Handover method and apparatus, communication device, and readable storage medium

Abstract

This application discloses a handover method, apparatus, communication device, and readable storage medium, belonging to the field of communication technology. The specific implementation includes: a source base station sending a first handover message to a target base station; the first handover message includes model configuration parameters of a first model, where the first model is a terminal service model trained in the source base station; the target base station constructs a second model based on the model configuration parameters of the first model. Thus, the target base station can efficiently process terminal services using the constructed second model.

Description

Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to a switching method, apparatus, communication equipment, and readable storage medium. Background Technology

[0002] With the development of communication technology, artificial intelligence (AI) models have gradually become an indispensable part of network architecture. In wireless networks, base stations can train or use service models for terminals. During this training or use process, if the terminal moves and the connected base station switches, the target base station will not have the terminal's model information, requiring model retraining. However, since model training takes time, the target base station will be unable to efficiently utilize the model for service processing. Summary of the Invention

[0003] The purpose of this application is to provide a handover method, apparatus, communication device, and readable storage medium to solve the problem that existing target base stations cannot efficiently utilize models for terminal service processing after handover due to terminal movement.

[0004] To solve the above-mentioned technical problems, this application is implemented as follows:

[0005] Firstly, a handover method is provided, applied to the source base station, including:

[0006] Send the first handover message to the target base station;

[0007] The first handover message includes model configuration parameters of a first model, which is a terminal service model trained in the source base station. The target base station constructs a second model based on the model configuration parameters of the first model.

[0008] Secondly, a handover method is provided, applied to a target base station, including:

[0009] Receive a first handover message from the source base station; wherein the first handover message includes model configuration parameters of a first model, and the first model is a terminal service model trained in the source base station;

[0010] Based on the model configuration parameters, a second model corresponding to the first model is constructed.

[0011] Thirdly, a handover device is provided for use at a source base station, comprising:

[0012] The first sending module is used to send a first handover message to the target base station;

[0013] The first handover message includes model configuration parameters of a first model, which is a terminal service model trained in the source base station. The target base station constructs a second model based on the model configuration parameters of the first model.

[0014] Fourthly, a handover device is provided, applied to a target base station, comprising:

[0015] The second receiving module is used to receive a first handover message from the source base station; wherein the first handover message includes model configuration parameters of a first model, and the first model is a terminal service model trained in the source base station.

[0016] A construction module is used to construct a second model corresponding to the first model based on the model configuration parameters.

[0017] Fifthly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect, or the steps of the method described in the second aspect.

[0018] In a sixth aspect, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect, or the steps of the method described in the second aspect.

[0019] In this embodiment of the application, when a handover occurs between base stations, a first handover message is sent from the source base station to the target base station. The first handover message includes model configuration parameters of a first model. The first model is a terminal service model trained in the source base station. This allows the target base station to construct a second model corresponding to the first model based on the received model configuration parameters, and to perform corresponding processing operations or continue model training based on the constructed second model, instead of retraining the model. This reduces model training time and computational load, thereby achieving efficient terminal service processing. Attached Figure Description

[0020] Figure 1 This is a flowchart of a switching method provided in an embodiment of this application;

[0021] Figure 2 This is a schematic diagram of model transfer in an embodiment of this application;

[0022] Figure 3 This is a flowchart of another switching method provided in the embodiments of this application;

[0023] Figure 4This is a flowchart of the switching process in a specific embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the structure of a switching device provided in an embodiment of this application;

[0025] Figure 6 This is a schematic diagram of another switching device provided in an embodiment of this application;

[0026] Figure 7 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and are not limited in number; for example, a first object can be one or more.

[0029] To facilitate understanding of the embodiments of this application, the following will be described first.

[0030] Transfer learning is the process of using a model developed for one task as a starting point to develop a model for another task. There are two fundamental concepts in transfer learning: domain and task. Transfer learning can be defined as an algorithm that, given a source domain, a source task, a target domain, and a target task, utilizes knowledge gained from solving the source task to solve the target task.

[0031] Model transfer learning, a method of transfer learning, leverages the similarity relationships between existing models to transfer model parameters to a new model, aiding in its training and accelerating learning and optimization. Therefore, models trained on base stations with abundant data samples can be transferred to other base stations for use or retraining, fully utilizing the advantages of transfer learning to apply learned knowledge to solve problems on other base stations and improve model training performance.

[0032] To address the problem that existing target base stations cannot efficiently utilize models for terminal service processing after handover due to terminal mobility, this application introduces transfer learning in its embodiments. If a handover is initiated while the source base station is training or using a terminal service model, the handover message (e.g., a handover request message) sent from the source base station to the target base station carries model configuration parameters for a first model. The first model is the terminal service model trained in the source base station. This allows the target base station to construct a second model corresponding to the first model based on the received model configuration parameters, and then perform corresponding processing operations or continue model training based on the constructed second model, thereby efficiently processing terminal services.

[0033] Optionally, the scenarios applicable to the embodiments of this application include, but are not limited to, the 5th generation (5 th Generation 5G communication systems can also be used in other communication systems, such as 6th generation (6G) communication systems. th Generation 6G communication systems, etc.

[0034] The switching method, apparatus, communication device, and readable storage medium provided in this application will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios.

[0035] Please see Figure 1 , Figure 1 This is a flowchart of a handover method provided in an embodiment of this application. The method is applied to the source base station, such as... Figure 1 As shown, the method includes the following steps:

[0036] Step 11: Send the first handover message to the target base station.

[0037] In this embodiment, the first handover message includes model configuration parameters for a first model. The first model is a terminal service model trained in the source base station. The model configuration parameters are used by the target base station to construct a second model corresponding to the first model; that is, the target base station constructs a corresponding second model based on the model configuration parameters of the first model.

[0038] In some embodiments, the first switching message is a switching request (such as a Handover Request) message.

[0039] Understandably, the first model and the second model are different models, located at the source base station and the target base station, respectively. The source base station is the source domain in transfer learning, i.e., the base station that provides the model parameters. The target base station is the target domain in transfer learning, i.e., the base station that receives the model parameters.

[0040] In some embodiments, the first model may be an AI model for terminal business data, such as a traffic prediction model or a movement trajectory prediction model.

[0041] In some embodiments, the first handover message can be transmitted via the Xn interface between the source base station and the target base station.

[0042] In some embodiments, such as Figure 2 As shown, when a user equipment (UE) switches from a source base station to a target base station due to location change, the source base station can send the model configuration parameters of its trained model to the target base station through a handover request, so that the target base station can build the corresponding model for use.

[0043] The handover method of this application embodiment, when a handover occurs between base stations of a terminal, sends a first handover message to the target base station through the source base station. The first handover message includes model configuration parameters of a first model. The first model is a terminal service model trained in the source base station. This allows the target base station to construct a second model corresponding to the first model based on the received model configuration parameters, and perform corresponding processing operations or continue model training based on the constructed second model, instead of retraining the model, thereby reducing model training time and computational load, and thus achieving efficient terminal service processing.

[0044] In a specific application scenario, when training or inferring AI models for UE service data, the base station transmits relevant decision results or parameters to the UE to improve its communication performance. When the UE moves, to ensure the target base station efficiently utilizes the model for service processing, the source base station needs to migrate its trained model to the target base station. For example, a base station trains a traffic prediction model to achieve traffic prediction, providing AI capabilities to meet diverse user needs. If the UE moves—that is, switches from one base station to an adjacent one—the change in UE location will interrupt the ongoing model training or inference at the source base station. Since the target base station lacks the necessary model parameters, it needs to retrain the model, thus affecting the efficiency of traffic prediction. In this case, the source base station can transmit the model configuration parameters of its trained model to the target base station, enabling the target base station to build and use the corresponding traffic prediction model based on the received model configuration parameters, thereby improving traffic prediction efficiency.

[0045] In this embodiment of the application, the model configuration parameters included in the first handover message sent by the source base station may include, but are not limited to, at least one of the following of the first model:

[0046] (1) Model structure configuration information; for example, model structure configuration information may include the number of layers n of the neural network in which the current structure is located, the calculation rules of each network structure, and the configuration parameters of each network structure category.

[0047] (2) Model training parameters, which are the parameters of the training model in the source base station.

[0048] (3) Input sample dimension; for example, the input sample dimension may include several types of samples and the type definition of each type of sample.

[0049] For example, taking a traffic prediction model as an example, the input sample categories can include data traffic matrix sequences, UE historical traffic, and the size of service packets, etc.

[0050] (4) Training mask; where the training mask is used to distinguish in which task scenario the current model was trained. For example, if the training task is traffic prediction, then the currently trained model and parameters are suitable for traffic prediction.

[0051] (5) The loss function is used to measure the degree of inconsistency between the model's predictions and the actual values.

[0052] For example, the loss function can be selected as the mean squared error loss function, the cross-entropy loss function, and / or the smooth L1 loss function.

[0053] (6) Hyperparameters are parameters whose values ​​are set before the model training process begins.

[0054] Optionally, the hyperparameters mentioned above may include, but are not limited to, at least one of the following parameters:

[0055] 1) Maximum number of training epochs (N): The process of training all training samples once is counted as a single training session. Training can be stopped when the number of training sessions reaches N. For example, the traffic prediction model stops training when it reaches the specified number of epochs.

[0056] 2) Current training epoch (n): The number of training epochs so far, for example, the number of epochs the traffic prediction model was trained before the UE switch.

[0057] 3) Number of samples per training batch (batchsize): Each training session will select batchsize samples for training, and the model weights will be updated once every batchsize samples.

[0058] 4) Gradient threshold: When the gradient is greater than the threshold, it can be converted to the threshold for parameter update calculation.

[0059] 5) Learning rate: A parameter that controls how fast the model learns.

[0060] 6) Loss function parameter: an adjustable parameter on the loss function.

[0061] 7) Optimizer: Used to update and compute network parameters that affect model training and model output, making them approach or reach the optimal value, thereby minimizing (or maximizing) the loss function, such as SGD, Adagrad, RMSprop, etc.

[0062] 8) Regularization: Used to modify learning algorithms to reduce generalization error rather than training error, such as L1 norm, L2 norm, etc.

[0063] In this embodiment, the first model can be a pre-trained model in the source base station or a model currently being trained in the source base station. Therefore, to facilitate the target base station's knowledge of the first model's status, the first handover message sent by the source base station may further include a first indicator bit. This first indicator bit indicates that the first model is in use, such as inference / prediction state, or it indicates that the first model is in training state. For example, when the first indicator bit is marked as 0, it can indicate that the corresponding model is in use; or, when the first indicator bit is marked as 1, it can indicate that the corresponding model is in training state.

[0064] Thus, when the first indicator bit indicates that the first model is in use, the target base station can directly use the second model to perform corresponding processing operations after constructing the second model corresponding to the first model; or, when the first indicator bit indicates that the first model is in training, the target base station needs to continue training the second model after constructing the second model corresponding to the first model.

[0065] Optionally, when the first indicator bit indicates that the first model is in training or usage mode, and the target base station does not have the required sample data, the target base station can request the required sample data. In this case, after sending a first handover message to the target base station, the source base station can receive a second handover message from the target base station. This second handover message includes confirmation information for the model configuration parameters and the sample categories required by the target base station. Then, the source base station sends the sample categories required by the target base station to the terminal, and the terminal sends the sample data corresponding to the required sample categories to the target base station. In this way, after receiving the sample data, the target base station can train the constructed second model or process the sample data according to the second model.

[0066] In some embodiments, the second handover message may be a handover request acknowledgment (e.g., a Handover Request ACK) message.

[0067] In some embodiments, regardless of whether the first model (i.e., the model trained in the source base station) is in use or in training state, the model configuration parameters of the first model can be added to the IE information of the handover request (e.g., Handover Request) message for transmission, to enable model migration between base stations when the UE moves. More specifically, migration management AI model information can be added to the handover request (e.g., Handover Request) message, and the model configuration parameters can be included in the transmission of the Mobile management AI model information.

[0068] In some embodiments, the confirmation feedback information of the received model configuration parameters and the sample category required by the target base station can be added to the IE information of the handover request confirmation (such as Handover Request ACK) message for transmission.

[0069] In some embodiments, the required sample category sent by the source base station to the target base station of the terminal can be added to the IE information of the Radio Resource Control (RRC) reconfiguration message for transmission.

[0070] Please see Figure 3 , Figure 3 This is a flowchart of a handover method provided in an embodiment of this application. The method is applied to a target base station, such as... Figure 3 As shown, the method includes the following steps:

[0071] Step 31: Receive the first handover message from the source base station.

[0072] In this embodiment, the first handover message includes model configuration parameters for the first model. The first model is a terminal service model trained in the source base station. The model configuration parameters are used by the target base station to construct a second model corresponding to the first model. For example, the first handover message is a handover request message (such as a Handover Request).

[0073] Step 32: Construct a second model corresponding to the first model based on the model configuration parameters.

[0074] Understandably, the first model and the second model are different models, located at the source base station and the target base station, respectively. The source base station is the source domain in transfer learning, i.e., the base station that provides the model parameters. The target base station is the target domain in transfer learning, i.e., the base station that receives the model parameters.

[0075] The handover method of this application embodiment, when a handover occurs between base stations of a terminal, receives a first handover message from the source base station. The first handover message includes model configuration parameters of a first model, which is a terminal service model trained in the source base station. Based on the model configuration parameters, a second model corresponding to the first model is constructed. This allows the target base station to perform corresponding processing operations or continue model training based on the constructed second model, instead of retraining the model, thereby reducing model training time and computational load, and achieving efficient terminal service processing.

[0076] Optionally, the first model mentioned above can be a pre-trained model in the source base station, or a model being trained in the source base station. Therefore, the first handover message sent by the source base station may also include a first indicator bit, which indicates that the first model is in use, such as inference / prediction state, or that the first indicator bit is in training state. For example, when the first indicator bit is marked as 0, it can indicate that the corresponding model is in use; or, when the first indicator bit is marked as 1, it can indicate that the corresponding model is in training state.

[0077] Optionally, when the first indicator bit indicates that the first model is in training or usage state, and the target base station does not have the required sample data, the target base station, after receiving the first handover message, can send a second handover message to the source base station. This second handover message includes confirmation information about the model configuration parameters and the sample categories required by the target base station. The source base station then sends the sample categories to the terminal, receives the sample data corresponding to the sample categories from the terminal, and uses the sample data to train the second model, or processes the sample data according to the second model. For example, the second handover message could be a handover request confirmation message.

[0078] Optionally, after constructing a second model corresponding to the first model, the target base station can use the second model to perform corresponding processing operations, or it can use the second model as a pre-trained model to continue model training, thereby shortening the training time, accelerating the convergence speed, and training more accurate model parameters.

[0079] Optionally, the above model configuration parameters may include, but are not limited to, at least one of the following of the first model:

[0080] Model structure configuration information;

[0081] Model training parameters;

[0082] Enter the sample category;

[0083] Training tasks;

[0084] Loss function;

[0085] Hyperparameters.

[0086] It should be noted that the specific contents of the model configuration parameters in this embodiment can be found in the above embodiments. To avoid redundant limitations, they will not be repeated here.

[0087] The following is combined with Figure 4 The switching process in a specific example of this application is explained.

[0088] In a specific example of this application, traffic prediction is used as the training task. Figure 4 As shown, the switching process includes the following steps:

[0089] Step 41: During the training or use of the traffic prediction model by the source base station (gNB), if the terminal location moves, the source gNB initiates a handover and sends a handover request (e.g., Handover Request) message to the target gNB via the Xn interface. In addition to the information required by the target gNB for the handover, the source gNB also needs to include the model configuration parameters of the traffic prediction model in the Request message.

[0090] Step 42: The target gNB constructs the corresponding traffic prediction model based on the received model configuration parameters in order to perform traffic prediction or continue model training.

[0091] Optionally, depending on the state of the traffic prediction model in the source base station—either in use / inference state or training state—the Request message may also include a first indicator bit (or training / inference indicator bit) to indicate whether the traffic prediction model is in inference or training state. When the first indicator bit indicates that the traffic prediction model is in inference state, the model configuration parameters may include, but are not limited to, at least one of the following parameters: model structure configuration information, model training parameters, input sample categories, and training tasks, to complete the parameter transfer of the model.

[0092] Alternatively, when the first indicator bit indicates that the traffic prediction model is in the training state, the model configuration parameters may include, but are not limited to, at least one of the following parameters: model structure configuration information, model training parameters, input sample category, training task, loss function, and hyperparameters, to complete the continued training of the transfer model. Hyperparameters may include, but are not limited to, at least one of the following: learning rate, gradient threshold, number of samples per training session, maximum number of training epochs, optimizer, current number of training epochs, parameters of the loss function, regularization, etc.

[0093] Optionally, if training the constructed traffic prediction model is required, the target gNB will determine whether it has the necessary sample data. If it does not have the necessary sample data, it can transmit the historical traffic data of the UE to be collected through steps 43 and 44, and transmit the historical traffic data of the UE through step 45 to complete the subsequent training or inference. If it has the necessary sample data, it can directly input the sample data into the model for model training or inference.

[0094] Step 43: The target gNB performs quasi-handover control, confirming the information transmitted by the source gNB via a handover request confirmation (e.g., Handover Request Acknowledge) message, and carrying the necessary RRC configuration parameters for the target gNB. If the target base station does not have the required sample data, it also needs to carry the sample categories of the UE that the target base station needs to collect.

[0095] Step 44: The source gNB sends a non-access stratum (NAS) handover command (such as a Handover Command) to the UE via RRC signaling (such as RRC reconfiguration). This command includes the RRC configuration of the target gNB and the sample categories that the target base station needs to collect (such as the target base station not having the required sample data).

[0096] Step 45: The UE switches to the new cell and notifies the target gNB of the RRC connection completion via an RRC reconfiguration complete message (e.g., RRC ReconfigurationComplete). If the target base station does not have the required sample data, the UE will transmit sample data to the target gNB according to the received sample category, so that the target gNB can input the received sample data into the already constructed model to continue model training or inference.

[0097] Optionally, if the first indicator bit indicates that the traffic prediction model is in inference mode, the target gNB can input sample data into the model for inference after building the model to predict the UE's traffic usage. After outputting the prediction results, the target gNB can modify the network configuration based on the prediction results, which may include bandwidth allocation, to improve user performance. If the predicted data is inaccurate, the module can be iteratively updated until the expected effect or convergence threshold is reached.

[0098] Optionally, if the first indicator bit indicates that the traffic prediction model is in a training state, the target gNB can input sample data into the model for training after building the model. Through continuous forward and backward propagation, the model is optimized and updated. For example, if the model is trained for the nth time, the target gNB will perform the nth forward propagation, calculate the descent direction of the gradient based on the loss function, perform backward propagation, calculate the new weights for each layer of the neural network, and then perform n+1 forward inferences. This process is repeated until the current number of training rounds reaches the maximum number of training rounds or the convergence threshold, at which point training stops, and model optimization is complete.

[0099] It should be noted that the switching method provided in this application embodiment can be executed by a switching device or a control module within the switching device for executing the switching method. This application embodiment uses the switching device executing the switching method as an example to illustrate the switching device provided in this application embodiment.

[0100] Please see Figure 5 , Figure 5 This is a schematic diagram of a handover device provided in an embodiment of this application. The device is applied to a source base station, such as... Figure 5 As shown, the switching device 50 includes:

[0101] The first sending module 51 is used to send a first handover message to the target base station;

[0102] The first handover message includes model configuration parameters for a first model, which is a terminal service model trained in the source base station. The target base station constructs a corresponding second model based on the model configuration parameters in the first model.

[0103] Optionally, the first switching message may further include a first indicator bit, which indicates that the first model is in use or that the first model is in training.

[0104] Optionally, the switching device 50 also includes:

[0105] A first receiving module is configured to receive a second handover message from the target base station when the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data; wherein the second handover message includes: the sample category required by the target base station;

[0106] The second sending module is used to send the sample category required by the target base station to the terminal, and the terminal sends the sample data corresponding to the sample category to the target base station.

[0107] Optionally, the model configuration parameters include at least one of the following of the first model:

[0108] Model structure configuration information; model training parameters; input sample categories; training task; loss function; hyperparameters.

[0109] The switching device 50 in this embodiment can achieve the above-mentioned... Figure 1 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0110] Please see Figure 6 , Figure 6 This is a schematic diagram of a handover device provided in an embodiment of this application. The device is applied to a target base station, such as... Figure 6 As shown, the switching device 60 includes:

[0111] The second receiving module 61 is used to receive a first handover message from the source base station; wherein the first handover message includes model configuration parameters of a first model, and the first model is a terminal service model trained in the source base station.

[0112] The construction module 62 is used to construct a second model corresponding to the first model according to the model configuration parameters.

[0113] Optionally, the first switching message may further include a first indicator bit, which indicates that the first model is in use or that the first model is in training.

[0114] Optionally, the switching device 60 further includes:

[0115] The third sending module is used to send a second handover message to the source base station when the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data; wherein, the second handover message includes the sample category required by the target base station, and the source base station sends the sample category to the terminal;

[0116] The third receiving module is used to receive sample data corresponding to the sample category from the terminal;

[0117] The processing module is used to train the second model using the sample data, or to process the sample data based on the second model.

[0118] Optionally, the switching device 60 further includes:

[0119] An execution module is used to perform at least one of the following:

[0120] The second model is used to perform the corresponding processing operations;

[0121] The second model is used as a pre-trained model for further model training.

[0122] Optionally, the model configuration parameters include at least one of the following of the first model:

[0123] Model structure configuration information; model training parameters; input sample categories; training task; loss function; hyperparameters.

[0124] The switching device 60 in this embodiment can achieve the above-mentioned... Figure 3 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0125] In addition, such as Figure 7 As shown, this application embodiment also provides a communication device 70, including a processor 71, a memory 72, and a program or instructions stored in the memory 72 and executable on the processor 71. For example, when the communication device 70 is a source base station, the program or instructions executed by the processor 71 implement the above-mentioned... Figure 1 The various processes of the switching method embodiment shown; or, when the communication device 70 is the target base station, the program or instructions executed by the processor 71 implement the above. Figure 3 The various processes of the switching method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0126] This application also provides a readable storage medium storing a program or instructions that, when executed by a processor, can achieve the above-described functions. Figure 1The various processes of the switching method embodiments shown, or the implementation of the above... Figure 3 The various processes of the switching method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0127] Computer-readable media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0128] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0129] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0130] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a service classification device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0131] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A handover method applied to a source base station, characterized in that, include: Send the first handover message to the target base station; The first handover message includes model configuration parameters for a first model, which is a terminal service model trained in the source base station. The target base station constructs a second model based on the model configuration parameters of the first model. The terminal service model includes a traffic prediction model or a movement trajectory prediction model. The first switching message also includes a first indicator bit, which is used to indicate that the first model is in use, or the first indicator bit is used to indicate that the first model is in training. When the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data, the method further includes: Receive a second handover message from the target base station; wherein the second handover message includes: the sample category required by the target base station; The terminal sends the sample category required by the target base station to the terminal, and the terminal sends the sample data corresponding to the sample category to the target base station.

2. The method according to claim 1, characterized in that, The model configuration parameters include at least one of the following for the first model: Model structure configuration information; Model training parameters; Enter the sample category; Training tasks; Loss function; Hyperparameters.

3. A handover method applied to a target base station, characterized in that, include: Receive a first handover message from the source base station; wherein the first handover message includes model configuration parameters of a first model, the first model being a terminal service model trained in the source base station; the terminal service model includes a traffic prediction model or a movement trajectory prediction model; Based on the model configuration parameters, construct a second model corresponding to the first model; The first switching message also includes a first indicator bit, which is used to indicate that the first model is in use, or the first indicator bit is used to indicate that the first model is in training. When the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data, the method further includes: A second handover message is sent to the source base station; wherein the second handover message includes the sample category required by the target base station, and the source base station sends the sample category to the terminal; The terminal receives sample data corresponding to the sample category and uses the sample data to train the second model, or processes the sample data based on the second model.

4. The method according to claim 3, characterized in that, After constructing the second model corresponding to the first model, the method further includes at least one of the following: The second model is used to perform the corresponding processing operations; The second model is used as a pre-trained model for further model training.

5. The method according to any one of claims 3 to 4, characterized in that, The model configuration parameters include at least one of the following for the first model: Model structure configuration information; Model training parameters; Enter the sample category; Training tasks; Loss function; Hyperparameters.

6. A handover device applied to a source base station, characterized in that, include: The first sending module is used to send a first handover message to the target base station; The first handover message includes model configuration parameters for a first model, which is a terminal service model trained in the source base station. The target base station constructs a second model based on the model configuration parameters of the first model. The terminal service model includes a traffic prediction model or a movement trajectory prediction model. The first switching message also includes a first indicator bit, which is used to indicate that the first model is in use, or the first indicator bit is used to indicate that the first model is in training. A first receiving module is configured to receive a second handover message from the target base station when the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data; wherein the second handover message includes: the sample category required by the target base station; The second sending module is used to send the sample category required by the target base station to the terminal, and the terminal sends the sample data corresponding to the sample category to the target base station.

7. The apparatus according to claim 6, characterized in that, The model configuration parameters include at least one of the following for the first model: Model structure configuration information; Model training parameters; Enter the sample category; Training tasks; Loss function; Hyperparameters.

8. A handover device applied to a target base station, characterized in that, include: The second receiving module is used to receive a first handover message from the source base station; wherein the first handover message includes model configuration parameters of a first model, the first model being a terminal service model trained in the source base station; the terminal service model includes a traffic prediction model or a movement trajectory prediction model. A construction module is used to construct a second model corresponding to the first model based on the model configuration parameters; The first switching message also includes a first indicator bit, which is used to indicate that the first model is in use, or the first indicator bit is used to indicate that the first model is in training. The third sending module is used to send a second handover message to the source base station when the first indicator bit indicates that the first model is in a training state or a usage state, and the target base station does not have the required sample data; wherein, the second handover message includes the sample category required by the target base station, and the source base station sends the sample category to the terminal; The third receiving module is used to receive sample data corresponding to the sample category from the terminal; The processing module is used to train the second model using the sample data, or to process the sample data based on the second model.

9. The apparatus according to claim 8, characterized in that, The device further includes: An execution module is used to perform at least one of the following: The second model is used to perform the corresponding processing operations; The second model is used as a pre-trained model for further model training.

10. The apparatus according to any one of claims 8 to 9, characterized in that, The model configuration parameters include at least one of the following for the first model: Model structure configuration information; Model training parameters; Enter the sample category; Training tasks; Loss function; Hyperparameters.

11. A communication device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the switching method as described in any one of claims 1-2, or the steps of the switching method as described in any one of claims 3-5.

12. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the switching method as described in any one of claims 1-2, or the steps of the switching method as described in any one of claims 3-5.

Images