or all AI-related operations. AI network elements can also be called AI nodes, AI devices, AI entities, AI modules, AI models, or AI units, etc. The AI ​​network element can be built into a network element within the communication system. For example, an AI network element can be an AI module built into: access network equipment, core network equipment, cloud server, or operation, administration, and maintenance (OAM) to implement AI-related functions. The OAM can act as the network management system for the core network equipment and / or the access network equipment. Alternatively, the AI ​​network element can also be a network element independently set up in the communication system. Optionally, the terminal or its built-in chip can also include an AI entity to implement AI-related functions.

[0104] Optionally, in communication systems, AI application cases may include, but are not limited to: channel state information (CSI) feedback enhancement, beam management enhancement, positioning accuracy enhancement, network energy saving, load balancing, and mobility optimization. These will be explained below.

[0105] 1. Enhanced CSI feedback

[0106] Channel quality information (CSI) is the channel attribute of a communication link, reported by the terminal device to the network device. By reporting this information, the terminal device can select an appropriate modulation and coding scheme (MCS) to adapt to changing wireless channels. For example, the terminal device might perform channel estimation based on the received channel state information-reference signal (CSI-RS) and then feed back the CSI-RS to the network device. This information serves as input to the network device's model, enabling AI model training. Applying AI to CSI feedback enhancement can reduce overhead, improve accuracy, and enhance predictive capabilities.

[0107] CSI-RS feedback enhancement may include at least one sub-function, such as: CSI compression, CSI prediction, and CSI-RS configuration signaling reduction. CSI compression may further include CSI compression in at least one domain: spatial, time, and frequency.

[0108] 2. Enhanced Beam Management

[0109] Enhanced beam management primarily aims to discover the strongest transmit / receive beam pairs. AI-based sparse beam prediction can improve accuracy. This can be achieved through both network-side and terminal-side AI sparse beam prediction, based on AI training and inference. Taking terminal-side AI sparse beam prediction as an example, the pre-trained AI model on the terminal device can be provided by the network or pre-stored on the terminal device. During training, the network device scans all possible beams and then reports the transmit beam pattern to the terminal device. Once training is complete, the network device only needs to scan a small subset of beams, and the terminal device then feeds back the inference results. AI-based beam management can achieve beam prediction in, for example, the temporal and / or spatial domains, reducing overhead and latency and improving beam selection accuracy.

[0110] Beam management enhancements may include at least one sub-function, such as: beam scan matrix prediction and optimal beam prediction.

[0111] 3. Enhanced positioning accuracy

[0112] In line-of-sight (LOS) or non-line-of-sight (NLOS) scenarios, AI-based positioning can improve positioning accuracy with a smaller number of TRP antennas. Positioning enhancement can include at least one sub-function, such as: positioning enhancement based on access network devices, positioning enhancement based on positioning management function network elements, and positioning enhancement based on terminal devices.

[0113] 4. Network energy saving

[0114] Network energy conservation can be achieved through cell activation / deactivation, load reduction, coverage improvement, or other RAN setting adjustments. AI technology can be used to optimize energy-saving decisions by leveraging data collected within the RAN network. AI algorithms can predict energy efficiency and load status for the next cycle, which can be used to assist in cell activation / deactivation decisions to save energy. Based on the predicted load, the system can dynamically configure energy-saving strategies to maintain a balance between system performance and energy efficiency, and reduce energy consumption.

[0115] 5. Load balancing

[0116] Load balancing can distribute the load evenly between cells and across different areas within a cell, or transfer some traffic from congested cells, or offload users across a single cell, carrier, or access standard, thereby improving network performance. Using AI models to enhance load balancing performance—such as inputting various measurements and feedback from terminal devices and network nodes, as well as historical data—can provide a higher quality user experience and increase system capacity.

[0117] 6. Mobility Management

[0118] Mobility management is a solution that ensures service continuity for mobile devices by minimizing dropped calls, radio link failures (RLFs), unnecessary handovers, and ping-pong effects. AI can enhance mobility management by, for example, reducing the probability of unexpected events, predicting device location / mobility / performance, and routing traffic.

[0119] It should be understood that the definitions of the above technical terms are merely illustrative. For example, as technology continues to develop, the scope of the above definitions may also change, and the embodiments of this application are not intended to limit the scope.

[0120] For example, an AI function may include multiple AI sub-functions.

[0121] Optionally, AI application cases are also called AI application scenarios or AI functions.

[0122] As described above regarding AI application examples, AI can be widely used to improve network performance in areas such as CSI feedback enhancement, beam management, positioning accuracy enhancement, energy saving, mobility enhancement, and load balancing. AI models can typically be deployed on the network side and / or the terminal device side. The training of AI models relies on the collection of training data, which can come from measurements and feedback from the terminal devices.

[0123] The following is a brief introduction to the concepts that may be involved in this application.

[0124] AI can endow machines with human-like intelligence, for example, allowing them to use computer hardware and software to simulate certain intelligent human behaviors. To achieve artificial intelligence, machine learning methods can be employed. In machine learning, machines learn (or train) a model using training data. This model represents the mapping between inputs and outputs. The learned model can be used for reasoning (or prediction), that is, it can be used to predict the output corresponding to a given input. This output can also be called the reasoning result (or prediction result).

[0125] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Unsupervised learning can also be called learning without supervision.

[0126] Supervised learning, based on collected sample values ​​and labels, uses machine learning algorithms to learn the mapping relationship between sample values ​​and labels, and then expresses this learned mapping relationship using an AI model. The process of training the machine learning model is the process of learning this mapping relationship. During training, sample values ​​are input into the model to obtain the model's predicted values, and the model parameters are optimized by calculating the error between the model's predicted values ​​and the sample labels (ideal values). After the mapping relationship is learned, it can be used to predict new sample labels. The mapping relationship learned in supervised learning can include linear or non-linear mappings. Based on the type of label, the learning task can be divided into classification tasks and regression tasks.

[0127] Unsupervised learning relies on collected sample values ​​to discover inherent patterns within the samples themselves. One type of unsupervised learning algorithm uses the samples themselves as supervisory signals, meaning the model learns the mapping relationship from sample to sample; this is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples themselves. Self-supervised learning can be used for signal compression and decompression recovery applications; common algorithms include autoencoders and generative adversarial networks.

[0128] Reinforcement learning, unlike supervised learning, is a type of algorithm that learns problem-solving strategies through interaction with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have explicit "correct" action labels. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust its decision actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user based on the total system throughput feedback from the wireless network, aiming to achieve a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and a better (e.g., optimal) decision action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action." Reinforcement learning training is achieved through iterative interaction with the environment.

[0129] Neural networks (NNs) are a specific model in machine learning techniques. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems rely on extensive expert knowledge to design communication modules, while deep learning communication systems based on neural networks can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0130] The idea behind neural networks comes from the neuronal structure of the brain. For example, each neuron performs a weighted summation of its input values ​​and outputs the result through an activation function.

[0131] like Figure 2a The diagram shown is a schematic representation of a neuron structure. Assume the neuron's input is x = [x0, x1, ..., x...]. n The weights corresponding to each input are w = [w0, w1, ..., w] n ], where n is a positive integer, w i and x i It can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. i As x i The weights are used to assign weights to x. i Weighting is applied. The bias for the weighted sum of the input values ​​is, for example, b. Activation functions can take many forms. Suppose the activation function of a neuron is: y = f(z) = max(0, z), then the output of that neuron is: For example, if the activation function of a neuron is y = f(z) = z, then the output of that neuron is: Here, b can be any possible type, such as a decimal, an integer (e.g., 0, a positive integer, or a negative integer), or a complex number. The activation functions of different neurons in a neural network can be the same or different.

[0132] Furthermore, neural networks generally consist of multiple layers, each of which may include one or more neurons. Increasing the depth and / or width of a neural network can improve its expressive power, providing more powerful information extraction and abstract modeling capabilities for complex systems. The depth of a neural network can refer to the number of layers it includes, and the number of neurons in each layer can be called the width of that layer. In one implementation, a neural network includes an input layer and an output layer. The input layer processes the received input information through neurons and passes the processing result to the output layer, which then obtains the output of the neural network. In another implementation, a neural network includes an input layer, hidden layers, and an output layer. The input layer processes the received input information through neurons and passes the processing result to the hidden layer. The hidden layer calculates the received processing result and passes the result to the output layer or the next adjacent hidden layer, ultimately obtaining the output of the neural network. A neural network may include one hidden layer or multiple sequentially connected hidden layers, without limitation.

[0133] Neural networks, for example, are deep neural networks (DNNs). Depending on how the network is constructed, DNNs can include feedforward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs).

[0134] Figure 2b This is a schematic diagram of a Free-Nearest Neural Network (FNN). A key characteristic of FNNs is that neurons in adjacent layers are completely connected pairwise. This characteristic makes FNNs typically require a large amount of storage space, leading to high computational complexity.

[0135] CNNs are neural networks specifically designed to process data with a grid-like structure. For example, time-series data (e.g., discrete sampling along a time axis) and image data (e.g., two-dimensional discrete sampling) can both be considered grid-like data. CNNs do not use all the input information at once for computation; instead, they use a fixed-size window to extract a portion of the information for convolution operations, which significantly reduces the computational cost of model parameters. Furthermore, depending on the type of information extracted by the window (e.g., people and objects in an image represent different types of information), each window can use different convolution kernels, allowing CNNs to better extract features from the input data.

[0136] Recurrent Neural Networks (RNNs) are a type of neural network that utilizes feedback time-series information. The input to an RNN includes the current input value and its own output value from the previous time step. RNNs are suitable for acquiring temporally correlated sequence features, and are applicable to applications such as speech recognition and channel coding / decoding.

[0137] In the model training process described above, a loss function can be defined. The loss function describes the difference between the model's output value and the ideal target value. The loss function can be expressed in various forms, and there are no restrictions on its specific form. The model training process can be viewed as follows: by adjusting some or all of the model's parameters, the value of the loss function is made to be less than a threshold or to meet the target requirement.

[0138] A model can also be called an AI model, a rule, or other names. An AI model can be considered a specific method for implementing AI functions. An AI model represents the mapping relationship or function between the model's input and output. AI functions can include one or more of the following: data collection, model training (or model learning), model information dissemination, model inference (or model reasoning, inference, or prediction, etc.), model monitoring or model validation, or inference result publication, etc. AI functions can also be called AI (related) operations or AI-related functions.

[0139] The implementation process of the neural network will be described below with reference to the accompanying drawings.

[0140] 1. Fully connected neural network, also known as multilayer perceptron (MLP).

[0141] like Figure 2c As shown, an MLP consists of an input layer (left side), an output layer (right side), and multiple hidden layers (middle). Each layer of an MLP contains several nodes, called neurons. Neurons in adjacent layers are connected pairwise.

[0142] Optionally, considering neurons in two adjacent layers, the output h of a neuron in the next layer is the weighted sum of all neurons x in the previous layer connected to it, processed by an activation function, and can be expressed as:

[0143] h = f(wx + b).

[0144] Where w is the weight matrix, b is the bias vector, and f is the activation function.

[0145] Alternatively, the output of the neural network can be recursively expressed as:

[0146] y = f z (w z f z-1 (…)+b z ).

[0147] Where z is the index of the neural network layer, z is greater than or equal to 1 and z is less than or equal to Z, where Z is the total number of layers in the neural network.

[0148] In other words, a neural network can be understood as a mapping from an input data set to an output data set. Neural networks are typically initialized randomly; the process of obtaining this mapping from random values ​​w and b using existing data is called training the neural network.

[0149] Optionally, the training method involves using a loss function to evaluate the output of the neural network.

[0150] like Figure 2d As shown, the error can be backpropagated, and the neural network parameters (including w and b) can be iteratively optimized using gradient descent until the output of the loss function reaches its minimum value. Figure 2d The term "relative advantage (e.g., optimal advantage)" is used. This is understandable. Figure 2d The neural network parameters corresponding to the "better points (e.g., the best points)" in the data can be used as neural network parameters in the trained AI model information.

[0151] Alternatively, the gradient descent process can be represented as:

[0152]

[0153] Where θ represents the parameters to be optimized (including w and b), L is the loss function, and η is the learning rate, controlling the step size of gradient descent. This represents the differentiation operation. This indicates taking the derivative of θ with respect to L.

[0154] Alternatively, the backpropagation process can utilize the chain rule for partial derivatives.

[0155] like Figure 2e As shown, the gradient of the parameters in the previous layer can be recursively calculated from the gradient of the parameters in the next layer, and can be expressed as:

[0156]

[0157] Among them, w ij Let s be the weight of the connection between node j and node i. i The weighted sum of the inputs at node i.

[0158] 2. Federated learning (FL).

[0159] The concept of federated learning effectively addresses the current challenges in the development of artificial intelligence. While fully protecting user data privacy and security, it enables various edge devices and central servers to collaborate efficiently to complete the model's learning task.

[0160] like Figure 2f As shown, the FL architecture is currently the most widely used training architecture in the FL field, and the FedAvg algorithm is the foundational algorithm of FL. The FedAvg algorithm flow is roughly as follows:

[0161] (1) Initialize the model to be trained at the center end. And broadcast it to all clients.

[0162] (2) In the t∈[1,T] round, the client k∈[1,K] is based on the local dataset. For the received global model Perform E epochs of training to obtain the local training results. Report it to the central node. Figure 2f In the example shown, the local training results sent by distributed nodes n, k, and m are denoted as G, respectively. n G k G m .

[0163] (3) The central node collects local training results from all (or some) clients. Assume the set of clients uploading local models in round t is... The central server will use the number of samples from the corresponding client as weights to calculate the new global model. The specific update rule is as follows: Then the central end will send the latest version of the global model. The broadcast is sent to all clients for a new round of training.

[0164] (4) Repeat steps (2) and (3) until the model finally converges or the number of training rounds reaches the upper limit.

[0165] Optionally, in addition to reporting the local model, the client can also... It can also train local gradients The central node averages the local gradients reported by all clients and updates the global model based on this average gradient.

[0166] As can be seen in the FL framework, the dataset resides on distributed nodes (such as clients). These distributed nodes collect their local datasets, perform local training, and report the local results (model or gradients) to the central node. The central node itself may not have a dataset; it can be responsible for fusing the training results from the distributed nodes to obtain a global model, which is then distributed back to the distributed nodes.

[0167] 3. Decentralized learning.

[0168] like Figure 2g The diagram shows a fully distributed system without a central node. The design goal f(x) of a decentralized learning system is generally the goal f of each node. i The mean of (x), i.e. Where n is the number of distributed nodes, and x is the parameter to be optimized; in machine learning, x is the parameter of the machine learning model (such as a neural network). Each node utilizes local data and its local target f. i (x) Calculate the local gradient Then it is sent to its communicatively reachable neighboring nodes. Upon receiving the gradient information from its neighbor, any node can update the parameters x of its local model according to the following formula:

[0169]

[0170] in, This represents the parameters of the local model after the (k+1)th update (k is a natural number) in the i-th node. This represents the parameters of the local model for the i-th node after the k-th update (if k is 0, then it represents...). (where α is the parameter of the local model of the i-th node that is not involved in the update) k N represents the tuning coefficient. i It is the set of neighboring nodes of node i, |N i | represents the number of elements in the set of neighboring nodes of node i, that is, the number of neighboring nodes of node i. Through information interaction between nodes, the decentralized learning system will eventually learn a unified model.

[0171] The technical solution provided in this application can be applied to communication systems (e.g.) Figure 1a or Figure 1b or Figure 1c In a communication system (as shown in the diagram), communication nodes typically possess both signal transmission and reception capabilities and computational capabilities. Taking a network device with computational capabilities as an example, the network device's computational capabilities primarily provide computing power support for signal transmission and reception capabilities (e.g., processing signals for transmission and reception) to enable the network device to perform communication tasks with other communication nodes.

[0172] With the development of communication technology, communication equipment in communication systems can now perform not only traditional communication services but also other new types of services, such as artificial intelligence (AI) services. Generally, systems capable of handling AI services, such as communication systems, can also be called AI systems.

[0173] Currently, communication devices can serve as participating nodes in AI systems, providing their own computing power and data. For example, a communication device can perform multiple model processing steps (such as model training, model updating, or model fine-tuning) on ​​a single model based on local data to obtain another model. One possible approach is to increase the complexity of the model (e.g., increasing the number of parameters or the number of neural network layers), which can effectively improve the model's performance.

[0174] However, as the complexity of the models increases, the processing time may become excessively long, and limitations in communication equipment resources (such as computing power and storage resources) may even prevent the processing from executing. Therefore, optimizing the model processing is a pressing technical problem that needs to be solved.

[0175] As an example, when communication devices are trained based on generic data, the model may become overfitted, becoming too closely matched to the training dataset, thus reducing its generalization ability during inference. In other words, the model may fail to adapt to the actual scenario and / or task during inference. For instance, the environmental map used during training for a radio frequency map (RF MAP) model may differ from the actual environmental map, and the location of the communication device used during training may also differ from the actual location of the communication device. Collecting training data according to the actual physical environment would incur significant data acquisition overhead, necessitating further fine-tuning or adjustments to the model based on the actual physical environment. In some scenarios, the resources (e.g., computing power or storage resources) of a single communication device allow it to train independently, which would degrade model performance. Conversely, in other scenarios, while the resources of a single communication device allow it to train independently, joint training with other communication devices increases the number of interactions, leading to higher transmission overhead.

[0176] To address the aforementioned problems, this application provides a communication method and related apparatus, which will be described in detail below with reference to the accompanying drawings.

[0177] Please see Figure 3-1 This is a schematic diagram of an implementation of the communication method provided in this application, which includes steps S301 and S302.

[0178] It should be noted that in the following text, Figure 3-1 The present application uses communication device 1-1 and communication device 2 as examples of the execution subjects of the interaction to illustrate the method, but this application does not limit the execution subjects of the interaction.

[0179] For example, a communication device can be a communication equipment (such as a terminal device or a network device), or a chip, baseband chip, modem chip, SoC chip (such as an SoC chip containing a modem core), SIP chip, communication module, chip system, processor, logic module or software in a communication equipment.

[0180] As an example, the communication device 1-1 can be a terminal device and the communication device 2 can be a network device.

[0181] As another example, communication device 1-1 can be a network device, and communication device 2 can be a terminal device.

[0182] As another example, both communication device 1-1 and communication device 2 are network devices.

[0183] It should be understood that the aforementioned network equipment can be access network equipment or ORAN equipment (including at least one of O-CU, O-DU, and O-RU).

[0184] As another example, both communication device 1-1 and communication device 2 are terminal devices, i.e. Figure 3-1 The proposed solution can be applied to sidelink communication scenarios.

[0185] Communication device 1-1 can be the first communication device described above, or it can be one or more communication devices in a set of communication devices. The set of communication devices may also include a communication device 1-2, which is different from communication device 1-1; communication device 1-2 can be the second communication device described above. The at least two communication devices, including the first communication device, described above can be all or some of the communication devices in the set of communication devices. The communication devices in the set of communication devices can be distributed nodes, where different distributed nodes can perform distributed model processing in a serial, parallel, or combined serial and parallel manner; this is not limited here. For example, the set of communication devices may include... Figure 2f The distributed nodes n, k, and m are shown. Communication devices 1-1 and 1-2 can be... Figure 2f The different distributed nodes among the multiple distributed nodes shown, and the at least two communication devices including the first communication device described above, may include Figure 2f All or some of the distributed nodes in the system. Optionally, the set of communication devices may include more or fewer communication devices.

[0186] Communication device 2 can be the third communication device described above. Communication device 2 can be a distributed node or a central node; this is not limited here. For example, communication device 2 can be... Figure 2f The central node is shown. In this application, communication device 2 can also be called a training supervision node, which is used to supervise the training process of communication devices in the communication device set.

[0187] Optionally, for any two different communication devices (i.e., any two different distributed nodes) in the communication device set, they may process different models to obtain model processing result information (for example, the central node can obtain the corresponding model processing result information obtained by multiple distributed nodes processing different models, which is beneficial to improving the processing efficiency of different models), or they may process the same model to obtain model processing result information (for example, the central node can obtain the corresponding model processing result information obtained by multiple distributed nodes processing the same model, which is beneficial to the central node to subsequently schedule nodes with better performance for processing based on the model processing result information, so as to improve processing efficiency). This is not limited here. For example, the models on different communication devices in the above communication device set can be different models or different sub-models within the same model (or a large model), or the same model or the same sub-model within the same model (or a large model).

[0188] S301, Communication device 2 sends first information, and correspondingly, communication device 1-1 receives the first information;

[0189] S302, the communication device 1-1 performs first model processing and / or second model processing on the first model based on the first information, wherein the first model processing includes model training of the first communication device, and the second model processing includes model training of at least two communication devices, wherein the at least two communication devices include the first communication device.

[0190] In this application, the model (e.g., the first model) may include an AI model, a neural network model, an AI neural network model, a machine learning model, or an AI processing model, etc.

[0191] In this application, the model (e.g., the first model) may include an AI model, a neural network model, an AI neural network model, a machine learning model, or an AI processing model, etc. In this application, model training may also refer to one or more of the following: fine-tuning, adjustment, updating, iteration, and optimization.

[0192] After receiving the first information, the communication device 1-1 can perform first model processing, including model training by the communication device 1-1, and / or second model processing, including model training by at least two communication devices, on the first model based on the first information.

[0193] The first model processing performed by communication device 1-1 on the first model can also be referred to as communication device 1-1 independently training the first model, or communication device 1-1 independently training the first model. The second model processing performed by communication device 1-1 on the first model can also be referred to as communication device 1-1 jointly training the first model with other communication devices, or communication device 1-1 jointly training the first model. In this application, the first model processing can be referred to as independent training or independent model training, and the second model processing can be referred to as joint training or joint model training.

[0194] Figure 3-2 by Figure 2f The example shown illustrates how distributed node n (hereinafter referred to as node n) performs independent training based on the first received information, and distributed node k (hereinafter referred to as node k) and distributed node m (hereinafter referred to as node m) perform joint training. Assume that the models deployed on nodes n, k, and m are model n, model k, and model m, respectively. The following section combines... Figure 3-2 Examples of independent training and joint training are given.

[0195] During the independent training of model n by node n, the data n (or samples n) used to train model n can be data specified by communication device 2. Communication device 2 can instruct node n on how to acquire data n, or node n can acquire data n itself. This data n can be stored locally or remotely on node n, and can be prepared before node n performs multiple rounds of independent iterative training on model n. Hereinafter, this data n will be referred to as local data n. Node n inputs local data n into model n to obtain output result n, and then can backpropagate the error based on the output result to update the parameters of model n. That is to say, during the independent training of model n by node n, there is no need for interaction with other nodes to jointly train data. Joint training data includes, for example, at least one of the model's output result, model parameters, and intermediate results.

[0196] During joint training of nodes k and m, node k is the predecessor node and node m is the successor node. When node k trains model k in conjunction with node m, the data k (or sample k) used to train model k can be data specified by communication device 2. Communication device 2 can instruct node k on how to acquire data n, or node k can acquire data n itself. This data k can exist locally or remotely on node k, and it can be prepared before node k performs multiple independent iterative training rounds on model k. Hereinafter, this data k is referred to as local data k. Node k inputs local data k into model k to obtain output result k, and then sends the output result k to node m, instead of directly updating the parameters of model k based on the backpropagation error of the output result k. When node m trains model m in conjunction with node k, the data m (or sample m) used to train model m includes the output result k sent by node k. Data m can also include local or remote data of node m. This data can be data specified by communication device 2, which can instruct node m on how to acquire data m, or node m can acquire data m itself. After node m obtains output result m based on output result k and / or data m, it backpropagates the error of output result m to update the parameters of model m and sends gradient update information to node k. Node k can update the parameters of model k based on output result k or output result k and the received gradient update information. In other words, during the joint training of their models, nodes k and m need to exchange the model's output results and gradient update information. Specifically, model k is jointly trained by nodes k and m, and model m is jointly trained by nodes k and m. Optionally, the information exchanged between nodes m and k may include model parameters and / or intermediate results, and the model parameters may include gradient update information.

[0197] In this application, the gradient update information can be a gradient accumulation factor. For example, node k can determine the gradient of model k or update model k based on the output of this layer (e.g., output data k) and the gradient accumulation factor of the previous time (e.g., the gradient accumulation factor sent by node m). Correspondingly, node m can send the gradient accumulation factor only to node k.

[0198] Communication device 1-1 can be Figure 3-2 The communication device 2 can be any of the nodes n, k, or m shown. Figure 2f The central node or distributed node in the process. Figure 3-2 Taking the example of a subset of nodes in a set of communication devices performing independent training while other nodes perform joint training, optionally, all nodes in the set of communication devices can train the model using the same model processing method, for example, all performing independent training or all performing joint training.

[0199] During the process of jointly training data, nodes k and m can communicate directly or forward data through other communication devices. Figure 3-3 Taking the example of nodes k and m forwarding joint training data through communication device 2, with communication device 2 as the central node, the process of nodes k and m performing joint training is illustrated. Figure 3-3 The process shown includes steps 1 through 6.

[0200] In steps 1 and 2, the central node sends the first information to node k and node m respectively. The first information is used to instruct the joint training to be performed.

[0201] After receiving the first information, node k can input data k into model k based on the first information and obtain the output result k.

[0202] Step 3: Node k sends the output result k to the central node;

[0203] Step 4: The central node sends the output result k to node m;

[0204] After receiving the first information and the output result k, node m can obtain the output result m based on the output result k, and backpropagate the error based on the output result m to update the parameters of model m.

[0205] Step 5: Node m sends the gradient update information of model m to the central node;

[0206] Step 6: The central node sends the gradient update information of model m to node k;

[0207] After receiving the gradient update information from model m, node k can update the parameters of model k based on the gradient update information.

[0208] This application does not limit the order of steps 1 and 2, as long as step 2 is executed before step 3 and step 1 is executed before step 4.

[0209] Communication device 2 can control the model processing method of communication device 1-1 by sending first information to communication device 1-1. For example, the first information can control communication device 1-1 to perform independent training, joint training, or time-sharing independent and joint training on the first model. This helps avoid communication device 1-1 training the first model in a fixed training method, thereby enabling communication device 1-1 to train the first model in a more suitable way and improve model training efficiency.

[0210] Optionally, the first information can be transmitted or carried via messages / signaling. For example, the message / signaling can be downlink control information (DCI), sidelink control information (SCI), radio resource control (RRC) messages, or medium access control control element (MAC-CE), etc. The communication device can transmit the first information in multiple ways, which helps to improve the flexibility of the solution implementation.

[0211] exist Figure 3-1 In one possible implementation of the method shown, the first information sent by the communication device 2 in step S301 may include at least one of the following: the first information is used to indicate a switch from processing the first model to processing the second model; the first information is used to indicate a switch from processing the second model to processing the first model; the first information is used to indicate the timing information for processing the first model and / or the second model. The communication device 2 can control the model processing method of the communication device 1-1 on the first model in various ways, which is beneficial to improving the flexibility of the solution implementation. In some examples, the timing information for processing the first model and / or the second model can also be predefined or preconfigured in the communication device 1-1, without needing to be indicated by the first information.

[0212] The first information can be used to instruct communication device 1-1 to switch model processing modes. Communication device 1-1 can switch model processing modes according to the first information, which helps to reduce the implementation complexity of communication device 1-1 and also helps communication device 2 to control communication device 1-1 to switch model processing modes with less latency.

[0213] Based on the first information, which can be used to indicate the timing of the first model processing and / or the second model processing, the communication device 1-1 can perform the corresponding model processing according to the timing of each model processing method indicated by the first information. This helps to reduce the number of interactions between the communication device 2 and the communication device 1-1, thereby saving costs.

[0214] Optionally, the communication device 1-1 can train the first model according to different model processing methods at different stages of training based on the first information. For example, the first model can be trained independently at the beginning and then jointly trained later. This is beneficial to combine the advantages of independent training and joint training, reducing overhead while ensuring training performance.

[0215] exist Figure 3-1In one possible implementation of the method shown, the timing information for the first model processing and / or the second model processing includes at least one of the following: the start time node of the first model processing, the end time node of the first model processing, the start time node of the second model processing, the end time node of the second model processing, determining a first condition for pausing the first model processing, determining a second condition for pausing the second model processing, determining a third condition for restarting the first model processing, and determining a fourth condition for restarting the second model processing. This allows the communication device to more precisely control the process of the communication device 1-1 training the first model, thereby improving the performance of the communication device 1-1 in training the first model.

[0216] In this application, "pause" can be used as a deactivation alternative, and "restart" can be used as an activation alternative.

[0217] In one possible implementation, the end time of the first model processing is the same as the start time of the second model processing. Figure 4 This schematically illustrates an example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on timing information of the first model processing and / or the second model processing. For example... Figure 4 As shown, communication device 1-1 determines, based on the time information of the first model processing and / or the second model processing, that training the first model begins at time T4-1, and the training method is independent training; that is, the start time of training the first model is the start time of independent training. Communication device 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, that independent training of the first model ends at time T4-2, and joint training of the first model begins; that is, the start time of independent training is the end time of independent training. Communication device 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, that training of the first model ends at time T4-3; that is, the time when communication device 1-1 ends training the first model (or the end time of training the first model) is the end time of joint training. Assume the time interval between time T4-1 and time T4-2 is t4-1, the time interval between time T4-3 and time T4-2 is t4-2, and the time interval between time T4-1 and time T4-3 is t4.

[0218] Optionally, the time information for the first model processing and / or the second model processing indicated by the first information may include timestamps T4-1, T4-2, and T4-3. Alternatively, the time information for the first model processing and / or the second model processing indicated by the first information may include a portion of timestamps T4-1, T4-2, and T4-3, and the communication device 1-1 may also predefine or preconfigure at least one duration t4-1, t4-2, and t4, allowing the communication device 1-1 to determine the time based on the predefined or preconfigured duration and the time information indicated by the first information. Figure 4 The information refers to various time points. For example, the first information may indicate time T4-1. Communication device 1-1 has predefined or preconfigured durations t4-1 and t4-2. Communication device 1-1 can determine the end time of independent training and the start time of joint training (i.e., time T4-2) based on time T4-1 and duration t4-1. Communication device 1-1 can also determine the end time of joint training and the end time of training the first model (i.e., time T4-3) based on the determined time T4-2 and duration t4-2. In this application, the duration can also be referred to as the time offset.

[0219] Figure 4 Taking the example of communication device 1-1 performing independent training followed by joint training based on the time information of the first model processing and / or the second model processing, according to different training stages, communication device 1-1 may perform joint training first and then independent training. Optionally, the end time node of the second model processing and the start time node of the first model processing may be the same.

[0220] In one possible implementation, the end time of the first model processing is different from the start time of the second model processing. Figure 5 This schematically illustrates another example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on timing information of the first model processing and / or the second model processing. For example... Figure 4As shown, communication device 1-1 determines, based on the timing information of the first model processing and / or the second model processing, to begin training the first model at time T5-1, and the training method is independent training; that is, the start time of training the first model is the start time of independent training. Communication device 1-1 can also determine, based on the timing information of the first model processing and / or the second model processing, to pause the independent training of the first model at time T5-2. Communication device 1-1 can also determine, based on the timing information of the first model processing and / or the second model processing, to restart the joint training of the first model at time T5-3. Communication device 1-1 can also determine, based on the timing information of the first model processing and / or the second model processing, to end the joint training of the first model at time T5-4, and the end time of joint training is the end time of training the first model. Assume the time interval between time T5-1 and time T5-2 is t5-1, the time interval between time T5-3 and time T5-2 is t5-2, the time interval between time T5-4 and time T5-3 is t5-3, and the time interval between time T5-1 and time T5-4 is t5.

[0221] Optionally, the time information for the first model processing and / or the second model processing indicated by the first information may include at least one of time information, duration information, and conditions. Optionally, the time information for the first model processing and / or the second model processing indicated by the first information may include times T5-1, T5-2, T5-3, and T5-4. Alternatively, optionally, the time information for the first model processing and / or the second model processing indicated by the first information may include a portion of times T5-1, T5-2, T5-3, and T5-4, and the communication device 1-1 may also predefine or preconfigure at least one duration t5-1, t5-2, t5-3, and t5. Alternatively, optionally, the time information for the first model processing and / or the second model processing indicated by the first information may include a portion of times T5-1, T5-2, T5-3, and T5-4 and a portion of durations t5-1, t5-2, t5-3, and t5. Time information not indicated by the first information may be predefined or preconfigured. Alternatively, the timing information for the first model processing and / or the second model processing indicated by the first information may include T5-1, a first condition for determining the pause of independent training, a fourth condition for determining the restart of joint training, and T5-4. Alternatively, T5-1 or T5-4 may be excluded, and T5-1 or T5-4 may be determined based on a predefined or preconfigured duration. Alternatively, the timing information for the first model processing and / or the second model processing indicated by the first information may include T5-1, a first condition for determining the pause of independent training, T5-3, and T5-4. Alternatively, at least one of T5-1, T5-3, and T5-4 may be excluded, and at least one of T5-1, T5-3, and T5-4 may be determined based on a predefined or preconfigured duration.

[0222] This application does not limit the specific content of the aforementioned first to fourth conditions. For example, if the first condition for pausing independent training is determined to be a performance threshold, then after the first communication device begins independent training of the first model, it determines T5-2 based on this performance threshold and the performance of the first model. For instance, if the first condition is that the performance of the first model is better than the performance threshold, then when the performance of the first model is better than the performance threshold, the independent training of the first model is automatically paused. In other words, communication devices in the communication device set can initially train the model independently, and then perform joint training. Communication devices with better performance can pause training, rest for a period of time, and then resume joint training with other communication devices. This helps to reduce training computation and power overhead while ensuring training performance.

[0223] The preceding text uses the example of communication device 1-1 pausing or restarting model training based on conditions or times indicated by the first information. Optionally, at least one of the aforementioned first to fourth conditions is predefined or pre-configured in communication device 1-1. Optionally, during the independent or joint training of the first model, communication device 1-1 can pause independent training, pause joint training, or restart independent or joint training according to the instructions issued by communication device 2. For example, after independently training the first model, communication device 1-1 can pause independent training according to the instructions issued by communication device 2, restart joint training according to the instructions issued by communication device 2, and begin joint training with other communication devices in the communication device set.

[0224] Figure 5 Taking the example of communication device 1-1 performing independent training followed by joint training based on the time information of the first model processing and / or the second model processing, according to different training stages, communication device 1-1 may perform joint training first and then independent training. Optionally, the end time node of the second model processing may be different from the start time node of the first model processing.

[0225] In one possible implementation, the timing information for the first model processing and / or the second model processing satisfies at least one of the following: the start time node of the first model processing is multiple time nodes, the end time node of the first model processing is multiple time nodes, the start time node of the second model processing is multiple time nodes, or the end time node of the second model processing is multiple time nodes. The communication device 1-1 can repeatedly perform independent training and joint training based on the first information. This facilitates the communication device 1-1 in performing multiple switching operations of model processing modes based on the first information, saving transmission overhead.

[0226] like Figure 4 and Figure 5 For example, the end time of the first model processing may be the same as or different from the start time of the second model processing. Taking the case where the end time of the first model processing is the same as and the start time of the second model processing is the same as both, as an example... Figure 6 This schematically illustrates another example of the timing of the first model processing and / or the second model processing determined by the communication device 1-1 based on timing information of the first model processing and / or the second model processing. For example... Figure 6As shown, the communication device 1-1 starts training the first model at time T6-1 based on the time information of the first model processing and / or the second model processing, and the training method is independent training. Then, at time T6-2, the independent training ends and joint training begins. Then, at time T6-3, the joint training ends and independent training begins. Then, at time T6-4, the independent training ends and joint training begins. Then, at time T6-5, the joint training ends, that is, the training of the first model ends.

[0227] like Figure 6 As shown, the time nodes for communication device 1-1 to perform the first model processing and / or the second model processing may include multiple start time nodes (or start times) for independent training, namely T6-1 and T6-3; multiple end time nodes (or end times) for independent training, namely T6-2 and T6-4; multiple start time nodes for joint training, namely T6-2 and T6-4; and multiple end time nodes for joint training, namely T6-3 and T6-5. Figure 6 As shown, the duration of the communication device 1-1 performing the first model processing and / or the second model processing may include the duration between adjacent time nodes, namely t6-1, t6-2, t6-3 and t6-4, and may also include the time interval between time T6-1 and time T6-5, namely t6.

[0228] Optionally, the timing information for the first model processing and / or the second model processing indicated by the first information may include... Figure 6 All time points shown. Alternatively, the time information for the first model processing and / or the second model processing indicated by the first information may include... Figure 6 The time points shown are a subset of the time points. Figure 6 All or part of the duration is predefined or preconfigured in communication device 1-1, and communication device 1-1 can determine the duration based on the predefined or preconfigured duration and the time node indicated by the first information. Figure 6 The start and end times of each training node are shown.

[0229] Figure 6 Taking the example of communication device 1-1 processing time information based on the first model and / or the second model, and performing independent training followed by joint training according to different training stages, communication device 1-1 may perform joint training first and then independent training.

[0230] exist Figure 3-1 In one possible implementation of the method shown, the method may further include S303.

[0231] S303, Communication device 1-1 sends the second information, and correspondingly, communication device 2 receives the second information;

[0232] Optionally, communication device 1-1 can send second information, and communication device 2 can receive the second information, wherein the second information is used to indicate the first performance of the first model and / or the second performance of the first model. This allows communication device 2 to determine the processing performance of the first model and / or the processing performance of the second model based on the first and / or second performance.

[0233] In the following text, the information used to indicate the first performance will be referred to as the performance information of the sub-model, and the information used to indicate the second performance will be referred to as the performance information of the complete model. The second information may include the performance information of the sub-model and / or the performance information of the complete model.

[0234] Optionally, the aforementioned at least two communication devices and other communication devices may include communication device 1-2. Optionally, the second performance may be determined based on data from at least two communication devices. Communication device 1-1 may determine the first performance based on first data from communication device 1-1, or determine the second performance based on second data from communication device 1-2, or determine the first performance based on first data from communication device 1-1 and determine the second performance based on second data from communication device 1-2.

[0235] Figure 7 The solid lines with arrows represent the process of each distributed node sending performance information of the sub-model. Communication device 1-1 can be... Figure 7 For any of the distributed nodes shown, communication device 2 can be... Figure 7 The central node is shown. (Reference) Figure 7 During the initial performance monitoring process, the distributed nodes, referring to the solid line marked 1-1, can input local data into their own models and obtain output data. Then, referring to the solid line marked 1-2, the distributed nodes can send the performance information of their sub-models to the central node. Optionally, the performance information sent by the distributed nodes may include the output data of their own models.

[0236] Communication device 2 can instruct a portion of the communication devices in the communication device set to measure ground truth related information. Optionally, communication device 2 can instruct communication device 1-1 to measure ground truth. The performance information of the sub-model transmitted by communication device 1-1 may also include measurement data 1 used to measure ground truth or ground truth 1 obtained based on measurement data 1. After receiving the performance information of the sub-model, communication device 2 automatically determines the first performance of the first model based on the performance information of the sub-model.

[0237] Optionally, during the process of determining the first performance, communication device 1-1 can obtain the first performance of the first model (denoted as performance index 1) based on the output data of the first model and the baseline truth 1, such as the accuracy of the first model. Referring to the solid lines marked with numbers 1-2, the performance information of the sub-models sent by the distributed nodes to the central node may include the performance index 1 determined by themselves.

[0238] The second data of communication device 1-2 upon which communication device 1-1 relies for supervising the second performance can be the output data obtained by communication device 1-2 after inputting the output data of the first model into its own model. Alternatively, the second data can be the output data obtained by communication device 1-2 after inputting the output data of the model on communication device 1-3 or the local data of communication device 1-2 into its own model. Here, communication device 1-1, communication device 1-2, and communication device 1-3 can be different communication devices within a set of communication devices.

[0239] Figure 7 The process of node m sending the performance information of the complete model is represented by a dashed line with an arrow. Communication device 1-1 can be... Figure 7 The node m shown can be connected to communication devices 1-2. Figure 7 As shown in the diagram, node k and communication device 2 can be... Figure 7 The central node is shown. (Reference) Figure 7 During the process of determining the second performance characteristic, referring to the dashed line marked 2-1, node m can input local data into model n to obtain the output data of model n (denoted as output data n). Node n sends output data n to node k to obtain the output data of model k (denoted as output data k). Then, referring to the dashed line marked 2-2, node k inputs output data k into model k to obtain the output data of model k (denoted as output data k), and sends output data k to node m. Then, referring to the dashed line marked 2-3, node m inputs output data k into model m to obtain the output data of model m (denoted as output data m). Finally, node m sends the complete model performance information to the central node. The complete model performance information may include output data m.

[0240] Optionally, node m can obtain baseline truth 2 based on measurement data 2. The performance information of the complete model sent by node m to the central node may also include measurement data 2 or baseline truth 2. The central node can determine the second performance (or the performance of the complete model) of model m based on the performance information of the complete model.

[0241] Alternatively, assuming that model m on node m is the last part of the complete model trained by the communication device set, since the output data m of model m is the output data of the complete model, node m can determine the second performance (or the performance of the complete model) of model m based on the baseline truth 2 and the output data m. Referring to the dashed lines marked 2-3, the performance information of the complete model sent by node m to the central node can include the second performance (or the performance of the complete model) of model m.

[0242] This application does not limit the first model on communication device 1-1 to the final part of the complete model. For example, communication device 1-1 and communication device 1-2 are respectively... Figure 7 Taking nodes k and m as an example, during the process of determining the second performance, communication device 1-1 can send the output data k of model k to node m. Node m inputs the received output data k into model m to obtain the output data m of model m. Afterwards, node m sends the output data m back to node k. The second data of communication device 1-2 (i.e., node m) may include the output data m of communication device 1-2 (i.e., node m).

[0243] Optionally, after receiving the second information, the communication device 2 can determine the first information in step S301 based on the second information. That is, after the communication device 1-1 sends the second information, the communication device 2 can receive the second information and determine the first information based on the first performance and / or second performance indicated by the second information. This is beneficial for more accurate control of the model processing method of the communication device 1-1 and improves the model processing efficiency.

[0244] refer to Figure 3-3 In step 0, optionally, nodes k and m can send the second information to the central node respectively.

[0245] For example, if the first performance is superior, communication device 1-1 can perform first model processing (or independent training) based on the first information, eliminating the need for it to train the first model in conjunction with other communication devices, thereby reducing transmission overhead. If the first performance is inferior, communication device 1-1 can perform second model processing (or joint training) in conjunction with other communication devices (e.g., including communication device 1-2) based on the first information, thereby improving the performance of communication device 1-1 in training the first model.

[0246] Continue to refer to Figure 7Referring to the solid lines marked 1-2, the central node can receive secondary information reported by each distributed node, indicating the first performance, which helps determine the performance of node n independently training model n, node k independently training model k, and node m independently training model m. Referring to the solid lines marked 2-3, the central node can receive secondary information reported by node m, indicating the second performance, which helps determine the performance of the model jointly trained by nodes n, k, and m (or the performance of the complete model). The central node can instruct each node to perform independent training, or instruct all nodes to perform joint training, or instruct a subset of nodes to perform joint training while the others perform independent training, based on the first performance of each node and the second performance of the complete model. For example, the central node can instruct nodes with performance above a threshold and nodes with performance below a threshold to perform joint training, or instruct nodes with better performance to help nodes with poorer performance during training.

[0247] It is evident that communication device 2 supervises the performance of the complete model and the model on the communication device. By combining the supervision results or performance indicators of different levels during the training process, the model processing method of the communication device in the communication device set can be determined, which is beneficial to ensuring the training performance.

[0248] In one possible implementation, the second information is carried on the first resource.

[0249] Optionally, the first resource includes one or more of the following: the starting time-domain position, the number of time-domain units, the ending time-domain position, the starting frequency-domain position, the number of frequency-domain units, and the ending frequency-domain position. This allows the communication device 1-1 to more accurately determine the first resource, thereby improving the success rate of receiving the second information.

[0250] Optionally, the starting time domain position can be implemented in various ways. For example, the starting time domain position can be the time unit in which the communication device 1-1 obtains model processing information (e.g., the time unit in which the eighth information is received later), or the starting time unit in which the communication device 1-1 processes the first model based on the model processing information, or other implementation methods, which are not limited here.

[0251] Optionally, the above duration can be achieved in other ways, such as offset, offset amount, time domain offset, time offset, etc.

[0252] Optionally, the communication device 1-1 can execute step S303 once or multiple times, that is, the communication device 1-1 can send the second information once or multiple times at different time units. The starting time domain positions corresponding to different second information (e.g., the time unit for receiving the eighth information, the starting time unit for processing the first model, etc.) can be the same or different.

[0253] Optionally, if the starting time domain positions corresponding to different second information are different, the starting time domain unit corresponding to the next second information can be the sending time unit of the previous second information (or the next time unit of the sending time unit of the previous second information).

[0254] It should be understood that the aforementioned time unit can be a minute, second, frame, subframe, time slot, or symbol, or other time unit defined by future communication networks.

[0255] It should be understood that the aforementioned frequency domain units can be resource blocks (RBs), physical resource blocks (PRBs), resource block groups (RBGs), or other frequency domain units defined by future communication networks.

[0256] In one possible implementation, the first resource is determined based on at least one of the following: the time of receiving model processing information of the first model, the start time of the first model processing, or the end time of the first model processing. This allows the communication device 1-1 to determine the first resource for sending the second information based on at least one of the time information among the time of receiving model processing information of the first model, the start time of the first model processing, and the end time of the first model processing, thereby improving the efficiency of the communication device 1-1 in sending the second information. Furthermore, it helps to avoid or reduce the need for additional signaling between the communication device 1-1 and the receiver of the second information to negotiate the first resource, thus reducing transmission overhead.

[0257] As described above, this application refers to the information used to indicate the first performance as the performance information of the sub-model, and the information used to indicate the second performance as the performance information of the complete model. The performance information of the sub-model and the performance information of the complete model can be carried on the same or different resources. For ease of description, this application refers to the resource carrying the performance information of the sub-model as the supervision resource of the sub-model, and the resource carrying the performance information of the complete model and / or other information related to that information (e.g., output data of other communication devices and / or output data sent to other communication devices) as the supervision resource of the complete model.

[0258] The following describes the possible ways in which communication device 1-1 determines the time units or time nodes (referred to as the supervision time nodes of the sub-model) in the supervision resources of the sub-model. Taking the execution of multiple steps S303 by communication device 1-1 as an example, there can be multiple supervision time nodes of the sub-model. Hereinafter, the two supervision time nodes of the sub-model will be referred to as supervision time node 1 and supervision time node 2 of the sub-model, respectively.

[0259] Method a1: Communication device 1-1 determines the supervision time node of the sub-model based on the receiving time of the model processing information of the first model and the time offset of the sub-model supervision.

[0260] The time bias of sub-model supervision can be indicated in the model processing information of the first model, or the time bias of sub-model supervision can be pre-configured or predefined in the communication device 1-1.

[0261] In one example, the time biases for sub-model supervision can be one or more time biases, and starting from the second supervision time node of the sub-model, the starting time corresponding to the time bias is the previous supervision time node of the sub-model. Multiple time biases can be the same or different. Taking multiple identical time biases as an example, each time bias is denoted as t8. Figure 8 As shown, communication device 1-1 receives model processing information of the first model at time T8-1, and can determine the supervision time node 1 of the sub-model (e.g., based on T8-1 and t8) Figure 8 As shown in T8-2), T8-2 can be T8-1 + t8. Communication device 1-1 can also determine the supervision time node 2 of the sub-model based on T8-2 and t8 (e.g., ...). Figure 8 (as shown in T8-3).

[0262] In one example, the time bias of sub-model supervision can include multiple different time biases, and the start time corresponding to each time bias is the time when the model processing information of the first model is received. In this application, two of the multiple time biases are denoted as t9-1 and t9-2, where t9-2 is greater than t9-1. For example... Figure 9 As shown, communication device 1-1 receives model processing information of the first model at time T9-1, and can determine the supervision time node 1 of the sub-model (e.g., t9-1) based on T9-1 and t9-1. Figure 9 As shown in T9-2), T9-2 can be T9-1 + t9-1. Communication device 1-1 can also determine the supervision time node 2 of the sub-model based on T9-1 and t9-2 (e.g., ...). Figure 9 (as shown in T9-3).

[0263] When the time interval between adjacent supervision time nodes of a sub-model is the same, the supervision time node of the sub-model can be considered to be periodic, and the time interval between adjacent supervision time nodes can also be called the supervision period of the sub-model (denoted as T1).

[0264] The frequency domain resources in the supervision resources of the sub-model can also be indicated in the model processing information of the first model, or the frequency domain resources in the supervision resources of the sub-model can be pre-configured or predefined in the communication device 1-1.

[0265] Method a2: Communication device 1-1 determines the supervision time node of the sub-model based on the start time of training the first model and the time bias of the sub-model supervision.

[0266] The temporal biases for sub-model supervision can be understood by referring to the relevant content in method a1. Simply replace the time of receiving the model processing information of the first model in method a1 with the start time of training the first model. For example, if the start time for each temporal bias of sub-model supervision is the time of receiving the model processing information of the first model, then... Figure 10 A schematic diagram illustrating mode a2 is shown. For example... Figure 10 As shown, communication device 1-1 starts training the first model at time T10-4, and can determine the supervision time node 1 of the sub-model based on T10-4 and the time bias t10-1 (e.g., Figure 10 As shown in T10-5), T10-5 can be T10-4 + t10-1. Communication device 1-1 can also determine the supervision time node 2 of the sub-model based on T10-4 and t10-2 (e.g., ...). Figure 10 (as shown in T10-6).

[0267] To facilitate communication device 2 in more accurately determining the start time of communication device 1-1's training of the first model, after receiving the model processing information of the first model at time T10-1, communication device 1-1 reports to communication device 2 the time when it started training the first model (i.e., time T10-4) at time T10-2. Optionally, after receiving the start time of training the first model, communication device 2 can indicate one or more time offsets for sub-model supervision to communication device 1-1. Communication device 1-1 can determine the supervision time node of the sub-model based on the start time of training the first model and the time offsets for sub-model supervision.

[0268] The following describes the possible ways in which communication device 1-1 determines the time units or time nodes (referred to as the supervision time nodes of the complete model) in the supervision resources of the complete model.

[0269] Method b1, communication device 1-1 determines the supervision time node of the complete model based on the start time of training of the complete model or the start time of supervision of the complete model and the time bias of supervision of the complete model.

[0270] The temporal bias of full model supervision can be understood by referring to the relevant content in method a1. Simply replace the information reception time of the first model in method a1 with the start time of full model training or the start time of full model supervision. When the time interval between adjacent supervision time nodes is the same, the supervision time nodes of the full model can be considered periodic, and the time interval between adjacent supervision time nodes can be called the supervision period of the full model (denoted as T2). Figure 11As shown, the supervision period T2 of the complete model can be greater than the supervision period T1 of the sub-model.

[0271] As previously described, the supervision resources for the complete model include resources carrying performance information of the complete model and / or other information related to that information (e.g., output data received from other communication devices and / or output data sent to other communication devices). For ease of description, the moment when communication device 1-1 sends its first model's output data to other communication devices is called the sending moment, and the moment when communication device 1-1 receives output data sent by other communication devices is called the receiving moment. Communication device 1-1 can send its first model's output data multiple times; correspondingly, there can be multiple sending moments and multiple receiving moments. Hereinafter, two of the multiple sending moments will be denoted as sending moment 1 and sending moment 2, and two of the multiple receiving moments will be denoted as receiving moment 1 and receiving moment 2.

[0272] Optionally, the transmission and reception times can be indicated by different offsets. The time offset used to determine reception time 1 is denoted as t12-1, and the time offset used to determine transmission time is denoted as t12-2. For example... Figure 12 As shown, communication device 1-1 starts training the complete model or is at the start of supervision of the complete model at time T12-1. Communication device 1-1 can determine the receiving time 1 based on T12-1 and t12-1 (e.g., Figure 12 As shown in T12-2), then, based on T12-2 and t12-2, the transmission time 1 is determined (e.g., Figure 12 As shown in T12-3), then, based on T12-2 and T2, the receiving time 2 is determined (e.g., Figure 12 As shown in T12-4), then, based on T12-4 and t12-2, the transmission time 2 is determined (e.g., Figure 12 (as shown in T12-5).

[0273] When the first model has multiple inputs and / or multiple outputs, a corresponding bias can be configured for each input and / or output so that the communication device 1-1 can determine the supervision time node for each input and / or output respectively.

[0274] When communication device 1-1 sends its own output data, it can forward the output data to other communication devices (such as subsequent training nodes) through communication device 2. Communication device 2 can configure different time offsets for different communication devices in the communication device set.

[0275] Optionally, the supervision time point of the complete model also includes the measurement time of the baseline truth. The communication device 2 can configure or dynamically indicate the measurement time of the baseline truth for some of the communication devices in the communication device set.

[0276] Method b2: Based on the performance information of the sub-model sent by communication device 1-1, communication device 2 dynamically indicates the supervision time node of the complete model to communication device 1-1.

[0277] Communication device 2 can receive performance information of sub-models reported by each communication device in the communication device set, and dynamically indicate the supervision time of the complete model to each communication device. The supervision time of the complete model can include the receiving time and the transmitting time.

[0278] The communication device 1-1 can determine the supervision resources of the complete model and the supervision resources of the sub-models through pre-configuration and / or dynamic indication. Combining different methods to determine the supervision resources helps to reduce signaling overhead.

[0279] exist Figure 3-1 In one possible implementation of the method shown, the method further includes S304.

[0280] S304, Communication device 2 sends third information, and correspondingly, communication device 1-1 receives the third information;

[0281] The third information is used to instruct the model processing information. In this way, the communication device 1-1 can process the first model through the model processing information indicated by the third information, thereby improving the model processing efficiency.

[0282] Optionally, the receiving time of the model processing information in the first model introduced above can be the time when the communication device 1-1 receives the third information.

[0283] Optionally, the model processing information indicates at least one of the following: dataset, resources for collecting the dataset, model structure, model parameters, model hyperparameters, or the format of the model parameters.

[0284] Optionally, when the model processing information indicates a dataset, the model processing information may include the dataset, configuration information (or collection configuration information) of the dataset, or an index of the dataset, or other implementation methods, which are not limited here.

[0285] Communication device 1-1 can acquire a dataset based on model processing information sent by communication device 2, such as acquiring the dataset indicated by the model processing information or acquiring the dataset according to the resources for collecting datasets. Then, communication device 1-1 can train a first model based on the acquired dataset or based on the acquired dataset and a local dataset.

[0286] The communication device 1-1 can determine the model structure indicated by the communication device 2 in a predefined or preconfigured manner. For example, the mapping relationship between the ID of the first model and the model structure of the first model can be predefined or preconfigured on the communication device 1-1, or the mapping relationship between the ID of the reference unit and the structure of the reference unit can be predefined or preconfigured, wherein the reference unit is the basic unit that makes up the first model.

[0287] Optionally, the model parameters indicated by the model processing information may include frozen parameters and / or non-frozen parameters, wherein frozen parameters remain unchanged during model processing (e.g., model training), while non-frozen parameters may change during model processing (e.g., model training).

[0288] Optionally, the format of the model parameters indicated by the model processing information can be understood as the quantization method of the model parameters, such as scalar quantization, vector quantization, quantization codebook, etc.

[0289] The information processed by the model can also include joint training information. (See also...) Figure 3-2 or Figure 3-3 In the scenario shown, the third information sent by communication device 2 to nodes k and m can respectively include joint training information. As described above, different communication devices can exchange joint training data during joint training. Joint training data includes, for example, at least one of the model's output results, model parameters, or intermediate results. The joint training information can include information about the exchange of joint training data between communication device 1-1 and other communication devices (e.g., communication device 1-2) during joint training (referred to as interaction information). This interaction information can include at least one of the following: time-frequency resources carrying the joint training data, the period of interaction of the joint training data, or the number of times the joint training data is interacted. Communication device 1-1 and other communication devices (e.g., including communication device 1-2) can perform joint training based on the received joint training information. Joint training of the first model and models on other communication devices by communication device 1-1 and other communication devices is beneficial for improving the performance of the complete model and indirectly realizing data sharing among different communication devices. The complete model can include the models on each communication device involved in the joint training.

[0290] Optionally, the communication device 1-1 sends its own status information; this status information is used to determine the third information. In this way, the communication device 2 can determine a first sub-model that matches the status information of the communication device 1-1, enabling the communication device 1-1 to perform model processing based on a model that matches its own capabilities. This avoids model processing failures caused by model-capability mismatch, thereby improving the success rate of model processing.

[0291] Optionally, the status information may include computing power information of communication device 1-1 (e.g., the computing power information may indicate one or more of total computing power, used computing power, and idle computing power), storage information (e.g., the storage information may indicate one or more of total storage space, used storage space, and idle storage space), AI performance information (e.g., the AI ​​performance information may indicate one or more of AI service latency and AI service accuracy), communication information (e.g., the communication information may indicate one or more of the following: antenna information of communication device 1-1, communication chip information, channel information between communication device 1-1 and other communication devices, latency, throughput, packet loss rate, and load), or one or more of other information.

[0292] Please see Figure 13 This application provides a communication device 1300, which can realize the functions of communication device 1-1 or communication device 2 in the above method embodiments, and thus also achieve the beneficial effects of the above method embodiments. In this application embodiment, the communication device 1300 can be communication device 1-1 or communication device 2, or it can be an integrated circuit or component inside communication device 1-1 or communication device 2, such as a chip, baseband chip, modem chip, SoC chip (e.g., an SoC chip containing a modem core), SIP chip, communication module, chip system, processor, etc.

[0293] It should be noted that the transceiver unit 1302 may include a transmitting unit and a receiving unit, which are used to perform transmitting and receiving respectively.

[0294] In one possible implementation, when the device 1300 is for performing Figure 3-1 When the method executed by the communication device 1-1 in the relevant embodiments is used, the device 1300 includes a processing unit 1301 and a transceiver unit 1302; the transceiver unit 1302 is used to receive first information, the processing unit 1301 is used to perform first model processing and / or second model processing on the first model based on the first information; the transceiver unit 1302 is also used to send second information; the transceiver unit 1302 is also used to receive third information.

[0295] In one possible implementation, when the device 1300 is for performing Figure 3-1 When the method executed by the communication device 2 in the related embodiments is performed, the device 1300 includes a processing unit 1301 and a transceiver unit 1302; the processing unit 1301 is used to determine first information; the transceiver unit 1302 is used to send the first information; the transceiver unit 1302 is also used to receive second information; the transceiver unit 1302 is also used to send third information.

[0296] In one possible design, when the communication device 1300 is a terminal device or a communication module within a terminal, the function of the processing unit 1301 can be implemented by one or more processors. Specifically, the processor may include a modem chip, a SoC chip (such as a SoC chip containing a modem core), or a SIP chip. The function of the transceiver unit 1302 can be implemented by transceiver circuitry.

[0297] In one possible design, when the communication device 1300 is a circuit or chip in a terminal responsible for communication functions, such as a modem chip, a SoC chip, or a SoC chip or SIP chip containing a modem core, the function of the processing unit 1301 can be implemented by a circuit system in the aforementioned chip that includes one or more processors or processor cores. The function of the transceiver unit 1302 can be implemented by the interface circuitry or data transceiver circuitry on the aforementioned chip.

[0298] It should be noted that the information execution process of the unit of the above-mentioned communication device 1300 can be specifically described in the method embodiment shown above in this application, and will not be repeated here.

[0299] Please see Figure 14 This is another schematic structural diagram of the communication device 1400 provided in this application. The communication device 1400 includes a logic circuit 1401 and an input / output interface 1402. The communication device 1400 can be a chip or an integrated circuit.

[0300] in, Figure 13 The transceiver unit 1302 shown can be a communication interface, which can be... Figure 14 The input / output interface 1402 may include an input interface and an output interface. Alternatively, the communication interface may also be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0301] In one possible implementation, when the device 1400 is for performing Figure 3-1 When the communication device 1-1 in the relevant embodiments executes the method, the input / output interface 1402 is used to receive first information, and the logic circuit 1401 is used to perform first model processing and / or second model processing on the first model based on the first information; the input / output interface 1402 is also used to send second information; the input / output interface 1402 is also used to receive third information.

[0302] In one possible implementation, when the device 1400 is for performing Figure 3-1When the communication device 2 in the related embodiments executes the method, the logic circuit 1401 is used to determine the first information, the input / output interface 1402 is used to send the first information, the input / output interface 1402 is used to receive the second information, and the input / output interface 1402 is used to send the third information.

[0303] The logic circuit 1401 and the input / output interface 1402 can also perform other steps performed by the communication device 1-1 or the communication device 2 in any embodiment and achieve corresponding beneficial effects, which will not be elaborated here.

[0304] In one possible implementation, Figure 13 The processing unit 1301 shown can be Figure 14 The logic circuit 1401 in the middle.

[0305] Optionally, the logic circuit 1401 can be a processing device, the functions of which can be partially or entirely implemented in software.

[0306] Optionally, the processing apparatus may include a memory and a processor, wherein the memory is used 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 of the method embodiments.

[0307] Optionally, the processing device may consist of only a processor. A memory for storing computer programs is located outside the processing device, and the processor is connected to the memory via circuitry / wires to read and execute the computer programs stored in the memory. The memory and processor may be integrated together or physically independent of each other.

[0308] Optionally, the processing device may be one or more chips, or one or more integrated circuits. For example, the processing device may be one or more field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), system on-chips (SoCs), central processing units (CPUs), network processors (NPs), digital signal processors (DSPs), microcontroller units (MCUs), programmable logic controllers (PLDs), or other integrated chips, or any combination of the above chips or processors.

[0309] Please see Figure 15 The communication device 1500 mentioned in the above embodiments provided for the purposes of this application can specifically be the communication device serving as a terminal device in the above embodiments. Figure 15 The example shown illustrates how a terminal device can be implemented through a terminal device (or a component within a terminal device).

[0310] The present invention provides a possible logical structure diagram of the communication device 1500, which may include, but is not limited to, at least one processor 1501 and a communication port 1502.

[0311] in, Figure 13 The transceiver unit 1302 shown can be a communication interface, which can be... Figure 15 The communication port 1502 may include an input interface and an output interface. Alternatively, the communication port 1502 may also be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0312] Further optionally, the device may also include at least one of a memory 1503 and a bus 1504. In the embodiments of this application, the at least one processor 1501 is used to control the operation of the communication device 1500.

[0313] Furthermore, the processor 1501 can 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 devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, etc. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0314] It should be noted that, Figure 15 The communication device 1500 shown can be used to implement the steps implemented by the terminal device in the aforementioned method embodiments, and to achieve the corresponding technical effects of the terminal device. Figure 15 The specific implementation of the terminal device shown can be referred to the description of communication device 1-1 or communication device 2 in the aforementioned method embodiments, and will not be repeated here.

[0315] Please see Figure 16 The above-described embodiment of the communication device 1600 is a schematic diagram illustrating the structure of the communication device 1600. Specifically, the communication device 1600 can be a network device as described in the above embodiment. Figure 16 The example shown illustrates a network device implemented through a network device (or a component within a network device). The structure of this communication device can be referenced. Figure 16 The structure shown.

[0316] The communication device 1600 includes at least one processor 1611 and at least one network interface 1614. Optionally, the communication device further includes at least one memory 1612, at least one transceiver 1613, and one or more antennas 1615. The processor 1611, memory 1612, transceiver 1613, and network interface 1614 are connected, for example, via a bus. In this embodiment, the connection may include various interfaces, transmission lines, or buses, etc., and this embodiment is not limited thereto. The antenna 1615 is connected to the transceiver 1613. The network interface 1614 enables the communication device to communicate with other communication devices through a communication link. For example, the network interface 1614 may include a network interface between the communication device and core network equipment, such as an S1 interface; the network interface may also include a network interface between the communication device and other communication devices (e.g., other network devices or core network equipment), such as an X2 or Xn interface.

[0317] in, Figure 13The transceiver unit 1302 shown can be a communication interface, which can be... Figure 16 The network interface 1614 may include an input interface and an output interface. Alternatively, the network interface 1614 may also be a transceiver circuit, which may include input interface circuitry and output interface circuitry.

[0318] The processor 1611 is primarily used to process communication protocols and communication data, control the entire communication device, execute software programs, and process data from the software programs, for example, to support the communication device in performing the actions described in the embodiments. The communication device may include a baseband processor and a central processing unit. The baseband processor is primarily used to process communication protocols and communication data, while the central processing unit is primarily used to control the entire terminal device, execute software programs, and process data from the software programs. Figure 16 The processor 1611 can integrate the functions of a baseband processor and a central processing unit. Those skilled in the art will understand that the baseband processor and the central processing unit can also be independent processors interconnected via technologies such as buses. Those skilled in the art will understand that a terminal device can include multiple baseband processors to adapt to different network standards, and a terminal device can include multiple central processing units to enhance its processing capabilities. The various components of the terminal device can be connected via various buses. The baseband processor can also be described as a baseband processing circuit or a baseband processing chip. The central processing unit can also be described as a central processing circuit or a central processing chip. The function of processing communication protocols and communication data can be built into the processor or stored in memory as a software program, with the processor executing the software program to implement the baseband processing function.

[0319] The memory is primarily used to store software programs and data. The memory 1612 can exist independently or be connected to the processor 1611. Optionally, the memory 1612 can be integrated with the processor 1611, for example, integrated within a single chip. The memory 1612 can store program code that executes the technical solutions of the embodiments of this application, and its execution is controlled by the processor 1611. The various types of computer program code being executed can also be considered as drivers for the processor 1611.

[0320] Figure 16 Only one memory and one processor are shown. In actual terminal devices, there may be multiple processors and multiple memories. Memory can also be called storage medium or storage device, etc. Memory can be a storage element on the same chip as the processor, i.e., an on-chip storage element, or it can be a separate storage element; this application does not limit this.

[0321] Transceiver 1613 can be used to support the reception or transmission of radio frequency (RF) signals between a communication device and a terminal. Transceiver 1613 can be connected to antenna 1615. Transceiver 1613 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 1615 can receive RF signals. The receiver Rx of transceiver 1613 is used to receive the RF signals from the antennas, convert the RF signals into digital baseband signals or digital intermediate frequency (IF) signals, and provide the digital baseband signals or IF signals to processor 1611 so that processor 1611 can perform further processing on the digital baseband signals or IF signals, such as demodulation and decoding. In addition, the transmitter Tx in transceiver 1613 is also used to receive modulated digital baseband signals or IF signals from processor 1611, convert the modulated digital baseband signals or IF signals into RF signals, and transmit the RF signals through one or more antennas 1615. Specifically, the receiver Rx can selectively perform one or more stages of downmixing and analog-to-digital conversion on the radio frequency signal to obtain a digital baseband signal or a digital intermediate frequency (IF) signal. The order of these downmixing and IF conversion processes is adjustable. The transmitter Tx can selectively perform one or more stages of upmixing and digital-to-analog conversion on the modulated digital baseband signal or digital IF signal to obtain a radio frequency signal. The order of these upmixing and IF conversion processes is also adjustable. The digital baseband signal and the digital IF signal can be collectively referred to as digital signals.

[0322] The transceiver 1613 can also be called a transceiver unit, transceiver, transceiver device, etc. Optionally, the device in the transceiver unit that performs the receiving function can be regarded as the receiving unit, and the device in the transceiver unit that performs the transmitting function can be regarded as the transmitting unit. That is, the transceiver unit includes a receiving unit and a transmitting unit. The receiving unit can also be called a receiver, input port, receiving circuit, etc., and the transmitting unit can be called a transmitter, transmitter, or transmitting circuit, etc.

[0323] It should be noted that, Figure 16 The communication device 1600 shown can be used to implement the steps implemented by the network device in the aforementioned method embodiments, and to achieve the corresponding technical effects of the network device. Figure 16 The specific implementation of the communication device 1600 shown can be referred to the description of communication device 1-1 or communication device 2 in the aforementioned method embodiments, and will not be repeated here.

[0324] Please see Figure 17 The above-described embodiments of the communication device provided in this application are schematic diagrams of the structure of the communication device.

[0325] It is understood that the communication device 1700 includes, for example, modules, units, elements, circuits, or interfaces, which are appropriately configured together to execute the technical solutions provided in this application. The communication device 1700 may be the terminal device or network device described above, or a component (e.g., a chip) within these devices, used to implement the methods described in the following method embodiments. The communication device 1700 includes one or more processors 1701. The processor 1701 may be a general-purpose processor or a dedicated processor, for example, a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication device (e.g., RAN node, terminal, or chip), execute software programs, and process data from the software programs.

[0326] Optionally, in one design, processor 1701 may include program 1703 (sometimes also referred to as code or instructions), which can be executed on processor 1701 to cause communication device 1700 to perform the methods described in the following embodiments. In yet another possible design, communication device 1700 includes circuitry (…). Figure 17 (Not shown).

[0327] Optionally, the communication device 1700 may include one or more memories 1702 storing a program 1704 (sometimes referred to as code or instructions), which can be run on the processor 1701 to cause the communication device 1700 to perform the methods described in the above method embodiments.

[0328] Optionally, the processor 1701 and / or memory 1702 may include AI modules 1707 and 1708, which are used to implement AI-related functions. The AI ​​modules can be implemented through software, hardware, or a combination of both. For example, the AI ​​module may include a radio intelligence control (RIC) module. For example, the AI ​​module may be a near real-time RIC or a non-real-time RIC.

[0329] Optionally, the processor 1701 and / or memory 1702 may also store data. The processor and memory may be configured separately or integrated together.

[0330] Optionally, the communication device 1700 may further include a transceiver 1705 and / or an antenna 1706. The processor 1701, sometimes referred to as a processing unit, controls the communication device (e.g., a RAN node or terminal). The transceiver 1705, sometimes referred to as a transceiver unit, transceiver, transceiver circuit, or transceiver, is used to realize the transmission and reception functions of the communication device through the antenna 1706.

[0331] in, Figure 13 The processing unit 1301 shown may be a processor 1701. Figure 13 The transceiver unit 1302 shown can be a communication interface, which can be... Figure 17 The transceiver 1705 may include an input interface and an output interface. Alternatively, the transceiver 1705 may also be a transceiver circuit, which may include an input interface circuit and an output interface circuit.

[0332] This application also provides a computer-readable storage medium for storing one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor performs the method as described in the possible implementations of communication device 1-1 or communication device 2 in the foregoing embodiments.

[0333] This application also provides a computer program product (or computer program) that, when executed by the processor, executes the method of the communication device 1-1 or the communication device 2 as described above.

[0334] This application also provides a chip system including at least one processor for supporting a communication device in implementing the functions involved in the possible implementations of the communication device described above. Optionally, the chip system further includes an interface circuit that provides program instructions and / or data to the at least one processor. In one possible design, the chip system may also include a memory for storing the program instructions and data necessary for the communication device. The chip system may be composed of chips or may include chips and other discrete devices, wherein the communication device may specifically be communication device 1-1 or communication device 2 in the aforementioned method embodiments.

[0335] This application also provides a communication system, which includes the communication device 1-1 in any of the above embodiments.

[0336] Optionally, the communication system may also include a communication device 2.

[0337] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms. Whether a function is implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0338] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0339] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A communication method, characterized in that, include: Receive the first message; Based on the first information, the first model is subjected to first model processing and / or second model processing. The first model processing includes model training of a first communication device, and the second model processing includes model training of at least two communication devices, wherein the at least two communication devices include the first communication device.

2. The method according to claim 1, characterized in that, The first information satisfies at least one of the following: The first information is used to indicate a switch from processing the first model to processing the second model. The first information is used to indicate a switch from processing the second model to processing the first model. or The first information is used to indicate the time information of the first model processing and / or the second model processing.

3. The method according to claim 1 or 2, characterized in that, The method further includes: Send a second message, the second message being used to indicate a first performance of the first model and / or a second performance of the first model, the first performance being determined based on first data from the first communication device, the second performance being determined based on second data from a second communication device, the second communication device being different from the first communication device.

4. The method according to claim 3, characterized in that, The first information is determined based on the first performance and / or the second performance.

5. The method according to claim 3 or 4, characterized in that, The second information is carried on a first resource, which is determined based on at least one of the following: The receiving time of the model processing information of the first model, the start time of the first model processing, or the termination time of the first model processing.

6. The method according to claim 5, characterized in that, The method further includes: Receive third information, which is used to instruct the model to process information.

7. The method according to claim 5 or 6, characterized in that, The first resource includes one or more of the following: start time domain position, number of time domain units, end time domain position, start frequency domain position, number of frequency domain units, or end frequency domain position.

8. The method according to any one of claims 2-7, characterized in that, The time information for the first model processing and / or the second model processing includes at least one of the following: The start time node of the first model processing, the end time node of the first model processing, the start time node of the second model processing, the end time node of the second model processing, determine the first condition for pausing the first model processing, determine the second condition for pausing the second model processing, determine the third condition for restarting the first model processing, or determine the fourth condition for restarting the second model processing.

9. The method according to claim 8, characterized in that, The time information of the first model processing and / or the second model processing satisfies any one of the following: The first model processing has multiple start time nodes and multiple end time nodes. The second model processing has multiple start time nodes and multiple end time nodes. The end time node of the first model processing is the same as the start time node of the second model processing. The end time node of the first model processing is different from the start time node of the second model processing. The end time node of the second model processing is the same as the start time node of the first model processing. Or, the end time node of the second model processing is different from the start time node of the first model processing.

10. The method according to any one of claims 1-9, characterized in that, The first information is carried by downlink control information (DCI), radio resource control (RRC) signaling, or media access control unit (MAC-CE).

11. A communication method, characterized in that, include: Send a first message, the first message being used to instruct a first model to undergo a first model processing and / or a second model processing, the first model processing including model training of a first communication device, the second model processing including model training of at least two communication devices, the at least two communication devices including the first communication device.

12. The method according to claim 11, characterized in that, The first information satisfies at least one of the following: The first information is used to indicate a switch from processing the first model to processing the second model. The first information is used to indicate a switch from processing the second model to processing the first model. or The first information is used to indicate the time information of the first model processing and / or the second model processing.

13. The method according to claim 11 or 12, characterized in that, The method further includes: Receive second information, the second information being used to indicate a first performance of the first model and / or a second performance of the first model, the first performance being determined based on first data from the first communication device, the second performance being determined based on second data from a second communication device, the second communication device being different from the first communication device.

14. The method according to claim 13, characterized in that, The first information is determined based on the first performance and / or the second performance.

15. The method according to claim 13 or 14, characterized in that, The second information is carried on a first resource, which is determined based on at least one of the following: The receiving time of the model processing information of the first model, the start time of the first model processing, or the termination time of the first model processing.

16. The method according to claim 15, characterized in that, The method further includes: Send a third message, which is used to instruct the model to process information.

17. The method according to claim 15 or 16, characterized in that, The first resource includes one or more of the following: start time domain position, number of time domain units, end time domain position, start frequency domain position, number of frequency domain units, or end frequency domain position.

18. The method according to any one of claims 12-17, characterized in that, The time information for the first model processing and / or the second model processing includes at least one of the following: The start time node of the first model processing, the end time node of the first model processing, the start time node of the second model processing, the end time node of the second model processing, determine the first condition for pausing the first model processing, determine the second condition for pausing the second model processing, determine the third condition for restarting the first model processing, or determine the fourth condition for restarting the second model processing.

19. The method according to claim 18, characterized in that, The time information of the first model processing and / or the second model processing satisfies any one of the following: The first model processing has multiple start time nodes and multiple end time nodes. The second model processing has multiple start time nodes and multiple end time nodes. The end time node of the first model processing is the same as the start time node of the second model processing. The end time node of the first model processing is different from the start time node of the second model processing. The end time node of the second model processing is the same as the start time node of the first model processing. Or, the end time node of the second model processing is different from the start time node of the first model processing.

20. The method according to any one of claims 11-19, characterized in that, The first information is carried by downlink control information (DCI), radio resource control (RRC) signaling, or media access control unit (MAC-CE).

21. A communication device, characterized in that, Includes a module for performing the method as described in any one of claims 1 to 20.

22. A communication device, characterized in that, It includes at least one processor, said at least one processor being used to perform the method as described in any one of claims 1 to 20.

23. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions that, when executed by a communication device, implement the method as described in any one of claims 1 to 20.

24. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a computer, implement the method as described in any one of claims 1 to 20.

Images