Communication method and related apparatus

Abstract

Disclosed in the present application are a communication method and a related apparatus. In the method, a first communication apparatus receives first information, and then performs first model processing and / or second model processing on a first model on the basis of the first information, wherein the first model processing comprises model training of the first communication apparatus, the second model processing comprises model training of at least two communication apparatuses, and the at least two communication apparatuses comprise the first communication apparatus. In this way, the method is conducive to preventing the first communication apparatus from training the first model according to a fixed training method, and the method thus facilitates the first communication apparatus in training the first model according to a more suitable training method, thereby improving the efficiency of model training.

Description

Communication method and related apparatus

[0001] The present application claims priority from the Chinese patent application No. 202411129192.6 filed on August 15, 2024, and entitled "A communication method and related apparatus", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0002] The present application relates to the field of communication, and in particular to a communication method and related apparatus. BACKGROUND

[0003] With the development of communication technology, in a communication system, in addition to the traditional communication service, the service performed by the communication device can also include other new services, such as artificial intelligence (AI) service. Generally, the communication system capable of processing AI service can also be referred to as AI system.

[0004] At present, the communication device can serve as a participating node of the AI system and provide its own computing power and data. For example, the communication device can perform one or more model processing (such as model training, model updating, or model fine-tuning, etc.) processes on a local model based on local data to obtain another model. In one possible way, the way to increase the complexity of the model (such as increasing the number of parameters of the model, increasing the number of neural network layers contained in the model, etc.) can effectively improve the performance of the model.

[0005] However, in the case of gradually increasing the complexity of the model, it is possible to cause the above-mentioned model processing process to consume a relatively long time, and even it is possible to cause the above-mentioned model processing process to be unable to be executed due to the limitation of the resources (such as computing power resources, storage resources, etc.) of the communication device. Therefore, how to optimize the model processing process is a technical problem to be solved. SUMMARY

[0006] The present application provides a communication method and related apparatus for improving the model processing performance.

[0007] The first aspect of the present application provides a communication method, which is performed by a first communication device. The first communication device can be a communication apparatus (e.g., a terminal device or a network device), or the first communication device can be a part of the communication apparatus (e.g., a circuit or a chip responsible for communication functions (e.g., a Modem chip (also referred to as a baseband chip), a system on chip (SoC) chip, such as an SoC chip including a modem core, or a system in package (SIP) chip), etc.), or the first communication device can also be a logic module or software capable of implementing all or part of the functions of the communication apparatus. In the method, the first communication device receives first information, which is used to indicate that the first model is subjected to first model processing and / or second model processing. Then, the first communication device can perform the first model processing and / or the 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, and the at least two communication devices include the first communication device.

[0008] After receiving the first information, the first communication device can perform model training of the first communication device and / or model training of the at least two communication devices on the first model according to the first information. That is, the sender (e.g., the third communication device) of the first information can control the model processing manner of the first communication device by sending the first information to the first communication device, for example, control the first communication device to perform the first model processing including model training of the first communication device (or control the first communication device to independently train the first model or control the first communication device to train the first model independently) through the first information, or control the first communication device to perform the first model processing including model training of the at least two communication devices (or control the first communication device to train the first model jointly with other communication devices or control the first communication device to train the first model jointly) through the first information, or control the first communication device to perform the first model processing including model training of the first communication device and model training of the at least two communication devices (or control the first communication device to perform independent model training and joint model training with other communication devices on the first model in time division) through the first information. In this way, it is beneficial to avoid the first communication device training the first model according to a fixed training manner, so as to facilitate the first communication device to train the first model according to a more suitable training manner, and improve the model training efficiency.

[0009] In the present application, independently training the first model can also be referred to as independently training the first model or training the first model independently.

[0010] In this application, the model (e.g., the first model) can 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 can also refer to one or more of fine-tuning, fine-tuning, updating, iteration, or optimization.

[0011] In this application, the first communication device jointly training the first model with other communication devices can refer to the first communication device training the first model based on the data of the other communication devices, rather than limiting the other communication devices to perform training of the first model. Taking the other communication device as a second communication device as an example, the data of the second communication device can include at least one of the output result of the model deployed by the second communication device, the parameters or intermediate results of the model. Therefore, the first communication device jointly training the first model with other communication devices can also be referred to as the first communication device jointly training with other communication devices, and in this process, each communication device in 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, and a single communication device needs to train its own model based on the data of other communication devices.

[0012] In a possible implementation manner of the first aspect, the first information includes or at least one of the following: the first information is used to indicate switching from the first model processing to the second model processing for the first model; the first information is used to indicate switching from the second model processing to the first model processing for the first model; or the first information is used to indicate time information of the first model processing and / or the second model processing. The sender of the first information (e.g., the third communication device) can control the model processing manner of the first communication device for the first model in multiple ways, which is beneficial to improve the flexibility of the scheme implementation. The first communication device can perform model training based on the model processing manner expected by the sender of the first information. Optionally, the first communication device uses the corresponding model processing manner for model training based on the indication of the other communication device, which can reduce the implementation complexity of the first communication device. In addition, the number of first communication devices can be one or more, and the one or more first communication devices can be distributed nodes. In the above scheme, the first communication device as a distributed node can perform model training based on the model processing manner indicated by the other node (e.g., the sender of the first information), which can improve the overall model training efficiency of the one or more distributed nodes.

[0013] It should be noted that in this application, the first information can carry the identifier of the first model, or the first information can also not carry the identifier of the first model. When the first information does not carry the identifier of the first model, the first model indicated by the first information can be any model deployed on the first communication device or the model being trained on the first communication device.

[0014] Based on the first information can be used to indicate the first communication device switching model processing manner, the first communication device can switch model processing manner according to the first information, which is conducive to reducing the implementation complexity of the first communication device, and also conducive to the sender of the first information (such as the third communication device) to control the first communication device to switch the model processing manner with less time delay.

[0015] Based on the first information can be used to indicate the time information of the first model processing and / or the second model processing, the first communication device can perform corresponding model processing according to the time information of each model processing method indicated by the first information, so as to reduce the interaction times between the sender of the first information (such as the third communication device) and the first communication device, and save the overhead.

[0016] In a possible implementation manner of the first aspect, in the method, the first communication device can further send second information, the second information being used to indicate the first performance of the first model and / or the second performance of the first model, the first performance being determined based on the first data of the first communication device, and the second performance being determined based on the second data of the second communication device, the second communication device being different from the first communication device. In this way, the receiver of the second information can determine the performance of the first model processing and / or the performance of the second model processing based on the first performance and / or the second performance.

[0017] Optionally, the at least two communication devices and / or other communication devices can include the second communication device. Optionally, the second performance can be determined based on the data of the at least two communication devices.

[0018] In a possible implementation manner of the first aspect, the first information is determined based on the first performance and / or the second performance. That is, after the first communication device sends the second information, the sender of the first information (such as the third communication device) can receive the second information and determine the first information based on the first performance and / or the second performance indicated by the second information, which is conducive to more accurately controlling the model processing manner of the first communication device and improving the model processing efficiency. For example, in the case that the first performance is better, the first communication device can perform the first model processing (or independent training) based on the first information, so that it does not need to train the first model jointly with other communication devices, to reduce the transmission overhead. In the case that the first performance is poor, the first communication device can perform the second model processing (or joint training) based on the first information jointly with other communication devices (such as including the second communication device), to improve the performance of the first communication device in training the first model.

[0019] In a possible implementation of the first aspect, the second information is carried in the first resource, and the first resource is determined based on at least one of the following: a time point at which the model processing information of the first model is received, a corresponding starting time point of the model processing of the first model, or a corresponding stopping time point of the model processing of the first model. In this way, the first communication apparatus can determine the first resource for sending the second information based on at least one of the time information of the time point at which the model processing information of the first model is received, the corresponding starting time point of the model processing of the first model, and the corresponding stopping time point of the model processing of the first model, thereby improving the efficiency of the first communication apparatus in sending the second information, and optionally, avoiding or reducing the negotiation of the first resource between the first communication apparatus and a receiver of the second information through additional messages or signaling, thereby reducing transmission overhead.

[0020] In this application, the term "time point" can be replaced by other terms, for example, the term "time point" can be replaced by "time node" or "starting time domain position" or "ending time domain position". In some examples, the "stopping" mentioned in this application can mean interruption, and subsequent restart, and "termination" can mean end. In some examples, the "stopping" mentioned in this application can be replaced by "termination" or "end". In some examples, the "termination" mentioned in this application can be replaced by "stopping" or "end". In some examples, the "end" mentioned in this application can be replaced by "stopping" or "termination".

[0021] In a possible implementation of the first aspect, the first communication apparatus can further receive third information, and the third information is used to indicate the model processing information. In this way, the first communication apparatus can process the first model based on the model processing information indicated by the third information, thereby improving the model processing efficiency.

[0022] Optionally, the time point at which the model processing information of the first model is received can be the time point at which the first communication apparatus receives the third information.

[0023] Optionally, the model processing information indicates at least one of the following: a data set, a resource for collecting the data set, a model structure, a model parameter, a model hyperparameter, or a format of the model parameter.

[0024] Optionally, the first communication apparatus sends state information of the first communication apparatus, and the state information is used to determine the third information. In this way, the sender of the third information (for example, the third communication apparatus) can determine the first sub-model that matches the state information of the first communication apparatus, so that the first communication apparatus can perform model processing based on the model that matches its own capability, and the model processing failure caused by the mismatch between the model and the capability can be avoided, thereby improving the success rate of model processing.

[0025] Optionally, the state information includes one or more of the following: computing power information of the first communication device (e.g., the computing power information can indicate one or more of total computing power, used computing power, and idle computing power), storage information (e.g., the storage information can indicate one or more of total storage space, used storage space, and idle storage space), AI performance information (e.g., the AI performance information can indicate one or more of AI service latency and AI service accuracy), communication information (e.g., the communication information can indicate one or more of 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 other information.

[0026] In a possible implementation of the first aspect, the first resource includes one or more of a starting time domain position, a time domain unit quantity, an ending time domain position, a starting frequency domain position, a frequency domain unit quantity, and an ending frequency domain position. In this way, the first communication device can determine the first resource more accurately, thereby improving the success rate of receiving the second information.

[0027] As introduced above, the first information received by the first communication device can be used to indicate time information of the first model processing and / or the second model processing, or in other words, the first information is used to indicate time information of the first model processing and / or the second model processing. The time information of the model processing can include time information of starting and / or ending and / or pausing and / or restarting of the model processing, and the time information can be a time node or a condition. The model processing can be the first model processing and / or the second model processing. In a possible implementation of the first aspect, the time information of the first model processing includes at least one of the following: a starting time node of the first model processing, an ending time node of the first model processing, a first condition for determining to pause the first model processing, or a third condition for determining to restart the first model processing. The time information of the second model processing can include at least one of the following: a starting time node of the second model processing, an ending time node of the second model processing, a second condition for determining to pause the second model processing, or a fourth condition for determining to restart the second model processing. In this way, the sender of the first information (e.g., the third communication device) can control the process of training the first model by the first communication device more finely, thereby improving the performance of training the first model by the first communication device.

[0028] In a possible implementation of the first aspect, the time information of the first model processing and / or the second model processing satisfies at least one of the following: the starting time node of the first model processing is a plurality of time nodes, the ending time node of the first model processing is a plurality of time nodes, the starting time node of the second model processing is a plurality of time nodes, the ending time node of the second model processing is a plurality of time nodes, the ending time node of the first model processing is the same as the starting time node of the second model processing, the ending time node of the first model processing is different from the starting time node of the second model processing, the ending time node of the second model processing is the same as the starting time node of the first model processing, or the ending time node of the second model processing is different from the starting time node of the first model processing. In other words, the time information of the first model processing and / or the second model processing satisfies at least one of the following: the starting time node of the first model processing is a plurality of time nodes, the ending time node of the first model processing is a plurality of time nodes, the starting time node of the second model processing is a plurality of time nodes, or the ending time node of the second model processing is a plurality of time nodes. And / or, the time information of the first model processing and / or the second model processing satisfies that the ending time node of the first model processing is the same as or different from the starting time node of the second model processing. And / or, the time information of the first model processing and / or the second model processing satisfies that the ending time node of the second model processing is the same as or different from the starting time node of the first model processing. In this way, the first communication device can perform the switching operation of the multiple model processing modes based on the first information, and transmission overhead is saved.

[0029] In a possible implementation of the first aspect, the first information can be transmitted or carried by a message / signaling, for example, the message / signaling can be downlink control information (DCI), sidelink control information (SCI), a radio resource control (RRC) message, or a medium access control control element (MAC-CE), and the like. The sender (for example, the third communication device) of the first information can transmit the first information in multiple ways, which is beneficial to improve the flexibility of the scheme implementation. In this application, the RRC signaling can replace the RRC message, and the DCI can also be referred to as the DCI message.

[0030] The second aspect of the present application provides a communication method, which is performed by a third communication device. The third communication device can be a communication device (e.g., a terminal device or a network device), or the third communication device can be a part of a communication device (e.g., a circuit or a chip responsible for communication functions (e.g., a Modem chip (also known as a baseband chip), a SoC chip, such as a SoC chip including a modem core, or a SIP chip, etc.), or the third communication device can also be a logic module or software that can implement all or part of the functions of the communication device. In the method, the third communication device sends first information, and the first information is used to indicate that the first model is subjected to a first model processing and / or a second model processing. 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, and the at least two communication devices include the first communication device. Optionally, the third communication device can send the first information to the first communication device.

[0031] Optionally, before sending the first information, the third communication device can determine the first information. Here, the “determination” can be “generation” or “acquisition”, etc.

[0032] After the third communication device sends the first information, the first communication device can train the first model according to the first information, including model training of the first communication device and / or model training of at least two communication devices. That is, the third communication device can control the model processing mode of the first communication device by sending the first information to the first communication device, for example, the third communication device can control the first communication device to train the first model by including model training of the first communication device (or control the first communication device to train the first model independently or control the first communication device to train the first model independently), or the third communication device can control the first communication device to train the first model by including model training of at least two communication devices (or control the first communication device to train the first model jointly with other communication devices or control the first communication device to train the first model jointly), or the third communication device can control the first communication device to train the first model by including model training of the first communication device and model training of at least two communication devices (or control the first communication device to train the first model independently and jointly with other communication devices in time division). In this way, it is beneficial to avoid the first communication device training the first model according to a fixed training mode, so as to train the first model according to a more suitable training mode, thereby improving the model training efficiency.

[0033] In a possible implementation of the second aspect, the first information comprises at least one of: the first information is used to indicate switching from the first model processing to the second model processing for the first model; the first information is used to indicate switching from the second model processing to the first model processing for the first model; or, the first information is used to indicate time information of the first model processing and / or the second model processing. The third communication device can control the model processing manner of the first model by the first communication device in multiple ways, which is beneficial to improve the flexibility of the scheme implementation.

[0034] In a possible implementation of the second aspect, the third communication device further receives 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 of the first communication device, and the second performance being determined based on second data of a second communication device, the second communication device being different from the first communication device. In this way, the third communication device can determine the performance of the first model processing and / or the performance of the second model processing based on the first performance and / or the second performance.

[0035] Optionally, the at least two communication devices and / or other communication devices can comprise the second communication device. Optionally, the second performance can be determined based on data of the at least two communication devices.

[0036] In a possible implementation of the second aspect, the first information is determined based on the first performance and / or the second performance. That is, the third communication device can receive the second information, and determine the first information based on the first performance and / or the second performance indicated by the second information, which is beneficial to more accurately control the model processing manner of the first communication device, and improve the model processing efficiency. For example, in the case that the first performance is better, the third communication device can control the first communication device to perform the first model processing (or independent training) by sending the first information, so that the first communication device does not need to train the first model jointly with other communication devices, to reduce the transmission overhead. In the case that the first performance is poor, the third communication device can control the first communication device to perform the second model processing (or joint training) jointly with other communication devices (for example, comprising the second communication device) by sending the first information, to improve the performance of the first communication device in training the first model.

[0037] In a possible implementation manner of the second aspect, the second information is carried in the first resource, and the first resource is determined based on at least one of the following: a time point at which the model processing information of the first model is received, a corresponding starting time point of the model processing of the first model, or a corresponding stopping time point of the model processing of the first model. In this way, the first communication device and the third communication device can determine the first resource used for sending the second information based on at least one of the time information of the time point at which the model processing information of the first model is received, the corresponding starting time point of the model processing of the first model, and the corresponding stopping time point of the model processing of the first model, thereby improving the efficiency of the third communication device in receiving the second information, and avoiding or reducing the extra signaling between the first communication device and the third communication device for negotiating the first resource, thereby reducing the transmission overhead.

[0038] In a possible implementation manner of the second aspect, the third communication device further sends third information, and the third information is used to indicate the model processing information. In this way, the third communication device can control the first communication device to process the first model according to the model processing information by sending the third information, thereby improving the model processing efficiency.

[0039] Optionally, the time point at which the model processing information of the first model is received can be a time point at which the third information is received by the first communication device.

[0040] Optionally, the model processing information indicates at least one of the following: a data set, a resource for collecting the data set, a model structure, a model parameter, a model hyperparameter, or a format of the model parameter.

[0041] Optionally, the third communication device receives state information of the first communication device, and the state information is used to determine the third information. In this way, the third communication device can determine a first submodel that matches the state information of the first communication device, so that the first communication device can perform model processing based on a model that matches the capability of the first communication device, thereby avoiding model processing failure caused by a mismatch between the model and the capability, and improving the success rate of model processing.

[0042] Optionally, the state information includes one or more of the following: computing power information (for example, the computing power information can indicate one or more of total computing power, used computing power, and idle computing power) of the first communication device, storage information (for example, the storage information can indicate one or more of total storage space, used storage space, and idle storage space) of the first communication device, AI performance information (for example, the AI performance information can indicate one or more of AI service delay and AI service accuracy) of the first communication device, communication information (for example, the communication information can indicate one or more of antenna information of the first communication device, communication chip information of the first communication device, channel information, delay, throughput, and packet loss rate between the first communication device and another communication device), or other information.

[0043] In a possible implementation manner of the second aspect, the first resource comprises one or more of a starting time domain position, a time domain unit quantity, a terminal time domain position, a starting frequency domain position, a frequency domain unit quantity, and a terminal frequency domain position. In this way, the first communication device can determine the first resource more accurately, and thus the success rate of receiving the second information can be improved.

[0044] As introduced above, the first information sent by the third communication device can be used to indicate time information of the first model processing and / or the second model processing, or in other words, the first information is used to indicate time information of the first model processing and / or the second model processing. The time information of the model processing can comprise time information of a start and / or an end and / or a pause and / or a restart of the model processing, and the time information can be a time node or a condition. The model processing can be the first model processing and / or the second model processing. In a possible implementation manner of the second aspect, the time information of the first model processing comprises at least one of a starting time node of the first model processing, an ending time node of the first model processing, a first condition for determining to pause the first model processing, or a third condition for determining to restart the first model processing. The time information of the second model processing can comprise at least one of a starting time node of the second model processing, an ending time node of the second model processing, a second condition for determining to pause the second model processing, or a fourth condition for determining to restart the second model processing. In this way, the third communication device can control the process of the first communication device training the first model more finely, and the performance of the first communication device training the first model can be improved.

[0045] In a possible implementation manner of the second aspect, the time information of the first model processing and / or the second model processing satisfies at least one of the following: the starting time node of the first model processing is a plurality of time nodes, the ending time node of the first model processing is a plurality of time nodes, the starting time node of the second model processing is a plurality of time nodes, and the ending time node of the second model processing is a plurality of time nodes. And / or, the time information of the first model processing and / or the second model processing satisfies that the ending time node of the first model processing is the same as or different from the starting time node of the second model processing. And / or, the time information of the first model processing and / or the second model processing satisfies that the ending time node of the second model processing is the same as or different from the starting time node of the first model processing. In this way, the third communication device can control the first communication device to perform a switching operation of a plurality of model processing modes by sending the first information, and transmission overhead can be saved.

[0046] In a possible implementation manner of the second aspect, the first information can be transmitted or carried by a message / signaling, for example, the message / signaling can be downlink control information (DCI), sidelink control information (SCI), a radio resource control (RRC) message, or a medium access control control element (MAC-CE), and the like. The sender of the first information (for example, the third communication apparatus) can transmit the first information in multiple ways, which is beneficial to improving the flexibility of the scheme implementation.

[0047] The third aspect of the present application provides a communication apparatus, which is a first communication apparatus, comprising a transceiver and a processing unit; the transceiver is configured to receive first information; and the processing unit is configured to perform first model processing and / or second model processing on a first model based on the first information, the first model processing comprising model training of the first communication apparatus, and the second model processing comprising model training of at least two communication apparatuses, the at least two communication apparatuses comprising the first communication apparatus.

[0048] In the third aspect of the present application, the constituent modules of the communication apparatus can also be configured to perform the steps performed in the various possible implementation manners of the first aspect and achieve the corresponding technical effects, which can be referred to the first aspect for details and will not be described here.

[0049] The fourth aspect of the present application provides a communication apparatus, which is a third communication apparatus, comprising a transceiver and a processing unit, the processing unit being configured to determine first information, and the transceiver being configured to transmit the first information, the first information being configured to indicate first model processing and / or second model processing on a first model, the first model processing comprising model training of a first communication apparatus, and the second model processing comprising model training of at least two communication apparatuses, the at least two communication apparatuses comprising the first communication apparatus.

[0050] In the fourth aspect of the present application, the constituent modules of the communication apparatus can also be configured to perform the steps performed in the various possible implementation manners of the second aspect and achieve the corresponding technical effects, which can be referred to the second aspect for details and will not be described here.

[0051] The fifth aspect of the present application provides a communication apparatus, comprising at least one processor coupled with a memory; the memory is configured to store programs or instructions; the at least one processor is configured to execute the programs or instructions, so that the apparatus implements the method in any possible implementation manner of any one of the first aspect to the second aspect. Optionally, the communication apparatus can comprise the memory.

[0052] The sixth aspect of the present application provides a communication apparatus, comprising at least one logic circuit and an input / output interface; the logic circuit is configured to execute the method in any possible implementation manner of any one of the first aspect to the second aspect.

[0053] The seventh aspect of the present application provides a communication system, comprising at least two communication apparatuses and a third communication apparatus, wherein the at least two communication apparatuses comprise a first communication apparatus, and optionally, the at least two communication apparatuses further comprise a second communication apparatus.

[0054] The eighth aspect of the present application provides a computer readable storage medium, configured to store one or more computer execution instructions; when the computer execution instructions are executed by a processor, the processor executes the method in any possible implementation manner of any one of the first aspect to the second aspect.

[0055] The ninth aspect of the present application provides a computer program product (or computer program), when the computer program in the computer program product is executed by a processor, the processor executes the method in any possible implementation manner of any one of the first aspect to the second aspect.

[0056] The tenth aspect of the present application provides a chip or chip system, comprising at least one processor, configured to support a communication apparatus to implement the method in any possible implementation manner of any one of the first aspect to the second aspect. For example, the chip can be a baseband chip, a modem chip, a SoC chip (such as a SoC chip comprising a modem core), a SIP chip, or a communication module, etc.

[0057] In a possible design, the chip or chip system can further comprise a memory, configured to store necessary programs and data of the communication apparatus. The chip system can be composed of a chip, or can comprise a chip and other discrete devices. Optionally, the chip system further comprises an interface circuit, configured to provide programs and / or data for the at least one processor.

[0058] The technical effects brought by any one of the third aspect to the tenth aspect can refer to the technical effects brought by different design manners of the first aspect to the second aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0059] FIGS. 1a to 1c are schematic diagrams of a communication system provided by the present application;

[0060] FIGS. 2a to 2g are schematic diagrams of an AI processing process related to the present application;

[0061] FIG. 3-1 is an interaction schematic diagram of a communication method provided by the present application;

[0062] FIG. 3-2 is a schematic diagram of the distributed node performing training shown in FIG. 2f;

[0063] FIG. 3-3 schematically shows the process of joint training performed by node k and node m shown in FIG. 2f;

[0064] FIG. 4 schematically shows an example of the time for performing the first model processing and / or the second model processing determined by the communication apparatus 1-1 based on the time information of the first model processing and / or the second model processing;

[0065] FIG. 5 schematically shows another example of the time for performing the first model processing and / or the second model processing determined by the communication apparatus 1-1 based on the time information of the first model processing and / or the second model processing;

[0066] FIG. 6 schematically shows another example of the time for performing the first model processing and / or the second model processing determined by the communication apparatus 1-1 based on the time information of the first model processing and / or the second model processing;

[0067] FIG. 7 schematically shows the process of the distributed nodes sending the second information;

[0068] FIGS. 8 to 10 respectively schematically show examples of the communication apparatus 1-1 determining the supervision time nodes of the sub-models;

[0069] FIG. 11 schematically shows an example of the communication apparatus 1-1 determining the supervision time nodes of the sub-models and the supervision time nodes of the complete model;

[0070] FIG. 12 schematically shows an example of the communication apparatus 1-1 determining the supervision time nodes of the complete model;

[0071] FIGS. 13 to 17 are schematic diagrams of a communication apparatus provided by the present application. DETAILED DESCRIPTION

[0072] First, some terms in the embodiments of the present application are explained and described, so as to facilitate the understanding of those skilled in the art.

[0073] (1) Terminal device: can be a wireless terminal device capable of receiving network device scheduling and indication information, the wireless terminal device can be a device that provides voice and / or data connectivity to a user, or a handheld device with a wireless connection function, or other processing devices connected to a wireless modem.

[0074] The terminal device can communicate with one or more core networks or the Internet through a radio access network (RAN), and the terminal device can be a mobile terminal device, such as a mobile phone (also known as a "cellular" phone, mobile phone), a computer, and a data card, for example, it can be a portable, pocket-sized, handheld, computer-built-in, or vehicle-mounted mobile device that exchanges voice and / or data with a radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablets, computers with wireless transceiver functions, and other devices. The wireless terminal device can also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a remote station, an access point, a remote terminal, an access terminal, a user terminal, a user agent, a subscriber station, a customer premises equipment, a terminal, a user equipment, a mobile terminal, and the like.

[0075] By way of example and not limitation, in embodiments of the present application, the terminal device can also be a wearable device. The wearable device can also be referred to as a smart wearable device or a smart wearable device, etc. It is a general term for devices that apply wearable technology to the intelligent design and development of daily wear, such as glasses, gloves, watches, clothing, and shoes, etc. The wearable device is a portable device that can be directly worn on the body or integrated into the user's clothes or accessories. The wearable device is not just a hardware device, but also a powerful function through software support and data interaction, cloud interaction. The general wearable smart device includes full function, large size, and can realize complete or partial functions without relying on a smart phone, such as smart watches or smart glasses, etc., and focuses only on a certain application function, and needs to cooperate with other devices such as a smart phone, such as various smart wristbands, smart helmets, smart jewelry, etc.

[0076] The terminal can also be a drone, a robot, a terminal in device-to-device (D2D) communication, a terminal in vehicle to everything (V2X), a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self driving, a wireless terminal in telemedicine or telehealth services, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, etc.

[0077] In addition, the terminal device can also be a terminal device of a communication system evolved after the 5th generation (5G) communication system, such as a terminal device of 5G Advanced or a future communication system, etc. For example, the morphology and function of the communication terminal can be further expanded, including but not limited to vehicles, cellular network terminals (integrating satellite terminal functions), drones, internet of things (IoT) devices.

[0078] In embodiments of the present application, the terminal device described above can also obtain artificial intelligence (AI) services provided by a network device. Optionally, the terminal device can also have AI processing capability.

[0079] (2) Network device: can be a device in a wireless network, for example, the network device can be a RAN node (or device) for accessing a terminal device to a wireless network, which can also be referred to as a base station. Currently, some examples of RAN devices are: base station (base station), evolved NodeB (eNodeB), base station gNB (gNodeB) in a 5G communication system, transmission reception point (TRP), evolved Node B (eNB), radio network controller (RNC), Node B (Node B, NB), home base station (for example, home evolved Node B, or home Node B, HNB), baseband unit (BBU), or wireless fidelity (Wi-Fi) access point (AP), etc. In addition, in one network structure, the network device can include a central unit (CU) node, or a distributed unit (DU) node, or a RAN device including a CU node and a DU node.

[0080] Optionally, the RAN node can also be a macro base station, a micro base station or an indoor station, a relay node or a donor node, or a wireless controller in a cloud radio access network (CRAN) scenario. The RAN node can also be a server, a wearable device, a vehicle or a vehicle-mounted device, etc. For example, the access network device in vehicle external connection (V2X) technology can be a road side unit (RSU).

[0081] In another possible scenario, multiple RAN nodes cooperate to assist a terminal to implement wireless access, and different RAN nodes respectively implement part of the functions of a base station. For example, the RAN node can be a CU, a DU, a CU-control plane (CP), a CU-user plane (UP), or a radio unit (RU), etc. The CU and the DU can be separately arranged, or can also be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or a radio frequency unit, such as a radio frequency remote unit (RRU), an active antenna processing unit (AAU), a radio head (RH), or a remote radio head (RRH).

[0082] In different systems, the CU (or CU-CP and CU-UP), DU or RU can also have different names, but those skilled in the art can understand their meanings. For example, in an open RAN (O-RAN or ORAN) system, the CU can also be referred to as an O-CU (open CU), the DU can also be referred to as an O-DU, the CU-CP can also be referred to as an O-CU-CP, the CU-UP can also be referred to as an O-CU-UP, and the RU can also be referred to as an O-RU. For the convenience of description, the CU, CU-CP, CU-UP, DU and RU are taken as examples for description in this application. Any one of the CU (or CU-CP, CU-UP), DU and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0083] The communication between the access network device and the terminal device complies with a certain protocol layer structure. The protocol layer can include a control plane protocol layer and a user plane protocol layer. The control plane protocol layer can include at least one of the following: a radio resource control (RRC) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, a media access control (MAC) layer, or a physical (PHY) layer, etc. The user plane protocol layer can include at least one of the following: a service data adaptation protocol (SDAP) layer, a PDCP layer, an RLC layer, a MAC layer, or a physical layer, etc.

[0084] For the correspondence between the network elements in the ORAN system and the protocol layer functions that can be implemented by the network elements, refer to Table 1 below.

[0085] Table 1

[0086] The network device can be another device that provides a wireless communication function for the terminal device. The embodiments of the present application do not limit the specific technology and specific device form adopted by the network device. For the convenience of description, the embodiments of the present application do not limit.

[0087] The network device can further include a core network device, for example, including a mobility management entity (MME) in a fourth generation (4G) network, a home subscriber server (HSS), a serving gateway (S-GW), a policy and charging rules function (PCRF), a public data network gateway (PDN gateway or P-GW), a network element such as an access and mobility management function (AMF), a user plane function (UPF), or a session management function (SMF) in a 5G network. In addition, the core network device can also include other core network devices in a 5G network and a next-generation network of the 5G network.

[0088] In the embodiments of the present application, the network device mentioned above can also be an AI-capable network node, which can provide AI services for terminals or other network devices. For example, it can be an AI node, a computing power node, an AI-capable RAN node, an AI-capable core network element, etc. on the network side (access network or core network).

[0089] In the embodiments of the present application, the device for implementing the function of the network device can be a network device or a device capable of supporting the network device to implement the function, such as a chip system, which can be arranged in the network device. In the technical solutions provided in the embodiments of the present application, the device for implementing the function of the network device is taken as an example to describe the technical solutions provided in the embodiments of the present application.

[0090] (3) Configuration and pre-configuration: in the present application, configuration and pre-configuration will be used simultaneously. Configuration refers to that the network device / server sends some parameter configuration information or parameter values to the terminal through messages or signaling, so that the terminal determines the communication parameters or transmission resources according to the values or information. Pre-configuration is similar to configuration, which can be parameter information or parameter values agreed by the network device / server and the terminal device in advance, or parameter information or parameter values adopted by the base station / network device or the terminal device according to the standard protocol, or parameter information or parameter values pre-stored in the base station / server or the terminal device. The present application does not limit this.

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

[0092] (4) The terms "system" and "network" in the embodiments of the present application can be used interchangeably. "Multiple" means two or more. "And / or" describes the association relationship of associated objects, which means that there can be three relationships, for example, A and / or B can represent the cases of A alone, A and B together, and B alone, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it. "At least one of the following" or similar expressions means any combination of these items, including any combination of single item or multiple items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. In addition, unless otherwise specified, the ordinal numbers "first", "second", etc. mentioned in the embodiments of the present application are used to distinguish multiple objects, and are not used to limit the order, time sequence, priority or importance of the multiple objects.

[0093] (5) In the embodiments of the present application, "sending" and "receiving" represent the direction of signal transmission. For example, "sending information to XX" can be understood as that the destination of the information is XX, which can include direct sending through the air interface, or indirect sending through the air interface by other units or modules. "Receiving information from YY" can be understood as that the source of the information is YY, which can include direct receiving from YY through the air interface, or indirect receiving from YY through the air interface by other units or modules. "Sending" can also be understood as "output" of the chip interface, and "receiving" can also be understood as "input" of the chip interface.

[0094] In other words, sending and receiving can be between devices, such as between network devices and terminal devices, or within devices, such as between components, modules, chips, software modules or hardware modules within a device through a bus, wire or interface.

[0095] It can be understood that the information may be processed as necessary between the source and the destination of the information transmission, such as encoding and modulation, but the destination can understand the valid information from the source. Similar expressions in the present application can be similarly understood, and will not be repeated here.

[0096] (6) In embodiments of the present application, “indication” can include direct indication and indirect indication, and can also include explicit indication and implicit indication. The information indicated by certain information (indication information described below) is referred to as to-be-indicated information. In the implementation process, there are many ways to indicate the to-be-indicated information, for example but not limited to, the to-be-indicated information can be directly indicated, such as the to-be-indicated information itself or an index of the to-be-indicated information. The to-be-indicated information can also be indirectly indicated by indicating other information, where the other information and the to-be-indicated information have an association relationship. The to-be-indicated information can also be indicated only by a part of the to-be-indicated information, and the other part of the to-be-indicated information is known or agreed in advance. For example, the indication of a specific information can be achieved by means of the arrangement order of each information agreed in advance (for example, protocol predefined or preconfigured), thereby reducing the indication overhead to a certain extent. The present application does not limit the specific manner of indication. It can be understood that the indication information can be used to indicate the to-be-indicated information for the sender of the indication information, and the indication information can be used to determine the to-be-indicated information for the receiver of the indication information.

[0097] In the present application, the same or similar parts between various embodiments can be mutually referred to, unless otherwise specified. In various embodiments of the present application, and various methods / designs / implementation manners in each embodiment, the terms and / or descriptions between different embodiments, and between various methods / designs / implementation manners in each embodiment are consistent and can be mutually referred to, unless otherwise specified and logically conflicted. The technical features in different embodiments, and in various methods / designs / implementation manners in each embodiment can be combined to form new embodiments, methods, or implementation manners according to their inherent logical relationship. The implementation manners of the present application described below do not constitute a limitation on the protection scope of the present application.

[0098] The present application can be applied to a long term evolution (LTE) system, a new radio (NR) system, or a future communication system after 5G. The communication system includes at least one network device and / or at least one terminal device.

[0099] Referring to FIG. 1a, an architecture diagram of a communication system 1000 to which embodiments of the present application are applied is shown. As shown in FIG. 1a, the communication system can include a radio access network (RAN) 100, and optionally, the communication system 1000 can further include a core network 200 and an Internet 300. The RAN 100 includes at least one RAN node (e.g., 110a and 110b in FIG. 1a, collectively referred to as 110), and can further include at least one terminal (e.g., 120a-120j in FIG. 1a, collectively referred to as 120). The RAN 100 can further include other RAN nodes, such as a wireless relay device and / or a wireless backhaul device (not shown in FIG. 1a). The terminal 120 is connected to the RAN node 110 in a wireless manner, and the RAN node 110 is connected to the core network 200 in a wireless or wired manner. The core network device in the core network 200 and the RAN node 110 in the RAN 100 can be independent and different physical devices, or can be the same physical device integrated with the logical functions of the core network device and the logical functions of the RAN node. Terminals and terminals, and RAN nodes and RAN nodes can be connected to each other in a wired or wireless manner.

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

[0101] As shown in FIG. 1b, taking a base station as an example of a network device, the base station can perform communication-related services and AI-related services between one or more terminal devices, and communication-related services and AI-related services can also be performed between different terminal devices.

[0102] As shown in FIG. 1c, taking a television and a mobile phone as examples of terminal devices, the television and the mobile phone can also perform communication-related services and AI-related services.

[0103] The technical solutions provided in the present application can be applied to a wireless communication system (for example, the system shown in FIG. 1a, FIG. 1b or FIG. 1c), for example, an AI network element can be introduced in the communication system provided in the present application to implement part or all of the AI related operations. The AI network element can also be referred to as an AI node, an AI device, an AI entity, an AI module, an AI model, or an AI unit, etc. The AI network element can be built-in in a network element of the communication system. For example, the AI network element can be an AI module built-in in an access network device, a core network device, a cloud server, or an operation, administration and maintenance (OAM) to implement AI related functions. The OAM can be a network management of the core network device and / or a network management of the access network device. Alternatively, the AI network element can also be a network element independently arranged in the communication system. Optionally, an AI entity can also be included in a terminal or a chip built-in in the terminal to implement AI related functions.

[0104] Optionally, in the communication system, the AI application cases can 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. The following will be described respectively.

[0105] 1. CSI feedback enhancement

[0106] CSI is the channel property of the communication link, and is the channel quality information reported by the terminal device to the network device. The terminal device reports the channel quality information to the network device, so as to select a suitable modulation and coding scheme (MCS) for the terminal device, so that the wireless channel can be adapted to the change. For example, the terminal device performs channel estimation according to the received channel state information-reference signal (CSI-RS), and then feeds back the channel quality information to the network device. The information is used as the input of the model of the network device, so that the network device can implement AI model training. By applying AI to CSI feedback enhancement, the overhead can be reduced, the accuracy can be improved, and prediction can be realized.

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

[0108] 2. Beam management enhancement

[0109] Beam management enhancement is mainly to find the strongest transmit / receive beam pair. AI-based sparse beam prediction can improve accuracy. According to AI training and inference, it can include AI sparse beam prediction on the network side and AI sparse beam prediction on the terminal device side. Taking AI sparse beam prediction on the terminal device side as an example, the pre-trained AI model on the terminal device side can be delivered by the network side or pre-stored on the terminal device side. In the training phase, the network device scans all possible beams, and then the network reports the transmit beam pattern to the terminal device. When the model training is completed, the network device only needs to scan a small part of the beam, and then the terminal device feeds back the inference result to the network device. AI-based beam management can realize, for example, beam prediction in time and / or spatial domain to reduce overhead and delay and improve beam selection accuracy.

[0110] Beam management enhancement can include at least one sub-function, such as: beam scanning matrix prediction, and optimal beam prediction, respectively.

[0111] 3. Positioning accuracy enhancement

[0112] In line of sight (LOS) or not 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: access network device-based positioning enhancement, positioning management function network element-based positioning enhancement, and terminal device-based positioning enhancement, respectively.

[0113] 4. Network energy saving

[0114] Network energy saving can be achieved through cell activation / deactivation, load reduction, improved coverage, or other RAN setting adjustment. AI technology can be used to optimize energy saving decisions by utilizing data collected in the RAN network. AI algorithms can predict the energy efficiency and load status of the next period, which can be used to assist in decision-making for cell activation / deactivation 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 make the load evenly distributed among cells and among areas within a cell, or divert part of the traffic from congested cells, or split users among cells, carriers or access technologies to improve network performance. AI model based load balancing can provide higher quality user experience and improve system capacity.

[0117] 6. Mobility management

[0118] Mobility management is a solution to ensure service continuity during terminal device movement by minimizing dropped calls, radio link failure (RLF), unnecessary handover and ping-pong effect. AI can enhance mobility management, such as reducing the probability of unexpected events, predicting terminal device location / mobility / performance, and traffic steering, etc.

[0119] It should be understood that the definitions of the above technical terms are only examples. For example, as technology continues to evolve, the scope of the above definitions can also change, and the embodiments of the present application are not limited.

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

[0121] Optionally, the AI application case is also referred to as an AI application scenario or an AI function.

[0122] As described above, AI can be widely used in CSI feedback enhancement, beam management, positioning accuracy enhancement, energy saving, mobility enhancement, load balancing and other aspects to improve network performance. AI models can be deployed on the network side and / or the terminal device side, and the training of AI models depends on the collection of training data, which can come from terminal device measurement and feedback.

[0123] The concepts that can be involved in the present application will be briefly introduced below.

[0124] AI can give machines human intelligence, for example, machines can use computer hardware and software to simulate some intelligent behaviors of humans. To achieve artificial intelligence, machine learning methods can be used. In machine learning methods, machines learn (or train) models using training data. The model represents the mapping between input and output. The learned model can be used for inference (or prediction), i.e., the model can be used to predict the output corresponding to a given input. The output can also be referred to as the inference result (or prediction result).

[0125] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Among them, unsupervised learning can also be referred to as non-supervised learning.

[0126] Supervised learning learns the mapping relationship from sample values to sample labels according to the collected sample values and sample labels, and uses an AI model to express the learned mapping relationship. The process of training a machine learning model is the process of learning such a mapping relationship. In the training process, the sample values are input into the model to obtain the predicted values of the model, and the model parameters are optimized by calculating the error between the predicted values of the model and the sample labels (ideal values). After the mapping relationship is learned, the learned mapping can be used to predict new sample labels. The learned mapping relationship of supervised learning can include linear mapping or nonlinear mapping. According to the type of label, the learned task can be divided into classification task and regression task.

[0127] Unsupervised learning uses algorithms to discover the internal patterns of samples according to the collected sample values. In unsupervised learning, a class of algorithms uses the sample itself as a supervision signal, that is, the model learns the mapping relationship from the sample to the sample, which is called self-supervised learning. In training, the model parameters are optimized by calculating the error between the predicted values of the model and the sample itself. Self-supervised learning can be used for signal compression and decompression recovery applications. Common algorithms include autoencoders and generative adversarial networks.

[0128] Reinforcement learning is different from supervised learning, and is a class of algorithms that learn strategies to solve problems by interacting with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have clear "correct" action label data. The algorithm needs to interact with the environment to obtain the reward signal of the environment feedback, and then adjust the decision action to obtain a larger reward signal value. In the following power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the system total throughput rate feedback by the wireless network, and then expects to obtain a higher system throughput rate. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal (for example, the optimal) decision action. However, because the "correct action" label 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 network (NN) is a specific model in machine learning technology. According to the universal approximation theorem, neural network can theoretically approximate any continuous function, so that neural network has the ability to learn any mapping. Traditional communication systems need to use rich expert knowledge to design communication modules, while deep learning communication systems based on neural network can automatically discover the implicit pattern structure from a large amount of data set, establish the mapping relationship between data, and obtain better performance than 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] Figure 2a shows a schematic diagram of a neuron structure. Assume the input to the neuron 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 calculation 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] The neural network is, for example, a deep neural network (DNN). According to the construction manner of the network, the DNN can include a feedforward neural network (FNN), a convolutional neural network (CNN) and a recurrent neural network (RNN).

[0134] FIG. 2b is a schematic diagram of a FNN network. The FNN network is characterized by that the neurons in adjacent layers are fully connected to each other. This feature makes the FNN usually need a large amount of storage space and lead to a high computational complexity.

[0135] The CNN is a neural network specially designed to process data with a similar grid structure. For example, time series data (e.g. time axis discrete sampling) and image data (e.g. two-dimensional discrete sampling) can be considered as data with a similar grid structure. The CNN does not use all the input information for operation at one time, but uses a fixed size window to extract part of the information for convolution operation, which greatly reduces the calculation amount of model parameters. In addition, according to the different types of information extracted by the window (such as people and objects in the same image are different types of information), each window can use different convolution kernel operations, which makes the CNN better extract the features of the input data.

[0136] The RNN is a kind of neural network that uses feedback time series information. The input of the RNN includes the new input value at the current time and the output value of itself at the previous time. The RNN is suitable for obtaining sequence features with temporal correlation, such as speech recognition, channel coding and decoding applications.

[0137] In the above model training process of machine learning, a loss function can be defined. The loss function describes the gap or difference between the output value of the model and the ideal target value. The loss function can be embodied in various forms, and the specific form of the loss function is not limited. The model training process can be regarded as the following process: by adjusting part or all of the parameters of the model, the value of the loss function is less than the threshold value or meets the target demand.

[0138] The model can also be referred to as an AI model, a rule, or other names, etc. The AI model can be considered as a specific method to implement an AI function. The AI model represents a mapping relationship or a function between the input and the output of the model. The AI function can include one or more of the following: data collection, model training (or model learning), model information publishing, model inference (or model reasoning, reasoning, or prediction, etc.), model monitoring or model verification, or inference result publishing, etc. The AI function can also be referred to as an AI (related) operation, or an AI-related function.

[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] As shown in FIG. 2c, an MLP includes an input layer (left side), an output layer (right side), and multiple hidden layers (middle). Each layer of the MLP includes a number of nodes, referred to as neurons. The neurons of adjacent two layers are connected to each other.

[0142] Optionally, considering the neurons of adjacent two layers, the output h of the neuron of the next layer is the weighted sum of all the neurons x of the previous layer connected to it and is processed by an activation function, which can be represented as: h = f(wx + b).

[0143] where w is a weight matrix, b is a bias vector, and f is an activation function.

[0144] Further optionally, the output of the neural network can be recursively expressed as: y = f z (w z f z-1 (…)+b z ).

[0145] where z is the index of the layer of the neural network, 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 of the neural network.

[0146] In other words, the neural network can be understood as a mapping relationship from a set of input data to a set of output data. Usually, the neural network is randomly initialized, and the process of obtaining this mapping relationship from the random w and b with the existing data is called training of the neural network.

[0147] Optionally, the specific way of training is to evaluate the output result of the neural network by using a loss function.

[0148] As shown in FIG. 2d, the error can be back-propagated, and the neural network parameters (including w and b) can be iteratively optimized by the method of gradient descent until the output of the loss function reaches a minimum value, i.e., the "better point (e.g., optimal point)" in FIG. 2d. It can be understood that the neural network parameters corresponding to the "better point (e.g., optimal point)" in FIG. 2d can be used as the neural network parameters in the trained AI model information.

[0149] Further optionally, the process of gradient descent can be represented as:

[0150] wherein θ is the parameter to be optimized (including w and b), L is the loss function, η is the learning rate, and controls the step size of gradient descent, represents the derivation operation, represents the derivative of L with respect to θ.

[0151] Further optionally, the process of back-propagation utilizes the chain rule of partial derivative.

[0152] As shown in FIG. 2e, the gradient of the parameters of the previous layer can be recursively calculated from the gradient of the parameters of the next layer, which can be represented as:

[0153] wherein w ij is the weight of node j connected to node i, and s i is the input weighted sum on node i.

[0154] 2. Federated learning (FL).

[0155] The concept of federated learning effectively solves the difficulties faced by the current development of artificial intelligence. Under the premise of fully guaranteeing the privacy and security of user data, the learning task of the model is efficiently completed by promoting the cooperation of various edge devices and central servers.

[0156] As shown in FIG. 2f, the FL architecture is the most widely used training architecture in the current FL field, and the FedAvg algorithm is the basic algorithm of FL. The algorithm process of FedAvg is roughly as follows:

[0157] (1) The central end initializes the model to be trained and broadcasts it to all clients.

[0158] (2) In the t-th round t∈[1, T], the client k∈[1, K] trains the received global model based on the local data set to obtain the local training result ​​Report it to the center node. In the example shown in Figure 2f, the local training results sent by the distributed nodes n, k, and m are respectively denoted as G n , k , m .

[0159] (3) The center node collects the local training results from all (or part) of the clients, assuming that the set of clients uploading the local model in the tthround is The center end will obtain a new global model by weighted averaging with the sample number of the corresponding client as the weight, and the specific updating rule is After that, the center end broadcasts the latest version of the global model to all clients for a new round of training.

[0160] (4) Repeat steps (2) and (3) until the model converges or the number of training rounds reaches the upper limit.

[0161] Optionally, in addition to reporting the local model , the client can also report the trained local gradient , and the center node will average all the local gradients reported by the clients and update the global model according to the average gradient.

[0162] As can be seen, in the FL framework, the data set exists in the distributed nodes (such as the client), that is, the distributed nodes collect the local data set and perform local training, and report the local results (model or gradient) obtained by training to the center node. The center node itself may not have a data set, and can be responsible for fusing the training results of the distributed nodes to obtain a global model and issuing it to the distributed nodes.

[0163] 3. Decentralized learning.

[0164] As shown in Figure 2g, it is a completely distributed system without a center node. The design goal f(x) of the decentralized learning system is generally the average of the goals f i (x) of each node, that is, 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 (such as neural network) model. Each node calculates the local gradient i using the local data and the local goal f and then sends it to the neighbor nodes that are communicatively reachable. After receiving the gradient information sent by the neighbor nodes, any node can update the parameter x of the local model according to the following formula:

[0165] where, denotes the parameter of the local model of the i th node after the k+1 th (k is a natural number) update, denotes the parameter of the local model of the i th node after the k th update (if k is 0, denotes the parameter of the local model of the i th node before the update) is the parameter of the local model of the i th node that does not participate in the update), and a k denotes an adjustment coefficient, and N i is a set of neighbor nodes of node i, and |N i denotes the number of elements in the set of neighbor nodes of node i, that is, the number of neighbor nodes of node i. Through information interaction between nodes, the decentralized learning system will eventually learn a unified model.

[0166] The technical solution provided in the present application can be applied in a communication system (such as the system shown in FIG. 1a or FIG. 1b or FIG. 1c). In the communication system, the communication nodes generally have signal transceiving capability and computing capability. Taking a network device with computing capability as an example, the computing capability of the network device is mainly to provide computing power support for the signal transceiving capability (such as sending and receiving processing of signals) to realize the communication task of the network device and other communication nodes.

[0167] With the development of communication technology, in the communication system, the services performed by the communication device can include other new services in addition to the traditional communication services, such as artificial intelligence (AI) services. Generally, a system capable of processing AI services, such as a communication system, can also be referred to as an AI system.

[0168] At present, the communication device can serve as a participating node of the AI system and provide its own computing power and data. For example, the communication device can perform a plurality of model processing (such as model training, model updating, or model fine-tuning) processes on a local model based on local data to obtain another model. In one possible manner, increasing the complexity of the model (such as increasing the number of parameters of the model or increasing the number of neural network layers included in the model) can effectively improve the performance of the model.

[0169] However, in the case where the complexity of the model gradually increases, the above-mentioned model processing process may consume a relatively long time, and even the above-mentioned model processing process may not be performed due to the limitation of the resources (such as computing power resources and storage resources) of the communication device. Therefore, how to optimize the model processing process is a technical problem to be solved.

[0170] As an example, when the communication device trains based on general data, it is possible that the model is over-fitted to the training data set, falling into over-fitting, thereby reducing the generalization of the model in inference. That is, the model cannot adapt to the actual scene and / or task in inference. For example, the environment map used by the radio frequency map (RF MAP) model in training may be different from the actual environment map, and the position of the communication device used by the RF MAP model in training may also be different from the actual position of the communication device. If the training data is collected according to the actual physical environment, the overhead of data collection will be very large, and therefore it is necessary to further fine-tune or fine-tune the model according to the actual physical environment. In some scenarios, the resources (such as computing resources or storage resources) of a single communication device can enable the communication device to independently complete training, and the independent training of the model by the communication device will reduce the performance of the model. In some scenarios, the resources of a single communication device can enable the communication device to independently complete training, and the joint training of the communication device with other communication devices will increase the number of interactions, resulting in a large transmission overhead.

[0171] To solve the above problems, the present application provides a communication method and related devices, which will be described in detail below with reference to the accompanying drawings.

[0172] Please refer to FIG. 3-1, which is an implementation schematic diagram of the communication method provided by the present application. The method includes steps S301 and S302.

[0173] It should be noted that in the following, the communication device 1-1 and the communication device 2 in FIG. 3-1 are taken as an example to illustrate the execution subject of the interaction, but the present application does not limit the execution subject of the interaction.

[0174] For example, the communication device can be a communication device (such as a terminal device or a network device), or a chip, a baseband chip, a modem chip, a SoC chip (such as a SoC chip containing a modem core), a SIP chip, a communication module, a chip system, a processor, a logic module or software in the communication device, etc.

[0175] As an example, the communication device 1-1 can be a terminal device and the communication device 2 can be a network device.

[0176] As another example, the communication device 1-1 can be a network device and the communication device 2 can be a terminal device.

[0177] As another example, the communication device 1-1 and the communication device 2 are both network devices.

[0178] It should be appreciated that the network device described above can be an access network device, an ORAN device (including at least one of an O-CU, an O-DU, and an O-RU).

[0179] As another example, both the communication device 1-1 and the communication device 2 are terminal devices, i.e., the scheme shown in FIG. 3-1 can be applied to a sidelink communication scenario.

[0180] The communication device 1-1 can be the first communication device described above, and the communication device 1-1 can be one or more of the communication devices in the set of communication devices. The set of communication devices can further include a communication device 1-2 different from the communication device 1-1, and the 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 a part of the communication devices in the set of communication devices. The communication devices in the set of communication devices can be distributed nodes, and different distributed nodes can perform distributed model processing in a serial manner, a parallel manner, or a combination of the serial manner and the parallel manner, which is not limited herein. For example, the set of communication devices can include the distributed node n, the distributed node k, and the distributed node m shown in FIG. 2f. The communication device 1-1 and the communication device 1-2 can be different distributed nodes in the plurality of distributed nodes shown in FIG. 2f, and the at least two communication devices including the first communication device described above can include all or a part of the distributed nodes in FIG. 2f. Alternatively, the set of communication devices can include more or fewer communication devices.

[0181] The communication device 2 can be the third communication device described above. The communication device 2 can be a distributed node or a central node, which is not limited herein. For example, the communication device 2 can be the central node shown in FIG. 2f. In this application, the communication device 2 can also be referred to as a training supervision node, which is used to supervise the training process of the communication devices in the set of communication devices.

[0182] Optionally, for any two different communication devices (i.e., any two different distributed nodes) in the set of communication devices, it is possible to process different models to obtain model processing result information (for example, the central node can obtain corresponding model processing result information obtained by processing different models by multiple distributed nodes, which is beneficial to improving the processing efficiency of different models), and it is also possible to process the same model to obtain model processing result information (for example, the central node can obtain corresponding model processing result information obtained by processing the same model by multiple distributed nodes, which is beneficial to the central node subsequently scheduling nodes with better performance based on the model processing result information for processing, so as to improve the processing efficiency), which is not limited here. Illustratively, the models on different communication devices in the set of communication devices described above can be different models or different sub-models in the same model (or large model), or be the same model or the same sub-model in the same model (or large model).

[0183] S301, the communication device 2 sends the first information, and correspondingly, the communication device 1-1 receives the first information;

[0184] 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 including the first communication device.

[0185] In this application, the model (for example, the first model) can include an AI model, a neural network model, an AI neural network model, a machine learning model, or an AI processing model.

[0186] In this application, the model (for example, the first model) can include an AI model, a neural network model, an AI neural network model, a machine learning model, or an AI processing model. In this application, model training can also refer to one or more of fine-tuning, fine-tuning, updating, iteration, and optimization.

[0187] After the communication device 1-1 receives the first information, it can perform first model processing including model training of the communication device 1-1 on the first model and / or second model processing including model training of at least two communication devices.

[0188] The first model processing by the communication device 1-1 can also be referred to as the independent training of the first model by the communication device 1-1, or the independent training of the first model by the communication device 1-1. The second model processing by the communication device 1-1 can also be referred to as the joint training of the first model by the communication device 1-1 with other communication devices, or the joint training of the first model by the communication device 1-1. 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.

[0189] FIG. 3-2 illustrates an example of the independent training by a distributed node n (hereinafter referred to as node n) based on the received first information, and the joint training by a distributed node k (hereinafter referred to as node k) and a distributed node m (hereinafter referred to as node m), as shown in FIG. 2f. It is assumed that the models deployed on the node n, the node k and the node m are model n, model k and model m, respectively. The independent training and the joint training will be described below with reference to FIG. 3-2.

[0190] During the independent training of the model n by the node n, the data n (or sample n) for training the model n can be specified by the communication device 2, and the communication device 2 can instruct the node n how to obtain the data n, or the node n can obtain the data n by itself. The data n can exist locally or remotely on the node n, and the data n can be prepared before the node n performs multiple rounds of independent iterative training on the model n. Hereinafter, the data n is referred to as local data n. The node n inputs the local data n into the model n to obtain an output result n, and then updates the parameters of the model n based on the error backpropagation of the output result. That is, during the independent training of the model n by the node n, the node n does not need to interact with other nodes for joint training data. The joint training data can include at least one of the output result of the model, the parameters of the model and the intermediate result.

[0191] In the joint training process, node k is the predecessor node and node m is the successor node. In the process of jointly training model m by node k, the data k (or sample k) used for training model k can be data designated by communication device 2, communication device 2 can instruct node k how to obtain data n, or node k can obtain data n by itself, and the data k can exist in the local or remote of node k, and the data k can be prepared before node k performs multiple rounds of independent iterative training on model k. Hereinafter, the data k is referred to as local data k. Node k inputs the 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 error of the output result k. In the process of jointly training model m by node k, the data m (or sample m) used for training model m includes the output result k sent by node k, and the data m can also include the local or remote data of node m, which can be data designated by communication device 2, and communication device 2 can instruct node m how to obtain data m, or node m can obtain data m by itself. After node m obtains output result m based on output result k and / or data m, the parameters of model m are updated based on the error of the output result m, and gradient update information is sent to node k. Node k can update the parameters of model k based on output result k or output result k and received gradient update information. That is, in the process of jointly training own model by node k and node m, the output result and the gradient update information of the model need to be interacted. That is, model k is jointly trained by node k and node m, and model m is jointly trained by node k and node m. Optionally, the information interacted by node m and node k can include the parameters and / or intermediate results of the model, and the parameters of the model can include the gradient update information.

[0192] 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 according to the output of the current layer (for example, output data k) and the last gradient accumulation factor (for example, the gradient accumulation factor sent by node m), and correspondingly, node m can only send the gradient accumulation factor to node k.

[0193] Communication device 1-1 can be node n or node k or node m shown in FIG. 3-2, and communication device 2 can be the center node or the distributed node in FIG. 2f. FIG. 3-2 takes part of the nodes in the communication device set performing independent training and the other nodes performing joint training as an example. Optionally, all nodes in the communication device set can train the model in the same model processing mode, for example, all nodes perform independent training or all nodes perform joint training.

[0194] The node k and the node m can directly communicate with each other in the process of interacting the jointly trained data, or the jointly trained data is forwarded through other communication devices. FIG. 3-3 illustrates the process of performing joint training by the node k and the node m taking the communication device 2 as the center node and forwarding the jointly trained data. The process shown in FIG. 3-3 includes steps 1-6.

[0195] The center node sends first information to the node k and the node m respectively in steps 1 and 2, and the first information is used to indicate performing joint training.

[0196] After receiving the first information, the node k can input the data k into the model k based on the first information to obtain an output result k.

[0197] The node k sends the output result k to the center node in step 3.

[0198] The center node sends the output result k to the node m in step 4.

[0199] After receiving the first information and the output result k, the node m can obtain an output result m based on the output result k, and the error of the output result m is back propagated to update the parameters of the model m.

[0200] The node m sends the gradient update information of the model m to the center node in step 5.

[0201] The center node sends the gradient update information of the model m to the node k in step 6.

[0202] After receiving the gradient update information of the model m, the node k can update the parameters of the model k based on the gradient update information.

[0203] The present application does not limit the time sequence of steps 1 and 2, as long as step 2 is performed before step 3 and step 1 is performed before step 4.

[0204] The communication device 2 can control the model processing mode of the communication device 1-1 by sending the first information to the communication device 1-1, for example, controlling the communication device 1-1 to independently train, jointly train, or independently train and jointly train the first model in time division manner through the first information. In this way, it is beneficial to avoid the communication device 1-1 training the first model according to a fixed training mode, thereby facilitating the communication device 1-1 to train the first model according to a more suitable training mode to improve the model training efficiency.

[0205] Optionally, the first information can be transmitted or carried by a message / signaling, e.g., the message / signaling can be a downlink control information (DCI), a sidelink control information (SCI), a radio resource control (RRC) message, or a medium access control control element (MAC-CE), etc. The communication device can transmit the first information in multiple ways, which is beneficial to improve the flexibility of the implementation of the scheme.

[0206] In a possible implementation of the method shown in FIG. 3-1, the first information transmitted by the communication device 2 in step S301 can include at least one of the following: the first information is used to indicate switching of the first model from the first model processing to the second model processing; the first information is used to indicate switching of the first model from the second model processing to the first model processing; and the first information is used to indicate time information of the first model processing and / or the second model processing. The communication device 2 can control the model processing manner of the communication device 1-1 for the first model in multiple ways, which is beneficial to improve the flexibility of the implementation of the scheme. In some examples, the time information of the first model processing and / or the second model processing can also be predefined or preconfigured in the communication device 1-1, and does not need to be indicated by the first information.

[0207] Based on the fact that the first information can be used to indicate the switching of the model processing manner of the communication device 1-1, the communication device 1-1 can switch the model processing manner according to the first information, which is beneficial to reduce the implementation complexity of the communication device 1-1, and also beneficial to the communication device 2 to control the communication device 1-1 to switch the model processing manner with a smaller time delay.

[0208] Based on the fact that the first information can be used to indicate the time information 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 time information of the respective model processing method indicated by the first information, which is beneficial to reduce the number of interactions between the communication device 2 and the communication device 1-1, so as to save the overhead.

[0209] Optionally, the communication device 1-1 can train the first model according to different model processing manners in different stages based on the first information, e.g., initially training the first model independently, and later jointly training the first model. In this way, the advantages of independent training and joint training can be combined, and the overhead can be reduced under the premise of ensuring the training performance.

[0210] In a possible implementation of the method shown in FIG. 3-1, the time information of the first model processing and / or the second model processing includes at least one of the following: a starting time node of the first model processing, an ending time node of the first model processing, a starting time node of the second model processing, an ending time node of the second model processing, a first condition for determining to suspend the first model processing, a second condition for determining to suspend the second model processing, a third condition for determining to restart the first model processing, and a fourth condition for determining to restart the second model processing. In this way, the communication apparatus can more finely control the process of training the first model by the communication apparatus 1-1, and improve the performance of training the first model by the communication apparatus 1-1.

[0211] In this application, “suspend” can be replaced by deactivation, and “restart” can be used instead of activation.

[0212] In a possible implementation, the ending time node of the first model processing is the same as the starting time node of the second model processing. FIG. 4 schematically shows an example of time for performing the first model processing and / or the second model processing determined by the communication apparatus 1-1 based on the time information of the first model processing and / or the second model processing. As shown in FIG. 4, the communication apparatus 1-1 determines, based on the time information of the first model processing and / or the second model processing, to start training the first model at time T4-1, and the training manner is independent training, that is, the starting time of training the first model is the starting time of independent training. The communication apparatus 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, to end the independent training of the first model at time T4-2, and to start the joint training of the first model, that is, the starting time of independent training is the ending time of independent training. The communication apparatus 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, to end the training of the first model at time T4-3, that is, the time (or the ending time of training the first model) at which the communication apparatus 1-1 ends the training of the first model is the ending time of joint training. It is assumed that 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.

[0213] Optionally, the time information of the first model processing and / or the second model processing indicated by the first information can comprise time points T4-1, T4-2 and T4-3. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can comprise a part of the time points T4-1, T4-2 and T4-3, and the communication device 1-1 further pre-defines or pre-configures at least one time duration t4-1, t4-2 and t4, and the communication device 1-1 can determine the time points in FIG. 4 based on the pre-defined or pre-configured time duration and the time information indicated by the first information. For example, the first information can indicate the time point T4-1, and the communication device 1-1 pre-defines or pre-configures the time duration t4-1 and t4-2, and the communication device 1-1 can determine the end time point of the independent training and the start time point of the joint training (i.e. the time point T4-2) based on the time point T4-1 and the time duration t4-1, and the communication device 1-1 can determine the end time point of the joint training and the end time point of the training of the first model (i.e. the time point T4-3) based on the determined time point T4-2 and the time duration t4-2. In this application, the time duration can also be referred to as time offset.

[0214] FIG. 4 takes the communication device 1-1 first performing the independent training and then performing the joint training according to different stages of the training as an example based on the time information of the first model processing and / or the second model processing. In some examples, the communication device 1-1 can first perform the joint training and then perform the independent training. Optionally, the end time node of the second model processing and the start time node of the first model processing can be the same.

[0215] In a possible implementation, the end time node of the first model processing is different from the start time node of the second model processing. FIG. 5 schematically shows another example of time of performing the first model processing and / or the second model processing determined by the communication apparatus 1-1 based on the time information of the first model processing and / or the second model processing. As shown in FIG. 4, the communication apparatus 1-1 determines, based on the time information of the first model processing and / or the second model processing, that the training of the first model is started at time T5-1, and the training manner is independent training, that is, the start time of training the first model is the start time of the independent training. The communication apparatus 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, that the independent training of the first model is paused at time T5-2. The communication apparatus 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, that the joint training of the first model is restarted at time T5-3. The communication apparatus 1-1 can also determine, based on the time information of the first model processing and / or the second model processing, that the joint training of the first model is ended at time T5-4, and the end time of the joint training is the end time of training the first model. It is assumed that the time interval between time T5-1 and time T5-2 before time T5-2 is t5-1, the time interval between time T5-3 and time T5-2 before time T5-2 is t5-2, the time interval between time T5-4 and time T5-3 before time T5-3 is t5-3, and the time interval between time T5-1 and time T5-4 before time T5-4 is t5.

[0216] Optionally, the time information of the first model processing and / or the second model processing indicated by the first information can comprise at least one of time point information, time length information and condition. Optionally, the time information of the first model processing and / or the second model processing indicated by the first information can comprise time points T5-1, T5-2, T5-3 and T5-4. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can comprise a part of the time points T5-1, T5-2, T5-3 and T5-4, and the communication device 1-1 further predefines or preconfigures at least one of time lengths t5-1, t5-2, t5-3 and t5. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can comprise a part of the time points T5-1, T5-2, T5-3 and T5-4 and a part of the time lengths t5-1, t5-2, t5-3 and t5. The time information not indicated by the first information can be predefined or preconfigured. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can comprise T5-1, the first condition for determining to suspend the independent training, the fourth condition for determining to restart the joint training and T5-4, or can not comprise T5-1 or T5-4, but determine T5-1 or T5-4 based on the predefined or preconfigured time length. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can comprise T5-1, the first condition for determining to suspend the independent training, T5-3 and T5-4, or can not comprise at least one of T5-1, T5-3 and T5-4, but determine at least one of T5-1, T5-3 and T5-4 based on the predefined or preconfigured time length.

[0217] The present application does not limit the specific content of the first condition to the fourth condition. As an example, the first condition for determining to suspend the independent training is a performance threshold, and correspondingly, the first communication device determines T5-2 based on the performance threshold and the performance of the first model after starting to independently train the first model. For example, the first condition is that the performance of the first model is better than the performance threshold, and correspondingly, when the performance of the first model is better than the performance threshold, the independent training of the first model is suspended by itself. That is, the communication devices in the set of communication devices can start to independently train the model, then jointly train the model, and the communication devices with better performance can suspend the training, and then jointly train with others after a period of rest. In this way, it is beneficial to reduce the training computation and power consumption on the premise of ensuring the training performance.

[0218] The foregoing takes the example that the communication device 1-1 suspends or restarts the model training based on the condition or time indicated by the first information. Optionally, at least one of the first condition to the fourth condition is predefined or preconfigured in the communication device 1-1. Optionally, the communication device 1-1 can suspend the independent training or the joint training or restart the independent training or the joint training according to the instruction issued by the communication device 2 during the independent training or the joint training of the first model. For example, the communication device 1-1 can suspend the independent training according to the instruction issued by the communication device 2 after the independent training of the first model, and restart the joint training according to the instruction issued by the communication device 2, and start the joint training with other communication devices in the communication device set.

[0219] FIG. 5 takes the example that the communication device 1-1 performs the independent training first and then performs the joint training according to the different stages of the training based on the time information of the first model processing and / or the second model processing. In some examples, the communication device 1-1 can perform the joint training first and then perform the independent training. Optionally, the end time node of the second model processing and the start time node of the first model processing can be different.

[0220] In a possible implementation, the time 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. The communication device 1-1 can perform the independent training and the joint training repeatedly based on the first information. In this way, it is beneficial to the communication device 1-1 to perform the switching operation of the multiple model processing modes based on the first information, and transmission overhead is saved.

[0221] As exemplarily shown in FIG. 4 and FIG. 5, 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. Taking the case that the end time node of the first model processing is the same as the start time node of the second model processing and the end time node of the first model processing is the same as the start time node of the second model processing as an example, FIG. 6 schematically shows another example of the time of performing the first model processing and / or the second model processing determined by the communication device 1-1 based on the time information of the first model processing and / or the second model processing. As shown in FIG. 6, 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 manner is independent training. Then, the communication device 1-1 ends the independent training and starts joint training at time T6-2. Then, the communication device 1-1 ends the joint training and starts independent training at time T6-3. Then, the communication device 1-1 ends the independent training and starts joint training at time T6-4. Then, the communication device 1-1 ends the joint training, i.e., ends the training of the first model at time T6-5.

[0222] As shown in FIG. 6, the time nodes of performing the first model processing and / or the second model processing by the communication device 1-1 can include multiple start time nodes (or start times) of independent training, i.e., T6-1 and T6-3, multiple end time nodes (or end times) of independent training, i.e., T6-2 and T6-4, multiple start time nodes of joint training, i.e., T6-2 and T6-4, and multiple end time nodes of joint training, i.e., T6-3 and T6-5. As shown in FIG. 6, the time length of performing the first model processing and / or the second model processing by the communication device 1-1 can include the time length between adjacent time nodes, i.e., t6-1, t6-2, t6-3 and t6-4, and the time interval before time T6-1 and time T6-5, i.e., t6.

[0223] Optionally, the time information of the first model processing and / or the second model processing indicated by the first information can include all the time nodes shown in FIG. 6. Alternatively, the time information of the first model processing and / or the second model processing indicated by the first information can include part of the time nodes shown in FIG. 6, and all or part of the time lengths in FIG. 6 are predefined or preconfigured in the communication device 1-1, and the communication device 1-1 can determine the start time nodes and the end time nodes of each training node shown in FIG. 6 based on the predefined or preconfigured time lengths and the time nodes indicated by the first information.

[0224] Figure 6 illustrates an example in which the communication device 1-1 performs the independent training first and then performs the joint training according to different stages of the training based on the time information of the first model processing and / or the second model processing. In some examples, the communication device 1-1 can perform the joint training first and then perform the independent training.

[0225] In a possible implementation of the method shown in Figure 3-1, the method can further include S303.

[0226] S303, the communication device 1-1 sends the second information, and the communication device 2 receives the second information accordingly.

[0227] Optionally, the communication device 1-1 can send the second information, and the communication device 2 can receive the second information, where the second information is used to indicate the first performance of the first model and / or the second performance of the first model. In this way, it is beneficial for the communication device 2 to determine the performance of the first model processing and / or the performance of the second model processing based on the first performance and / or the second performance.

[0228] Hereinafter, the information used to indicate the first performance is referred to as the performance information of the sub-model, and the information used to indicate the second performance is referred to as the performance information of the complete model. The second information can include the performance information of the sub-model and / or the performance information of the complete model.

[0229] Optionally, the at least two communication devices and the other communication devices can include the communication device 1-2. Optionally, the second performance can be determined based on the data of the at least two communication devices. The communication device 1-1 can determine the first performance based on the first data of the communication device 1-1, or determine the second performance based on the second data of the communication device 1-2, or determine the first performance based on the first data of the communication device 1-1 and determine the second performance based on the second data of the communication device 1-2.

[0230] Figure 7 illustrates the process in which the performance information of the sub-model is sent by the distributed nodes with solid lines with arrows. The communication device 1-1 can be any distributed node shown in Figure 7, and the communication device 2 can be the center node shown in Figure 7. Referring to Figure 7, the distributed node inputs the local data into its own model in the process of supervising the first performance, as indicated by the solid line with reference numeral 1-1, to obtain output data. Then, as indicated by the solid line with reference numeral 1-2, the distributed node can send the performance information of the sub-model to the center node. Optionally, the performance information of the sub-model sent by the distributed node can include the output data of its own model.

[0231] The communication device 2 can instruct part of the communication devices in the set of communication devices to measure the ground truth of the related information. Alternatively, the communication device 2 can instruct the communication device 1-1 to measure the ground truth. The performance information of the sub-model sent by the communication device 1-1 can further include the measurement data 1 for measuring the ground truth or the ground truth 1 obtained based on the measurement data 1. After the communication device 2 receives the performance information of the sub-model, the communication device 2 determines the first performance of the first model according to the performance information of the sub-model.

[0232] Alternatively, in the process of determining the first performance, the communication device 1-1 can obtain the first performance (denoted as performance indicator 1) of the first model according to the output data of the first model and the ground truth 1, for example, the accuracy of the first model. Referring to the solid line with reference numeral 1-2, the performance information of the sub-model sent by the distributed node to the center node can include the performance indicator 1 determined by the distributed node.

[0233] The second data of the communication device 1-2 used by the communication device 1-1 in the process of supervising the second performance can be the output data obtained by the communication device 1-2 after inputting the output data of the first model into the model of the communication device 1-2. Alternatively, the second data can be the output data obtained by the communication device 1-2 after inputting the output data of the model on the communication device 1-3 or the local data of the communication device 1-2 into the model of the communication device 1-2. The communication device 1-1, the communication device 1-2 and the communication device 1-3 can be different communication devices in the set of communication devices.

[0234] Fig. 7 represents the process of sending the performance information of the complete model by the node m with a dashed line with an arrow. The communication device 1-1 can be the node m shown in Fig. 7, the communication device 1-2 can be the node k shown in Fig. 7, and the communication device 2 can be the center node shown in Fig. 7. Referring to Fig. 7, in the process of determining the second performance, referring to the dashed line with reference numeral 2-1, the node n can input the local data into the model n to obtain the output data of the model n (denoted as output data n), the node n sends the output data n to the node k to obtain the output data of the model k (denoted as output data k), then, referring to the dashed line with reference numeral 2-2, the node k inputs the output data k into the model k to obtain the output data of the model k (denoted as output data k), and sends the output data k to the node m, then, referring to the dashed line with reference numeral 2-3, the node m inputs the output data k into the model m to obtain the output data of the model m (denoted as output data m), and then the node m sends the performance information of the complete model to the center node. The performance information of the complete model can include the output data m.

[0235] Optionally, the node m can obtain the ground truth 2 based on the measurement data 2. The performance information of the complete model sent by the node m to the center node can further include the measurement data 2 or the ground truth 2. The center node can determine the second performance of the model m (or the performance of the complete model) based on the performance information of the complete model.

[0236] Alternatively, optionally, assuming that the model m on the node m is the last part of the complete model trained by the set of communication devices, since the output data m of the model m is the output data of the complete model, the node m can determine the second performance of the model m (or the performance of the complete model) according to the ground truth 2 and the output data m. Referring to the dashed line marked with the numeral 2-3, the performance information of the complete model sent by the node m to the center node can include the second performance of the model m (or the performance of the complete model).

[0237] The present application does not limit the first model on the communication device 1-1 to be the last part of the complete model. For example, taking the communication device 1-1 and the communication device 1-2 as the node k and the node m shown in FIG. 7, respectively, in the process of determining the second performance, the node k can send the output data k of the model k to the node m, the node m inputs the received output data k into the model m to obtain the output data m of the model m, and then the node m sends the output data m to the node k. The second data of the communication device 1-2 (i.e., the node m) can include the output data m of the communication device 1-2 (i.e., the node m).

[0238] 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 the second performance indicated by the second information, which is beneficial to more accurately control the model processing manner of the communication device 1-1 and improve the model processing efficiency.

[0239] Referring to step 0 of FIG. 3-3, optionally, the node k and the node m can send the second information to the center node, respectively.

[0240] For example, in the case that the first performance is better, the communication device 1-1 can perform the first model processing (or independent training) based on the first information, so that it does not need to train the first model jointly with other communication devices, thereby reducing the transmission overhead. In the case that the first performance is poor, the communication device 1-1 can perform the second model processing (or joint training) based on the first information jointly with other communication devices (for example, including the communication device 1-2), thereby improving the performance of the communication device 1-1 in training the first model.

[0241] With reference to FIG. 7, the center node can receive the second information reported by each distributed node for indicating the first performance, respectively, according to the solid line of reference sign 1-2, so as to facilitate determining the performance of the node n independent training model n, the performance of the node k independent training model k, and the performance of the node m independent training model m. The center node can receive the second information reported by the node m for indicating the second performance, according to the solid line of reference sign 2-3, so as to facilitate determining the performance of the joint training model of the node n, the node k, and the node m (or the performance of the complete model). The center node can indicate each node to perform independent training according to the first performance of each node and the second performance of the complete model, or indicate all nodes to perform joint training, or indicate a part of nodes to perform joint training and the other nodes to perform independent training. For example, the center node can indicate the nodes with performance higher than a threshold and the nodes with performance lower than the threshold to perform joint training, and indicate the nodes with better performance to help the nodes with worse performance to perform training.

[0242] It can be seen that the communication device 2 supervises the performance of the complete model and the performance of the model on the communication device, and determines the model processing manner of the communication device in the set of communication devices in combination with the supervision results or performance indicators of different levels in the process of training, which is beneficial to guarantee the performance of training.

[0243] In a possible implementation, the second information is carried in the first resource.

[0244] Optionally, the first resource includes one or more of a starting time domain position, a time domain unit quantity, an ending time domain position, a starting frequency domain position, a frequency domain unit quantity, and an ending frequency domain position. In this way, it is beneficial for the communication device 1-1 to determine the first resource more accurately, and thus to improve the success rate of receiving the second information.

[0245] Optionally, the starting time domain position can be implemented in various ways. For example, the starting time domain position can be a time unit in which the communication device 1-1 obtains the model processing information (for example, a time unit in which the eighth information is received in the following), or can be a starting time unit in which the communication device 1-1 processes the first model based on the model processing information, or can be implemented in other ways, which are not limited herein.

[0246] Optionally, the duration can be implemented in other ways, such as bias, bias quantity, time domain bias, time bias, and the like.

[0247] Optionally, the communication device 1-1 can perform step S303 one or more times, that is, the communication device 1-1 can send the second information one or more times in different time units. The starting time domain positions (for example, the time unit in which the eighth information is received, the starting time unit in which the first model is processed, and the like) corresponding to different second information can be the same or different.

[0248] Optionally, in the case that the starting time domain positions corresponding to different second information are different, the starting time domain unit corresponding to the next time of sending second information can be the sending time unit of the last time of sending second information (or the next time unit of the sending time unit of the last time of sending second information).

[0249] It should be understood that the above-mentioned time units can be minutes, seconds, frames, subframes, time slots, or symbols, or other time units defined in future communication networks.

[0250] It should be understood that the above-mentioned frequency domain units can be resource blocks (RBs), physical resource blocks (PRBs), or resource block groups (RBGs), or other frequency domain units defined in future communication networks.

[0251] In a possible implementation, the first resource is determined based on at least one of the following: a receiving time of the model processing information of the first model, a starting time corresponding to the first model processing, or an ending time corresponding to the first model processing. In this way, the communication device 1-1 can determine the first resource for sending the second information based on at least one of the following time information: the receiving time of the model processing information of the first model, the starting time corresponding to the first model processing, and the ending time corresponding to the first model processing, thereby improving the efficiency of the communication device 1-1 in sending the second information, and avoiding or reducing the communication device 1-1 and the receiver of the second information from negotiating the first resource through additional signaling, thereby reducing the transmission overhead.

[0252] As introduced above, the information used to indicate the first performance is referred to as the performance information of the sub-model, and the information used to indicate the second performance is referred to 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, the resource carrying the performance information of the sub-model is referred to as the supervision resource of the sub-model, and the resource carrying the performance information of the complete model and / or other information (for example, the output data of the other communication device and / or the output data sent to the other communication device) related to the information is referred to as the supervision resource of the complete model.

[0253] The possible way of determining the time unit or time node (referred to as the supervision time node of the sub-model) in the supervision resource of the sub-model by the communication device 1-1 is introduced below. Taking the case that the communication device 1-1 performs multiple steps S303 as an example, there can be multiple supervision time nodes of the sub-model. Hereinafter, the two supervision time nodes of the sub-model are referred to as the supervision time node 1 of the sub-model and the supervision time node 2 of the sub-model, respectively.

[0254] The communication device 1-1 determines the supervision time node of the sub-model according to the time offset of the sub-model supervision and the receiving time of the model processing information of the first model.

[0255] The time offset of the sub-model supervision can be indicated in the model processing information of the first model, or the time offset of the sub-model supervision can be pre-configured or pre-defined in the communication device 1-1.

[0256] In an example, the time offset of the sub-model supervision can be one time offset or multiple time offsets, and the starting time corresponding to each time offset is the previous supervision time node of the sub-model. The multiple time offsets can be the same or different. Taking multiple time offsets as multiple same time offsets as an example, each time offset is denoted as t8. As shown in FIG. 8, the communication device 1-1 receives the model processing information of the first model at time T8-1, and can determine the supervision time node 1 (T8-2 shown in FIG. 8) of the sub-model according to T8-1 and t8. T8-2 can be T8-1+t8. The communication device 1-1 can also determine the supervision time node 2 (T8-3 shown in FIG. 8) of the sub-model according to T8-2 and t8.

[0257] In an example, the time offset of the sub-model supervision can include multiple different time offsets, and the starting time corresponding to each time offset is the receiving time of the model processing information of the first model. In this application, two time offsets in the multiple time offsets are denoted as t9-1 and t9-2 respectively, and t9-2 is greater than t9-1. As shown in FIG. 9, the communication device 1-1 receives the model processing information of the first model at time T9-1, and can determine the supervision time node 1 (T9-2 shown in FIG. 9) of the sub-model according to T9-1 and t9-1. T9-2 can be T9-1+t9-1. The communication device 1-1 can also determine the supervision time node 2 (T9-3 shown in FIG. 9) of the sub-model according to T9-1 and t9-2.

[0258] When the time interval between adjacent supervision time nodes of the sub-model is the same, it can be considered that the supervision time nodes of the sub-model are periodic, and the time interval between adjacent supervision time nodes can also be referred to as the supervision period (denoted as T1) of the sub-model.

[0259] The frequency domain resource in the supervision resource of the sub-model can also be indicated in the model processing information of the first model, or the frequency domain resource in the supervision resource of the sub-model can be pre-configured or pre-defined in the communication device 1-1.

[0260] The communication device 1-1 determines the supervision time node of the sub-model according to the time offset of the sub-model supervision and the receiving time of the model processing information of the first model.

[0261] The time bias of the sub-model supervision can be understood with reference to the related content in mode a1, and only the receiving time of the model processing information of the first model in mode a1 is replaced by the starting time of training the first model. Taking the starting time corresponding to each time bias of the sub-model supervision as the receiving time of the model processing information of the first model as an example, FIG. 10 schematically shows a schematic diagram of mode a2. As shown in FIG. 10, the communication apparatus 1-1 starts training the first model at time T10-4, and can determine the supervision time node 1 (T10-5 shown in FIG. 10) of the sub-model according to T10-4 and the time bias t10-1. T10-5 can be T10-4+t10-1. The communication apparatus 1-1 can also determine the supervision time node 2 (T10-6 shown in FIG. 10) of the sub-model according to T10-4 and t10-2.

[0262] In order to facilitate the communication apparatus 2 to more accurately determine the starting time of the communication apparatus 1-1 training the first model, after the communication apparatus 1-1 receives the model processing information of the first model at time T10-1, the communication apparatus 1-1 reports the starting time (i.e., time T10-4) of training the first model to the communication apparatus 2 at time T10-2. Optionally, after the communication apparatus 2 receives the starting time of training the first model, the communication apparatus 2 can indicate one or more time biases of the sub-model supervision to the communication apparatus 1-1, and the communication apparatus 1-1 can determine the supervision time node of the sub-model according to the starting time of training the first model and the time bias of the sub-model supervision.

[0263] The following describes possible modes of the communication apparatus 1-1 determining the time unit or time node (referred to as the supervision time node of the complete model) in the supervision resource of the complete model.

[0264] Mode b1, the communication apparatus 1-1 determines the supervision time node of the complete model according to the starting time of training the complete model or the starting time of supervision of the complete model and the time bias of the supervision of the complete model.

[0265] The time bias of the supervision of the complete model can be understood with reference to the related content in mode a1, and only the receiving time of the model processing information of the first model in mode a1 is replaced by the starting time of training the first model. When the time interval between adjacent supervision time nodes is the same, the supervision time node of the complete model can be considered to be periodic, and the time interval between adjacent supervision time nodes can also be referred to as the supervision period (denoted as T2) of the complete model. As shown in FIG. 11, the supervision period T2 of the complete model can be greater than the supervision period T1 of the sub-model.

[0266] As introduced above, the supervision resource of the complete model includes resources carrying performance information of the complete model and / or other information (e.g. received output data of other communication devices and / or output data sent to other communication devices) related to the information. For the convenience of description, the time when the communication device 1-1 sends the output data of its own first model to other communication devices is referred to as sending time, and the time when the communication device 1-1 receives the output data sent by other communication devices is referred to as receiving time. The communication device 1-1 can send the output data of its own first model multiple times, and correspondingly, the sending time can be multiple, and the receiving time can be multiple. Hereinafter, two sending times in the multiple sending times are respectively referred to as sending time 1 and sending time 2, and two receiving times in the multiple receiving times are respectively referred to as receiving time 1 and receiving time 2.

[0267] Optionally, the sending time and the receiving time can be indicated by different biases respectively. The time bias used to determine the receiving time 1 is referred to as t12-1, and the time bias used to determine the sending time is referred to as t12-2. As shown in FIG. 12, the communication device 1-1 starts to train the complete model or supervise the starting time of the complete model at time T12-1, and the communication device 1-1 can determine the receiving time 1 (T12-2 as shown in FIG. 12) according to T12-1 and t12-1, then determine the sending time 1 (T12-3 as shown in FIG. 12) according to T12-2 and t12-2, then determine the receiving time 2 (T12-4 as shown in FIG. 12) according to T12-2 and T2, and then determine the sending time 2 (T12-5 as shown in FIG. 12) according to T12-4 and t12-2.

[0268] 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 determines the supervision time node of each input and / or output respectively.

[0269] When the communication device 1-1 sends its own output data, the output data can be forwarded to other communication devices (e.g. subsequent training nodes) by the communication device 2. The communication device 2 can configure different time biases for different communication devices in the set of communication devices.

[0270] Optionally, the supervision time node of the complete model also includes the measurement time of the benchmark truth, and the communication device 2 can configure or dynamically indicate the measurement time of the benchmark truth for part of the communication devices in the set of communication devices.

[0271] Mode b2, the communication device 2 dynamically indicates the supervision time node of the complete model for the communication device 1-1 according to the performance information of the sub-model sent by the communication device 1-1.

[0272] The communication device 2 can receive the performance information of the sub-models reported by the communication devices in the set of communication devices, and dynamically instruct the communication devices to perform supervision at the supervision time of the complete model. The supervision time of the complete model can include a receiving time and a sending time.

[0273] The communication device 1-1 can determine the supervision resource of the complete model and the supervision resource of the sub-models by pre-configuration and / or dynamic indication. The different manners of determining the supervision resource can help reduce signaling overhead.

[0274] In a possible implementation of the method shown in FIG. 3-1, the method further includes S304.

[0275] S304, the communication device 2 sends third information, and the communication device 1-1 receives the third information accordingly.

[0276] The third information is used to indicate model processing information. In this way, the communication device 1-1 can process the first model according to the model processing information indicated by the third information, so as to improve the model processing efficiency.

[0277] Optionally, the receiving time of the model processing information of the first model introduced in the foregoing can be the time when the communication device 1-1 receives the third information.

[0278] Optionally, the model processing information indicates at least one of the following: a data set, a resource for collecting the data set, a model structure, a model parameter, a model hyperparameter, or a format of the model parameter.

[0279] Optionally, when the model processing information indicates the data set, the model processing information can include the data set, or can include configuration information (or collection configuration information) of the data set, or an index of the data set, or other implementation manners, which are not limited herein.

[0280] The communication device 1-1 can obtain the data set according to the model processing information sent by the communication device 2, for example, obtain the data set indicated by the model processing information or obtain the data set according to the resource for collecting the data set. Then, the communication device 1-1 can train the first model according to the obtained data set or according to the obtained data set and a local data set.

[0281] The communication device 1-1 can determine the model structure indicated by the communication device 2 according to a predefined or preconfigured manner, for example, a mapping relationship between an ID of the first model and a model structure of the first model can be predefined or preconfigured on the communication device 1-1, or a mapping relationship between a reference unit ID and a reference unit structure can be predefined or preconfigured, wherein the reference unit is a basic unit of the first model.

[0282] Optionally, the model parameters indicated by the model processing information can comprise frozen parameters and / or unfrozen parameters, wherein the frozen parameters remain unchanged in the model processing (e.g., model training), and the unfrozen parameters can be changed in the model processing (e.g., model training).

[0283] Optionally, the format of the model parameters indicated by the model processing information can be understood as a quantization manner of the model parameters, such as scalar quantization, vector quantization, quantization codebook, etc.

[0284] The model processing information can further comprise joint training information. With reference to the scenario shown in FIG. 3-2 or FIG. 3-3, the third information sent by the communication device 2 to the node k and the node m can comprise joint training information, respectively. As introduced before, the joint training can involve the interaction of joint training data between different communication devices, which can comprise at least one of the output result of the model, the parameter of the model, or the intermediate result. The joint training information can comprise information (referred to as interaction information) of the interaction of joint training data between the communication device 1-1 and other communication devices (e.g., the communication device 1-2) involved in the joint training, which can comprise at least one of the time-frequency resource carrying the joint training data, the period of interaction of the joint training data, or the number of interactions of the joint training data, etc. The communication device 1-1 and other communication devices (e.g., comprising the communication device 1-2) can perform joint training according to the received joint training information, and the communication device 1-1 and other communication devices jointly train the first model and the models on other communication devices, which is beneficial to improve the performance of the complete model and indirectly realize the data sharing of different communication devices. The complete model can comprise the models on the communication devices involved in the joint training.

[0285] Optionally, the communication device 1-1 sends state information of the communication device 1-1; and the state information is used to determine the third information. In this way, it is beneficial for the communication device 2 to determine the first sub-model matching the state information of the communication device 1-1, so that the communication device 1-1 can perform model processing based on the model matching its own capability, which can avoid the failure of model processing due to the mismatch between the model and the capability, so as to improve the success rate of model processing.

[0286] Optionally, the state information includes one or more of computing power information of the communication device 1-1 (for example, the computing power information can indicate one or more of total computing power, used computing power, and idle computing power), storage information (for example, the storage information can indicate one or more of total storage space, used storage space, and idle storage space), AI performance information (for example, the AI performance information can indicate one or more of AI service latency and AI service accuracy), communication information (for example, the communication information can indicate one or more of antenna information of the communication device 1-1, communication chip information, channel information between the communication device 1-1 and other communication devices, latency, throughput, packet loss rate, and load), or other information.

[0287] Referring to FIG. 13, the embodiment of the present application provides a communication device 1300, which can implement the functions of the communication device 1-1 or the communication device 2 in the above-mentioned method embodiments, and thus can also implement the beneficial effects possessed by the above-mentioned method embodiments. In the embodiment of the present application, the communication device 1300 can be the communication device 1-1 or the communication device 2, or an integrated circuit or element etc. inside the communication device 1-1 or the communication device 2, such as a chip, a baseband chip, a modem chip, an SoC chip (such as an SoC chip containing a modem core), a SIP chip, a communication module, a chip system, a processor, etc.

[0288] It should be noted that the transceiver unit 1302 can include a sending unit and a receiving unit, which are respectively used for performing sending and receiving.

[0289] In a possible implementation, when the device 1300 is used to perform the method performed by the communication device 1-1 in FIG. 3-1 and related embodiments, the device 1300 includes a processing unit 1301 and a transceiver unit 1302; the transceiver unit 1302 is configured to receive first information, and the processing unit 1301 is configured to perform first model processing and / or second model processing on a first model based on the first information; the transceiver unit 1302 is further configured to send second information; and the transceiver unit 1302 is further configured to receive third information.

[0290] In a possible implementation, when the device 1300 is used to perform the method performed by the communication device 2 in FIG. 3-1 and related embodiments, the device 1300 includes a processing unit 1301 and a transceiver unit 1302; the processing unit 1301 is configured to determine first information; the transceiver unit 1302 is configured to send the first information; the transceiver unit 1302 is further configured to receive second information; and the transceiver unit 1302 is further configured to send third information.

[0291] In a possible design, when the communication apparatus 1300 is a communication module in a terminal device or terminal, the function of the processing unit 1301 can be implemented by one or more processors. Specifically, the processor can include a modem chip, a SoC chip (such as a SoC chip including a modem core), or a SIP chip. The function of the transceiver unit 1302 can be implemented by a transceiver circuit.

[0292] In a possible design, when the communication apparatus 1300 is a circuit or chip responsible for communication functions in a terminal, such as a modem chip or a SoC chip or a SoC chip including a modem core or a SIP chip, the function of the processing unit 1301 can be implemented by circuitry including one or more processors or processor cores in the chip. The function of the transceiver unit 1302 can be implemented by an interface circuit or data transceiver circuit on the chip.

[0293] It should be noted that the information processing process and the like of the units of the communication apparatus 1300 described above can be specifically refer to the description in the method embodiments described above, and will not be described here.

[0294] Please refer to FIG. 14, which is another schematic structural diagram of a communication apparatus 1400 provided in the present application, the communication apparatus 1400 includes a logic circuit 1401 and an input / output interface 1402. Wherein, the communication apparatus 1400 can be a chip or an integrated circuit.

[0295] Wherein, the transceiver unit 1302 shown in FIG. 13 can be a communication interface, which can be the input / output interface 1402 in FIG. 14, and the input / output interface 1402 can include an input interface and an output interface. Alternatively, the communication interface can also be a transceiver circuit, which can include an input interface circuit and an output interface circuit.

[0296] In a possible implementation, when the apparatus 1400 is configured to perform the method performed by the communication apparatus 1-1 in FIG. 3-1 and related embodiments, the input / output interface 1402 is configured to receive first information, and the logic circuit 1401 is configured to perform first model processing and / or second model processing on a first model based on the first information; the input / output interface 1402 is further configured to send second information; and the input / output interface 1402 is further configured to receive third information.

[0297] In a possible implementation, when the apparatus 1400 is configured to perform the method performed by the communication apparatus 2 in FIG. 3-1 and related embodiments, the logic circuit 1401 is configured to determine first information, and the input / output interface 1402 is configured to send the first information; the input / output interface 1402 is configured to receive second information; and the input / output interface 1402 is configured to send third information.

[0298] The logic circuit 1401 and the input and output interface 1402 can also perform other steps and achieve corresponding beneficial effects performed by the communication device 1-1 or the communication device 2 in any embodiment, which will not be repeated here.

[0299] In a possible implementation, the processing unit 1301 shown in FIG. 13 can be the logic circuit 1401 in FIG. 14.

[0300] Optionally, the logic circuit 1401 can be a processing device, and the functions of the processing device can be partially or entirely implemented by software.

[0301] Optionally, the processing device can include a memory and a processor, where the memory is configured to store a computer program, and the processor is configured to read and execute the computer program stored in the memory to perform the corresponding processing and / or steps in any one of the method embodiments.

[0302] Optionally, the processing device can only include the processor. The memory for storing the computer program is located outside the processing device, and the processor is connected with the memory through a circuit / wire to read and execute the computer program stored in the memory. The memory and the processor can be integrated together or can be physically independent of each other.

[0303] Optionally, the processing device can be one or more chips or one or more integrated circuits. For example, the processing device can be one or more field-programmable gate arrays (FPGA), application specific integrated circuits (ASIC), system on chips (SoC), central processing units (CPU), network processors (NP), digital signal processors (DSP), micro controller units (MCU), programmable logic controllers (PLD) or other integrated chips, or any combination of the above chips or processors, etc.

[0304] Referring to FIG. 15, the communication device 1500 involved in the above embodiments provided by the embodiments of the present application can be specifically the communication device as the terminal device in the above embodiments, and the example shown in FIG. 15 is implemented by the terminal device (or components in the terminal device).

[0305] In the communication device 1500, at least one processor 1501 and a communication port 1502 can be included.

[0306] The transceiver unit 1302 in FIG. 13 can be a communication interface, which can be the communication port 1502 in FIG. 15, and the communication port 1502 can include an input interface and an output interface. Alternatively, the communication port 1502 can also be a transceiver circuit, which can include an input interface circuit and an output interface circuit.

[0307] Further, the device can further include at least one of a memory 1503 and a bus 1504, and in the embodiments of the present application, the at least one processor 1501 is configured to control and process actions of the communication device 1500.

[0308] In addition, 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 device, transistor logic device, hardware component, or any combination thereof. It can implement or execute various exemplary logical blocks, modules and circuits described in combination with the disclosure. The processor can also be a combination of computing functions, such as one or more microprocessor combinations, digital signal processor and microprocessor combinations, etc. Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0309] It should be noted that the communication device 1500 shown in FIG. 15 can be specifically used to implement the steps implemented by the terminal device in the foregoing method embodiments, and achieve the corresponding technical effects of the terminal device. The specific implementation of the terminal device shown in FIG. 15 can refer to the description of the communication device 1-1 or the communication device 2 in the foregoing method embodiments, which will not be described here.

[0310] Please refer to FIG. 16, which is a structural schematic diagram of the communication device 1600 involved in the foregoing embodiments according to an embodiment of the present application. The communication device 1600 can be specifically the communication device as the network device in the foregoing embodiments, and the example shown in FIG. 16 is that the network device is implemented by the network device (or components in the network device), wherein the structure of the communication device can refer to the structure shown in FIG. 16.

[0311] The communication device 1600 comprises at least one processor 1611 and at least one network interface 1614. Further optionally, the communication device further comprises at least one memory 1612, at least one transceiver 1613 and one or more antennas 1615. The processor 1611, the memory 1612, the transceiver 1613 and the network interface 1614 are connected, for example, through a bus, which may, in embodiments of the present application, comprise various types of interfaces, transmission lines or buses, etc., and the present embodiments do not limit the same. The antenna 1615 is connected to the transceiver 1613. The network interface 1614 is configured to enable the communication device to communicate with other communication devices through a communication link. For example, the network interface 1614 can comprise a network interface between the communication device and a core network device, such as an S1 interface, and the network interface can comprise a network interface between the communication device and other communication devices (such as other network devices or core network devices), such as an X2 or Xn interface.

[0312] The transceiver unit 1302 shown in FIG. 13 can be a communication interface, which can be the network interface 1614 in FIG. 16, and the network interface 1614 can comprise an input interface and an output interface. Alternatively, the network interface 1614 can also be a transceiver circuit, which can comprise an input interface circuit and an output interface circuit.

[0313] The processor 1611 is mainly configured to process communication protocols and communication data, and control the whole communication device, execute software programs, process data of the software programs, for example, to support the communication device to perform the actions described in the embodiments. The communication device can comprise a baseband processor and a central processor, the baseband processor is mainly configured to process communication protocols and communication data, and the central processor is mainly configured to control the whole terminal device, execute software programs, and process data of the software programs. The processor 1611 in FIG. 16 can integrate the functions of the baseband processor and the central processor, and those skilled in the art can understand that the baseband processor and the central processor can also be independent processors interconnected through a bus. Those skilled in the art can understand that the terminal device can comprise a plurality of baseband processors to adapt to different network modes, and the terminal device can comprise a plurality of central processors to enhance the processing capability, and various components of the terminal device can be connected through various buses. The baseband processor can also be referred to as a baseband processing circuit or a baseband processing chip. The central processor can also be referred to as a central processing circuit or a central processing chip. The function of processing communication protocols and communication data can be built in the processor, or stored in the memory in the form of software programs, and the processor executes the software programs to realize the baseband processing function.

[0314] The memory is mainly used for storing software programs and data. The memory 1612 can exist independently of the processor 1611. Alternatively, the memory 1612 can be integrated with the processor 1611, for example, integrated in a chip. The memory 1612 is capable of storing program codes for implementing the technical solutions of the embodiments of the present application, and the processor 1611 controls the execution. The executed computer programs of various types can also be regarded as a driver of the processor 1611.

[0315] FIG. 16 only shows one memory and one processor. In actual terminal equipment, there can be multiple processors and multiple memories. The memory can also be referred to as a storage medium or a storage device, etc. The memory can be a storage element on the same chip as the processor, i.e., an on-chip storage element, or an independent storage element, and the embodiments of the present application do not limit this.

[0316] The transceiver 1613 can be used to support the reception or transmission of radio frequency signals between the communication device and the terminal. The transceiver 1613 can be connected to the antenna 1615. The transceiver 1613 includes a transmitter Tx and a receiver Rx. Specifically, one or more antennas 1615 can receive radio frequency signals, the receiver Rx of the transceiver 1613 is used to receive the radio frequency signals from the antenna and convert the radio frequency signals into digital baseband signals or digital intermediate frequency signals, and provide the digital baseband signals or digital intermediate frequency signals to the processor 1611 for further processing of the digital baseband signals or digital intermediate frequency signals by the processor 1611, such as demodulation processing and decoding processing. In addition, the transmitter Tx in the transceiver 1613 is also used to receive modulated digital baseband signals or digital intermediate frequency signals from the processor 1611, and convert the modulated digital baseband signals or digital intermediate frequency signals into radio frequency signals, and transmit the radio frequency signals through one or more antennas 1615. Specifically, the receiver Rx can selectively perform one or more levels of down-mixing and analog-to-digital conversion to obtain digital baseband signals or digital intermediate frequency signals. The order of the down-mixing and analog-to-digital conversion can be adjusted. The transmitter Tx can selectively perform one or more levels of up-mixing and digital-to-analog conversion to obtain radio frequency signals. The order of the up-mixing and digital-to-analog conversion can be adjusted. The digital baseband signals and the digital intermediate frequency signals can be collectively referred to as digital signals.

[0317] The transceiver 1613 can also be referred to as a transceiving unit, a transceiver, a transceiving device, etc. Optionally, a device in the transceiving unit for implementing a receiving function can be regarded as a receiving unit, and a device in the transceiving unit for implementing a sending function can be regarded as a sending unit, i.e., the transceiving unit includes the receiving unit and the sending unit, the receiving unit can also be referred to as a receiver, an input port, a receiving circuit, etc., and the sending unit can be referred to as a transmitter, a transmitter, or a transmitting circuit, etc.

[0318] It should be noted that the communication apparatus 1600 shown in FIG. 16 can be specifically used to implement the steps implemented by the network device in the foregoing method embodiments, and achieve the corresponding technical effects of the network device. The specific implementation of the communication apparatus 1600 shown in FIG. 16 can be referred to the description of the communication apparatus 1-1 or the communication apparatus 2 in the foregoing method embodiments, which will not be repeated here.

[0319] Please refer to FIG. 17, which is a structural schematic diagram of the communication apparatus involved in the foregoing embodiments provided by the embodiments of the present application.

[0320] It can be understood that the communication apparatus 1700 includes, for example, modules, units, elements, circuits, or interfaces, etc., which are properly configured together to execute the technical solutions provided by the present application. The communication apparatus 1700 can be a terminal device or a network device as described above, or a component (such as a chip) of these devices, to implement the methods described in the following method embodiments. The communication apparatus 1700 includes one or more processors 1701. The processor 1701 can be a general-purpose processor or a special-purpose processor, etc. For example, it can be a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication apparatus (such as a RAN node, a terminal, or a chip, etc.), execute software programs, and process data of the software programs.

[0321] Optionally, in one design, the processor 1701 can include a program 1703 (which can also be referred to as code or instructions at times), which can be run on the processor 1701, so that the communication apparatus 1700 executes the methods described in the following embodiments. In another possible design, the communication apparatus 1700 includes a circuit (not shown in FIG. 17).

[0322] Optionally, the communication apparatus 1700 can include one or more memories 1702, which have a program 1704 (which can also be referred to as code or instructions at times) stored thereon, and the program 1704 can be run on the processor 1701, so that the communication apparatus 1700 executes the methods described in the foregoing method embodiments.

[0323] Optionally, the processor 1701 and / or the memory 1702 can include an AI module 1707, 1708, which is configured to implement AI-related functions. The AI module can be implemented in software, hardware, or a combination of software and hardware. For example, the AI module can include a radio intelligence control (RIC) module. For example, the AI module can be a near-real-time RIC or a non-real-time RIC.

[0324] Optionally, the processor 1701 and / or the memory 1702 can also store data. The processor and the memory can be separately arranged or integrated together.

[0325] Optionally, the communication device 1700 can also include a transceiver 1705 and / or an antenna 1706. The processor 1701 can also be referred to as a processing unit, which controls the communication device (e.g., a RAN node or a terminal). The transceiver 1705 can also be referred to as a transceiving unit, a transceiver, a transceiving circuit, or a transceiver, which is configured to perform the transceiving function of the communication device through the antenna 1706.

[0326] In the above, the processing unit 1301 in FIG. 13 can be the processor 1701. The transceiving unit 1302 in FIG. 13 can be a communication interface, which can be the transceiver 1705 in FIG. 17. The transceiver 1705 can include an input interface and an output interface. Alternatively, the transceiver 1705 can be a transceiving circuit, which can include an input interface circuit and an output interface circuit.

[0327] The embodiments of the present application also provide a computer readable storage medium for storing one or more computer-executable instructions, which, when executed by a processor, cause the processor to perform the method described in the possible implementation manners of the communication device 1-1 or the communication device 2.

[0328] The embodiments of the present application also provide a computer program product (or a computer program), which, when executed by a processor, causes the processor to perform the method of the possible implementation manners of the communication device 1-1 or the communication device 2.

[0329] The embodiments of the present application further provide a chip system, which comprises at least one processor for supporting the communication device to implement the functions involved in the possible implementation manners of the communication device. Optionally, the chip system further comprises an interface circuit for providing program instructions and / or data for the at least one processor. In a possible design, the chip system can further comprise a memory for storing the necessary program instructions and data of the communication device. The chip system can be composed of a chip, or can comprise a chip and other discrete components, and the communication device can be the communication device 1-1 or the communication device 2 in the foregoing method embodiments.

[0330] The embodiments of the present application further provide a communication system, which comprises the communication device 1-1 in any of the foregoing embodiments.

[0331] Optionally, the communication system further comprises the communication device 2.

[0332] In the several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms. The exact method for executing the functions described above will depend 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, and such implementation should not be considered beyond the scope of the present application.

[0333] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e. can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0334] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit. When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or in the form of a contribution, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

Claims

1. A communication method characterized by comprising: Comprising: receiving first information; performing first model processing and / or second model processing on a first model based on the first information, the first model processing comprising model training of a first communication device, and the second model processing comprising model training of at least two communication devices, the at least two communication devices comprising the first communication device.

2. The method of claim 1, wherein, The first information satisfies at least one of: the first information is used to indicate switching of the first model from the first model processing to the second model processing; the first information is used to indicate switching of the first model from the second model processing to the first model processing; or the first information is used to indicate 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 comprises: sending 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 of the first communication device, and the second performance being determined based on second data of a second communication device, the second communication device being different from the first communication device.

4. The method of claim 3, wherein, 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 in a first resource, the first resource being determined based on at least one of: a time of receiving model processing information of the first model, a starting time corresponding to the first model processing, or an ending time corresponding to the first model processing.

6. The method of claim 5, wherein, The method further comprises: receiving third information, the third information being used to indicate the model processing information.

7. The method according to claim 5 or 6, characterized in that, The first resource comprises one or more of a starting time domain position, a time domain unit quantity, an ending time domain position, a starting frequency domain position, a frequency domain unit quantity, or an ending frequency domain position.

8. The method according to any one of claims 2-7, characterized in that, The time information of the first model processing and / or the second model processing comprises at least one of: a starting time node of the first model processing, an ending time node of the first model processing, a starting time node of the second model processing, an ending time node of the second model processing, a first condition for determining to pause the first model processing, a second condition for determining to pause the second model processing, a third condition for determining to restart the first model processing, or a fourth condition for determining to restart the second model processing.

9. The method of claim 8, wherein, The time information of the first model processing and / or the second model processing satisfies any one of: the starting time node of the first model processing is a plurality of time nodes, the ending time node of the first model processing is a plurality of time nodes, the starting time node of the second model processing is a plurality of time nodes, the ending time node of the second model processing is a plurality of time nodes, the ending time node of the first model processing is the same as the starting time node of the second model processing, the ending time node of the first model processing is different from the starting time node of the second model processing, the ending time node of the second model processing is the same as the starting time node of the first model processing, or the ending time node of the second model processing is different from the starting 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 a medium access control control element (MAC-CE).

11. A communication method, comprising: The method comprises: sending first information, the first information being used for indicating first model processing and / or second model processing on a first model, the first model processing comprising model training of a first communication device, and the second model processing comprising model training of at least two communication devices, the at least two communication devices comprising the first communication device.

12. The method of claim 11, wherein, The first information satisfies at least one of: the first information is used for indicating switching from the first model processing to the second model processing on the first model; the first information is used for indicating switching from the second model processing to the first model processing on the first model; or the first information is used for indicating 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 comprises: receiving second information, the second information being used for indicating 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 of the first communication device, and the second performance being determined based on second data of a second communication device, the second communication device being different from the first communication device.

14. The method of claim 13, wherein, 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, the first resource being determined based on at least one of: a time instance of receiving model processing information of the first model, a corresponding start time instance of the first model processing, or a corresponding stop time instance of the first model processing.

16. The method of claim 15, wherein, The method further comprises: sending third information, the third information being used for indicating the model processing information.

17. The method according to claim 15 or 16, characterized in that, The first resource comprises one or more of a start time domain position, a time domain unit quantity, an end time domain position, a start frequency domain position, a frequency domain unit quantity, or an end frequency domain position.

18. The method according to any one of claims 12-17, characterized by, The time information of the first model processing and / or the second model processing comprises at least one of: a start time node of the first model processing, an end time node of the first model processing, a start time node of the second model processing, an end time node of the second model processing, a first condition for determining to pause the first model processing, a second condition for determining to pause the second model processing, a third condition for determining to restart the first model processing, or a fourth condition for determining to restart the second model processing.

19. The method of claim 18, wherein, The time information of the first model processing and / or the second model processing satisfies any one of: The starting time node of the first model processing is a plurality of time nodes, the ending time node of the first model processing is a plurality of time nodes, the starting time node of the second model processing is a plurality of time nodes, the ending time node of the second model processing is a plurality of time nodes, the ending time node of the first model processing is the same as the starting time node of the second model processing, the ending time node of the first model processing is different from the starting time node of the second model processing, the ending time node of the second model processing is the same as the starting time node of the first model processing, or the ending time node of the second model processing is different from the starting time node of the first model processing.

20. The method of any one of claims 11-19, wherein, The first information is carried by downlink control information (DCI), radio resource control (RRC) signaling, or a medium access control control element (MAC-CE).

21. A communications device, characterized by The apparatus comprises a module for performing the method of any one of claims 1-10, or a module for performing the method of any one of claims 11-20.

22. A communications device, characterized by The apparatus comprises at least one processor configured to perform the method of any one of claims 1-10, or the method of any one of claims 11-20.

23. The communication apparatus according to claim 22, wherein, The apparatus further comprises a memory storing computer programs or instructions.

24. The communication apparatus according to claim 22 or 23, wherein, The communication device is a chip or a chip system.

25. A computer readable storage medium, characterized in that, The computer readable storage medium stores computer programs or instructions, which, when executed by a communication device, implement the method of any one of claims 1-10; or implement the method of any one of claims 11-20.

26. A computer program product, characterised in that, The computer readable storage medium stores computer programs or instructions, which, when executed by a computer, implement the method of any one of claims 1-10; or implement the method of any one of claims 11-20. The computer readable storage medium stores computer programs or instructions, which, when executed by a computer, implement the method of any one of claims 1-10; or implement the method of any one of claims 11-20.

Images