Switching processing method, device and system

Abstract

The invention provides a switching processing method, device and system, which are applied to a first access network device and a communication device to jointly execute a training task, the training task refers to a task of performing AI training based on model parameters or measurement data to obtain an AI model, and the AI model is used for the first access network device and the communication device to jointly execute model reasoning. The method includes determining that the communication apparatus is to switch from a first access network device to a second access network device, terminating or pausing a first portion of a training task to be performed by the first access network device, and triggering the communication apparatus to terminate or pause a second portion of the training task to be performed by the communication apparatus. Based on the scheme, resource waste caused by switching of the communication device in the AI model training process can be reduced.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a switching processing method, device and system. Background Art

[0002] In order to improve the intelligence and automation level of the network, reasoning models, such as artificial intelligence (AI) models and machine learning (ML) models, are used in more and more technical fields. Models are usually obtained through training, such as joint training of terminals and networks. The terminals and networks train their own AI models based on local data, that is, the terminal trains the encoder model and the network trains the decoder model. The two exchange model parameters and / or training data and other information during the model training process through the air interface, thereby achieving joint collaborative training.

[0003] However, if the AI ​​model training has not converged yet and the terminal device moves or the base station / cell switches, it will result in a waste of resources on the terminal device side and the network side. Summary of the invention

[0004] The present application provides a switching processing method, device and system, which can reduce the resource waste caused by switching of communication devices during AI model training.

[0005] In a first aspect, a handover processing method is provided, which can be executed by a first access network device, or by a component (such as a chip or circuit) of the first access network device. For convenience, the first access network device is used as an example for illustration. As an example, the first access network device can be a source base station.

[0006] The method includes: determining that a communication device will switch from a first access network device to a second access network device; terminating or pausing a first part of a training task that is to be executed by the first access network device, and triggering the communication device to terminate or suspend a second part of the training task that is to be executed by the communication device.

[0007] It should be understood that this technical solution is applicable to the process in which the first access network device and the communication device jointly perform a training task. The training task refers to a task of obtaining an AI model through AI training based on model parameters or measurement data. The AI ​​model is used by the first access network device and the communication device to jointly perform model reasoning.

[0008] Based on the above scheme, when it is determined that the communication device will switch, the first access network device terminates or suspends the first part of the training task that is to be executed by the first access network device, and triggers the communication device to terminate or suspend the second part of the training task that is to be executed by the communication device, thereby avoiding the communication device from continuing training and data reporting, that is, continuing to execute the second part of the training task, which causes a waste of resources, and avoiding the first access network device from continuing to execute the first part of the training task, which causes a waste of resources.

[0009] In combination with the first aspect, in certain implementations of the first aspect, triggering the communication device to terminate or pause the second part of the training task that is to be executed by the communication device includes: sending a training task termination message or a training task pause message, the training task termination message or the training task pause message includes an identifier and a reason value of the second part, and the reason value is used to indicate that a switch has occurred in the communication device.

[0010] Based on the above solution, the first access network device instructs the communication device to terminate or suspend the second part of the training task to be executed by the communication device by sending training task termination information or training task suspension information.

[0011] In combination with the first aspect, in certain implementations of the first aspect, before determining that the communication device will switch from the first access network device to the second access network device, the method also includes: receiving first AI capability information, the first AI capability information indicating the AI ​​training capability supported by the communication device; determining model information for training the AI ​​model based on the first AI capability information and the second AI capability information, the second AI capability information indicating the AI ​​training capability supported by the first access network device; sending training task start information, the training task start information including the model information of the AI ​​model and the identifier of the second part, so that the communication device executes the second part of the training task that is to be executed by the communication device; based on the model information of the AI ​​model, executing the first part of the training task that is to be executed by the first access network device.

[0012] Based on the above scheme, the first access network device receives the first AI capability information and determines the model information in combination with the second AI capability information, and then notifies the communication device to start task training, so that the first access network device and the communication device jointly perform the training task. The training task refers to the task of obtaining an AI model by performing AI training based on model parameters or measurement data.

[0013] In combination with the first aspect, in some implementations of the first aspect, the first AI capability information includes at least one of the following: a first AI switch, a supported first AI algorithm type, supported first AI framework information, or first computing power information, wherein the first computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

[0014] In combination with the first aspect, in certain implementations of the first aspect, the second AI capability information includes at least one of the following: a second AI switch, a supported second AI algorithm type, supported second AI framework information, or second computing power information, wherein the second computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

[0015] Based on the above scheme, the first computing power information is added to the first AI capability information reported by the communication device, so that the first access network device can determine the AI ​​model training task that is adapted to the communication device capability based on the first computing power information and its own second AI capability information, thereby appropriately reducing the AI ​​model training time.

[0016] In combination with the first aspect, in certain implementations of the first aspect, the model information of the AI ​​model includes at least one of the following: model structure, hyperparameters, model accuracy, or number of training rounds.

[0017] Based on the above scheme, based on the model information of the AI ​​model determined by the first access network device, the first access network device and the communication device can perform training tasks in a targeted manner to improve the training efficiency of the AI ​​model.

[0018] In combination with the first aspect, in certain implementations of the first aspect, before terminating or pausing the first part of the training task that is to be executed by the first access network device, the method also includes: receiving third AI capability information, the third AI capability information indicating the AI ​​training capability supported by the second access network device; determining, based on the measurement report and the third AI capability information, that the communication device will switch from the first access network device to the second access network device, the measurement report being used to indicate characteristics of the network environment in which the communication device is located.

[0019] Based on the above scheme, by receiving the third AI capability information of an adjacent access network device (for example, a second access network device), when the first access network device determines that the communication device will be switched based on the received measurement report, it can determine that the second access network device is the target access network device for switching the communication device in combination with the third AI capability information.

[0020] In combination with the first aspect, in certain implementations of the first aspect, manufacturer information is received, and the manufacturer information indicates the manufacturer providing the second access network device; wherein, based on the measurement report and the third AI capability information, it is determined that the communication device will switch from the first access network device to the second access network device, including: based on the measurement report, the third AI capability information and the manufacturer information, it is determined that the communication device will switch from the first access network device to the second access network device.

[0021] Based on the above solution, in the case where the communication device may switch to other access network devices that belong to different manufacturers as the first access network device, the first access network device needs to further consider the manufacturer information when making a communication device switching decision, and give priority to selecting the second access network device that belongs to the same manufacturer as the first access network device as the target access network device, so as to avoid resource exposure due to selecting access network devices from different manufacturers and ensure model security.

[0022] In combination with the first aspect, in certain implementations of the first aspect, the third AI capability information includes at least one of the following: a third AI switch, a supported third AI algorithm type, supported third AI framework information, third computing power information, or retraining capability information, the retraining capability information indicates whether the second access network device supports retraining of an already trained AI model, and the third computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

[0023] Based on the above scheme, retraining capability information is added to the third AI capability information reported by the second access network device, so that the first access network device considers whether the second access network device supports retraining of the trained model when making a communication device switching decision, so that after the communication device switching is completed, the second access network device and the communication device can continue to perform the currently unfinished AI training tasks, avoiding resource waste while improving the training efficiency of the AI ​​model.

[0024] In combination with the first aspect, in certain implementations of the first aspect, triggering the communication device to terminate or suspend the second part of the training task that is to be executed by the communication device includes: determining, based on the third AI capability information, that the second access network device does not support the training of the AI ​​model; triggering the communication device to terminate the second part of the training task that is to be executed by the communication device.

[0025] Based on the above scheme, when the first access network device determines that the second access network device does not support the training of the AI ​​model, it can instruct the communication device to terminate the second part of the training task that is to be executed by the communication device, thereby avoiding waste of resources caused by continued training or data reporting.

[0026] In combination with the first aspect, in certain implementations of the first aspect, a non-AI mode switching indication is sent to the communication device, where the non-AI mode switching indication is used to instruct the communication device to switch to the non-AI mode.

[0027] Based on the above scheme, when the communication device terminates the second part of the training task, the first access network device can also send a non-AI mode switching instruction to the communication device, so that the communication device switches from the AI ​​mode to the non-AI model, that is, the communication device no longer performs AI model training and can use traditional non-AI business algorithms to continue to complete the corresponding business.

[0028] In combination with the first aspect, in certain implementations of the first aspect, triggering the communication device to terminate or suspend the second part of the training task that is to be executed by the communication device includes: determining, based on third AI capability information, that the second access network device supports the training of the AI ​​model; triggering the communication device to suspend the second part of the training task that is to be executed by the communication device.

[0029] Based on the above scheme, when the first access network device determines that the second access network device supports the training of the AI ​​model, it can instruct the communication device to suspend the second part of the training task that is to be executed by the communication device, so that after the switching is completed, the communication device can continue to execute the previously unfinished AI training task with the second access network device, thereby avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0030] In combination with the first aspect, in certain implementations of the first aspect, the intermediate training results of the AI ​​model are saved, where the intermediate training results are obtained by the first part of the training task that is performed by the first access network device; and the intermediate training results and the identifier of the first part are sent to the second access network device.

[0031] Based on the above scheme, the first access network device can save the intermediate training results of the AI ​​model obtained by executing the first part of the training task that belongs to the first access network device, and then notify the second access device of the intermediate training results, so that after the switching is completed, the communication device can continue to execute the previously unfinished AI training tasks with the second access network device, avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0032] In combination with the first aspect, in some implementations of the first aspect, first switching request information is sent to the second access network device, and the first switching request information includes an identifier of the first part and an intermediate training result.

[0033] In combination with the first aspect, in certain implementations of the first aspect, model training information of a communication device is received, the model training information is the model parameters and / or training data most recently reported by the communication device to the first access network device, the model training information of the communication device is used for the second access network device to construct an AI model, and the constructed AI model is used for the second access network device to continue to execute the first part of the training task that belongs to the first access network device; an identifier of the first part and the model training information of the communication device are sent to the second access network device.

[0034] Based on the above scheme, by sending the model training information of the communication device and the identifier of the first part to the second access network device, the second access network device can construct the NW side AI model of the second access network device based on the model training information of the communication device, so that after the subsequent communication device completes the switching, the second access network device and the communication device can continue to perform the previously unfinished AI training tasks, avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0035] In combination with the first aspect, in some implementations of the first aspect, second switching request information is sent to the second access network device, and the second switching request information includes an identifier of the first part and model training information of the communication device.

[0036] In combination with the first aspect, in certain implementations of the first aspect, the first switching request information or the second switching request information also includes at least one of the following: first AI capability information of the communication device, or model information of the AI ​​model.

[0037] Based on the above scheme, the first access network device can request the second access network device to switch the communication device to the second access network device through the first switching request information or the second switching request information, respectively, carrying the model information or intermediate training results of the AI ​​model, so that when the second access network device determines that the communication device is allowed to switch to the second access network device, the second access network device and the communication device can continue to perform the previously unfinished AI training tasks, thereby avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0038] In combination with the first aspect, in certain implementations of the first aspect, when the communication device switches back to the first access network device, the first part of the training task that is to be executed by the first access network device continues to be executed.

[0039] Based on the above scheme, considering that the first access network device suspends the first part of the training task that belongs to the first access network device, when the communication device subsequently switches back to the first access network device, the communication device can continue to execute the previously unfinished AI training task with the first access network device, avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0040] In combination with the first aspect, in certain implementations of the first aspect, terminating or pausing the first part of the training task that is to be executed by the first access network device includes: receiving switching notification information, the switching notification information including an identifier of the communication device; and terminating or pausing the first part of the training task that is to be executed by the first access network device based on the switching notification information.

[0041] Based on the above solution, the first mobile intelligent function corresponding to the first access network device can receive the switching notification information from the first access network device through the G1 interface, and terminate or suspend the first part of the training task that is to be executed by the first access network device based on the switching notification information. It should be understood that the AI ​​model training function in this implementation is deployed in the first mobile intelligent function.

[0042] In combination with the first aspect, in certain implementations of the first aspect, the switching notification information also includes identification information of at least one candidate access network device, the at least one candidate access network device includes a second access network device, and the at least one candidate access network device is determined based on a measurement report, and the measurement report is used to indicate characteristics of the network environment in which the communication device is located.

[0043] Before sending the first switching request information to the second access network device, or before sending the second switching request information to the second access network device, the method also includes: receiving at least one fourth AI capability information corresponding to at least one candidate access network device, the fourth AI capability information indicating the AI ​​training capability supported by each candidate access network device; and determining, based on the at least one fourth AI capability information, that the second access network device is the optimal access network device for switching the communication device.

[0044] Based on the above scheme, the first mobile intelligent function corresponding to the first access network device determines that the second access network device is the optimal access network device for switching the communication device according to the fourth AI capability information, that is, the first access network device or the first mobile intelligent function in the technical scheme of the present application can determine the target access network device for switching the communication device.

[0045] In combination with the first aspect, in certain implementations of the first aspect, determining, based on at least one fourth AI capability information, that the second access network device is the optimal access network device for switching the communication device includes: determining, based on at least one fourth AI capability information and the vendor information of each candidate access network device, that the second access network device is the optimal access network device for switching the communication device.

[0046] Based on the above scheme, in the case where the communication device may switch to other access network devices that belong to different manufacturers as the first access network device, the first mobile intelligent function corresponding to the first access network device can further consider the manufacturer information when making a decision to switch the communication device, and give priority to selecting the second access network device that belongs to the same manufacturer as the first access network device as the target access network device, thereby avoiding resource exposure due to selecting access network devices from different manufacturers and ensuring model security.

[0047] In combination with the first aspect, in certain implementations of the first aspect, when the communication device has successfully switched to the second access network device, a switching success notification message is received, and the switching success notification message includes an identifier of the communication device and an identifier of the second access network device.

[0048] In combination with the first aspect, in certain implementations of the first aspect, AI training task information is sent to a second access network device, the AI ​​training task information including an intermediate training result of the AI ​​model or model training information of the communication device, and an identifier of the first part, and the intermediate training result or the model training information of the communication device is used by the second access network device to continue to execute the first part of the training task that is to be executed by the first access network device.

[0049] Based on the above scheme, after the communication device completes the switching, the first mobile intelligent function corresponding to the first access network device can send AI training task information to the second mobile intelligent function corresponding to the second access network device, so that the second mobile intelligent function can continue to execute the previously unfinished AI training task with the communication device based on the AI ​​training task information, thereby avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0050] In a second aspect, a handover processing method is provided, which can be executed by a second access network device, or by a component (such as a chip or circuit) of the second access network device. For convenience, the second access network device is used as an example for illustration. As an example, the second network device can be a target base station.

[0051] The method includes: when a communication device is about to switch from a first access network device to a second access network device, receiving an intermediate training result of an artificial intelligence (AI) model or model training information of the communication device, and an identifier of a first part of a training task that is to be executed by the first access network device, wherein the model training information is the model parameters and / or training data that the communication device last reported to the first access network device; based on the intermediate training result, continuing to execute the first part of the training task that is to be executed by the first access network device; or, constructing an AI model based on the model training information of the communication device; and based on the constructed AI model, continuing to execute the first part of the training task that is to be executed by the first access network device.

[0052] It should be understood that this technical solution is applicable to the process in which the first access network device and the communication device jointly perform a training task. The training task refers to a task of obtaining an AI model through AI training based on model parameters or measurement data. The AI ​​model is used by the first access network device and the communication device to jointly perform model reasoning.

[0053] Based on the above scheme, when it is determined that the communication device will switch, the second access network device continues to execute the first part of the training task according to the intermediate training results, or builds an AI model and continues to execute the first part of the training task according to the constructed AI model. That is, after the communication device switches to the second access network device, the communication device and the second access network device can continue to complete the training task, thereby improving the efficiency of AI training while reducing resource waste during the AI ​​training process.

[0054] In combination with the second aspect, in certain implementations of the second aspect, third AI capability information is sent, the third AI capability information indicates the AI ​​training capability supported by the second access network device, and the third AI capability is used to determine that the communication device needs to switch to the second access network device.

[0055] In combination with the second aspect, in certain implementations of the second aspect, the third AI capability information includes at least one of the following: a third AI switch, a supported third AI algorithm type, supported third AI framework information, third computing power information, or retraining capability information, wherein the retraining capability information indicates whether the second access network device supports retraining of an already trained AI model, and the third computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

[0056] In combination with the second aspect, in certain implementations of the second aspect, a switching request message is received, the switching request message is used to request that the communication device be switched to a second access network device, the switching request message includes an intermediate training result of the AI ​​model or model training information of the communication device, and an identifier of the first part; access control of the communication device is completed according to the switching request information.

[0057] In combination with the second aspect, in certain implementations of the second aspect, manufacturer information is sent, the manufacturer information indicates a manufacturer providing the second access network device, and the manufacturer information is used to determine that the communication device will switch from the first access network device to the second access network device.

[0058] In combination with the second aspect, in certain implementations of the second aspect, receiving switching request information includes: when the first access network device and the second access network device are provided by the same manufacturer, receiving first switching request information, wherein the first switching request information includes the intermediate training result of the AI ​​model and the identifier of the first part.

[0059] In combination with the second aspect, in certain implementations of the second aspect, receiving switching request information includes: when the first access network device and the second access network device are provided by different manufacturers, receiving second switching request information, wherein the second switching request information includes model training information of the communication device and an identifier of the first part.

[0060] In combination with the second aspect, in certain implementations of the second aspect, the switching request information also includes an identifier of the second part. Before continuing to execute the first part of the training task that is to be executed by the first access network device, the method also includes: sending training task startup information to the communication device, and the training task startup information includes the identifier of the second part.

[0061] The beneficial effects of the above-mentioned second aspect and some implementation methods of the second aspect can be referred to the relevant description of the first aspect, which will not be repeated here.

[0062] In a third aspect, a handover processing method is provided, which can be executed by a communication device or by a component (such as a chip or circuit) of the communication device. For convenience, the communication device is used as an example for description. As an example, the communication device can be a terminal device or a user equipment.

[0063] The method includes: when a communication device is about to switch from a first access network device to a second access network device, receiving training task pause information or training task termination information, the training task pause information or the training task termination information including an identifier of a second part of the training task to be executed by the communication device; when receiving the training task pause information, receiving first training task start information, the first training task start information including an identifier of the second part; based on the identifier of the second part, continuing to execute the second part of the training task to be executed by the communication device; when receiving the training task termination information, terminating the execution of the second part of the training task to be executed by the communication device.

[0064] It should be understood that this technical solution is applicable to the process in which the first access network device and the communication device jointly perform a training task. The training task refers to a task of obtaining an AI model through AI training based on model parameters or measurement data. The AI ​​model is used by the first access network device and the communication device to jointly perform model reasoning.

[0065] Based on the above scheme, when it is determined that the communication device will switch, the communication device receives training task pause information or training task termination information to avoid wasting resources due to continuing training or reporting model information or measurement data to the first access network device due to switching.

[0066] In combination with the third aspect, in certain implementations of the third aspect, the training task termination information or the training task suspension information further includes a reason value, and the reason value is used to indicate that a switch occurs in the communication device.

[0067] In combination with the third aspect, in certain implementations of the third aspect, before receiving training task pause information or training task termination information, the method also includes: sending first AI capability information, the first AI capability information indicating the AI ​​training capability supported by the communication device; receiving second training task start information, the second training task start information including model information of the AI ​​model and an identifier of the second part, the model information is determined based on the first AI capability information and the second AI capability information, the second AI capability information indicating the AI ​​training capability supported by the first access network device; based on the model information of the AI ​​model and the identifier of the second part, executing the second part of the training task that is to be executed by the communication device.

[0068] In combination with the third aspect, in some implementations of the third aspect, the first AI capability information includes at least one of the following: a first AI switch, a supported first AI algorithm type, supported first AI framework information, or first computing power information, wherein the first computing power information includes at least one of the following: hardware type, computing power, or resource utilization;

[0069] In combination with the third aspect, in certain implementations of the third aspect, the second AI capability information includes at least one of the following: a second AI switch, a supported second AI algorithm type, supported second AI framework information, or second computing power information, wherein the second computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

[0070] In combination with the third aspect, in certain implementations of the third aspect, the model information of the AI ​​model includes at least one of the following: model structure, hyperparameters, model accuracy, or number of training rounds.

[0071] In combination with the third aspect, in certain implementations of the third aspect, when the communication device switches back to the first access network device, the second part of the training task that is to be performed by the communication device continues to be executed.

[0072] The beneficial effects of the third aspect and certain implementation methods of the third aspect can be referred to the relevant description of the first aspect, which will not be repeated here.

[0073] In a fourth aspect, a switching processing device is provided, such as a first access network device, comprising: a processing unit, used to determine that a communication device will switch from the first access network device to a second access network device; the processing unit is also used to terminate or suspend a first part of a training task that is to be executed by the first access network device, and to trigger the communication device to terminate or suspend a second part of a training task that is to be executed by the communication device.

[0074] The transceiver unit can perform the reception and transmission processing in the aforementioned first aspect, and the processing unit can perform other processing except reception and transmission in the aforementioned first aspect.

[0075] In a fifth aspect, a switching processing device is provided, such as a second access network device, including: a transceiver unit, used to receive the intermediate training results of the artificial intelligence AI model or the model training information of the communication device, and the identifier of the first part of the training task that belongs to the first access network device to be executed when the communication device is to switch from the first access network device to the second access network device, the model training information is the model parameters and / or training data most recently reported by the communication device to the first access network device; a processing unit, used to continue to execute the first part of the training task that belongs to the first access network device to be executed according to the intermediate training results; or, a processing unit, used to construct an AI model according to the model training information of the communication device; and continue to execute the first part of the training task that belongs to the first access network device to be executed according to the constructed AI model.

[0076] The transceiver unit can perform the reception and transmission processing in the aforementioned second aspect, and the processing unit can perform other processing except reception and transmission in the aforementioned second aspect.

[0077] In a sixth aspect, a switching processing device is provided, such as a communication device (such as a terminal device), comprising: a transceiver unit, used to receive training task pause information or training task termination information when the communication device is to switch from a first access network device to a second access network device, the training task pause information or the training task termination information including an identifier of a second part of the training task to be executed by the communication device; when the transceiver unit receives the training task pause information, the transceiver unit is also used to receive first training task start information, the first training task start information including an identifier of the second part; a processing unit, used to continue executing the second part of the training task to be executed by the communication device based on the identifier of the second part; when the transceiver unit receives the training task termination information, the processing unit is also used to terminate the execution of the second part of the training task to be executed by the communication device.

[0078] The transceiver unit can perform the receiving and sending processing in the fifth aspect, and the processing unit can perform other processing except receiving and sending in the fifth aspect.

[0079] In a seventh aspect, a switching processing device is provided, comprising a processor, wherein the processor is used to call and run the computer program from a memory, so that the switching processing device executes the method in any possible implementation of the first to third aspects above.

[0080] Optionally, there are one or more processors.

[0081] Optionally, the switching processing device further includes one or more memories, and the memories may be integrated with the processor, or the memories may be separately arranged from the processor.

[0082] Optionally, the switching processing device further includes a transmitter (transmitter) and a receiver (receiver).

[0083] In an eighth aspect, a communication system is provided, comprising a first access network device and a second access network device, the first access network device being used to execute the method in any possible implementation manner of the above-mentioned first aspect, and the second access network device being used to execute the method in any possible implementation manner of the above-mentioned second aspect.

[0084] Optionally, the communication system may also include a first mobile intelligent function, a second mobile intelligent function and a communication device, the first mobile intelligent function is used to execute the method in any possible implementation of the first aspect above, the second mobile intelligent function is used to execute the method in any possible implementation of the second aspect above, and the communication device is used to execute the method in any possible implementation of the third aspect above.

[0085] In the ninth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program or code, and when the computer program or code is run on a computer, the computer executes a method in any possible implementation of the first to third aspects above.

[0086] In the tenth aspect, a chip is provided, comprising at least one processor, wherein the at least one processor is coupled to a memory, the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that a device equipped with the chip system executes a method in any possible implementation of the first to third aspects above.

[0087] The chip may include an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0088] In the eleventh aspect, a computer program product is provided, the computer program product comprising: a computer program code, when the computer program code is executed by a device, the device executes a method in any possible implementation of the first to third aspects above. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a schematic diagram of a network architecture 100 provided in an embodiment of the present application;

[0090] Figure 2 is a schematic diagram of a network architecture 200 provided in an embodiment of the present application;

[0091] Figure 3 is a schematic diagram of a network architecture 300 provided in an embodiment of the present application;

[0092] Figure 4 is a flowchart of a switching processing method 400 provided in an embodiment of the present application;

[0093] Figure 5 is a flowchart of a switching processing method 500 provided in an embodiment of the present application;

[0094] Figure 6 is a schematic block diagram of a switching processing device 1000 provided in an embodiment of the present application;

[0095] Figure 7 is a schematic block diagram of a switching processing device 2000 provided in an embodiment of the present application;

[0096] Figure 8 It is a schematic block diagram of a switching processing device 3000 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0097] The technical solution in this application will be described below in conjunction with the accompanying drawings.

[0098] Figure 1 1 is a schematic diagram of a communication system 100 provided in an embodiment of the present application. Figure 1 As shown, the communication system 100 includes a network device 10 and a terminal device 20. In the communication system, the terminal device 20 can send uplink data / signals / information to the network device 10, and the network device 10 can send downlink data / signals / information to the terminal device 20.

[0099] The handover processing method provided in the embodiment of the present application may also involve Figure 1 The devices or transmission nodes not shown in the figure are not limited in the embodiments of the present application.

[0100] The above-mentioned communication system applied to the embodiment of the present application is only an example, and the communication system applicable to the embodiment of the present application is not limited to this. Any communication system that can realize the functions of the above-mentioned devices is applicable to the embodiment of the present application.

[0101] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as: long term evolution (LTE) system, advanced long term evolution (LTE-A) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), NR system and other fifth generation (5G) systems, or future evolved communication systems (for example, 6G mobile communication system). The technical solutions of the embodiments of the present application can be applied to various scenarios such as vehicle-to-everything (V2X), Internet of Vehicles, machine type communication (MTC), Internet of Things (IoT), long term evolution-machine (LTE-M), machine to machine (M2M), etc.

[0102] The terminal device may be a wireless terminal device capable of receiving access network device scheduling and indication information. The terminal device may be a device that provides voice and / or data connectivity to a user, or a handheld device with wireless connection function, or other processing device connected to a wireless modem.

[0103] The terminal device in the embodiments of the present application may also be referred to as a terminal, an access terminal, a user unit, a user equipment (UE), a user station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a wireless communication device, a user agent or a user device. The terminal device is a device that includes a wireless communication function (providing voice / data connectivity to the user). For example, a handheld device with a wireless connection function, or a vehicle-mounted device, etc. The terminal device in the embodiments of the present application can be a mobile phone, a tablet computer, a computer with wireless transceiver function, a train, an airplane, a mobile Internet device (MID), a virtual reality (VR) terminal, an augmented reality (AR) terminal, a wireless terminal in industrial control (such as a robot, etc.), a wireless terminal in the Internet of Vehicles (for example, vehicle-mounted equipment, vehicle equipment, vehicle-mounted modules, vehicles, etc.), a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a wearable device, etc. In addition, the terminal device can also be used in scenarios such as unmanned driving, telemedicine, smart grid, transportation safety, smart city, smart city, smart home, etc.

[0104] The network device may be a device in a wireless network. For example, the network device may be a device deployed in a wireless network to provide wireless communication functions for a terminal device. For example, the network device may be a radio access network (RAN) node that connects the terminal device to the wireless network. The RAN may be connected to a core network (for example, a core network of long term evolution (LTE) or a core network of 5G).

[0105] The network device in the embodiment of the present application may be an access network device, including but not limited to: various base stations, such as the next generation Node B (gNodeB, gNB), evolved Node B (evolved Node B, eNB) or base station equipment in the future evolved communication system, or a server, a wearable device, a vehicle-mounted device, a wireless relay node, a wireless backhaul node, a transmission point (TP) or a transmission and reception point (TRP), etc., or one or a group of antenna panels (including multiple antenna panels) of a base station, or a network node constituting a base station, such as a baseband unit (BBU), a centralized unit (CU), or a distributed unit (DU), etc. Among them, the base station may be a macro base station, a micro base station, a micro-micro base station, a small station, a relay station or a balloon station, etc.

[0106] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0107] In some deployments, the network device in the embodiments of the present application may refer to a CU or a DU, or the network device includes a CU and a DU. The network device may also include an active antenna unit (AAU). As an example, see Figure 1 , Figure 1A schematic diagram of the CU and DU architecture is shown. The CU implements some functions of the gNB, and the DU implements some functions of the gNB. For example, the CU is responsible for processing non-real-time protocols and services, and implements the functions of the radio resource control (RRC) layer and the packet data convergence protocol (PDCP) layer. The DU is responsible for processing physical layer protocols and real-time services, and implements the functions of the radio link control (RLC) layer, the medium access control (MAC) layer, and the physical (PHY) layer. The AAU implements some physical layer processing functions, RF processing, and related functions of active antennas. Since the information of the RRC layer will eventually become the information of the PHY layer, or be converted from the information of the PHY layer, therefore, under this architecture, high-level signaling, such as RRC layer signaling, can also be considered to be sent by the DU, or, sent by the DU+AAU. It can be understood that the network device can be a device including one or more of a CU node, a DU node, and an AAU node. In addition, the CU may be classified as a network device in an access network (radio access network, RAN), or the CU may be classified as a network device in a core network (core network, CN), which is not limited in the present application.

[0108] Further, the CU can also be divided into a central unit-control plane (CU-CP) of the control plane and a central unit-user plane (CU-UP) of the user plane. Among them, the CU-CP and the CU-UP can also be deployed on different physical devices. The CU-CP is responsible for the control plane function, mainly including the RRC layer and the PDCP-C layer. The PDCP-C layer is mainly responsible for the encryption and decryption, integrity protection, data transmission, etc. of the control plane data. The CU-UP is responsible for the user plane function, mainly including the service data adaptation protocol (SDAP) layer and the PDCP-U layer. Among them, the SDAP layer is mainly responsible for processing the data of the core network and mapping the flow to the bearer. The PDCP-U layer is mainly responsible for at least one function of encryption and decryption, integrity protection, header compression, sequence number maintenance, data transmission, etc. of the data plane. Specifically, the CU-CP and the CU-UP are connected through a communication interface (e.g., an E1 interface). The CU-CP represents that the network device is connected to the core network device through a communication interface (e.g., an Ng interface), and is connected to the DU through a communication interface (e.g., an F1-C (control plane) interface). The CU-UP is connected to the DU via a communication interface (eg, F1-U (user plane) interface).

[0109] It can be understood that the above protocol layer divisions of CU and DU, and CU-CP and CU-UP are only examples, and there may be other division methods, which are not limited to this. The network device involved in the embodiment of the present application may be a device including a CU, or a DU, or a device including a CU and a DU, or a control plane CU node (CU-CP node) and a user plane CU node (CU-UP node) and a DU node.

[0110] In different systems, CU (including CU-CP or CU-UP) or DU may also have different names, but those skilled in the art can understand their meanings. For example, in an open radio access network (O-RAN) system, CU may also be referred to as O-CU (open CU), DU may also be referred to as O-DU, CU-CP may also be referred to as O-CU-CP, and CU-UP may also be referred to as O-CU-UP. For the convenience of description, this application takes CU, CU-CP, CU-UP, and DU as examples for description.

[0111] The network device in the embodiment of the present application may also be a core network device, including but not limited to: Specifically, the core network device refers to a device in the core network (CN) that provides service support for the terminal device. The core network device may include one or more core network elements. Taking the 5G core network as an example, the 5G core network includes an access and mobility management function (AMF) network element responsible for services such as mobility management and access management, a session management function (SMF) network element responsible for session management, a user plane function (UPF) network element responsible for data packet routing and forwarding and quality of service (QoS) control on the user plane, and a policy control function (PCF) network element. The above-mentioned core network elements can work independently or be combined to implement certain control functions, such as: AMF, SMF and PCF can be combined together as a core network device.

[0112] It can be understood that all or part of the functions of the network device or terminal device in the present application can also be implemented through software functions running on hardware, or through virtualization functions instantiated on a platform (such as a cloud platform).

[0113] Figure 2 2 is a schematic diagram of a communication system 200 provided in an embodiment of the present application. Figure 2As shown, gNBs are connected through Xn interfaces, and gNBs and UEs are connected through Uu interfaces. Among them, gNB includes centralized units CU and distributed units DU. One CU can be connected to one or more DUs, or multiple DUs can share one CU. CU and DU can be understood as the division of gNB from the perspective of logical functions. CU and DU can be physically separated or deployed together. CU and DU can be connected through an interface, such as F1 interface. CU and DU can be divided according to the protocol layer of the wireless network. One possible division method is that CU processes non-real-time RAN high-level protocol stack functions, such as functions for executing the radio resource control (RRC) layer, service data adaptation protocol (SDAP) layer, and packet data convergence protocol (PDCP) layer, and DU is used to process RAN low-level protocol stack functions with higher real-time requirements, such as functions for executing the radio link control (RLC) layer, media access control (MAC) layer, physical layer, etc.

[0114] It can be understood that the division of the processing functions of CU and DU according to the protocol layer is only an example, and CU and DU can also be divided in other ways. For example, CU or DU can be divided into functions of more protocol layers than the above division method. For example, CU or DU can also be divided into part of the processing functions of the protocol layer of the above division method. In one design, the functions of CU or DU can be divided according to service type or other system requirements. For example, according to the delay division, the function whose processing time meets the delay requirement is set to DU implementation, and the function that does not meet the delay requirement is set to CU implementation. In another design, CU may also have one or more functions of the core network. One or more CUs can be set centrally or separately. For example, CU can be set on the network side for convenient centralized management. DU can have multiple radio frequency functions, or the radio frequency function can be set remotely.

[0115] The functions of CU can be implemented by one entity or by different entities. For example, the functions of CU can be further divided, for example, the control plane (CP) and the user plane (UP) are separated, that is, the control plane (CU-CP) of CU and the user plane (CU-UP) of CU. The control plane CU-CP of CU also includes a further divided architecture, that is, the CU-CP is further divided into CU-CP1 and CU-CP2. Among them, CU-CP1 includes various radio resource management functions, and CU-CP2 only includes RRC functions and PDCP-C functions (that is, the basic functions of control plane signaling at the PDCP layer).

[0116] In different systems, CU (including CU-CP or CU-UP) or DU may also have different names, but those skilled in the art can understand their meanings. For example, in an open radio access network (O-RAN) system, CU may also be referred to as O-CU (open CU), DU may also be referred to as O-DU, CU-CP may also be referred to as O-CU-CP, and CU-UP may also be referred to as O-CU-UP. For the convenience of description, this application takes CU, CU-CP, CU-UP, and DU as examples for description.

[0117] The AI ​​module is a module with machine learning computing capabilities. In a wireless communication system, the AI ​​module can be located in operations, administration and maintenance (OAM), in access network equipment (for example, if the network equipment is a separated architecture, it can be located in the CU), in the terminal equipment, or as a separate network element entity AI control layer (AI control layer, AIC). The main function of the AI ​​module in the wireless communication system is to perform a series of AI calculations such as model building, training approximation, and reinforcement learning based on input data. The trained model provided by the AI ​​module has a prediction function for network changes on the RAN side, which can be used for load prediction, terminal device path prediction, etc. In addition, the AI ​​module can also perform policy reasoning from the perspectives of network energy saving and mobility optimization based on the prediction results of the trained model for RAN network performance, so as to obtain reasonable and efficient energy saving strategies, mobility optimization strategies, etc. When the AI ​​model is located in the CU and the CP and UP of the CU are separated, the CP can be responsible for receiving the AI ​​model and the subsequent AI reasoning and policy generation functions. When CU-CP is further divided into CU-CP1 and CU-CP2, CU-CP1 can be responsible for receiving the AI ​​model and the reasoning function of the subsequent AI model, and generating specific interaction signaling, which will be sent by CU-CP2.

[0118] In addition, the gNB also deploys the model training function, which can be used as a new independent functional module in the gNB or deployed on the CU or DU. Figure 2 ,exist Figure 3 The model training function in the communication system 300 shown is deployed on a newly added independent mobile intelligent function (MIF). MIF is the AI ​​intelligent business processing logic function entity of the RAN network, which is used to implement the intelligent data analysis function of the RAN domain. MIF is connected to the gNB through the G1 interface. Among them, the MIF is deployed with model training functions and model evaluation functions related to the AI ​​model. Optionally, the MIF may also be deployed with a computing power perception function. The AI ​​model training function is responsible for model training, and an ML model is generated after the model training is completed. The computing power perception function is responsible for the perception of the computing power status of the hardware inside the MIF and the computing power status of the CU / DU devices, and can determine the computing power requirements of the AI ​​model training task based on the training and reasoning requirements of the AI ​​model. Optionally, MIF can also be deployed inside the gNB, and this application does not make specific limitations on this.

[0119] The above-mentioned network architecture applied to the embodiment of the present application is only an example, and the network architecture applicable to the embodiment of the present application is not limited to this. Any network architecture that can realize the functions of the above-mentioned network elements is applicable to the embodiment of the present application, that is, the network architecture and business scenarios described in the embodiment of the present application are for the purpose of more clearly illustrating the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. Ordinary technicians in this field can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0120] It can be understood that the network elements or devices listed in the above network architecture are only exemplary, and the network architecture applicable to the present application may also include other network elements or devices, which are not limited in the present application. The naming of the above network elements or devices is only defined for the convenience of distinguishing different functions and should not constitute any limitation to the present application. The present application does not exclude the possibility of adopting other names in 4G networks, 5G networks and other networks in the future. For example, in a 6G network, some or all of the above network elements may use the terminology in 4G / 5G, or may use other names.

[0121] The communication device in the embodiment of the present application includes a hardware layer, an operating system layer running on the hardware layer, and an application layer running on the operating system layer. The hardware layer includes hardware such as a central processing unit (CPU), a memory management unit (MMU), and a memory (also called main memory). The operating system can be any one or more computer operating systems that implement business processing through a process (Process), for example, a Linux operating system, a Unix operating system, an Android operating system, an iOS operating system, or a windows operating system. The application layer includes applications such as browsers, address books, word processing software, and instant messaging software. In addition, the embodiment of the present application does not specifically limit the specific structure of the execution subject of the method provided in the embodiment of the present application, as long as it can communicate according to the method provided in the embodiment of the present application by running a program that records the code of the method provided in the embodiment of the present application, for example, the execution subject of the method provided in the embodiment of the present application can be a terminal device or a network device, or a functional module in a terminal device or a network device that can call a program and execute a program.

[0122] In addition, various aspects or features of the embodiments of the present application can be implemented as methods, devices or products using standard programming and / or engineering techniques. The term "product" used in this application covers computer programs that can be accessed from any computer-readable device, carrier or medium. For example, computer-readable media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks or tapes, etc.), optical disks (e.g., compact discs (CDs), digital versatile discs (DVDs), etc.), smart cards and flash memory devices (e.g., erasable programmable read-only memory (EPROM), cards, sticks or key drives, etc.). In addition, the various storage media described herein may represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media capable of storing, containing and / or carrying instructions and / or data.

[0123] For air interface AI, typical use cases include beam management, positioning, and channel state information (CSI) feedback. Usually, AI model reasoning is performed by the gNB or UE alone. This AI model usage mode is called unilateral AI, and the corresponding AI model is a unilateral AI model. For example, use cases such as beam management and positioning use unilateral AI models. In comparison, the AI ​​model consists of two parts. The gNB and UE each need to use a part of the AI ​​model for reasoning, and the model output of the first part is the model input of the second part (such as: the model output on the UE side is the model input on the gNB side). This AI model usage mode is called bilateral AI, and the corresponding AI model is a bilateral AI model. For example, the CSI feedback use case uses a bilateral AI model. For example, in a frequency division multiplexing network, the gNB needs to obtain the downlink CSI fed back by the UE to perform precoding and achieve performance gain. In order to reduce the overhead of CSI feedback, the AutoEncoder network can be used to build a CSI feedback and reconstruction mechanism. AutoEncoder is a typical bilateral AI model consisting of an encoder model and a decoder model. The encoder model is the first part of the bilateral AI model. The UE performs model reasoning and is used to compress the original CSI matrix. The input is the original CSI and the output is the compressed CSI. The decoder model is the second part of the bilateral AI model. The gNB performs model reasoning and is used to restore the received compressed CSI to the original CSI using a neural network. The input is the compressed CSI and the output is the restored original CSI. In other words, for the training of the bilateral AI model, one possible implementation method is the joint training of the UE and the gNB. The UE side and the gNB side train their respective AI models based on local data, that is, the UE side trains the encoder model and the gNB trains the decoder model. The UE and the gNB exchange the model parameters and / or training data in the intermediate process of the air interface model training, thereby realizing joint collaborative training. During the entire training process, the UE and the gNB need to exchange model parameters / data and other information for multiple rounds to converge the bilateral AI model. Due to the limited AI computing power of the UE device and the need for multiple rounds of interaction and iteration during training, the training convergence time is generally long. Considering that if the AI ​​model training has not converged yet, the UE moves or the base station / cell is switched, it will lead to a waste of resources on the UE side and the gNB side.

[0124] In response to the above-mentioned technical problems, the embodiments of the present application provide a switching processing method, device and system. When it is determined that a communication device will switch from a first access network device to a second access network device, the first access network device terminates or suspends the first part of the training task that belongs to the first access network device, and simultaneously triggers the communication device to terminate or suspend the second part of the training task that belongs to the communication device, thereby avoiding the communication device from continuing training and data reporting, that is, continuing to execute the second part of the training task. Waste of resources caused by, and avoiding the first access network device continuing to execute the first part of the training task. Waste of resources caused by.

[0125] It should be noted that in an embodiment of the present application, the first part of the training task that is to be executed by the first access network device corresponds to the first part of the model in the bilateral AI model, that is, the first access network device executes the first part of the training task, and the first part of the model of the bilateral AI model can be obtained; similarly, the second part of the training task that is to be executed by the communication device corresponds to the second part of the model in the bilateral AI model, that is, the first access network device executes the second part of the training task, and the second part of the model of the bilateral AI model can be obtained.

[0126] To facilitate understanding of the embodiments of the present application, the following points are explained:

[0127] First, in this application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships.

[0128] It can be understood that some optional features in the embodiments of the present application may not depend on other features in some scenarios, and may also be combined with other features in some scenarios, without limitation.

[0129] It can be understood that the schemes in the various embodiments of the present application can be used in combination, and the explanations or descriptions of the various terms, similar operations or steps appearing in the embodiments can be referenced or explained to each other in the various embodiments, and the present application does not limit this.

[0130] Second, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. In the text description of this application, the character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can mean: a, or b, or c, or a and b, or a and c, or b and c, or a, b and c. Where a, b and c can be single or multiple, respectively.

[0131] Third, in the present application, "first", "second" and various numerical references are used to distinguish for the convenience of description and are not used to limit the scope of the embodiments of the present application. For example, different messages are distinguished, rather than used to describe a specific order or sequence. It should be understood that the objects described in this way can be interchanged where appropriate so as to be able to describe solutions other than the embodiments of the present application.

[0132] Fourth, in this application, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0133] Fifth, in this application, "used for indication" may include being used for direct indication and being used for indirect indication. When describing that a certain indication information is used for indicating A, it may include that the indication information directly indicates A or indirectly indicates A, but it does not mean that the indication information must carry A.

[0134] Sixth, in this application, "sending information to XX (device)" can be understood as the destination of the information being the device. It can include sending information to the device directly or indirectly. "Receiving information from XX (device) or receiving information from XX (device)" can be understood as the source of the information being the device, which can include receiving information from the device directly or indirectly. The information may be processed as necessary between the source and destination of the information, such as format changes, but the destination can understand the valid information from the source.

[0135] Seventh, in this application, "storage" or "saving" may refer to storage in one or more memories. The one or more memories may be separately set or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories may also be partially separately set and partially integrated in a decoder, a processor, or a communication device. The type of memory may be any form of storage medium, which is not limited by this application.

[0136] Eighth, the “protocol” involved in the embodiments of the present application may refer to a standard protocol in the field of communications, for example, it may include a fifth generation (5G) network protocol, a new radio (NR) protocol, and related protocols used in future communication systems. This application does not limit this.

[0137] The following will describe in detail the handover processing method provided by the embodiment of the present application in conjunction with the accompanying drawings. The embodiment provided by the present application can be applied to any communication scenario in which a transmitting device and a receiving device communicate, for example, it can be applied to the above Figures 1 to 3 In order to avoid the waste of resources caused by continuing training and data reporting during the switching process, the present application proposes the following Figure 4 and Figure 5 In the method shown, when the first access network device determines that the communication device will be switched, the first part of the training task that is to be executed by the first access network device can be suspended or terminated, and the communication device can be triggered to suspend or terminate the second part of the training task that is to be executed by the communication device. Furthermore, in order to improve the efficiency of AI model training, after the communication device is switched to the second access network device, the communication device and the second access network device can continue to execute the training task. It should be understood that

[0138] Figure 4 is a flow chart of a switching processing method 400 provided in an embodiment of the present application, which can be applied to the above Figures 1 to 3 In the communication system shown. Exemplarily, the method flow can be executed by a communication device (e.g., UE), a first access network device (e.g., a source base station), and a second access network device (e.g., a target base station), or by a module and / or device (e.g., a chip or an integrated circuit, etc.) with corresponding functions installed in the communication device, the first access network device, and the second access network device, and the present application does not limit it. For ease of description, the following description is based on the communication device, the first access network device, and the second access network device as the execution subjects, such as Figure 4 As shown, the method includes the following steps, and the incomplete parts can refer to the existing protocol.

[0139] It should be understood that this technical solution is applicable to the process in which the first access network device and the communication device jointly perform a training task. The training task refers to a task of obtaining an AI model by performing AI training based on model parameters or measurement data. The AI ​​model is used by the first access network device and the communication device to jointly perform model reasoning.

[0140] S410: The first access network device determines that the communication apparatus will be switched from the first access network device to the second access network device.

[0141] Optionally, before executing step S410, the method 400 also includes steps S401 to S403, namely, initial capability reporting of the communication device and an AI model training task startup process.

[0142] S401: The communication device sends first AI capability information to a first access network device. Correspondingly, the first access network device receives the first AI capability information.

[0143] The first AI capability information indicates the AI ​​training capability supported by the communication device.

[0144] Exemplarily, the first AI capability information includes at least one of the following:

[0145] (1) First AI switch.

[0146] For example, used to indicate whether the communication device supports AI, the value may include 1 or 0, or YES or NO, or true or false, wherein a value of 1 or YES or true is used to indicate that the communication device supports AI, and a value of 0 or NO or false is used to indicate that the communication device does not support AI.

[0147] (2) The first AI algorithm type supported.

[0148] For example, it is used to indicate the type of AI algorithm supported by the communication device, including but not limited to: decision tree or neural network, etc.

[0149] (3) Information about the first AI framework supported.

[0150] For example, it is used to indicate the AI ​​framework supported by the communication device, including but not limited to: tensorflow or keras.

[0151] (4) First computing power information.

[0152] For example, it is used to indicate the computing power supported by the communication device, including but not limited to: hardware type, computing power, or resource utilization.

[0153] (a) Hardware types include one or more of the following: central processing unit (CPU), graphics processing unit (GPU), neural network processing unit (NPU), etc.

[0154] (b) Computing power: This can generally be measured by the number of floating-point operations per second (FLOPS) supported.

[0155] (c) Resource utilization: indicates the current computing resource utilization, for example, 80%.

[0156] S402: The first access network device determines model information for training an AI model according to the first AI capability information and the second AI capability information.

[0157] The second AI capability information indicates the AI ​​training capability supported by the first access network device.

[0158] Exemplarily, the second AI capability information includes at least one of the following:

[0159] (1) Second AI switch.

[0160] For example, used to indicate whether the first access network device supports AI, the value may include 1 or 0, or YES or NO, or true or flase, wherein a value of 1 or YES or true is used to indicate that the first access network device supports AI, and a value of 0 or NO or flase is used to indicate that the first access network device does not support AI.

[0161] (2) Second AI algorithm type supported.

[0162] For example, it is used to indicate the AI ​​algorithm types supported by the first access network device, including but not limited to: decision tree or neural network, etc.

[0163] (3) Supported second AI framework information.

[0164] For example, it is used to indicate the AI ​​framework supported by the first access network device, including but not limited to: tensorflow or keras.

[0165] (4) Second computing power information.

[0166] For example, the second computing power information includes at least one of the following: hardware type, computing power, or resource utilization. For specific explanations, please refer to the above related descriptions and will not be repeated here.

[0167] Exemplarily, the model information of the AI ​​model includes at least one of the following:

[0168] (1) Model structure

[0169] The model structure refers to the model structure determined before the AI ​​model is trained, such as a fully connected neural network structure, a convolutional neural network structure, or a random forest structure.

[0170] (2) Super parameters.

[0171] Hyperparameters are also called hyperparameters, which refer to a set of configuration parameters used to determine the AI ​​model together with the model structure before the AI ​​model is trained. For example, if the model structure adopts a fully connected neural network structure, the hyperparameters include but are not limited to: learning rate, number of layers, number of neurons in each layer, etc. It should be understood that hyperparameters are generally determined based on experience.

[0172] (3) Model accuracy.

[0173] Model accuracy is used to indicate how close the model output value is to the true value. Generally speaking, the greater the error between the output value and the true value, the lower the model accuracy, and the smaller the error between the output value and the true value, the higher the model accuracy. For example, the expected CSI model accuracy is 90%.

[0174] (4) Number of training rounds.

[0175] The number of training rounds can be understood as the training requirements of the source gNB based on the AI ​​model training. In order to ensure the effect of model training, the source gNB puts forward the requirement for the number of training rounds, for example, 5 rounds.

[0176] Optionally, the present application does not limit the number of AI models, and the AI ​​model can be one or more.

[0177] In one example, the first access network device determines the model information based on the training requirements, the second AI capability information and the first AI capability information. For example, the first AI capability information indicates that the communication device supports AI, the supported AI algorithm type is a neural network, the supported AI framework is tensorflow, and the resource utilization rate is 60%. Similarly, the second AI capability information indicates that the first access network device supports AI, the supported AI algorithm type is a neural network, the supported AI framework is tensorflow, and the resource utilization rate is 80%. Before the AI ​​model is trained, the model information determined by the first access network device includes: the model structure adopts a convolutional neural network structure, the model accuracy is 90%, the number of training rounds is 10, etc.

[0178] Among them, the training requirements can be understood as the requirements for the AI ​​model trained by the first access network device and the communication device, for example: it is necessary to train a bilateral AI model to complete CSI feedback with an accuracy of 90%. The first access network device determines the training task in combination with the training requirements to obtain an AI model that meets the training requirements. It should be understood that this application is applicable to the training process of the AI ​​model, including a series of AI calculations such as AI model establishment, training approximation, and reinforcement learning.

[0179] S403, the first access network device sends training task start information to the communication apparatus, and correspondingly, the communication apparatus receives the training task start information from the first access network device.

[0180] Among them, the training task startup information includes model information of the AI ​​model and the identifier of the second part, so that the communication device executes the second part of the training task that is to be executed by the communication device.

[0181] In one example, the first access network device sends the identifier of the second part and the model information determined in the above step S402 to the communication device, and instructs the communication device to start the AI ​​model training task. The identifier of the second part can also be called a training ID, which is used to identify the second part of the AI ​​model training task that is to be executed by the communication device, and the model information can determine the specific implementation method of the training task of the AI ​​model.

[0182] Based on the above steps S401 to S403, the first access network device executes the first part of the training task that is to be executed by the first access network device according to the model information of the AI ​​model. At the same time, the communication device executes the second part of the training task that is to be executed by the communication device according to the model information of the AI ​​model.

[0183] Optionally, before executing step S410, the method 400 also includes steps S404 and / or S405, that is, the second access network device reports AI capability information and / or manufacturer information to the first access network device, so that the first access network device can make a switching decision for the communication device based on the AI ​​capability information of the second access network device.

[0184] S404: The second access network device sends third AI capability information to the first access network device. Correspondingly, the first access network device receives the third AI capability information from the second access network device.

[0185] Among them, the third AI capability information indicates the AI ​​training capability supported by the second access network device.

[0186] Exemplarily, when the Xn interface is established, the second access network device sends the third AI capability information to the first access network device through the Xn interface, and correspondingly, the first access network device receives the third AI capability information from the second access network device through the Xn interface.

[0187] Exemplarily, the third AI capability information includes at least one of the following:

[0188] (1) Third AI switch.

[0189] For example, used to indicate whether the second access network device supports AI, the value may include 1 or 0, or YES or NO, or true or flase, wherein a value of 1 or YES or true is used to indicate that the second access network device supports AI, and a value of 0 or NO or flase is used to indicate that the second access network device does not support AI.

[0190] (2) Supported third AI algorithm types.

[0191] For example, it is used to indicate the AI ​​algorithm type supported by the second access network device, including but not limited to: decision tree or neural network, etc.

[0192] (3) Supported third-party AI framework information.

[0193] For example, it is used to indicate the AI ​​framework supported by the second access network device, including but not limited to: tensorflow or keras.

[0194] (4) Third computing power information.

[0195] For example, the third computing power information includes at least one of the following: hardware type, computing power, or resource utilization. For specific explanations, please refer to the above related descriptions and will not be repeated here.

[0196] (5) Retraining capability information.

[0197] For example, it is used to indicate whether the second access network device supports retraining of an already trained AI model, and the value may include 1 or 0, or YES or NO, or true or false. Among them, a value of 1 or YES or true is used to indicate that the second access network device supports retraining, and a value of 0 or NO or false is used to indicate that the second access network device does not support retraining. In an embodiment of the present application, the retraining capability information can be specifically used to indicate whether the second access network device supports continuing to execute the first part of the training task that was originally executed by the first access network device.

[0198] Optionally, based on the received third AI capability information, the first access network device may create a third AI capability information context. It should be understood that the present application does not limit the number of adjacent access network devices, and the adjacent access network device may be one or more, and the adjacent access network device includes a second access network device. For example, assuming that multiple adjacent access network devices (e.g., gNB#1, gNB#2, and gNB#3) perform the above step S404, correspondingly, the first access network device creates AI capability information contexts for gNB#1, gNB#2, and gNB#3, respectively.

[0199] S405: The second access network device sends the manufacturer information to the first access network device, and correspondingly, the first access network device receives the manufacturer information from the second access network device.

[0200] The manufacturer information indicates the manufacturer of the second access network device, and the manufacturer information may be manufacturer name information, such as Huawei, Ericsson, Nokia, etc.

[0201] Exemplarily, when the Xn interface is established, the second access network device sends the manufacturer information to the first access network device through the Xn interface, and correspondingly, the first access network device receives the manufacturer information from the second access network device through the Xn interface.

[0202] It should be understood that there may be a scenario where the communication device switches to a target access network device of a different manufacturer. Therefore, in order to ensure network communication security, the first access network device needs to further consider the manufacturer information of the second access network device when making a switching decision. Optionally, an access network device of the same manufacturer as the first access network device is preferentially selected as the target access network device.

[0203] Optionally, the above-mentioned manufacturer information and the third AI capability information may be sent simultaneously or independently. In addition, this application does not limit the execution order of the above-mentioned steps S404, S405, and steps S401 to S403.

[0204] Next, a specific implementation manner in which the first access network device determines that the communication apparatus is to be switched from the first access network device to the second access network device in the above step S410 is described.

[0205] In a first implementation, the communication device performs AI training based on the above step S403, executes the second part of the training task, and performs network measurement to obtain a measurement report (MR). Then the communication device sends the MR to the first access network device. Correspondingly, the first access network device receives the MR from the communication device, which is used to indicate the characteristics of the network environment in which the communication device is located. The first access network device can determine the signal quality of the communication device based on the MR. If the signal quality is lower than a preset threshold, it determines that the communication device needs to be switched, otherwise the communication device does not need to be switched. When the first access network device determines that the communication device will be switched, further, the first access network device determines that the communication device can be switched from the first access network device to the second access network device based on the third AI capability information received in the above step S404.

[0206] Exemplarily, the MR may include one or more of the following: RSRP and reference signal received quality (RSRQ) measurement value of the first access network device, an identifier of an adjacent access network device (including a second access network device), and RSRP and RSRQ measurement values ​​of the adjacent access network device.

[0207] It should be understood that the adjacent access network device may be located within the measurement range of the communication device, or the distance between the adjacent access network device and the communication device may be less than or equal to a preset threshold.

[0208] Exemplarily, when the difference between the RSRP of the adjacent access network device #1 and the RSRP of the first access network device is greater than or equal to a preset threshold, the first access network device triggers the communication device to switch to the adjacent access network device #1, then the adjacent access network device #1 (i.e., the second access network device) is the target access network device to which the communication device is about to switch.

[0209] Optionally, if there are multiple adjacent access network devices sending AI capability information to the first access network device in the above step S404, the first access network device can determine a list of candidate access network devices to which the communication device switches based on the MR, and the list of candidate access network devices includes the second access network device. Further, combined with the AI ​​capability information of each candidate access network device, the second access network device is finally determined to be the target access network device to which the communication device is about to switch from the first access network device.

[0210] Optionally, a difference between the RSRP of each candidate access network device in the candidate access network device list and the RSRP of the first access network device is greater than or equal to a preset threshold.

[0211] Optionally, a candidate access network device with better signal quality is preferentially selected as the target access network device to be switched by the communication device.

[0212] Optionally, the candidate access network device having the largest difference in RSRP with the first access network device is preferentially selected as the target access network device to which the communication device is to switch.

[0213] Optionally, a candidate access network device that supports retraining capabilities and has sufficient computing resources is preferentially selected as the target access network device to which the communication device is about to switch.

[0214] Exemplarily, assuming that there are two candidate access network devices, including access network device #1 and access network device #2, if the RSRP difference #1 between access network device #1 and the first access network device is greater than a preset threshold, and the RSRP difference #2 between access network device #2 and the first access network device is less than the preset threshold, then the first access network device can determine that the access network device #1 is the target access network device for the communication device to switch. If the RSRP differences between access network device #1, access network device #2 and the first access network device are all greater than the preset threshold, and the RSRP difference #2 of access network device #2 is greater than the RSRP difference #1 of access network device #1, then the first access network device can determine that the access network device #2 (i.e., the second access network device) is the target access network device for the communication device to switch.

[0215] Exemplarily, assuming that there are three adjacent access network devices, such as gNB#1, gNB#2 and gNB#3, if the difference between the RSRP of gNB#1, gNB#2 and gNB#3 and the first access network device is greater than a preset threshold, the first access network device can determine that the list of candidate access network devices for the communication device to switch includes gNB#1, gNB#2 and gNB#3, and if the order of RSRP from large to small is gNB#3, gNB#1 and gNB#2, the first access network device can determine that gNB#3 is the target access network device for the communication device to switch. Alternatively, the first access network device traverses the list of candidate access network devices in the order of RSRP from large to small, and the first candidate access network device that supports retraining capability and has sufficient computing resources is gNB#3, then the first access network device determines gNB#3 as the target access network device, that is, the second access network device.

[0216] In a second implementation, the communication device performs AI training based on the above step S403, executes the second part of the training task, and performs network measurement to obtain MR at the same time. Then the communication device sends MR to the first access network device. Correspondingly, the first access network device receives MR from the communication device, and the MR is used to indicate the characteristics of the network environment in which the communication device is located. The first access network device can determine the signal quality of the communication device based on the MR. If the signal quality is lower than a preset threshold, it determines that the communication device needs to be switched, otherwise the communication device does not need to be switched. When the first access network device determines that the communication device will be switched, further, the first access network device determines that the communication device can be switched from the first access network device to the second access network device based on the third AI capability information received in the above step S404 and the manufacturer information received in the above step S405.

[0217] Optionally, a candidate access network device that belongs to the same manufacturer as the first access network device is preferentially selected as the target access network device.

[0218] Exemplarily, assume that the list of candidate access network devices contains three candidate access network devices, such as gNB#1, gNB#2 and gNB#3. If the differences in RSRP between gNB#1, gNB#2 and gNB#3 and the first access network device are all greater than a preset threshold, and if the order of RSRP from large to small is gNB#3, gNB#1 and gNB#2, the first access network device traverses the list of candidate access network devices in order of RSRP from large to small, and the first candidate access network device that supports retraining capability, has sufficient computing resources, and belongs to the same manufacturer as the first access network device is gNB#1. The first access network device determines gNB#1 as the target access network device, i.e., the second access network device.

[0219] S420, the first access network device terminates or suspends the first part of the training task that is to be executed by the first access network device, and triggers the communication device to terminate or suspend the second part of the training task that is to be executed by the communication device.

[0220] Exemplarily, the first access network device queries the training task status of the AI ​​model related to the communication device, that is, the training task status of the first part of the training task that is to be executed by the first access network device, including but not limited to: training identification (such as the ID of the second part) and whether the training task is completed. If the first part of the training task is not completed, the first access network device can terminate or suspend the first part, and at the same time instruct the communication device to terminate or suspend the second part of the training task.

[0221] Optionally, the method further includes the following step S406, which can be performed after step S420, that is, the first access network device saves the current intermediate training result after terminating or pausing the first part of the training task.

[0222] S406: The first access network device saves the intermediate training results of the AI ​​model.

[0223] The intermediate training result is obtained by executing the first part of the training task that is to be executed by the first access network device.

[0224] Exemplarily, the saved intermediate training results of the AI ​​model may be model files, model parameters, training data and other information. The model files may include files of types such as .pkl and .m. The model parameters may be saved in the form of files or in structured form. This application does not limit this.

[0225] Next, the second part of the training task that is executed by the communication device and is triggered by the first access network device to terminate or suspend the communication device is specifically described, including the following method A (suspending the second part of the training task) and method B (terminating the second part of the training task). Method A includes the following steps S430 to S460b, and method B includes the following steps S470 and S480.

[0226] Method A:

[0227] S430, the first access network device sends training task suspension information to the communication apparatus, and correspondingly, the communication apparatus receives the training task suspension information from the first access network device.

[0228] Among them, the training task suspension information includes the identifier and reason value of the second part, and the reason value is used to indicate the reason for pausing the training task of the AI ​​model, such as switching of the communication device.

[0229] In one example, the first access network device sends a request message #1 to the communication device, the request message #1 is used to request the communication device to suspend the second part of the training task, the request message #1 carries the training task suspension information, and optionally, the request message #1 may also carry a reason value. It should be understood that based on the training task suspension information, the communication device suspends the second part of the training task that is to be executed by the communication device.

[0230] Optionally, the training task pause information can also be replaced by the pause training interaction information, and the pause training interaction information is used to indicate that the communication device can continue to execute the second part of the training task, but there is no need to report the training information of the second part of the training task to the first access network device. At this time, the model information obtained by the communication device training or the training information of the second part of the training task can be reported to the second access network device after the communication device switches to the second access network device, and this application does not limit this. It should be pointed out that before the communication device switches to the second access network device, the communication device can continue to report MR to the first access network device, such as periodically reporting MR.

[0231] In another example, the first access network device sends a request message #2 to the communication device, the request message #2 is used to request the communication device to suspend reporting the training information corresponding to the second part of the training task, the request message #2 carries the suspension training interaction information, and optionally, the request message #2 may also carry a reason value. It should be understood that based on the suspension training interaction information, the communication device can continue to execute the second part of the training task belonging to the communication device, but there is no need to report the training information of the second part of the training task. For example, the communication device can locally save the model information obtained by training or the training information of the second part of the training task.

[0232] In another example, the first access network device determines that the second access network device supports the training of the AI ​​model based on the third AI capability information, or the second access network device supports the continued execution of the first part of the training task that belongs to the first access network device. It can be understood that the first access network device selects a suitable target access network device (i.e., the second access network device) from the list of candidate access network devices, and then the first access network device sends a training task pause message or a pause training interaction message to the communication device. Further, the second access network device and the communication device can continue to execute the first and second parts of the training task respectively.

[0233] It should be noted that if step S430 is not performed, that is, the first access network device does not trigger the communication device to suspend the second part of the training task of the AI ​​model, or does not notify the communication device to suspend the training interaction, the communication device will continue to execute the second part of the training task of the AI ​​model, and when one round of training is completed, the model training information is reported to the first access network device. At this time, the following three situations may occur:

[0234] (1) The communication device is in a connected state with the first access network device, that is, the communication device has not yet switched. In this case, the first access network device can instruct the communication device to save the current model training information and report the model training information to the second access network device after the switch is successful.

[0235] (2) The communication device is in a switching state, but the communication device fails to find the target access network device, then the communication device suspends the training task and reporting of the AI ​​model, and saves the current model training information.

[0236] Optionally, the communication device may set a waiting time, for example, 5 minutes. If the switch is successful within the waiting time, the communication device reports the model training information to the target access network device.

[0237] (3) If the communication device has been successfully switched, the communication device can report the current model training information to the target access network device.

[0238] Next, after executing the above step S430, two implementation methods are described for switching the communication device to the second access network device, and the second access network device and the communication device continuing to perform the training task.

[0239] Method 1:

[0240] S440a, the first access network device sends the intermediate training result and the identifier of the first part to the second access network device, and correspondingly, the second access network device receives the intermediate training result and the identifier of the first part from the first access network device.

[0241] In one example, the first access network device sends a first handover request message to the second access network device, and correspondingly, the second access network device receives the first handover request message from the first access network device. The first handover request message includes an identifier of the first part and an intermediate training result. Further, the second access network device performs access control, sends a handover confirmation to the first access network device, and completes random access with the communication device.

[0242] Optionally, the first switching request information may also carry one or more of an identifier of the communication device, first AI capability information of the communication device (see step S401), or model information (see step S402).

[0243] S450a: The second access network device continues to execute the first part of the training task that should be executed by the first access network device according to the intermediate training result.

[0244] Exemplarily, after the communication device successfully switches to the second access network device and the second access network device randomly accesses the communication device, the second access network device can send a training task start message or a training interaction start message to the communication device, carrying an identifier of the first part of the training task to be executed by the communication device, and then the communication device and the second access network device continue to execute the second part and the first part of the training task of the AI ​​model respectively, avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0245] Method 2:

[0246] S440b, the first access network device sends the model training information of the communication device and the identifier of the first part to the second access network device, and correspondingly, the first access network device receives the model training information of the communication device and the identifier of the first part from the first access network device.

[0247] It should be understood that the model training information is the model parameters and / or training data that the communication device reported to the first access network device most recently.

[0248] In one example, the first access network device sends a second switching request message to the second access network device, and correspondingly, the second access network device receives the second switching request message from the first access network device, wherein the second switching request message includes an identifier of the first part and model training information of the communication device.

[0249] Optionally, the second switching request information may also carry one or more of an identifier of the communication device, first AI capability information of the communication device (see step S401), or model information (see step S402).

[0250] S450b, the second access network device builds an AI model based on the model training information of the communication device.

[0251] It should be understood that the AI ​​model constructed by the second access network device is a network NW side model in the second access network device. The constructed AI model is used for the second access network device to continue to execute the first part of the training task that is to be executed by the first access network device.

[0252] Exemplarily, the second access network device determines the model structure, hyperparameters, number of training rounds, model accuracy and other information of the second access network device based on the model training information of the communication device received in step S440b, combined with the training experience, and performs model training using the data collected by the second access network device and the training data included in the model training information of the communication device, thereby obtaining the network NW side model in the second access network device. The specific interpretation of the model structure, hyperparameters, number of training rounds, and model accuracy determined by the second access network device can be referred to the relevant description of the above step S402, and for the sake of brevity, it will not be repeated here.

[0253] It should be understood that when the second access network device determines information such as model structure, hyperparameters, number of training rounds, model accuracy, etc., it is usually necessary to consider the compatibility between the second access network device and the model training information of the communication device. For example, based on the model structure contained in the model information determined in step S402 above, indicating that the model structure of the communication device is a neural network, the model structure of the second access network device is a neural network, and the number and type of input nodes of the second access network device are consistent with the number and type of model output nodes indicated by the model training information of the communication device. Optionally, other parameters can be determined based on experience, and this application does not limit this.

[0254] Optionally, the model structure, hyper-parameters, number of training rounds, model accuracy and other information determined by the second access network device may be the same as or different from the model structure, hyper-parameters, number of training rounds, model accuracy and other information determined by the first access network device in the above step S402, and this application does not limit this.

[0255] S460b: The second access network device continues to execute the first part of the training task that is to be executed by the first access network device based on the constructed AI model.

[0256] Exemplarily, after the communication device successfully switches to the second access network device and the second access network device randomly accesses the communication device, the second access network device can send a training task start message or a training interaction start message to the communication device, carrying an identifier of the first part of the training task to be executed by the communication device, and then the communication device and the second access network device continue to execute the second part and the first part of the training task of the AI ​​model respectively, avoiding the waste of previous training resources and ensuring the training efficiency of the AI ​​model.

[0257] Optionally, the training task start information or the training interaction start information may also carry modified model information, that is, the model structure, hyperparameters, number of training rounds, model accuracy and other information determined by the second access network device in the above step S450b.

[0258] It should be noted that compared to method 1, the second access network device in method 2 can be a target access network device from a different manufacturer than the first access network device. At this time, the first access network device may not send the intermediate training results of its own AI model to the second access network device, but instead send the model training information of the communication device to the second access network device, which can avoid exposure of resources on the first access network device side and ensure model security.

[0259] Method B:

[0260] S470, the first access network device sends training task termination information to the communication apparatus, and correspondingly, the communication apparatus receives the training task termination information from the first access network device.

[0261] Among them, the training task termination information includes the identifier and reason value of the second part, and the reason value is used to indicate the termination reason of the training task of the AI ​​model, such as switching of the communication device, or failure to select a suitable target access network device with AI capabilities.

[0262] In one example, the first access network device sends a request message #3 to the communication device, the request message #3 is used to request the communication device to terminate the second part of the training task, the request message #3 carries the training task termination information, and optionally, the request message #3 may also carry a reason value. It should be understood that based on the training task termination information, the communication device terminates the second part of the training task that is to be executed by the communication device.

[0263] In another example, the first access network device determines that the second access network device does not support the training of the AI ​​model based on the third AI capability information, or the second access network device does not support the continuation of the first part of the training task that belongs to the first access network device. It can be understood that the first access network device has not selected a suitable target access network device (i.e., the second access network device) from the list of candidate access network devices. For example, if the target access network device does not support AI capabilities, the first access network device sends a training task termination message to the communication device. Further, the second access network device and the communication device can subsequently use traditional non-AI business algorithms to perform training tasks.

[0264] Optionally, the first access network device sends a non-AI mode switching indication to the communication device, and correspondingly, the communication device receives the non-AI mode switching indication from the first access network device. The non-AI mode switching indication is used to instruct the communication device to switch to the non-AI mode.

[0265] Optionally, the non-AI mode switching indication can be sent separately, or carried in the training task termination information, or carried in the request information #3, which is not limited in this application.

[0266] Next, after executing step S470, a specific implementation method of switching the communication device to the second access network device is described, that is, executing steps S480a to S480c.

[0267] S480a, the first access network device sends a switching request message to the second access network device, and correspondingly, the second access network device receives the switching request message from the first access network device.

[0268] The switching request information is used to request to switch the communication device to the second access network device, and the switching request information includes an identifier of the communication device.

[0269] S480b: The second access network device performs access control.

[0270] S480c, the second access network device sends switching confirmation information to the first access network device, and correspondingly, the first access network device receives the switching confirmation information from the second access network device.

[0271] Exemplarily, after receiving the switching request information sent by the first access network device, the second access network device can determine whether to allow access to the communication device based on the third AI capability information and its own current state. If the communication device is allowed to switch to the second access network device, the second access network device performs access control and switching confirmation. For example, if the AI ​​resources of the second access network device are sufficient, a switching confirmation message allowing the communication device to switch is sent to the first access network device; for another example, if the AI ​​resources of the second access network device are insufficient, the switching request of the first access network device may be rejected. Optionally, the second access network device sends a rejection message to the first access network device, and the rejection message carries a reason value for indicating the reason for rejecting the switching of the communication device, such as insufficient AI resources of the second access network device, or the second access network device is busy.

[0272] S490, the communication device and the second access network device perform random access.

[0273] Exemplarily, based on the above steps S480a to S480c, the communication device successfully switches to the second access network device, and then the communication device no longer performs AI model training, and no longer continues to execute the second part of the training task, and can use traditional non-AI business algorithms to continue to complete the training task.

[0274] Optionally, when the communication device switches back to the first access network device, the communication device and the first access network device may continue to execute the second part and the first part of the training task respectively.

[0275] Based on the above scheme, when the first access network device determines that the communication device will switch from the first access network device to the second access network device, the first part of the training task that belongs to the first access network device is terminated or suspended, and the communication device is triggered to terminate or suspend the second part of the training task that belongs to the communication device. The waste of resources caused by the continued training and data transmission of the first access network device and the communication device can be avoided. Furthermore, the first access network device sends the intermediate training results of the AI ​​model or the model training information of the communication device to the second access network device, so that the second access network device continues to execute the first part of the training task that belongs to the first access network device, and the communication device can continue to execute the second part of the training task that belongs to the communication device after switching to the second access network device, thereby reducing resource waste and improving the training efficiency of the AI ​​model.

[0276] In an embodiment of the present application, an AI model training function is deployed in the access network device for performing AI model training. Among them, the AI ​​model training function can be deployed on a newly added MIF. Optionally, the MIF can be deployed inside the access network device or independently of the access network device. For example, the MIF is connected to the gNB through the G1 interface, which is not limited in this application. Figure 5 , for the architecture where MIF is independent of access network equipment, an example is given in which MIF performs AI model training.

[0277] Figure 5 is a flow chart of a switching processing method 500 provided in an embodiment of the present application, which can be applied to the above Figures 1 to 3 In the communication system shown. Exemplarily, the method flow can be executed by a communication device (e.g., UE), a first access network device (e.g., source gNB), a second access network device (e.g., target gNB), a first mobile intelligent function (e.g., MIF1), and a second mobile intelligent function (e.g., MIF2), or by modules and / or devices (e.g., chips or integrated circuits, etc.) with corresponding functions installed in the communication device, the first access network device, the second access network device, the first mobile intelligent function, and the second mobile intelligent function, which is not limited in this application. For the sake of ease of description, the following description is based on the communication device, the first access network device, the second access network device, the first mobile intelligent function, and the second mobile intelligent function as the execution subjects, such as Figure 5 As shown, the method includes the following steps, and the incomplete parts can refer to the existing protocol.

[0278] It should be noted that the difference between method 500 and the above-mentioned method 400 is that MIF is deployed outside the access network device, that is, independent of the access network device. Specifically, the management of AI model training and training tasks in method 500 is completed by MIF, including determining the model information of the AI ​​model, terminating or pausing the first part of the training task that is to be executed by MIF1, and triggering the communication device to terminate or suspend the second part of the training task that is to be executed by the communication device. For example, the information interaction between MIF1 and the UE is transparently transmitted through the source gNB, and the information interaction between MIF2 and the UE is transparently transmitted through the target gNB. It should be understood that Figure 5 and Figure 4 The method embodiments shown can be combined with each other. Figure 4 and Figure 5 The steps in the method embodiments shown can refer to each other. For example, in the embodiments of the present application, Figure 5 The method embodiment shown can be regarded as implementing Figure 4 A way of implementing the functions of the method embodiment shown.

[0279] It should be understood that this technical solution is applicable to the process in which the first mobile intelligent function and the communication device jointly perform a training task. The training task refers to a task of obtaining an AI model through AI training based on model parameters or measurement data. The AI ​​model is used by the first mobile intelligent function and the communication device to jointly perform model reasoning.

[0280] S501, the communication device sends first AI capability information to the first mobile intelligent function, and correspondingly, the first mobile intelligent function receives the first AI capability information from the communication device.

[0281] Exemplarily, the communication device sends the first AI capability information to the first mobile intelligent function through the first access network device, and correspondingly, the first mobile intelligent function receives the first AI capability information through the first access network device.

[0282] Among them, the parameters and specific interpretations included in the first AI capability information can be referred to the relevant description of step S401 of the above method 400. For the sake of brevity, they will not be repeated here.

[0283] S502, the first access network device sends second AI capability information to the first mobile intelligent function, and correspondingly, the first mobile intelligent function receives the second AI capability information from the first access network device.

[0284] Exemplarily, when the G1 interface is established, the first access network device and the first mobile intelligent function exchange their respective AI capability information through the G1 interface, that is, the first access network device sends the second AI capability information to the first mobile intelligent function, and the first mobile intelligent function sends the fourth AI capability information to the first access network device.

[0285] In one example, the first mobile intelligent function sends a query message to the first access network device to query the second AI capability information, and the first access network device reports the second AI capability information to the first mobile intelligent function based on the query message.

[0286] In another example, the first access network device actively sends the second AI capability information to the first mobile intelligent function.

[0287] Among them, the parameters and specific interpretations included in the second AI capability information can be referred to the relevant description of step S402 of the above method 400, and will not be repeated here for the sake of brevity.

[0288] S503, the first mobile intelligent function determines model information for AI model training based on the first AI capability information, the second AI capability information and the fourth AI capability information.

[0289] In one example, the first mobile intelligent function determines model information based on training requirements, first AI capability information, second AI capability information, and fourth AI capability information. The specific implementation method can refer to the relevant description of the above step S402. For the sake of brevity, it will not be repeated here.

[0290] Among them, the fourth AI capability information indicates the AI ​​training capability supported by the first mobile intelligent function.

[0291] Exemplarily, the fourth AI capability information includes at least one of the following:

[0292] (1) Fourth AI switch.

[0293] For example, used to indicate whether the first mobile intelligent function supports AI, the value may include 1 or 0, or YES or NO, or true or flase, wherein a value of 1 or YES or true is used to indicate that the first mobile intelligent function supports AI, and a value of 0 or NO or flase is used to indicate that the first mobile intelligent function does not support AI.

[0294] (2) The fourth AI algorithm type supported.

[0295] For example, it is used to indicate the type of AI algorithm supported by the first mobile intelligent function, including but not limited to: decision tree or neural network, etc.

[0296] (3) Information about the fourth AI framework supported.

[0297] For example, the AI ​​framework used to indicate the AI ​​framework supported by the first mobile intelligent function includes but is not limited to: tensorflow or keras.

[0298] (4) Fourth computing power information.

[0299] For example, it is used to indicate the computing power supported by the first mobile intelligent function, including but not limited to: hardware type, computing power, or resource utilization. The specific interpretation can be referred to the relevant description of the above step S402, which will not be repeated here.

[0300] S504, the first mobile intelligent function sends a training task start message #1 to the communication device.

[0301] S505: The communication device sends the MR to the first access network device, and correspondingly, the first access network device receives the MR from the communication device.

[0302] S506, the second mobile intelligent function sends the fifth AI capability information to the second access network device, and correspondingly, the first access network device receives the fifth AI capability information from the second mobile intelligent function.

[0303] Exemplarily, when the G1 interface is established, the second access network device and the second mobile intelligent function exchange their respective AI capability information through the G1 interface, that is, the second access network device sends the third AI capability information to the second mobile intelligent function, and the second mobile intelligent function sends the fifth AI capability information to the second access network device.

[0304] In one example, the second access network device sends a query message to the second mobile intelligent function to query the fifth AI capability information, and the second mobile intelligent function reports the fifth AI capability information to the second access network device based on the query message.

[0305] In another example, the second mobile intelligent function actively sends the fifth AI capability information to the second access network device.

[0306] Among them, the parameters and specific interpretations included in the third AI capability information can be referred to the relevant description of step S404 of the above method 400. For the sake of brevity, they will not be repeated here.

[0307] Among them, the fifth AI capability information indicates the AI ​​training capability supported by the second mobile intelligent function.

[0308] Exemplarily, the fifth AI capability information includes at least one of the following:

[0309] (1) Fifth AI switch.

[0310] For example, used to indicate whether the second mobile intelligent function supports AI, the value may include 1 or 0, or YES or NO, or true or flase, wherein a value of 1 or YES or true is used to indicate that the second mobile intelligent function supports AI, and a value of 0 or NO or flase is used to indicate that the second mobile intelligent function does not support AI.

[0311] (2) The fifth AI algorithm type supported.

[0312] For example, it is used to indicate the type of AI algorithm supported by the second mobile intelligent function, including but not limited to: decision tree or neural network, etc.

[0313] (3)Supported fifth AI framework information.

[0314] For example, the AI ​​framework used to indicate the AI ​​framework supported by the second mobile intelligent function includes but is not limited to: tensorflow or keras.

[0315] (4) Fifth computing power information.

[0316] For example, it is used to indicate the computing power supported by the second mobile intelligent function, including but not limited to: hardware type, computing power, or resource utilization. The specific interpretation can be referred to the relevant description of the above step S402, which will not be repeated here.

[0317] (5) Retraining capability information.

[0318] For example, it is used to indicate whether the second mobile intelligent function supports retraining of an already trained AI model, and the value may include 1 or 0, or YES or NO, or true or false. Among them, a value of 1 or YES or true is used to indicate that the second mobile intelligent function supports retraining, and a value of 0 or NO or false is used to indicate that the second mobile intelligent function does not support retraining. In an embodiment of the present application, the retraining capability information can be specifically used to indicate whether the second mobile intelligent function supports continuing to execute the first part of the training task that was originally executed by the first mobile intelligent function.

[0319] S507: The second access network device sends the fifth AI capability information and the third AI capability information to the first access network device. Correspondingly, the first access network device receives the fifth AI capability information and the third AI capability information from the second access network device.

[0320] Among them, the parameters and specific interpretations included in the third AI capability information can be referred to the relevant description of step S404 of the above method 400. For the sake of brevity, they will not be repeated here.

[0321] Exemplarily, when the Xn interface is established, the second access network device sends the third AI capability information and the fifth AI capability information to the first access network device through the Xn interface.

[0322] Optionally, the second access network device may also send the manufacturer information of the second access network device to the first access network device through the Xn interface, so as to indicate the manufacturer providing the second access network device.

[0323] S508: The first access network device determines that the communication apparatus will be switched from the first access network device to the second access network device.

[0324] Among them, the specific implementation methods of the above steps S504, S505, S507, and S508, and the meaning of the interactive information can refer to the relevant descriptions of steps S403, S404, and S410 of the above method 400, and will not be repeated here for the sake of brevity.

[0325] S509, the first access network device sends a switching notification to the first mobile intelligent function, and correspondingly, the first mobile intelligent function receives the switching notification from the first access network device.

[0326] The switching notification includes an identifier of the communication device, such as a UE ID. For example, when the first access network device determines that the communication device needs to be switched, the switching notification is sent to MIF1.

[0327] Optionally, the switching notification may further include a candidate access network device list, where the candidate access network device list includes the second access network device.

[0328] S510, the first mobile intelligent function terminates or pauses the first part of the training task that is executed by the first mobile intelligent function, and triggers the communication device to terminate or pause the second part of the training task that is executed by the communication device. The specific implementation method can refer to step S420 of the above method 400. For the sake of brevity, it will not be repeated here.

[0329] Optionally, the first access network device saves the intermediate training results of the AI ​​model. The specific implementation method can refer to the relevant description of step S406 of the above method 400. For the sake of brevity, it will not be repeated here.

[0330] Next, the second part of the training task that is executed by the communication device and is terminated or paused by the communication device when the first mobile intelligent function triggers the communication device is specifically described, including the following method A (pausing the second part of the training task) and method B (terminating the second part of the training task). Method A includes the following steps S511 to S517, and method B includes the following steps S518 and S522.

[0331] Method 1:

[0332] S511, the first mobile intelligent function sends training task pause information to the communication device, and correspondingly, the communication device receives the training task pause information from the first mobile intelligent function.

[0333] S512, the first access network device sends a switching request message #1 to the second access network device, and correspondingly, the second access network device receives the switching request message #1 from the first access network device.

[0334] S513: The second access network device performs access control.

[0335] S514, the second access network device sends a switching confirmation message #1 to the first access network device, and correspondingly, the first access network device receives the switching confirmation message #1 from the second access network device.

[0336] S515, the second access network device sends a switching success notification to the first mobile intelligent function, and correspondingly, the first mobile intelligent function receives the switching success notification from the second access network device.

[0337] Exemplarily, when the communication device has successfully switched to the second access network device, the first access network device sends a switching success notification to the first mobile intelligent function, wherein the switching success notification includes an identifier of the communication device and an identifier of the second access network device.

[0338] Optionally, the first mobile intelligent function sends query information to the second mobile intelligent function to query the fifth AI capability information. Furthermore, the first mobile intelligent function determines the target access network device (i.e., the second access network device) based on the fifth AI capability information and the third AI capability information. Further optionally, the first mobile intelligent function can also determine the target access network device based on the manufacturer information of the second access network device. Then, the first mobile intelligent function sends a notification message to the first access network device to indicate the target access network device determined by the first mobile intelligent function. The specific determination method can refer to the relevant description of the above step S508. For the sake of brevity, it will not be repeated here. In other words, determining the target access network device to which the communication device switches can be performed by the first access network device or by the first mobile intelligent function, and this application does not make specific limitations on this.

[0339] S516, the first mobile intelligent function sends AI training task information to the second mobile intelligent function, and correspondingly, the second mobile intelligent function receives the AI ​​training task information from the first mobile intelligent function.

[0340] The AI ​​training task information includes model information of the communication device or an intermediate training result of the AI ​​model, and an identifier of the first part. For example, the first mobile intelligent function sends the AI ​​training task information to the second mobile intelligent function by multiplexing the Xn interface.

[0341] Optionally, if the first access network device saves the intermediate training results of the AI ​​model, the AI ​​training task information may include the intermediate training results of the AI ​​model and the identifier of the first part; if the first access network device does not save the intermediate training results of the AI ​​model, the AI ​​training task information may include the model information of the communication device and the identifier of the first part. The specific implementation method can refer to the relevant description of method 1 or method 2 of the above method 400. For the sake of brevity, it will not be repeated here.

[0342] Optionally, the AI ​​training task information may also include one or more of the first AI capability information, the identification of the communication device, and the model information of the AI ​​model.

[0343] S517, the communication device randomly accesses the second access network device, and a training task is initiated between the communication device and the second mobile intelligent function.

[0344] Among them, the specific implementation method of the above steps S511 to S517 and the meaning of the interactive information can refer to the relevant description of steps S430 to S450a of the above method 400, and will not be repeated here for the sake of brevity.

[0345] Method 2:

[0346] S518, the first mobile intelligent function sends training task termination information to the communication device, and correspondingly, the communication device receives the training task termination information from the first mobile intelligent function.

[0347] S519, the first access network device sends a switching request message #2 to the second access network device, and correspondingly, the second access network device receives the switching request message #2 from the first access network device.

[0348] S520: The second access network device performs access control.

[0349] S521, the second access network device sends a switching confirmation message #2 to the first access network device, and correspondingly, the first access network device receives the switching confirmation message #2 from the second access network device.

[0350] S522, the communication device randomly accesses the second access network device, and a training task is initiated between the communication device and the second mobile intelligent function.

[0351] Exemplarily, after the communication device switches to the second access network device, the communication device negotiates the AI ​​training task with the second mobile intelligent function through the second access network device and starts the AI ​​model training task. The specific implementation method can refer to the relevant description of step S490 of the above method 400. For the sake of brevity, it will not be repeated here.

[0352] Among them, the specific implementation method of the above steps S518 to S522 and the meaning of the interactive information can refer to the relevant description of steps S470 to S490 of the above method 400, and will not be repeated here for the sake of brevity.

[0353] Based on the above scheme, when the first access network device determines that the communication device will switch from the first access network device to the second access network device, it sends a switching notification to the first mobile intelligent function, so that the first mobile intelligent function terminates or suspends the first part of the training task that belongs to the first mobile intelligent function, and triggers the communication device to terminate or suspend the second part of the training task that belongs to the communication device, which can avoid the waste of resources caused by the continued training and data transmission of the first mobile intelligent function. Furthermore, the first mobile intelligent function can send the intermediate training results of the AI ​​model or the model training information of the communication device to the second mobile intelligent function, so that the second mobile intelligent function continues to execute the first part of the training task that belongs to the first mobile intelligent function, and the communication device can continue to execute the second part of the training task that belongs to the communication device after switching to the second access network device, thereby reducing resource waste and improving the training efficiency of the AI ​​model.

[0354] Combined with the above Figures 1 to 5 Describes the method embodiment of the present application, and the following is combined with Figures 6 to 8The device embodiment of the embodiment of the present application is described. It can be understood that the description of the method embodiment and the description of the device embodiment can correspond to each other, so the parts not described can refer to the previous method embodiment.

[0355] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the interaction between various network elements. It can be understood that each network element, such as a transmitting end device or a receiving end device, includes a hardware structure and / or software module corresponding to the execution of each function in order to realize the above functions. Those skilled in the art should be aware that, in combination with the units and algorithm steps of each example described in the embodiments disclosed in this document, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0356] The embodiment of the present application can divide the functional modules of the transmitting end device or the receiving end device according to the above method example. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation. The following is an example of dividing each functional module corresponding to each function.

[0357] Figure 6 1 is a schematic block diagram of a switching processing device 1000 provided in an embodiment of the present application. Figure 6 As shown, the device 1000 includes a transceiver module 1010 and / or a processing module 1020. The transceiver module 1010 can implement corresponding communication functions, and the processing module 1020 is used to perform data processing, or in other words, the transceiver module 1010 is used to perform operations related to receiving and sending, and the processing module 1020 is used to perform other operations besides receiving and sending. The transceiver module 1010 can also be called a communication interface or a communication unit.

[0358] Optionally, the device 1000 may also include a storage module 1030, which can be used to store instructions and / or data. The processing module 1020 can read the instructions and / or data in the storage module so that the device implements the actions of the device or network element in the aforementioned method embodiments.

[0359] In a first design, the apparatus 1000 may correspond to the communication apparatus (eg, Figure 4 or Figure 5 A communication device in a communication device), or a component of a communication device (such as a chip or a circuit).

[0360] The device 1000 can implement the steps or processes executed by the communication device in the above method embodiment, wherein the transceiver module 1010 can be used to execute the transceiver related operations of the communication device in the above method embodiment, and the processing module 1020 can be used to execute the processing related operations of the communication device in the above method embodiment.

[0361] In a second design, the apparatus 1000 may correspond to the first access network device (eg, Figure 4 or Figure 5 The first access network device in the embodiment of the present invention may be a first access network device in the embodiment of the present invention, or a component (such as a chip or circuit) of the first access network device.

[0362] The device 1000 can implement the steps or processes executed by the first access network device in the above method embodiment, wherein the transceiver module 1010 can be used to execute the transceiver related operations of the first access network device in the above method embodiment, and the processing module 1020 can be used to execute the processing related operations of the first access network device in the above method embodiment.

[0363] In a third design, the apparatus 1000 may correspond to the second access network device (eg, Figure 4 or Figure 5 The second access network device in the embodiment of the present invention may be a component of the second access network device (such as a chip or circuit).

[0364] The device 1000 can implement the steps or processes executed by the second access network device in the above method embodiment, wherein the transceiver module 1010 can be used to execute the transceiver related operations of the second access network device in the above method embodiment, and the processing module 1020 can be used to execute the processing related operations of the second access network device in the above method embodiment.

[0365] In a fourth design, the device 1000 may correspond to the first mobile smart function (eg, Figure 5 The first mobile intelligent function in the mobile intelligent function), or a component of the first mobile intelligent function (such as a chip or circuit).

[0366] The device 1000 can implement steps or processes corresponding to the execution of the first mobile intelligent function in the above method embodiment, wherein the transceiver module 1010 can be used to execute operations related to the transmission and reception of the first mobile intelligent function in the above method embodiment, and the processing module 1020 can be used to execute operations related to the processing of the first mobile intelligent function in the above method embodiment.

[0367] In a fifth design, the device 1000 may correspond to the second mobile intelligent function (eg Figure 5 The second mobile intelligent function in the mobile intelligent function), or a component of the second mobile intelligent function (such as a chip or circuit).

[0368] The device 1000 can implement steps or processes corresponding to the execution of the second mobile intelligent function in the above method embodiment, wherein the transceiver module 1010 can be used to execute operations related to the transmission and reception of the second mobile intelligent function in the above method embodiment, and the processing module 1020 can be used to execute operations related to the processing of the second mobile intelligent function in the above method embodiment.

[0369] It should be understood that the specific process of each module executing the above corresponding steps has been described in detail in the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0370] It should also be understood that the device 1000 here is embodied in the form of a functional module. The term "module" here may refer to an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a proprietary processor or a group processor, etc.) and a memory for executing one or more software or firmware programs, a merged logic circuit and / or other suitable components that support the described functions. In an optional example, those skilled in the art may understand that the device 1000 may be specifically the switching processing device in the above-mentioned embodiment, and may be used to execute the various processes and / or steps corresponding to the switching processing device in the above-mentioned method embodiments; or, the device 1000 may be specifically the network device in the above-mentioned embodiment, and may be used to execute the various processes and / or steps corresponding to the network device in the above-mentioned method embodiments. To avoid repetition, it will not be repeated here.

[0371] The device 1000 of each of the above schemes has the function of implementing the corresponding steps performed by the device (such as a communication device, a first access network device, or a second access network device) in the above method. This function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions; for example, the transceiver module can be replaced by a transceiver (for example, the sending unit in the transceiver module can be replaced by a transmitter, and the receiving unit in the transceiver module can be replaced by a receiver), and other units, such as processing modules, can be replaced by processors to respectively perform the transceiver operations and related processing operations in each method embodiment.

[0372] In addition, the above-mentioned transceiver module 1010 can also be a transceiver circuit (for example, can include a receiving circuit and a sending circuit), and the processing module can be a processing circuit.

[0373] Figure 7 FIG. 2 is a schematic diagram of another switching processing device 2000 provided in an embodiment of the present application. Figure 7 As shown, the device 2000 includes a processor 2010 and a transceiver 2020. The processor 2010 and the transceiver 2020 communicate with each other through an internal connection path, and the processor 2010 is used to execute instructions to control the transceiver 2020 to send signals and / or receive signals.

[0374] Optionally, the device 2000 may further include a memory 2030, and the memory 2030 communicates with the processor 2010 and the transceiver 2020 through an internal connection path. The memory 2030 is used to store instructions, and the processor 2010 may execute the instructions stored in the memory 2030.

[0375] As a solution, the device 2000 is used to implement the operations performed by the communication device (eg, UE) in the above various method embodiments.

[0376] As another solution, the apparatus 2000 is used to implement the operations performed by the first access network device (e.g., the source gNB) in the above method embodiments.

[0377] As another solution, the apparatus 2000 is used to implement the operations performed by the second access network device (e.g., the target gNB) in the above method embodiments.

[0378] As another solution, the device 2000 is used to implement the operations performed by the first mobile intelligent function (eg, MIF1) in the above method embodiments.

[0379] As another solution, the device 2000 is used to implement the operations performed by the second mobile intelligent function (eg, MIF2) in the above method embodiments.

[0380] It should be understood that the processor mentioned in the embodiments of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0381] It should also be understood that the memory mentioned in the embodiments of the present application may be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM). For example, RAM can be used as an external cache. By way of example and not limitation, RAM includes the following forms: static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0382] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, the memory (storage module) can be integrated into the processor.

[0383] It should also be noted that the memory described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0384] Figure 8 FIG. 3 is a schematic diagram of a chip system 3000 provided in an embodiment of the present application. Figure 8 As shown, the chip system 3000 (or may also be called a processing system) includes a logic circuit 3010 and an input / output interface (input / output interface) 3020.

[0385] Among them, the logic circuit 3010 can be a processing circuit in the chip system 3000. The logic circuit 3010 can be coupled to the storage unit and call the instructions in the storage unit so that the chip system 3000 can implement the methods and functions of each embodiment of the present application. The input / output interface 3020 can be an input / output circuit in the chip system 3000, outputting information processed by the chip system 3000, or inputting data or signaling information to be processed into the chip system 3000 for processing.

[0386] As a solution, the chip system 3000 is used to implement the operations performed by the communication device in the above method embodiments.

[0387] As another solution, the chip system 3000 is used to implement the operations performed by the first access network device in the above method embodiments.

[0388] As another solution, the chip system 3000 is used to implement the operations performed by the second access network device in the above method embodiments.

[0389] As another solution, the chip system 3000 is used to implement the operations performed by the first mobile intelligent function in the above method embodiments.

[0390] As another solution, the chip system 3000 is used to implement the operations performed by the first mobile intelligent function in the above method embodiments.

[0391] For example, the logic circuit 3010 is used to implement the processing-related operations performed by the receiving device in the above method embodiment, such as Figure 4 or Figure 5 The processing-related operations performed by the receiving device in the embodiment shown; the input / output interface 3020 is used to implement the sending and / or receiving-related operations performed by the receiving device in the above method embodiment, such as Figure 4 or Figure 5 The receiving device in the illustrated embodiment performs operations related to sending and / or receiving.

[0392] An embodiment of the present application also provides a computer-readable storage medium on which computer instructions for implementing the methods executed by the device in the above-mentioned method embodiments are stored.

[0393] An embodiment of the present application also provides a computer program product, comprising instructions, which, when executed by a computer, implement the methods performed by the device in the above-mentioned method embodiments.

[0394] The embodiment of the present application further provides a communication system, comprising the aforementioned application server and data channel server. Optionally, the system further comprises the aforementioned first access network device and second access network device.

[0395] The explanation of the relevant contents and beneficial effects of any of the above-mentioned devices can be referred to the corresponding method embodiments provided above, which will not be repeated here.

[0396] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0397] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0398] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0399] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0400] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0401] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0402] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A handover processing method, characterized in that: Applied to a first access network device, the first access network device and a communication device jointly perform a training task of a bilateral artificial intelligence AI model, the method comprising: Determining that the communication device will switch from the first access network device to the second access network device; Terminate or suspend a first part of the training task that is to be executed by the first access network device, and trigger the communication device to terminate or suspend a second part of the training task that is to be executed by the communication device.

2. The method according to claim 1, characterized in that The triggering the communication device to terminate or pause the second part of the training task that is to be executed by the communication device includes: Sending training task termination information or training task suspension information, wherein the training task termination information or the training task suspension information includes an identifier and a cause value of the second part, and the cause value is used to indicate that a switch occurs in the communication device.

3. The method according to claim 1 or 2, characterized in that: Before determining that the communication device is to be switched from the first access network device to the second access network device, the method further includes: receiving first AI capability information, where the first AI capability information indicates an AI training capability supported by the communication device; Determining, according to the first AI capability information and the second AI capability information, model information for training the AI ​​model, wherein the second AI capability information indicates an AI training capability supported by the first access network device; Sending training task startup information, wherein the training task startup information includes model information of the AI ​​model and an identifier of the second part, so that the communication device executes the second part of the training task that is to be executed by the communication device; Based on the model information of the AI ​​model, execute the first part of the training task that is to be executed by the first access network device.

4. The method according to claim 3, characterized in that The first AI capability information includes at least one of the following: a first AI switch, a supported first AI algorithm type, supported first AI framework information, or first computing power information, wherein the first computing power information includes at least one of the following: hardware type, computing power, or resource utilization; The second AI capability information includes at least one of the following: a second AI switch, a supported second AI algorithm type, supported second AI framework information, or second computing power information, wherein the second computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

5. The method according to claim 3 or 4, characterized in that: The model information of the AI ​​model includes at least one of the following: model structure, hyperparameters, model accuracy, or number of training rounds.

6. The method according to any one of claims 1 to 5, characterized in that Before terminating or pausing the first part of the training task that is to be executed by the first access network device, the method further includes: receiving third AI capability information, where the third AI capability information indicates an AI training capability supported by the second access network device; According to the measurement report and the third AI capability information, it is determined that the communication device will switch from the first access network device to the second access network device, and the measurement report is used to indicate the characteristics of the network environment in which the communication device is located.

7. The method according to claim 6, characterized in that The method further comprises: receiving manufacturer information, where the manufacturer information indicates a manufacturer providing the second access network device; Wherein, determining, according to the measurement report and the third AI capability information, that the communication device is to be switched from the first access network device to the second access network device includes: According to the measurement report, the third AI capability information and the vendor information, it is determined that the communication device will switch from the first access network device to the second access network device.

8. The method according to claim 6 or 7, characterized in that: The third AI capability information includes at least one of the following: a third AI switch, a supported third AI algorithm type, supported third AI framework information, third computing power information, or retraining capability information, where the retraining capability information indicates whether the second access network device supports retraining of the trained AI model, and the third computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

9. The method according to any one of claims 6 to 8, characterized in that The triggering the communication device to terminate or pause the second part of the training task that is to be executed by the communication device includes: Determining, based on the third AI capability information, that the second access network device does not support training of the AI ​​model; The communication device is triggered to terminate the second part of the training task that is performed by the communication device.

10. The method according to claim 9, characterized in that The method further comprises: A non-AI mode switching indication is sent to the communication device, where the non-AI mode switching indication is used to instruct the communication device to switch to the non-AI mode.

11. The method according to any one of claims 6 to 8, characterized in that The triggering the communication device to terminate or pause the second part of the training task that is to be executed by the communication device includes: Determining, based on the third AI capability information, that the second access network device supports training of the AI ​​model; The communication device is triggered to pause the second part of the training task that is executed by the communication device.

12. The method according to any one of claims 1 to 11, characterized in that The method further comprises: Saving an intermediate training result of the AI ​​model, where the intermediate training result is obtained by executing a first part of the training task that is to be executed by the first access network device; Send the intermediate training result and the identifier of the first part to the second access network device.

13. The method according to claim 12, characterized in that The method further comprises: Sending first switching request information to the second access network device, where the first switching request information includes an identifier of the first part and the intermediate training result.

14. The method according to any one of claims 1 to 11, characterized in that The method further comprises: Receive model training information of the communication device, where the model training information is the model parameters and / or training data most recently reported by the communication device to the first access network device, and the model training information of the communication device is used by the second access network device to construct an AI model, and the constructed AI model is used by the second access network device to continue to execute the first part of the training task that belongs to the first access network device; Sending the identifier of the first part and the model training information of the communication device to the second access network equipment.

15. The method according to claim 14, characterized in that The method further comprises: Sending second switching request information to the second access network device, where the second switching request information includes an identifier of the first part and model training information of the communication device.

16. The method according to claim 13 or 15, characterized in that The first switching request information or the second switching request information further includes at least one of the following: first AI capability information of the communication device, or model information of the AI ​​model.

17. The method according to any one of claims 1 to 16, characterized in that The method further comprises: When the communication device switches back to the first access network device, the first part of the training task that is to be executed by the first access network device continues to be executed.

18. The method according to claim 13 or 15, characterized in that The terminating or pausing the first part of the training task that is to be executed by the first access network device includes: receiving switching notification information, wherein the switching notification information includes an identifier of the communication device; According to the switching notification information, the first part of the training task that is to be executed by the first access network device is terminated or suspended.

19. The method according to claim 18, characterized in that The switching notification information further includes identification information of at least one candidate access network device, the at least one candidate access network device includes the second access network device, the at least one candidate access network device is determined based on a measurement report, and the measurement report is used to indicate characteristics of a network environment in which the communication device is located; Wherein, before sending the first switching request information to the second access network device, or sending the second switching request information to the second access network device, the method further includes: receiving at least one fourth AI capability information corresponding to the at least one candidate access network device, where the fourth AI capability information indicates an AI training capability supported by each candidate access network device; According to the at least one fourth AI capability information, it is determined that the second access network device is the optimal access network device for switching the communication device.

20. The method according to any one of claims 1 to 19, characterized in that The method further comprises: In the case that the communication device has successfully switched to the second access network device, a switching success notification message is received, where the switching success notification message includes an identifier of the communication device and an identifier of the second access network device.

21. The method according to claim 20, characterized in that The method further comprises: Send AI training task information to the second access network device, the AI ​​training task information including the intermediate training result of the AI ​​model or the model training information of the communication device, and the identifier of the first part, the intermediate training result or the model training information of the communication device is used by the second access network device to continue to execute the first part of the training task that belongs to the first access network device.

22. A switching processing method, characterized in that: Applied to the second access network device, the communication device and the first access network device jointly perform a training task of a bilateral artificial intelligence AI model, and the method includes: In a case where the communication device is to switch from the first access network device to the second access network device, receiving an intermediate training result of an artificial intelligence AI model or model training information of the communication device, and an identifier of a first part of the training task that is to be executed by the first access network device, the model training information being the model parameters and / or training data most recently reported by the communication device to the first access network device; According to the intermediate training result, continue to execute the first part of the training task that is to be executed by the first access network device; or, Building an AI model based on the model training information of the communication device; According to the constructed AI model, continue to execute the first part of the training task that is to be executed by the first access network device.

23. The method according to claim 22, characterized in that The method further comprises: Sending third AI capability information, where the third AI capability information indicates an AI training capability supported by the second access network device, and the third AI capability is used to determine that the communication device needs to be switched to the second access network device.

24. The method according to claim 23, characterized in that The third AI capability information includes at least one of the following: a third AI switch, a supported third AI algorithm type, supported third AI framework information, third computing power information, or retraining capability information, wherein the retraining capability information indicates whether the second access network device supports retraining of the trained AI model, and the third computing power information includes at least one of the following: hardware type, computing power, or resource utilization.

25. The method according to any one of claims 22 to 24, characterized in that The method further comprises: Receiving switching request information, the switching request information is used to request to switch the communication device to the second access network device, the switching request information includes the intermediate training result of the AI ​​model or the model training information of the communication device, and the identifier of the first part; Access control of the communication device is completed according to the switching request information.

26. The method according to any one of claims 22 to 25, characterized in that The method further comprises: Sending vendor information, wherein the vendor information indicates a vendor providing the second access network device, and the vendor information is used to determine that the communication device will switch from the first access network device to the second access network device.

27. The method according to claim 25 or 26, characterized in that The receiving of the switching request information includes: In a case where the first access network device and the second access network device are provided by the same manufacturer, first switching request information is received, wherein the first switching request information includes an intermediate training result of the AI ​​model and an identifier of the first part.

28. The method according to any one of claims 25 to 27, characterized in that The receiving of the switching request information includes: In a case where the first access network device and the second access network device are provided by different manufacturers, second switching request information is received, wherein the second switching request information includes model training information of the communication apparatus and an identifier of the first part.

29. The method according to any one of claims 25 to 28, characterized in that The switching request information further includes an identifier of the second part. Before continuing to execute the first part of the training task that is to be executed by the first access network device, the method further includes: Send training task initiation information to the communication device, where the training task initiation information includes an identifier of the second part.

30. A switching processing method, characterized in that: Applied to a communication device, the communication device and a first access network device jointly perform a training task of a bilateral artificial intelligence AI model, the method comprising: In a case where the communication device is to switch from the first access network device to the second access network device, receiving training task suspension information or training task termination information, wherein the training task suspension information or the training task termination information includes an identifier of a second part of the training task to be executed by the communication device; In case of receiving the training task pause information, receiving first training task start information, where the first training task start information includes an identifier of the second part; Based on the identifier of the second part, continue to execute the second part of the training task that is to be executed by the communication device; When the training task termination information is received, the execution of the second part of the training task that is to be executed by the communication device is terminated.

31. The method according to claim 30, characterized in that The training task termination information or the training task suspension information further includes a cause value, and the cause value is used to indicate that a switch occurs in the communication device.

32. The method according to claim 30 or 31, characterized in that Before receiving the training task suspension information or the training task termination information, the method further includes: Sending first AI capability information, where the first AI capability information indicates an AI training capability supported by the communication device; receiving second training task startup information, where the second training task startup information includes model information of the AI ​​model and an identifier of the second part, where the model information is determined according to the first AI capability information and the second AI capability information, where the second AI capability information indicates an AI training capability supported by the first access network device; Based on the model information of the AI ​​model and the identifier of the second part, execute the second part of the training task that is to be executed by the communication device.

33. The method according to any one of claims 30 to 32, characterized in that The method further comprises: When the communication device switches back to the first access network device, the second part of the training task that is to be executed by the communication device continues to be executed.

34. A switching processing method, characterized in that: The first access network device and the communication device jointly perform a training task of a bilateral artificial intelligence AI model, and the method includes: The first access network device determines that the communication apparatus is to be switched from the first access network device to the second access network device; The first access network device terminates or suspends a first part of the training task that is to be executed by the first access network device, and triggers the communication device to terminate or suspend a second part of the training task that is to be executed by the communication device; The first access network device sends the intermediate training result of the AI ​​model or the model training information of the communication device, and the identifier of the first part to the second access network device, wherein the model training information is the model parameters and / or training data most recently reported by the communication device to the first access network device; The second access network device receives the intermediate training result or the model training information of the communication device, and an identifier of the first part; The second access network device continues to execute the first part of the training task that is to be executed by the first access network device according to the intermediate training result; or, The second access network device constructs an AI model according to the model training information of the communication device; The second access network device continues to execute the first part of the training task that is to be executed by the first access network device based on the constructed AI model.

35. The method according to claim 34, characterized in that The triggering the communication device to terminate or pause the second part of the training task that is to be executed by the communication device includes: The first access network device sends training task suspension information or training task termination information to the communication device, where the training task suspension information or the training task termination information includes an identifier of the second part; The communication device receives the training task suspension information or the training task termination information from the first access network device; When the communication device receives the training task pause information, the second access network device sends the first training task start information to the communication device, and the communication device receives the first training task start information from the second access network device, where the first training task start information includes the identifier of the second part; The communication device continues to execute the second part of the training task that is to be executed by the communication device based on the identifier of the second part; When the communication device receives the training task termination information, the communication device terminates the execution of the second part of the training task that is to be executed by the communication device.

36. The method according to claim 34 or 35, characterized in that Before the first access network device determines that the communication apparatus will be switched from the first access network device to the second access network device, the method further includes: The communication device sends first AI capability information to the first access network device, and the first access network device receives the first AI capability information from the communication device, where the first AI capability information indicates an AI training capability supported by the communication device; The first access network device determines, according to the first AI capability information and the second AI capability information, model information for training the AI ​​model, where the second AI capability information indicates an AI training capability supported by the first access network device; The first access network device sends second training task start information to the communication device, and the communication device receives the second training task start information from the first access network device, where the second training task start information includes model information of the AI ​​model and an identifier of the second part; The communication device executes the second part of the training task that is to be executed by the communication device according to the model information of the AI ​​model and the identifier of the second part; The first access network device executes the first part of the training task that is to be executed by the first access network device according to the model information of the AI ​​model.

37. The method according to any one of claims 34 to 36, characterized in that The first access network device determines that the communication device will be switched from the first access network device to the second access network device, including: The second access network device sends third AI information to the first access network device, and the first access network device receives the third AI capability information from the second access network device, where the third AI capability information indicates an AI training capability supported by the second access network device; The first access network device determines, based on the measurement report and the third AI capability information, that the communication device will switch from the first access network device to the second access network device, wherein the measurement report is used to indicate characteristics of a network environment in which the communication device is located.

38. The method according to any one of claims 34 to 37, characterized in that The method further comprises: When the communication device switches back to the first access network device, the communication device continues to execute the second part of the training task that is assigned to the communication device, and the first access network device continues to execute the first part of the training task that is assigned to the first access network device.

39. A switching processing device, characterized in that: include: A module for executing the method of any one of claims 1 to 21, or a module for executing the method of any one of claims 22 to 29, or a module for executing the method of any one of claims 30 to 33.

40. A switching processing device, characterized in that: include: A processor, configured to execute a computer program stored in a memory so that the switching processing device performs a method as described in any one of claims 1 to 21, or so that the switching processing device performs a method as described in any one of claims 22 to 29, or so that the switching processing device performs a method as described in any one of claims 30 to 33.

41. A computer-readable storage medium, characterized in that include: The computer-readable storage medium stores a computer program; when the computer program runs on a computer, the computer executes the method as described in any one of claims 1 to 21, or the computer executes the method as described in any one of claims 22 to 29, or the computer executes the method as described in any one of claims 30 to 33.

Images