Model switching method and related apparatus

Abstract

The present application provides a model switching method and a related apparatus, capable of avoiding downloading a complete AI model every time a network coverage area or a task is switched, thereby reducing signaling overhead. The method comprises: determining an AI model, wherein the AI model comprises multiple levels of sub-models, and each level of sub-model among the multiple levels of sub-models is applicable to at least one object; acquiring an identifier of a target object; on the basis of the identifier of the target object, determining a target sub-model at the level corresponding to the target object; and switching a sub-model at the level corresponding to the target object among the multiple levels of sub-models to the target sub-model.

Description

Model switching method and related device

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on December 27, 2023, with application number 202311840732.7 and application name “Model Switching Method and Related Devices”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of communications, and in particular to a model switching method and related devices. Background Art

[0003] In communication devices, processing modules based on artificial intelligence (AI) models can optimize channel coding, modulation, waveform modules in the physical layer, as well as upper-layer beam management, channel state information (CSI) compression, positioning and other functions according to scenarios through data-driven training, thereby achieving better transmission signal design and reception performance.

[0004] Currently, different AI models may correspond to different network coverage areas or tasks (for example, positioning tasks, beam prediction tasks). Network devices can train AI models for specific network coverage areas and specific tasks. Terminals can download trained AI models from network devices. When the terminal moves, causing the network coverage area to switch, or the task to switch, the terminal needs to download the AI ​​model applicable to the switched network coverage area or task from the network device.

[0005] However, downloading the complete AI model from the network device side each time the network coverage area or task is switched will incur large signaling overhead. Summary of the Invention

[0006] The present application provides a model switching method and related devices, which can avoid downloading the complete AI model every time the network coverage area or task is switched, which is conducive to reducing signaling overhead.

[0007] In the first aspect, a model switching method is provided, which can be executed by a first communication device. The first communication device can be a terminal or a network device, or a component configured in the terminal or network device (such as a processor, chip, or chip system, etc.), or a logic module or software that can realize all or part of the functions of the first communication device. This application does not limit this.

[0008] The method includes: determining an AI model, which includes a multi-level sub-model, and each level sub-model in the multi-level sub-model is applicable to at least one object; obtaining the identifier of the target object; based on the identifier of the target object, determining the target sub-model of the level corresponding to the target object; and switching the sub-model of the level corresponding to the target object in the multi-level sub-model to the target sub-model.

[0009] In the present application, the AI ​​model has a hierarchical structure, including multiple levels of sub-models, and each level of sub-model has at least one applicable object, which can be a network coverage area or a task.

[0010] The objects applicable to multiple sub-models are nested level by level, or in other words, at least one object applicable to the upper-level sub-model includes at least one object applicable to the lower-level sub-model. For example, the hierarchical structure of the AI ​​model includes levels 1, 2, and 3, where level 1 is the upper level of level 2, and level 2 is the upper level of level 3. At least one object applicable to the sub-model of level 1 includes at least one object applicable to the sub-model of level 2, and at least one object applicable to the sub-model of level 2 includes at least one object applicable to the sub-model of level 3.

[0011] Taking the three levels above as an example, a sub-model at level 1 can correspond to one or more sub-models at level 2, and a sub-model at level 2 can correspond to one or more sub-models at level 3. A sub-model at level 1, a sub-model at level 2, and a sub-model at level 3 can be cascaded to form a complete AI model.

[0012] Based on the technical solution of this application, when switching objects, the first communication device can simply switch the submodel at the level corresponding to the target object in the multi-level submodel to the target submodel. The target submodel is the submodel applicable to the target object, while submodels at other levels do not need to be switched if they are still applicable to the target object. This eliminates the need to switch the entire AI model, thereby avoiding downloading the complete AI model every time the network coverage area or task is switched, which helps reduce signaling overhead.

[0013] In combination with the first aspect, in certain implementations of the first aspect, before determining the AI ​​model, the method also includes: receiving a first message, the first message indicating parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each level of sub-model, the multi-level sub-model being determined from the AI ​​model based on the hierarchical structure of the AI ​​model.

[0014] In combination with the first aspect, in certain implementations of the first aspect, the hierarchical structure of the AI ​​model indicates the starting neural network layer and the ending neural network layer of each level of sub-model in the AI ​​model.

[0015] In combination with the first aspect, in certain implementations of the first aspect, based on the identification of the target object, determining the target sub-model of the hierarchy corresponding to the target object includes: based on the identification of the target object, determining whether at least one object applicable to the sub-model of the hierarchy corresponding to the target object in the multi-level sub-model includes the target object; if at least one object applicable to the sub-model of the hierarchy corresponding to the target object in the multi-level sub-model does not include the target object, obtaining parameters for constructing the target sub-model; and constructing the target sub-model based on the parameters for constructing the target sub-model.

[0016] In combination with the first aspect, in certain implementations of the first aspect, the parameters used to construct the target submodel are carried through the following messages: radio resource control (RRC) message, medium access control (MAC) control element (CE), downlink control information (DCI) or uplink control information (UCI).

[0017] In combination with the first aspect, in some implementations of the first aspect, an identifier of the target object is carried in a switching message, and the switching message is used to indicate switching to the target object.

[0018] In this application, by reusing the switching messages in the existing process and combining the binding relationship between the sub-model and the identifier of the target object, the target sub-model suitable for the target object can be determined without sending the identifier of the target sub-model additionally, thereby reducing signaling overhead.

[0019] In combination with the first aspect, in certain implementations of the first aspect, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on network coverage areas of multiple different granularities.

[0020] In conjunction with the first aspect, in certain implementations of the first aspect, the object is a network coverage area, and the network coverage areas of multiple different granularities are: the network coverage area of ​​a cell group, the network coverage area of ​​a cell, and the network coverage area of ​​a beam. The multi-level sub-model includes: a first-level sub-model, a second-level sub-model, and a third-level sub-model. The network coverage area applicable to the first-level sub-model is the network coverage area of ​​the cell group, the network coverage area applicable to the second-level sub-model is the network coverage area of ​​at least one cell in the cell group, and the network coverage area applicable to the third-level sub-model is the network coverage area of ​​at least one beam within the at least one cell.

[0021] In conjunction with the first aspect, in certain implementations of the first aspect, the method further includes storing the first-level sub-model and / or the second-level sub-model. In this way, when switching between different network coverage areas applicable to the first sub-model or the second sub-model, there is no need to receive parameters for constructing the first sub-model and / or the second sub-model, thereby reducing signaling overhead.

[0022] In combination with the first aspect, in some implementations of the first aspect, the identifier of the target object includes a cell group identifier, a cell identifier, and a beam identifier.

[0023] In conjunction with the first aspect, in certain implementations of the first aspect, the method further includes: training / evaluating, based on data collected in the first network coverage area, a sub-model at a level corresponding to the first network coverage area in the multi-level sub-model. This helps improve the accuracy of sub-model training / evaluation.

[0024] In combination with the first aspect, in certain implementations of the first aspect, before training / evaluating the sub-model of the level corresponding to the first network coverage area in the multi-level sub-model based on the data collected in the first network coverage area, the method also includes: marking the data collected in the first network coverage area with the identifier of the first network coverage area.

[0025] In combination with the first aspect, in certain implementations of the first aspect, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on multiple data processing functions, and the multiple data processing functions are functions provided by the AI ​​model based on input data.

[0026] In this application, the AI ​​model can provide multiple data processing functions based on the same input data. For multiple different tasks in the same task group, the input data of the AI ​​model is the same input data when implementing the multiple different tasks.

[0027] For example, Task 1 and Task 2 belong to the same task group A1. For all tasks in task group A1, that is, Task 1 and Task 2, the input data of the AI ​​model is CSI.

[0028] In conjunction with the first aspect, in certain implementations of the first aspect, the object is a task, and the multiple data processing functions are: a feature extraction function associated with a task group corresponding to the input data, a feature extraction function associated with at least one task in the task group, and an output function associated with one of the at least one tasks. The multi-level sub-model includes: a first-level sub-model, a second-level sub-model, and a third-level sub-model, wherein the first-level sub-model is applicable to the task group, the second-level sub-model is applicable to the at least one task, and the third-level sub-model is applicable to one of the at least one task.

[0029] In conjunction with the first aspect, in certain implementations of the first aspect, the method further includes: storing the first-level sub-model and / or the second-level sub-model. In this way, when switching between different tasks applicable to the first-level sub-model or the second-level sub-model, there is no need to receive parameters for constructing the first-level sub-model and / or the second-level sub-model, thereby reducing signaling overhead.

[0030] In combination with the first aspect, in some implementations of the first aspect, the identifier of the target object includes an identifier of the task group to which the target object belongs.

[0031] In a second aspect, a communication device is provided, including: a module for executing the method in any possible implementation of the first aspect. Specifically, the device includes a module for executing the method in any possible implementation of the first aspect.

[0032] In one design, the device may include a module that executes the methods / operations / steps / actions described in the first aspect above. The module may be a hardware circuit, software, or a combination of hardware circuit and software.

[0033] In another design, the device is a communication chip, which may include an input circuit or interface for sending information or data, and an output circuit or interface for receiving information or data.

[0034] In another design, the apparatus is a terminal or a network device, which may include a transmitter for sending information or data and a receiver for receiving information or data.

[0035] In another design, the apparatus is used to execute the method in any possible implementation of the first aspect above, and the apparatus can be configured in a terminal or a network device.

[0036] In a third aspect, a communication device is provided, comprising a processor configured to call and run a computer program from a memory, so that the device executes the method in any possible implementation of the first aspect.

[0037] Optionally, the device further comprises a memory, which can be used to store instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the above aspects can be implemented.

[0038] Optionally, the device further includes: a transmitter (emitter) and a receiver (receiver), and the transmitter and the receiver can be separately provided or integrated together, and are referred to as a transceiver (transceiver).

[0039] In a fourth aspect, another communication device is provided, comprising a processor, wherein the processor is configured to execute a computer program or instruction in a memory to implement a method as in any possible implementation of the first aspect described above.

[0040] In a fifth aspect, a computer program product is provided, comprising: a computer program (also referred to as code, or instructions), which, when executed, enables a computer to execute the method in any possible implementation of the first aspect.

[0041] In a sixth aspect, a computer-readable storage medium is provided for storing a computer program or instruction, which, when executed on a computer, enables the method in any possible implementation of the first aspect to be executed.

[0042] In the seventh aspect, the present application provides a chip system, which includes at least one processor for supporting the functions involved in implementing any possible implementation method of any of the above aspects, such as receiving or processing the data involved in the above method.

[0043] In one possible design, the chip system further includes a memory, which is used to store program instructions and data, and the memory is located inside or outside the processor.

[0044] Optionally, the chip system may consist of a chip, or may include a chip and other discrete devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] FIG1 is a schematic diagram of a neuron structure;

[0046] FIG2 is a schematic diagram of the structure of a neural network;

[0047] FIG3 is a schematic diagram of a communication system provided in an embodiment of the present application;

[0048] FIG4 is a schematic diagram of a model download provided in an embodiment of the present application;

[0049] FIG5 is a schematic diagram of a model switching provided by an embodiment of the present application;

[0050] FIG6 is a schematic flow chart of a model switching method provided in an embodiment of the present application;

[0051] FIG7 is a schematic diagram of a hierarchical structure of an AI model provided in an embodiment of the present application;

[0052] FIG8 is a schematic flow chart of another model switching method provided in an embodiment of the present application;

[0053] FIG9 is a schematic diagram of another model switching provided in an embodiment of the present application;

[0054] FIG10 is a schematic structural diagram of a MAC CE provided in an embodiment of the present application;

[0055] FIG11 is a schematic diagram of a hierarchical structure of data provided in an embodiment of the present application;

[0056] FIG12 is a schematic diagram of a data identifier provided in an embodiment of the present application;

[0057] 13 to 16 are schematic block diagrams of communication devices provided in embodiments of the present application. DETAILED DESCRIPTION

[0058] The technical solution in this application will be described below with reference to the accompanying drawings.

[0059] Before introducing the model switching method and related devices provided in the embodiments of the present application, the following points are explained.

[0060] First, in the embodiments described below, various terms and abbreviations, such as AI model and multi-level sub-model, are provided for ease of description and should not constitute any limitation on this application. This application does not exclude the possibility of defining other terms in existing or future protocols that can achieve the same or similar functions.

[0061] Second, the first, second and various numerical numbers in the embodiments shown below are only used for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0062] Third, "at least one" means one or more, and "more" 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. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items 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, c can be single or multiple.

[0063] Fourth, “sending” and “receiving” in this application indicate the direction of signal transmission. For example, “sending a model request message to the base station” can be understood as the destination end of the model request message is the base station, which can include direct sending through the air interface, and also includes indirect sending through the air interface by other units or modules. “Receiving the identifier of the target object from the terminal” can be understood as the source end of the identifier of the target object is the terminal, which can include direct receiving from the terminal through the air interface, and also includes indirect receiving from the terminal through the air interface from other units or modules. “Sending” can also be understood as the “output” of the chip interface, and “receiving” can also be understood as the “input” of the chip interface.

[0064] In other words, sending and receiving can be performed between devices, for example, between a terminal and a base station; or it can be performed within a device, for example, sending or receiving between components, modules, chips, software modules or hardware modules within the device through a bus, wiring or interface.

[0065] The following is an introduction to the relevant technologies and concepts involved in this application.

[0066] Artificial intelligence (AI) can imbue machines with human intelligence. For example, it can use computer hardware and software to simulate certain intelligent human behaviors. Machine learning methods can be used to achieve AI. In machine learning, a machine uses training data to learn (or train) a model. This model represents the mapping from input to output. The learned model can be used for inference (or prediction), meaning that the model can be used to predict the output corresponding to a given input. This output can also be called an inference result (or prediction result).

[0067] Machine learning can include supervised learning, unsupervised learning, and reinforcement learning. Among them, unsupervised learning can also be called unsupervised learning.

[0068] A neural network (NN) is a specific model in machine learning technology. According to the universal approximation theorem, NNs can theoretically approximate any continuous function, enabling them to learn arbitrary mappings. Traditional communication systems require extensive expert knowledge to design communication modules. However, deep learning communication systems based on neural networks can automatically discover implicit patterns in massive data sets and establish mapping relationships between data, achieving performance superior to traditional modeling methods.

[0069] The concept of neural networks originates from the neuronal structure of the brain. Figure 1 illustrates a schematic diagram of a neuron structure. Each neuron performs a weighted sum operation on its input values ​​and outputs the result through an activation function. The activation functions of different neurons in a neural network can be the same or different.

[0070] Neural networks typically consist of multiple layers, each of which may include one or more neurons. Increasing the depth and / or width of a neural network can improve its expressive power, providing more powerful information extraction and abstract modeling capabilities for complex systems. The depth of a neural network can refer to the number of layers, while the number of neurons in each layer can be referred to as the width of that layer.

[0071] Figure 2 is a schematic diagram of the structure of a neural network. The neural network shown in Figure 2 includes an input layer, a hidden layer, and an output layer. The input layer of the neural network processes the received input information through neurons and passes the processing results to the intermediate hidden layer. The hidden layer calculates the received processing results to obtain a calculation result. The hidden layer then passes the calculation result to the output layer or the next adjacent hidden layer. The output layer ultimately obtains the output result of the neural network. A neural network can include one hidden layer or multiple hidden layers connected in sequence, without limitation.

[0072] The neural network is, for example, a deep neural network (DNN). Depending on how the network is constructed, DNNs can include feedforward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs).

[0073] During the machine learning model training process, a loss function can be defined. This function describes the gap or discrepancy between the model's output and the ideal target value. Loss functions can be expressed in various forms, and there are no restrictions on their specific form. The model training process can be viewed as adjusting some or all of the model's parameters to keep the loss function below a threshold or meet the target.

[0074] A model may also be referred to as an AI model, rule, or other name. An AI model can be considered a specific method for implementing an AI function. An AI model represents a mapping relationship or function between the input and output of a model. AI functions may include one or more of the following: data collection, model training (or model learning), model information release, model inference (or model reasoning, inference, or prediction, etc.), model monitoring or model verification, or inference result release, etc. AI functions may also be referred to as AI (related) operations, or AI-related functions.

[0075] FIG3 is a schematic diagram of a communication system 1000 provided in an embodiment of the present application. As shown in FIG3 , the communication system 1000 includes a radio access network (RAN) 100 and a core network (CN) 200. The RAN 100 includes at least one RAN node (such as 110a and 110b in FIG1 , collectively referred to as 110) and at least one terminal (such as 120a-120j in FIG1 , collectively referred to as 120). Optionally, the communication system 1000 also includes the Internet 300. It should be understood that FIG3 is a schematic diagram of a possible, non-limiting communication system, and the communication system may also include more or fewer devices, which is not limited by the present application.

[0076] RAN 100 may also include other RAN nodes, such as wireless relay equipment and / or wireless backhaul equipment (not shown in FIG1 ). Terminal 120 is wirelessly connected to RAN node 110. RAN node 110 is wirelessly or wiredly connected to core network 200. The core network equipment in core network 200 and RAN node 110 in RAN 100 may be separate physical devices, or they may be a single physical device that integrates core network logical functions and radio access network logical functions.

[0077] The RAN 100 may be a cellular system related to the Third Generation Partnership Project (3GPP), such as a fourth generation (4G) mobile communication system, a fifth generation (5G) mobile communication system, or a future-oriented evolutionary system, such as a sixth generation (6G). The RAN 100 may also be an open access network (O-RAN or ORAN) or a cloud radio access network (CRAN). The RAN 100 may also be a communication system that integrates two or more of the above systems.

[0078] RAN node 110, sometimes also referred to as access network equipment, RAN entity, or access node, constitutes part of the communication system and facilitates wireless access for terminals. Multiple RAN nodes 110 in communication system 1000 can be of the same type or different types. In some scenarios, the roles of RAN node 110 and terminal 120 are relative. For example, network element 120i in Figure 3 can be a helicopter or drone, which can be configured as a mobile base station. For terminal 120j accessing RAN 100 via network element 120i, network element 120i is a base station; however, for base station 110a, network element 120i is a terminal. RAN node 110 and terminal 120 are sometimes referred to as communication devices. For example, network elements 110a and 110b in Figure 3 can be understood as communication devices with base station functionality, and network elements 120a-120j can be understood as communication devices with terminal functionality.

[0079] In one possible scenario, a RAN node may be a base station, an evolved NodeB (eNodeB), an access point (AP), a transmission reception point (TRP), a next-generation NodeB (gNB), a next-generation base station in a 6G mobile communication system, a base station in a future mobile communication system, etc. A RAN node may be a macro base station (such as 110a in Figure 1 ), a micro base station or an indoor station (such as 110b in Figure 1 ), a relay node or a donor node, or a wireless controller in a CRAN scenario. Optionally, a RAN node may also be a server, a wearable device, a vehicle, or an onboard device, etc. For example, the access network device in vehicle to everything (V2X) technology may be a road side unit (RSU).

[0080] In another possible scenario, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, and different RAN nodes respectively implement part of the functions of the base station. For example, the RAN node can be a centralized unit (CU), a distributed unit (DU), a CU-control plane (CP), a CU-user plane (UP) or a radio unit (RU). The CU and DU can be set separately, or they can be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU) or a remote radio head (RRH).

[0081] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, CU may also be called O-CU (Open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. For the convenience of description, this application uses CU, CU-CP, CU-UP, DU and RU as examples for description. Any unit of CU (or CU-CP, CU-UP), DU and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0082] A terminal may also be referred to as a terminal device, user equipment (UE), mobile station, mobile terminal, etc. Terminals can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), Internet of Things (IOT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, etc. A terminal may be a mobile phone, tablet computer, computer with wireless transceiver function, wearable device, vehicle, drone, helicopter, airplane, ship, robot, robotic arm, smart home device, etc. The embodiments of the present application do not limit the device form of the terminal.

[0083] The communication system provided in this application (the communication system 1000 shown in FIG3 ) can introduce an AI network element to implement some or all AI-related operations. The AI ​​network element can also be referred to as an AI node, AI device, AI entity, AI module, AI model, or AI unit, etc. The AI ​​network element can be a network element built into the communication system. For example, the AI ​​network element can be an AI module built into: an access network device, a core network device, a cloud server, or a network management (OAM) to implement AI-related functions. The OAM can be a network management device as a core network device and / or as a network management device as an access network device. Alternatively, the AI ​​network element can also be an independently set network element in the communication system. Alternatively, the terminal or the chip built into the terminal can also include an AI model to implement AI-related functions. A processing module based on an AI model. For ease of description, this application takes the AI ​​model as an example for description.

[0084] In communication systems, AI-based processing modules, through data-driven model training, can optimize physical layer channel coding, modulation, waveform modules, and upper-layer beam management, CSI compression, positioning, and other functions based on the scenario, thereby achieving better transmission signal design and reception performance. Different AI models may correspond to different network coverage areas or tasks (for example, positioning tasks, beam prediction tasks).

[0085] Network devices, terminals, or other entities can train AI models for specific network coverage areas and specific tasks, and the network devices can store the AI ​​models. As shown in Figure 4, UE 1 can download a trained first AI model from base station 1 that is applicable to the current network coverage area or task. When the terminal moves, causing the network coverage area to switch, or the task to switch, the terminal needs to switch the AI ​​model, as shown in Figure 5.

[0086] As shown in Figure 5, UE 1 is within network coverage area 1 of base station 1 and downloads a first AI model applicable to network coverage area 1 from base station 1. When UE 1 switches from base station 1's network coverage area to base station 2's network coverage area 2, and the first AI model is no longer applicable to network coverage area 2, after UE 1 accesses base station 2, base station 2 indicates the identifier of a second AI model to UE 1, indicating to UE 1 that the model applicable to network coverage area 2 is the second AI model. UE 1 then downloads the second AI model applicable to network coverage area 2 from base station 2.

[0087] Due to the limited storage capacity of the terminal, it may not be possible to save all AI models in advance. Therefore, in each switching scenario, the terminal may need to download the AI ​​model from the network device. However, downloading the complete AI model from the network device will incur a large signaling overhead.

[0088] In view of this, an embodiment of the present application provides a model switching method, in which the AI ​​model is divided into multiple levels of sub-models, and each level of sub-model has its applicable network coverage area or task. In the scenario where the network coverage area or task is switched, the communication device only needs to switch the sub-model of the level corresponding to the switched network coverage area or task in the multi-level sub-model. This can avoid downloading the complete AI model during switching, thereby helping to reduce signaling overhead.

[0089] The model switching method provided in the embodiment of the present application is described in detail below with reference to Figures 6 to 12.

[0090] Figure 6 is a schematic flow chart of a model switching method 600 provided in an embodiment of the present application. This embodiment of the present application relates to model switching in a switching scenario, which may include switching network coverage areas or switching tasks. Switching network coverage areas or tasks can be initiated by either a terminal or a network device.

[0091] The method of the embodiment of the present application can be performed by a first communication device, which is a terminal or a network device. When the first communication device is a terminal, the second communication device is a network device. When the first communication device is a network device, the second communication device is a terminal.

[0092] The network device may be, for example, the RAN node 100 in FIG. 3 , and may have the possible forms described for the RAN node 100 , such as a base station in 4G or 5G, or a next-generation base station in a 6G system, or a base station in a future mobile communication system. Alternatively, the RAN node may be a CU (CU-CP, CU-UP), a DU, a RU, a near-real-time radio network intelligent controller (RAN intelligent controller, RIC), or a non-real-time RIC, which is not limited in this application.

[0093] The method 600 includes steps S601 to S604, and the specific steps are as follows:

[0094] S601: Determine an AI model, where the AI ​​model includes multiple levels of sub-models. Each level of the sub-models is applicable to at least one object, which is a network coverage area or a task.

[0095] In this step, when the first communication device needs to optimize a certain wireless function (for example, channel coding, modulation, or waveform design) in the current network coverage area, or needs to perform a certain task, it can determine the AI ​​model. The AI ​​model has multiple levels of sub-models, and each level of sub-model is applicable to at least one object, and the at least one object includes the current network coverage area of ​​the first communication device or the task that currently needs to be performed.

[0096] For ease of description, the current network coverage area of ​​the first communication device or the current task to be performed is referred to as the initial object, and the submodel in the multi-level submodel that applies to the initial object is referred to as the initial submodel. The current network coverage area of ​​the first communication device can also be described as the network coverage area before the first communication device switches, and the current task to be performed by the first communication device can also be described as the task to be performed before the first communication device switches.

[0097] In this step, the AI ​​model is an AI model applicable to the initial object, that is, at least one object applicable to each level of sub-models in the multi-level sub-models of the AI ​​model includes the initial object.

[0098] Optionally, before S601, the first communication device receives a first message from the second communication device, the first message indicating parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each level of sub-model, and the multi-level sub-model is determined from the AI ​​model based on the hierarchical structure of the AI ​​model.

[0099] Optionally, the first message is an RRC message.

[0100] S601 may include: the first communication device determining the AI ​​model based on the first message. Furthermore, the first communication device may determine, based on the first message, multiple levels of sub-models of the AI ​​model and at least one object applicable to each level of sub-model.

[0101] Based on the instructions of the first message, the AI ​​model can be divided into multiple levels, each level has its corresponding sub-model. In other words, the AI ​​model can be divided into multiple modules, each module has its corresponding sub-model. The objects applicable to the sub-models at multiple levels are nested with each other, as described below with reference to FIG. 7.

[0102] Figure 7 is a schematic diagram of the hierarchical structure of an AI model provided in an embodiment of the present application. Exemplarily, the AI ​​model is divided into three levels, including level 1, level 2, and level 3. The sub-models of level 1 include sub-model 1, the sub-models of level 2 include sub-model 2 and sub-model 3, and the sub-models of level 3 include sub-model 4, sub-model 5, sub-model 6, and sub-model 7.

[0103] Among them, at least one object applicable to the sub-model of level 1 includes at least one object applicable to the sub-model of level 2, and at least one object applicable to the sub-model of level 2 includes at least one object applicable to the sub-model of level 3. Specifically, at least one object applicable to sub-model 1 includes at least one object applicable to sub-model 2 and at least one object applicable to sub-model 3, at least one object applicable to sub-model 2 includes at least one object applicable to sub-model 4 and at least one object applicable to sub-model 5, and at least one object applicable to sub-model 3 includes at least one object applicable to sub-model 6 and at least one object applicable to sub-model 7.

[0104] For sub-models at the same level, different sub-models may be applicable to different objects. For example, sub-model 2 may be applicable to at least one object that is different from that of sub-model 3, and sub-model 4 may be applicable to at least one object that is different from that of sub-model 5.

[0105] It should be noted that the cascading of sub-models at multiple levels can form a complete AI model.

[0106] For example, sub-model 1, sub-model 2, and sub-model 4 are cascaded to form a complete AI model.

[0107] For another example, sub-model 1, sub-model 2, and sub-model 5 are cascaded to form a complete AI model.

[0108] For another example, sub-model 1, sub-model 2, and sub-model 5 are cascaded to form a complete AI model.

[0109] For another example, sub-model 1, sub-model 3, and sub-model 6 are cascaded to form a complete AI model.

[0110] For another example, sub-model 1, sub-model 3, and sub-model 7 are cascaded to form a complete AI model.

[0111] It should be noted that the AI ​​model determined in this step is, for example, an AI model composed of the cascade of the above-mentioned sub-model 1, sub-model 2 and sub-model 4. In other words, the multi-level sub-model in this step includes sub-model 1, sub-model 2 and sub-model 4.

[0112] S602: Obtain the identifier of the target object.

[0113] When the first communication device needs to switch from the initial object to the target object, the first communication device obtains the identifier of the target object and then determines the target sub-model of the hierarchy corresponding to the target object. For example, due to terminal movement, the terminal needs to access the target cell due to cell switching. In this case, the terminal can obtain the identifier of the target cell.

[0114] For ease of description, the layer in the multiple layers corresponding to the target object is referred to as the target layer. In the aforementioned multi-level sub-model, the sub-model of the target layer applicable to the initial object can be referred to as the initial sub-model of the target layer. The target layer can be one or more of the multiple layers.

[0115] The first communication device obtains the identifier of the target object in the following two situations:

[0116] Case 1: The first communication device determines the target object and obtains the identifier of the target object.

[0117] Case 2: The first communication device receives the identifier of the target object from the second communication device. In this case, the second communication device determines the target object and sends the identifier of the target object to the first communication device.

[0118] S603: Based on the identifier of the target object, determine a target sub-model of a level corresponding to the target object. The target sub-model is applicable to the target object.

[0119] After obtaining the identification of the target object, the first communication device knows at least one object applicable to each level of sub-model. Therefore, after knowing the identification of the target object, the first communication device can determine whether the at least one object applicable to the sub-model of the target level includes the target object from the correspondence between each level of sub-model and at least one applicable object.

[0120] In one possible implementation, S603 may include: based on the identification of the target object, determining that at least one object to which the sub-model of the target level applies does not include the target object; obtaining parameters for constructing the target sub-model; and constructing the target sub-model based on the parameters for constructing the target sub-model.

[0121] In combination with the above example of Figure 7, the multi-level sub-models of the AI ​​model applicable to the initial object include sub-model 1, sub-model 2 and sub-model 4. If switching from the initial object to the target object, the level corresponding to the target object is level 3, and the first communication device determines that at least one object to which sub-model 4 applies does not include the target object, then the first communication device needs to obtain parameters for constructing the target sub-model to construct the target sub-model. The target sub-model is, for example, sub-model 5 in level 3.

[0122] It should be noted that the target level can be one or more. When there are multiple target levels, each target level will correspond to a target sub-model. Taking Figure 7 as an example, when the target level includes level 1, the target sub-model of level 1 (such as the cell group layer or common feature extraction layer below) can be referred to as target sub-model 1, when the target level includes level 2, the target sub-model of level 2 (such as the cell layer or task feature extraction layer below) can be referred to as target sub-model 2, and when the target level includes level 3, the target sub-model of level 3 (such as the beam layer or task output layer below) can be referred to as target sub-model 3.

[0123] The specific process of determining the target level can be found in the description below and will not be described in detail here.

[0124] Optionally, the first communication device acquires parameters for constructing the target sub-model, including: the first communication device receives the parameters for constructing the target sub-model from the second communication device.

[0125] Optionally, when the first communication device is a terminal and the second communication device is a network device, parameters for constructing the target sub-model can be carried through RRC messages, MAC CE or DCI.

[0126] Optionally, when the first communication device is a network device and the second communication device is a terminal, parameters for constructing the target sub-model may be carried through the UCI.

[0127] Further description of the first communication device receiving the parameters for constructing the target sub-model from the second communication device is as follows:

[0128] In conjunction with scenario 1 in S602, before the first communication device receives the parameters for constructing the target sub-model from the second communication device, the first communication device may send the identifier of the target object to the second communication device. After receiving the identifier of the target object, the second communication device determines the target sub-model corresponding to the target object based on the stored correspondence information between the sub-models and the objects. Furthermore, the second communication device may send the parameters for constructing the target sub-model to the first communication device. Accordingly, the first communication device receives the parameters for constructing the target sub-model.

[0129] In conjunction with scenario 2 in S602, after determining the target object, the second communication device determines the target sub-model corresponding to the target object from the stored information about the correspondence between sub-models and objects. Furthermore, the second communication device sends parameters for constructing the target sub-model to the first communication device. Correspondingly, the first communication device receives the parameters for constructing the target sub-model from the second communication device.

[0130] Optionally, the second communication device sending parameters for constructing the target sub-model to the first communication device includes: the second communication device sending the parameters for constructing the target sub-model to the first communication device based on the first request message. In this implementation, after receiving the identifier of the target object from the second communication device, the first communication device determines, based on the identifier of the target object, that at least one object applicable to the sub-model of the target level does not include the target object, and then the first communication device sends the first request message to the second communication device, where the first request message is used to request the sub-model applicable to the target object.

[0131] In another possible implementation, S603 may include: determining, based on the identifier of the target object, that at least one object to which the sub-model of the target layer applies includes the target object; and determining the initial sub-model of the target layer as the target sub-model of the target layer. In this implementation, the target sub-model of the target layer determined by the first communication device is the initial sub-model of the target layer.

[0132] In combination with the above example of Figure 7, the multi-level sub-models of the AI ​​model applicable to the initial object include sub-model 1, sub-model 2 and sub-model 4. If switching from the initial object to the target object, the level corresponding to the target object is level 3, and the first communication device determines that at least one object applicable to sub-model 4 of level 3 includes the target object, then the first communication device determines sub-model 4 as the target sub-model.

[0133] S604: Switch the sub-model of the level corresponding to the target object in the multi-level sub-model to the target sub-model.

[0134] In this step, the submodel of the target level is the initial submodel of the target level, that is, the submodel applicable to the initial object in the multi-level submodel. The first communication device switches the initial submodel applicable to the initial object in the target level to the target submodel applicable to the target object.

[0135] As can be seen from the description above, the objects applicable to the sub-models of multiple levels are nested with each other. Therefore, when the first communication device switches to the target object, there may be a situation where the initial sub-models of some levels among the multiple levels are not applicable to the target object, but the initial sub-models of other levels are still applicable to the target object.

[0136] For example, referring to the example of Figure 7 above, the multi-level sub-models of the AI ​​model applicable to the initial object include sub-model 1, sub-model 2 and sub-model 4, the objects applicable to sub-model 1 of level 1 include object 1, object 2, object 3 and object 4, the objects applicable to sub-model 2 of level 2 include object 1 and object 2, the objects applicable to sub-model 4 of level 3 include object 1, the target object is object 2, and the target level is level 3.

[0137] In this example, the objects applicable to sub-model 4 at level 3 do not include object 2, but the objects applicable to sub-model 4 at level 3 include object 2, and the objects applicable to sub-model 4 at level 2 also include object 2. Therefore, in the scenario of switching objects, the first communication device only needs to switch the sub-model at level 3 to the sub-model applicable to object 2, without having to obtain the target sub-model at level 1 and the target sub-model at level 2.

[0138] In an embodiment of the present application, the first communication device switches the sub-model of the target level in the multi-level sub-model to the target sub-model, and the initial sub-models of other levels in the multiple levels are still applicable to the target object, so there is no need to obtain the target sub-models of other levels. This helps avoid downloading the complete AI model, thereby helping to reduce signaling overhead.

[0139] The above-mentioned object may be a network coverage area or a task. The following first describes the model switching process when the object is a network coverage area with reference to FIG8 to FIG12.

[0140] Optionally, when the above-mentioned object is a network coverage area, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on network coverage areas of multiple different granularities.

[0141] In one possible implementation, the network coverage areas of various granularities include the network coverage area of ​​the cell group, the network coverage area of ​​the cell, and the network coverage area of ​​the beam. In this approach, the AI ​​model is divided into three levels: the cell group level, the cell level, and the beam level.

[0142] In another possible implementation, the network coverage areas of different granularities include cell coverage areas and beam coverage areas. In this approach, the AI ​​model is divided into two levels: the cell level and the beam level.

[0143] In another possible implementation, the network coverage areas of various granularities include the network coverage area of ​​the cell group, the network coverage area of ​​the cell, the network coverage area of ​​the sector, and the network coverage area of ​​the beam. In this approach, the AI ​​model is divided into four levels: the cell group level, the cell level, the sector level, and the beam level.

[0144] The following uses network coverage areas of various granularities, including the network coverage area of ​​a cell group, the network coverage area of ​​a cell, and the network coverage area of ​​a beam, as examples for introduction.

[0145] The above-mentioned multi-level sub-model includes a first-level sub-model (e.g., sub-model 1 in FIG7 ), a second-level sub-model (e.g., sub-model 2 in FIG7 ), and a third-level sub-model (e.g., sub-model 4 in FIG7 ). The first-level sub-model is a sub-model at the cell group layer, and the network coverage area applicable to the first-level sub-model is the network coverage area of ​​the cell group; the second-level sub-model is a sub-model at the cell layer, and the network coverage area applicable to the second-level sub-model is the network coverage area of ​​at least one cell in the cell group; the third-level sub-model is a sub-model at the beam layer, and the network coverage area applicable to the third-level sub-model is the network coverage area of ​​at least one beam within the at least one cell.

[0146] Figure 8 is a schematic flow chart of another model switching method 800 provided in an embodiment of the present application. In conjunction with the above-described method 500, method 800 is described using a first communication device terminal, a second communication device being a base station, and an object being a network coverage area as an example. In method 800, the aforementioned various network coverage areas of varying granularity include the network coverage area of ​​a cell group, the network coverage area of ​​a cell, and the network coverage area of ​​a beam.

[0147] The method 800 includes steps S801 to S806, and the specific steps are as follows:

[0148] S801: A terminal sends a model request message to a base station, where the model request message is used to request an AI model. In response, the base station receives the model request.

[0149] For example, after a terminal accesses a base station, it is within the network coverage area of ​​a beam (denoted as beam C1) of a cell (denoted as cell B1) of the base station. Cell B1 is a cell within a cell group (denoted as cell group A1). Based on this, the AI ​​model sent by the base station to the terminal is the AI ​​model applicable to the network coverage area of ​​beam C1 of cell B1 in cell group A1.

[0150] S802: The base station sends a model response message to the terminal, where the model response message indicates parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each sub-model. Accordingly, the terminal receives the model response message.

[0151] The model response message in this step may correspond to the first message mentioned above. The introduction of the hierarchical structure of the AI ​​model can refer to the description of S601 above, which will not be repeated here.

[0152] With reference to the description of S801, the AI ​​model includes a first-level sub-model applicable to the network coverage area of ​​cell group A1, a second-level sub-model applicable to the network coverage area of ​​cell B1, and a third-level sub-model applicable to the network coverage area of ​​beam C1. Since the network coverage area of ​​cell group A1 includes the network coverage area of ​​cell B1, and the network coverage area of ​​cell B1 includes the network coverage area of ​​beam C1, the network coverage area applicable to the first-level sub-model includes the network coverage area applicable to the second-level sub-model, and the network coverage area applicable to the second-level sub-model includes the network coverage area applicable to the third-level sub-model.

[0153] Optionally, since the first sub-model and the second sub-model are applicable to more network coverage areas, the first communication device can store the first sub-model and / or the second sub-model. In this way, when switching between different network coverage areas applicable to the first sub-model or the second sub-model, there is no need to receive parameters for constructing the first sub-model and / or the second sub-model, which is conducive to reducing signaling overhead.

[0154] Optionally, the hierarchical structure of the AI ​​model may indicate the identifiers of the starting neural network layer and the ending neural network layer of the first-level sub-model, the second-level sub-model, and the third-level sub-model in the AI ​​model, respectively.

[0155] Taking the AI ​​model including a 30-layer neural network as an example, Table 1 shows the correspondence between a multi-level sub-model and the network coverage area. The first-level sub-model includes the 1st to 10th layers of the 30-layer neural network and is applicable to the network coverage area of ​​cell group A1. The second-level sub-model includes the 11th to 20th layers of the 30-layer neural network and is applicable to the network coverage area of ​​cell group B1. The third-level sub-model includes the 21st to 30th layers of the 30-layer neural network and is applicable to the network coverage area of ​​beam C1.

[0156] Table 1

[0157] Different network coverage areas are distinguished by different identifiers. Optionally, the identifier of the network coverage area includes an identifier of a cell group, an identifier of a cell, and an identifier of a beam.

[0158] In an example 1, the identifier of the network coverage area of ​​the beam C1 of the cell B1 in the cell group A1 may be: A1-B1-C1.

[0159] In another example 2, the identifier of the network coverage area of ​​beam C2 of cell B1 in cell group A1 may be: A1-B1-C2.

[0160] In another example 3, the identifier of the network coverage area of ​​beam C3 of cell B2 in cell group A1 can be: A1-B2-C3.

[0161] In another example 4, the identifier of the network coverage area of ​​beam C4 of cell B3 in cell group A2 may be: A2-B3-C4.

[0162] Optionally, the level corresponding to the target object may be determined based on the identification of the target object.

[0163] For example, the identifier of the initial object is A1-B1-C1, and the identifier of the target object is A1-B1-C2. It can be seen that the cell group identifier and the cell identifier remain unchanged, while the beam identifier changes from C1 to C2. The level corresponding to the target object includes the beam layer.

[0164] For another example, the initial object is identified as A1-B1-C1, and the target object is identified as A1-B2-C3. It can be seen that the cell group identifier remains unchanged, while the cell identifier changes from B1 to B2, and the beam identifier changes from C1 to C3. The layers corresponding to the target object include the cell layer and the beam layer.

[0165] For another example, the initial object is identified as A1-B1-C1, and the target object is identified as A2-B3-C4. It can be seen that the cell group identifier changes from A1 to A2, the cell identifier changes from B2 to B3, and the beam identifier changes from C1 to C4. The levels corresponding to the target object include the cell group layer, the cell layer, and the beam layer.

[0166] The above representation of the identification of the network coverage area is only an example. In addition, other forms can be used to represent the identification of the network coverage area, and this application does not limit this. For example, if the identification of a cell group is 000, the identification of a cell within the cell group is 01, and the identification of a beam within the cell is 0, then the identification of the network coverage area of ​​the beam of the cell within the cell group can be expressed as: 000-01-0.

[0167] The above-mentioned method of marking the network coverage area can be called a method based on absolute identification. In addition, the network coverage area can also be marked in a method based on relative identification.

[0168] For example, the beam of cell B1 in cell group A1 and the beam of cell B2 in cell group A1 use the same identifier, for example, both are beam C. That is, the beam in cell B1 in cell group A1 and the beam in cell B2 in cell group A1 may be the same, but the network coverage areas of the two beams C are different. Therefore, when switching from cell B1 to cell B2, the identifier of the initial object is A1-B1-C, and the identifier of the target object is A1-B2-C, where the beam identifier does not change. However, cell switching usually causes beam switching, so the layers corresponding to the target object include the cell layer and the beam layer.

[0169] S803-1, in the first scenario, the terminal sends a handover message to the base station, where the handover message is used to instruct the terminal to be handed over to the target network coverage area. Correspondingly, the base station receives the handover message.

[0170] In a possible case, the terminal needs to switch to the target network coverage area due to movement, so the terminal can send a switching message to the base station, where the switching message carries an identifier of the target network coverage area.

[0171] In one example, the initial network coverage area of ​​the terminal is the network coverage area of ​​beam C1 of cell B1 in cell group A1, and the target network coverage area is the network coverage area of ​​beam C2 of cell B1 in cell group A1. The identifier of the target network coverage area carried in the handover message can be: A1-B1-C2.

[0172] S803-2, in the second scenario, the base station sends a handover message to the terminal, where the handover message is used to instruct the terminal to handover to the target network coverage area. Correspondingly, the terminal receives the handover message.

[0173] In one possible case, the base station determines that the signal quality of the target network coverage area meets the condition, and therefore the base station may instruct the terminal to switch to the target network coverage area through a switching message.

[0174] S804: The base station determines whether it is necessary to send parameters for constructing a target sub-model to the terminal based on the identifier of the target network coverage area.

[0175] When the terminal needs to switch to the target network coverage area, the terminal needs to switch the current AI model to an AI model suitable for the target network coverage area. Based on the above introduction to the hierarchical structure of the AI ​​model, it can be seen that when switching the model, it is only necessary to switch the sub-model of the target hierarchy to the target sub-model suitable for the target network coverage area. Therefore, after determining the identifier of the target network coverage area, the base station can determine whether it is necessary to send parameters for constructing the target sub-model to the terminal based on the stored correspondence information between the sub-model and the network coverage area.

[0176] If the initial sub-model of the target level in the above multi-level sub-model is also applicable to the target network coverage area, the base station may not send the parameters for constructing the target sub-model to the terminal device.

[0177] If the initial sub-model of the target level in the multi-level sub-model is not applicable to the target network coverage area, the base station executes S805 .

[0178] S805: The base station sends parameters for constructing the target sub-model to the terminal. Correspondingly, the terminal receives the parameters for constructing the target sub-model.

[0179] Optionally, the parameters for constructing the target sub-model may be carried by the following messages: RRC message, MAC CE or DCI.

[0180] Optionally, the base station may indicate the parameters for constructing the AI ​​model through an RRC message. The parameters may include the identifier of the model function, the model structure, the parameters of each layer of the neural network, etc. A possible configuration signaling structure of the AI ​​model in the RRC message is as follows:

[0181] Among them, "nnAPIIdex" indicates the identifier of the model function of the AI ​​model, "Layer index" indicates the identifier of a certain layer of neural network, "Layer type" indicates the type of a certain layer of neural network, "Layer parameter" indicates the parameters of a certain layer of neural network, "neuralnetwork weights" indicates the weights of neurons, and "compression parameters" indicate the compression parameters of the neural network.

[0182] Optionally, the base station may indicate the parameters for constructing the target sub-model of the target layer through MAC CE. FIG10 is a schematic diagram of the structure of a MAC CE provided by an embodiment of the present application. In FIG10, "nnID" indicates the identifier of the AI ​​model, "nnAPIIdex" indicates the identifier of the model function of the AI ​​model, and the layer identification (LID) indicates a certain layer of the neural network in the sub-model of the target layer. i Indicates the i-th parameter of the neural network layer, i ranges from 0 to 7, W1...W m Indicates the weights of the neurons in this layer.

[0183] Optionally, the base station may indicate parameters for constructing a target sub-model of the target layer through DCI. The base station may explicitly indicate the identifier of the target layer, or the base station may implicitly indicate the identifier of the target layer through cyclic redundancy check (CRC) scrambling, or the base station may bundle the DCI and the physical downlink shared channel (PDSCH) for transmission, indicate the identifier of the target layer through DCI, and carry the weight value for updating through PDSCH.

[0184] Optionally, when a sub-model of a certain level needs to be updated, the base station may indicate a portion of the weight of the sub-model of the level by lengthening the DCI, and the portion of the weight is used to update the sub-model of the level.

[0185] S806: The terminal switches the model.

[0186] Based on the description of S804, if the initial sub-model of the target level in the multi-level sub-model is not applicable to the target network coverage area, the base station determines to send parameters for constructing the target sub-model to the terminal device. Accordingly, the terminal receives the parameters for constructing the target sub-model to construct the target sub-model and switches the initial sub-model of the target level in the multi-level sub-model to the target sub-model.

[0187] If the initial submodel of the target layer in the above multi-level submodel is also applicable to the target network coverage area, the submodel of the target layer in the multi-level submodel before switching is the same as the submodel after switching.

[0188] In an embodiment of the present application, when switching network coverage areas, the terminal can switch the sub-model of the target layer according to the hierarchical structure of the AI ​​model, which is beneficial to avoid excessive signaling overhead caused by downloading the complete AI model during switching.

[0189] In addition, when the terminal switches the network coverage area, it indicates the model switch according to the switching message used to request the switching of the network coverage area. Then, the base station can determine the target sub-model suitable for the target network coverage area based on the correspondence information between the network coverage area and the sub-model. In this way, there is no need for the base station to send the model identifier, which is conducive to reducing signaling overhead.

[0190] In conjunction with method 800, FIG9 is a schematic diagram of another model switching provided by an embodiment of the present application. As shown in FIG9, after the terminal initially accesses a cell of the base station, there may be the following switching scenarios:

[0191] Scenario 1: Beam switching within the same cell. In this scenario, the target layer includes the beam layer, and the terminal can switch the sub-model of the beam layer to the target sub-model.

[0192] Referring to Example 1 and Example 2 in S802 above, the initial network coverage area is the network coverage area of ​​beam C1 of cell B1 of cell group A1. When switching from beam C1 to beam C2, the target network coverage area is the network coverage area of ​​beam C2 of cell B1 of cell group A1. Since beam C1 and beam C2 belong to the same cell B1, only the switching of the sub-model of the beam layer is involved. The terminal then determines the target sub-model of the beam layer and switches the sub-model of the beam layer to the target sub-model. Among them, at least one object to which the target sub-model applies includes the network coverage area of ​​beam C2 of cell B1 of cell group A1.

[0193] Scenario 2: Beam switching between different cells. In this scenario, the target layer includes the cell layer and the beam layer. The terminal can switch the submodels of the cell layer and the beam layer to the target submodel. It should be understood that there are multiple target submodels in this case, including target submodel 2 at the cell layer and target submodel 3 at the beam layer.

[0194] Referring to Example 1 and Example 3 in S802 above, the initial network coverage area is the network coverage area of ​​beam C1 of cell B1 of cell group A1. When switching from beam C1 to beam C3, since beam C3 is the beam of cell B2 and belongs to a different cell than beam C1, switching from beam C1 to beam C3 will also trigger switching from cell B1 to cell B2, that is, the target network coverage area is the network coverage area of ​​beam C3 of cell B2 of cell group A1. The terminal determines the target submodel 2 of the cell layer and the target submodel 3 of the beam layer, and switches the submodel of the cell layer to the target submodel 2 and the submodel of the beam layer to the target submodel 3.

[0195] It should be understood that the target sub-model 2 of the cell layer and the target sub-model 3 of the beam layer are different. At least one object to which the target sub-model 2 of the cell layer applies includes the network coverage area of ​​cell B2 of cell group A1, and at least one object to which the target sub-model 3 of the beam layer applies includes the network coverage area of ​​beam C3 of cell B2 of cell group A1.

[0196] Scenario 3 involves cell handover within the same cell group, which typically results in beam handover. In this scenario, the target layer includes the cell layer and the beam layer. The terminal can switch the submodels of the cell layer and the beam layer to the target submodel. Similar to the description for Scenario 2, there are multiple target submodels, including target submodel 2 at the cell layer and target submodel 3 at the beam layer.

[0197] Similar to the example in scenario 2 above, when switching from cell B1 to cell B2, the beam is correspondingly switched from beam C1 of cell B1 to beam C3 of cell B2, and the target network coverage area is the network coverage area of ​​beam C3 of cell B2 of cell group A1. The terminal needs to determine target sub-model 2 at the cell layer and target sub-model 3 at the beam layer, where target sub-model 2 applies to the network coverage area of ​​cell B2 of cell group A1, and target sub-model 3 applies to the network coverage area of ​​beam C3 of cell B2 of cell group A1.

[0198] Scenario 4: Cell handover between different cell groups. This means that cell group handover, cell handover, and beam handover are triggered simultaneously. In this scenario, the target layer includes the cell group layer, cell layer, and beam layer. The terminal can switch the sub-models of the cell group layer, cell layer, and beam layer to the target sub-model.

[0199] Referring to Example 1 and Example 4 in S802 above, when switching from cell B1 to cell B3, since cell B1 belongs to cell group A1 and cell B3 belongs to cell group A2, and the beam of cell B1 is beam C1 and the beam of cell B3 is beam C4, switching from cell group A1 to cell group A2 and switching from beam C1 to beam C4 will also occur accordingly, that is, the target network coverage area is the network coverage area of ​​beam C4 of cell B3 of cell group A2. The terminal determines the target submodel 1 of the cell group layer, the target submodel 2 of the cell layer, and the target submodel 3 of the beam layer, and switches the submodel of the cell group layer to the target submodel 1, switches the submodel of the cell layer to the target submodel 2, and switches the target submodel of the beam layer to the target submodel 3.

[0200] The cell handover between different cell groups in Scenario 4 can also be understood as a cell group handover. When a cell group handover occurs, it may be necessary to switch the sub-models at the cell group layer, the cell layer, and the beam layer. This means switching the AI ​​model applicable to the initial target network coverage area to the AI ​​model applicable to the target network coverage area.

[0201] In order to train / evaluate the sub-models of the levels corresponding to different network coverage areas, it is necessary to bind the data collected in a certain network coverage area with the sub-model of the level corresponding to the network coverage area. The purpose of binding is to use the data collected in the network coverage area to train / evaluate the sub-model of the level corresponding to the network coverage area.

[0202] Corresponding to the cell group layer, cell layer, and beam layer of the aforementioned AI model, data collected within different cell groups is called cell group data, data collected within a cell is called intra-cell data, and data collected within a beam is called intra-beam data. As shown in Figure 11, intra-cell group data includes intra-cell data, and intra-cell data includes intra-beam data.

[0203] In terms of scope of application, intra-beam data can be used to train complete AI models. In other words, intra-beam data is suitable for training cell-group-level sub-models, cell-level sub-models, and beam-level sub-models. Inter-beam data can be used to jointly train cell-level sub-models and cell-group-level sub-models, or to evaluate (including monitoring or verifying) the performance of multiple beam-level sub-models, evaluate the performance of cell-level sub-models, or evaluate the performance of cell-level sub-models. Intra-cell data can be used to train cell-level sub-models and cell-group-level sub-models, and inter-cell data can be used to train cell-group models. In addition, the performance of the cell-group-level sub-model can be evaluated in combination with the beam-level sub-model or the cell-level sub-model.

[0204] In a possible implementation, data collected in a certain network coverage area may be marked with an identifier of the network coverage area, thereby binding the data collected in the network coverage area to a sub-model of a level corresponding to the network coverage area.

[0205] Figure 12 is a schematic diagram of a data identification provided by an embodiment of the present application. It can be seen that the data identification also has a hierarchical structure, including a cell group identification, a cell identification, and a beam identification cascaded to form the data identification.

[0206] For example, the identifier of the data collected in the network coverage area of ​​beam C1 of cell B1 in cell group A1 can be: A1-B1-C1, which can also be understood as the identifier of the network coverage area of ​​beam C1 of cell B1 in cell group A1 is A1-B1-C1, and the data collected in the network coverage area of ​​beam C1 of cell B1 in cell group A1 is marked with A1-B1-C1. In this way, before training, the first communication device can determine the data corresponding to the network coverage area of ​​beam C1 of cell B1 in cell group A1 based on the identifier of the data, that is, determine the data collected in the network coverage area of ​​beam C1 of cell B1 in cell group A1, and then use the data collected in the network coverage area of ​​beam C1 of cell B1 in cell group A1 to train / evaluate the hierarchical sub-model corresponding to the network coverage area of ​​beam C1 of cell B1 in cell group A1.

[0207] When training a model based on data identification, the data collected within the beam can be used to train a complete AI model or to evaluate the performance of a sub-model at the beam layer.

[0208] In the above-described method 800, the target is a network coverage area. When the first communication device switches to the target network coverage area, a model switch is triggered. During the model switch, the first communication device only needs to switch the submodel at the target level of the multi-level submodel to the target submodel, which is a submodel suitable for the target network coverage area. The following describes the hierarchical structure of the AI ​​model when the target is a task.

[0209] Optionally, the tasks can be positioning, environmental object detection, beam prediction, trajectory prediction, etc. Multiple tasks can be grouped according to task type, and tasks of the same type can be classified into the same task group. For example, environmental object detection and trajectory prediction are both perception types and can be classified into the same task group.

[0210] Different tasks have different task identifiers, and different task groups have different task group identifiers. When a task belongs to a task group, the identifier of the task group to which it belongs can be added when identifying the task. In other words, a task identifier can include the identifier of the task group to which it belongs and the identifier of the task itself.

[0211] In an example 5, task group A1 includes task 1, task 2, task 3 and task 4, and task 1 is identified as: A1-1; task 2 is identified as: A1-2; task 3 is identified as: A1-3; task 4 is identified as: A1-4.

[0212] In another example 6, task group A2 includes task 5, task 6, and task 7, and the identifier of task 5 is: A2-5; the identifier of task 6 is: A2-6, and the identifier of task 7 is: A2-7.

[0213] Task group A1 or task group A2 can be further divided into multiple sub-task groups.

[0214] In Example 7, task group A1 is divided into sub-task group B1 and sub-task group B2. Sub-task group B1 includes tasks 1 and 2, and sub-task group B2 includes tasks 3 and 4. Task 1 can be identified by A1-B1-1; task 2 can be identified by A1-B1-2; task 3 can be identified by A1-B2-3; and task 4 can be identified by A1-B2-4.

[0215] In another example 8, task group A2 is divided into sub-task group B3 and sub-task group B4, where sub-task group B3 includes tasks 5 and 6, and sub-task group B4 includes task 7. Task 5 can be identified by: A2-B3-5; task 6 can be identified by: A2-B3-6; and task 7 can be identified by: A2-B4-7.

[0216] Optionally, when the above-mentioned object is a task, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on multiple data processing functions, and the multiple data processing functions are functions provided by the AI ​​model based on input data.

[0217] In one possible implementation, the multiple data processing functions include: a feature extraction function related to the task group corresponding to the input data (hereinafter referred to as the common feature extraction function), a feature extraction function related to at least one task in the task group (hereinafter referred to as the task feature extraction function), and an output function related to one task in the at least one task (hereinafter referred to as the output function).

[0218] In this implementation, as shown in Figure 7, the three layers of the AI ​​model are the common feature extraction layer, the task feature extraction layer, and the task output layer. The submodels in the common feature extraction layer are referred to as first-level submodels, which have the common feature extraction function described above; the submodels in the task feature extraction layer are referred to as second-level submodels, which have the task feature extraction function described above; and the submodels in the task output layer are referred to as third-level submodels, which have the output function described above.

[0219] It should be noted that the tasks applicable to the first-level sub-model, the second-level sub-model, and the third-level sub-model are nested in layers. The first-level sub-model applies to all tasks in a task group, the second-level sub-model applies to at least one task in the task group, and the third-level sub-model applies to one task in the at least one task.

[0220] Optionally, since the first-level sub-model and the second-level sub-model are applicable to more tasks, the first communication device can store the first-level sub-model and / or the second-level sub-model, so that there is no need to receive parameters for constructing the first-level sub-model and / or the second-level sub-model, thereby reducing signaling overhead.

[0221] The nesting of tasks applicable to the first-level sub-model, second-level sub-model, and third-level sub-model is further explained as follows: The first-level sub-model can extract a first feature from the input data corresponding to a task group, and this first feature is applicable to all tasks in the task group. The first-level sub-model then outputs this first feature to the second-level sub-model, which extracts a second feature from the first feature, and this second feature is applicable to at least one task in the task group. Furthermore, the second-level sub-model outputs the second feature to the third-level sub-model, which obtains a prediction result based on the second feature, and this prediction result is applicable to one task in the at least one task.

[0222] For example, a task group includes positioning tasks and motion trajectory prediction tasks. The input data corresponding to this task group is CSI. That is to say, for all tasks in this task group, the input data of the AI ​​model is CSI.

[0223] It is assumed that the AI ​​model obtained by cascading the first-level sub-model, the second-level sub-model, and the third-level sub-model is an AI model suitable for positioning tasks. When implementing the positioning task, first, the CSI is input into the first-level sub-model of the AI ​​model, and the CSI is feature extracted by the first-level sub-model to obtain the channel information features. Among them, the feature extraction of CSI is, for example, dimensionality reduction of CSI or extraction of multipath information. Afterwards, the first-level sub-model outputs the channel information features to the second-level sub-model, and the second-level sub-model extracts positioning-related features, such as position features, from the channel information features. Furthermore, the second-level sub-model outputs the features related to positioning to the third-level sub-model, and the third-level sub-model predicts the positioning results from the features related to positioning, for example, outputs the terminal coordinates.

[0224] If the target task is a motion trajectory prediction task, and at least one task applicable to the third-level sub-model does not include the motion trajectory prediction task, the first communication device can switch the sub-model of the task output layer to the target sub-model, so that the first-level sub-model, the second-level sub-model and the target sub-model are cascaded to form an AI model. Then, when implementing the motion trajectory prediction task, first, the CSI is input into the first-level sub-model of the AI ​​model, and the CSI is feature extracted by the first-level sub-model to obtain the channel information feature. Afterwards, the first-level sub-model outputs the channel information feature to the second-level sub-model, and the second-level sub-model extracts features related to motion trajectory prediction from the channel information feature, such as position features. Furthermore, the second-level sub-model outputs features related to motion trajectory prediction to the third-level sub-model, and the third-level sub-model predicts the object's motion trajectory from the features related to motion trajectory prediction, for example, the moving direction of the output terminal.

[0225] In the above example, the data input to the AI ​​model for both the positioning and trajectory prediction tasks is CSI. This means that the input data for positioning and trajectory prediction tasks within the same task group is the same. Furthermore, the first-level sub-model extracts channel information features from the input CSI for both positioning and trajectory prediction tasks. Therefore, the first-level sub-model has a common feature extraction function related to the task group.

[0226] Referring to Example 7 above for task identification, taking the example where the first-level sub-model applies to task group A1, the second-level sub-model applies to sub-task group B1 within task group A1, and the third-level sub-model applies to task 1 within sub-task group B1, when task switching occurs, the following switching scenarios may exist:

[0227] Scenario 5: Task switching within subtask group B1. For example, the initial task is Task 1, and the target task is Task 2. Task 1 and Task 2 both belong to subtask group B1. In this scenario, the target layer includes the task output layer. The terminal can determine the target submodel 3 of the task output layer and switch the submodel of the task output layer to the target submodel 3.

[0228] Scenario 6: Task switching between subtask group B1 and subtask group B2. For example, the initial task is Task 1, which belongs to subtask group B1, and the target task is Task 3, which belongs to subtask group B2. In this scenario, the target layer includes the task feature extraction layer and the task output layer. The terminal can determine the target submodel 2 of the task feature extraction layer and the target submodel 3 of the task output layer, and switch the submodel of the task feature extraction layer to the target submodel 2 and the submodel of the task output layer to the target submodel 3.

[0229] Scenario 7, task switching between task group A1 and task group A2. Referring to Examples 7 and 8 above for task identification, the initial task is task 1, which belongs to subtask group B1 in task group A1, and the target task is 5, which belongs to subtask group B3 in task group A2. In this scenario, the target layer includes a common feature extraction layer, a task feature extraction layer, and a task output layer. The terminal can obtain the target sub-model 1 of the common feature extraction layer, the target sub-model 2 of the task feature extraction layer, and the target sub-model 3 of the task output layer, and switch the sub-model of the common feature extraction layer to target sub-model 1, switch the sub-model of the task feature extraction layer to target sub-model 2, and switch the sub-model of the task output layer to target sub-model 3.

[0230] The process of triggering model switching based on task switching is similar to the description of the above method 800 and will not be repeated here.

[0231] In the above method 800, the terminal initiates the switching of the network coverage area and performs the model switching. In addition, the following situations are also included:

[0232] Case 1: The base station is a fixed base station, meaning it is fixed and cannot be moved. In this case, when the terminal initiates a network coverage area switch, the base station can receive the target object identifier from the terminal and determine the target sub-model based on the target object identifier to perform the model switch.

[0233] Case 2: The base station is a mobile base station, such as a drone base station or a vehicle-mounted base station. In this case, the base station can initiate a handover of the network coverage area, and the terminal can determine the target sub-model based on the network coverage area identifier sent by the base station and perform the model handover.

[0234] Case 3: The base station is a mobile base station. The base station can initiate a handover of the network coverage area and determine the target sub-model based on the identifier of the target network coverage area and then perform model switching.

[0235] When switching tasks, the base station and the terminal are peers, and both the terminal and the base station switch the sub-model of the target layer.

[0236] The specific method of determining the target sub-model and switching the model can be found in the description above and will not be repeated here.

[0237] It should be understood that the size of the serial numbers of the above processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0238] The above describes in detail the model switching method according to the embodiment of the present application in combination with Figures 6 to 12. The following describes in detail the communication device according to the embodiment of the present application in combination with Figures 13 and 14.

[0239] FIG13 is a schematic block diagram of a communication device 1300 according to an embodiment of the present application. The device 1300 includes a processing module 1310 and optionally a transceiver module 1320.

[0240] The processing module 1310 is used to perform data processing. The transceiver module 1320 can implement corresponding communication functions. The transceiver module 1320 can also be called a communication interface or a communication module.

[0241] Optionally, the device 1300 may further include a storage module, which may be used to store data and / or to store computer programs or instructions. The processing module 701 may read the computer programs / instructions and / or data in the storage module so that the device 1300 implements the above-mentioned method embodiment.

[0242] Apparatus 1300 can be configured to execute the actions performed by the first communication device in the aforementioned method embodiment. Apparatus 1300 can be the first communication device or a component configured in the first communication device. Processing module 1310 is configured to execute processing-related operations on the first communication device in the aforementioned method embodiment. Transceiver module 1320 is configured to execute reception-related operations on the first communication device in the aforementioned method embodiment.

[0243] Optionally, the transceiver module 1320 may include a sending module and a receiving module. The sending module is used to perform the sending operation in the above method embodiment. The receiving module is used to perform the receiving operation in the above method embodiment.

[0244] It should be noted that the apparatus 1300 may include a sending module but not a receiving module. Alternatively, the apparatus 1300 may include a receiving module but not a sending module. This may depend on whether the above solution executed by the apparatus 1300 includes both a sending action and a receiving action.

[0245] Optionally, the device 1300 is configured to execute the actions executed by the first communication device in the embodiment shown in Figure 6 or Figure 8. For details, please refer to the relevant introduction of the embodiment shown in Figure 6 or Figure 8, which will not be repeated here.

[0246] In one possible embodiment, the apparatus 1300 is configured to perform the following scheme:

[0247] The processing module 1310 is used to: determine an AI model, which includes a multi-level sub-model, and each level sub-model in the multi-level sub-model is applicable to at least one object, which is a network coverage area or a task; obtain the identifier of the target object; based on the identifier of the target object, determine the target sub-model of the level corresponding to the target object, and the target sub-model is applicable to the target object; and, switch the sub-model of the level corresponding to the target object in the multi-level sub-model to the target sub-model.

[0248] Optionally, the transceiver module 1320 is used to: receive a first message, which indicates the parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each level of sub-model, and the multi-level sub-model is determined from the AI ​​model based on the hierarchical structure of the AI ​​model.

[0249] Optionally, the hierarchical structure of the AI ​​model indicates the starting neural network layer and the ending neural network layer of each level of sub-model in the AI ​​model.

[0250] Optionally, the processing module 1310 is used to: determine, based on the identification of the target object, whether at least one object applicable to the sub-model of the level corresponding to the target object in the multi-level sub-model includes the target object; if at least one object applicable to the sub-model of the level corresponding to the target object in the multi-level sub-model does not include the target object, obtain parameters for constructing the target sub-model; and, construct the target sub-model based on the parameters for constructing the target sub-model.

[0251] Optionally, the parameters used to construct the target sub-model are carried through the following messages: RRC message, MAC CE, DCI or UCI.

[0252] Optionally, the identifier of the target object is carried in a switching message, and the switching message is used to indicate switching to the target object.

[0253] Optionally, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on network coverage areas of multiple different granularities.

[0254] Optionally, the object is a network coverage area, and the network coverage areas of various granularities are: the network coverage area of ​​a cell group, the network coverage area of ​​a cell, and the network coverage area of ​​a beam. The multi-level sub-model includes: a first-level sub-model, a second-level sub-model, and a third-level sub-model. The network coverage area applicable to the first-level sub-model is the network coverage area of ​​the cell group, the network coverage area applicable to the second-level sub-model is the network coverage area of ​​at least one cell in the cell group, and the network coverage area applicable to the third-level sub-model is the network coverage area of ​​at least one beam within the at least one cell.

[0255] Optionally, the storage module is used to store the first-level sub-model and / or the second-level sub-model.

[0256] Optionally, the identifier of the target object includes a cell group identifier, a cell identifier, and a beam identifier.

[0257] Optionally, the processing module 1310 is configured to: train / evaluate the sub-model of the level corresponding to the first network coverage area in the multi-level sub-model based on data collected in the first network coverage area.

[0258] Optionally, the processing module 1310 is configured to mark data collected in the first network coverage area with an identifier of the first network coverage area.

[0259] Optionally, the multi-level sub-model includes a multi-level sub-model obtained by dividing the AI ​​model based on multiple data processing functions, and the multiple data processing functions are functions provided by the AI ​​model based on input data.

[0260] Optionally, the object is a task, and the multiple data processing functions are: a feature extraction function associated with a task group corresponding to the input data, a feature extraction function associated with at least one task in the task group, and an output function associated with one of the at least one task. The multi-level sub-model includes: a first-level sub-model, a second-level sub-model, and a third-level sub-model, wherein the first-level sub-model is applicable to the task group, the second-level sub-model is applicable to the at least one task, and the third-level sub-model is applicable to one of the at least one task.

[0261] Optionally, the storage module is used to store the first-level sub-model and / or the second-level sub-model.

[0262] Optionally, the identifier of the target object includes an identifier of the task group to which the target object belongs.

[0263] In this embodiment, device 1300 may specifically be the first communication device in method 600 described above, and the functions of the first communication device in method 600 described above may be integrated into device 1300. The functions described above may be implemented via hardware, or by hardware executing corresponding software. The hardware or software may include one or more modules corresponding to the functions described above. For example, the transceiver module may be a communication interface, such as a transceiver interface. Device 1300 may be configured to execute the various processes and / or steps in method 600 described above corresponding to the first communication device.

[0264] In another possible embodiment, the apparatus 1300 is configured to execute the following solution:

[0265] The transceiver module 1320 is used to send a model request message, which is used to request an AI model.

[0266] The transceiver module 1320 is also used to: receive a model response message, which indicates the parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each level of sub-model.

[0267] The transceiver module 1320 is further configured to: send a handover message. Alternatively, the transceiver module 1320 is further configured to: receive a handover message. The handover message is used to instruct handover to the target network coverage area.

[0268] The transceiver module 1320 is further used to receive the parameters for constructing the target sub-model.

[0269] The processing module 1310 is used to perform model switching.

[0270] In this embodiment, device 1300 may be specifically the first communication device (e.g., the first communication device is a terminal) in method 800 described above, and the functions of the first communication device (e.g., the first communication device is a terminal) in method 800 described above may be integrated into device 1300. The above functions may 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 above transceiver module may be a communication interface, such as a transceiver interface. Device 1300 may be used to execute the various processes and / or steps corresponding to the first communication device (e.g., the first communication device is a terminal) in method 800 described above.

[0271] In another possible embodiment, the apparatus 1300 is configured to execute the following solution:

[0272] The transceiver module 1320 is used to receive a model request message, where the model request message is used to request an AI model.

[0273] The transceiver module 1320 is also used to: send a model response message, which indicates the parameters used to construct the AI ​​model, the hierarchical structure of the AI ​​model, and at least one object applicable to each level of sub-model.

[0274] The transceiver module 1320 is further configured to receive a handover message. Alternatively, the transceiver module 1320 is further configured to send a handover message. The handover message is used to indicate handover to the target network coverage area.

[0275] The transceiver module 1320 is further configured to determine, based on the identifier of the target network coverage area, whether it is necessary to send parameters for constructing the target sub-model to the terminal.

[0276] It should be understood that the specific process of each module executing the above corresponding process has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

[0277] The processing module 1310 in the above embodiment can be implemented by at least one processor or processor-related circuits. The transceiver module 1320 can be implemented by a transceiver or transceiver-related circuits. The transceiver module 1320 can also be referred to as a communication module or communication interface. The storage module can be implemented by at least one memory.

[0278] In this embodiment, the device 1300 may be specifically the second communication device in the method 800 (for example, the second communication device is a base station), and the functions of the second communication device in the method 800 (for example, the second communication device is a base station) may be integrated into the device 1300. The above functions may 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 may be a communication interface, such as a transceiver interface. The device 1300 may be used to execute the various processes and / or steps corresponding to the second communication device (for example, the second communication device is a base station) in the method 800.

[0279] The apparatus 1300 is, for example, a RAN node, a terminal, a core network device or other network device in FIG3 , or a component (eg, a chip) in these devices, used to implement the method described in the above method embodiment.

[0280] Figure 14 is a schematic block diagram of another communication device 1400 provided in an embodiment of the present application. It is understood that the device 1400 includes, for example, modules, units, elements, circuits, or interfaces, which are appropriately configured together to perform the method of the present application. The device 1400 can be the RAN node, terminal, core network device, or other network device in Figure 3, or a component (e.g., a chip) in these devices, used to implement the method described in the above method embodiment. The device 1400 includes one or more processors 111. The processor 111 can be a general-purpose processor or a dedicated processor. For example, it can be a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication device (e.g., network device, terminal, or chip), execute software programs, and process software program data.

[0281] Optionally, in one design, the processor 111 may include a program 113 (sometimes also referred to as code, computer program, or instructions), which may be executed on the processor 111 to enable the apparatus 1400 to perform the methods described in the above embodiments.

[0282] Optionally, the device 1400 may include one or more memories 112 on which a program 114 (sometimes also referred to as code, computer program or instruction) is stored. The program 114 can be executed on the processor 111 so that the device 1400 performs the method described in the above method embodiment.

[0283] Optionally, the processor 111 and / or the memory 112 may include AI modules 117 and 118, each configured to implement AI-related functions. The AI ​​module may be implemented using software, hardware, or a combination of software and hardware. For example, the AI ​​module may include a real-time information processing (RIC) module. For example, the AI ​​module may be a near-real-time RIC or a non-real-time RIC.

[0284] Optionally, data may be stored in the processor 111 and / or the memory 112. The processor and the memory may be provided separately or integrated together.

[0285] Optionally, the device 1400 may further include a transceiver 115 and / or an antenna 116. The processor 111 may also be referred to as a processing unit, and controls the communication device 1400 (e.g., a network device or terminal). The transceiver 115 may also be referred to as a transceiver unit, a transceiver, a transceiver circuit, or a transceiver, and is configured to implement the transceiver functions of the communication device via the antenna 116.

[0286] The present application further provides an apparatus 1500, which may be a terminal, a processor in a terminal, or a chip. The apparatus 1500 may be configured to execute the operations performed by the first communication apparatus (eg, the terminal) in the above method embodiment.

[0287] When device 1500 is a terminal, FIG15 shows a simplified schematic diagram of the terminal structure. As shown in FIG15 , the terminal includes a processor, a memory, and a transceiver. The memory can store computer program code, and the transceiver includes a transmitter 1531, a receiver 1532, a radio frequency circuit (not shown in FIG15 ), an antenna 1533, and input / output devices (not shown in FIG15 ).

[0288] The processor is mainly used to process communication protocols and communication data; control the terminal, execute software programs and process software program data, etc.

[0289] Memory is mainly used to store software programs and data.

[0290] Radio frequency circuits are mainly used for conversion between baseband signals and radio frequency signals and for processing radio frequency signals.

[0291] Antennas are mainly used to send and receive radio frequency signals in the form of electromagnetic waves.

[0292] The input and output device may include a touch screen, a display screen, or a keyboard. The input and output device is mainly used to receive data input by the user and output data to the user. It should be noted that some types of terminals may not have input and output devices.

[0293] When data needs to be sent, the processor performs baseband processing on the data to be sent and outputs the baseband signal to the RF circuit. The RF circuit then performs RF processing on the baseband signal and transmits it outward in the form of electromagnetic waves via the antenna. When data is sent to the terminal, the RF circuit receives the RF signal via the antenna. The RF circuit converts the RF signal into a baseband signal and outputs the baseband signal to the processor. The processor converts the baseband signal into data and processes the data. For ease of illustration, Figure 15 shows only one memory, processor, and transceiver. In actual terminal products, there may be one or more processors and one or more memories. The memory may also be referred to as a storage medium or storage device, etc. The memory may be set independently of the processor or integrated with the processor, and this is not limited in the embodiments of the present application.

[0294] In the embodiment of the present application, the antenna and radio frequency circuit with transceiver functions can be regarded as the transceiver module of the terminal, and the processor with processing function can be regarded as the processing module of the terminal.

[0295] As shown in FIG15 , the terminal includes a processor 1510, a memory 1520, and a transceiver 1530. The processor 1510 may also be referred to as a processing unit, a processing board, a processing module, or a processing device. The transceiver 1530 may also be referred to as a transceiver unit, a transceiver, or a transceiver device.

[0296] Alternatively, the device implementing the receiving function in transceiver 1530 is considered a receiving module, and the device implementing the transmitting function in transceiver 1530 is considered a transmitting module. That is, transceiver 1530 includes a receiver and a transmitter. A transceiver may also be sometimes referred to as a transceiver, a transceiver module, or a transceiver circuit. A receiver may also be sometimes referred to as a receiver, a receiving module, or a receiving circuit. A transmitter may also be sometimes referred to as a transmitter, a transmitting module, or a transmitting circuit.

[0297] The processor 1510 is configured to execute the processing actions of the first communication device in the embodiment shown in Figure 6 or Figure 8. The transceiver 1530 is configured to execute the transceiver actions of the first communication device in the embodiment shown in Figure 6 or Figure 8.

[0298] It should be understood that FIG15 is merely an example and not a limitation, and the terminal including the transceiver module and the processing module may not rely on the structures shown in FIG13 , FIG14 or FIG15 .

[0299] When the device 1500 is a chip, in one possible design, the chip includes a processor, a memory, and a transceiver. The transceiver may be an input / output circuit or a communication interface. The processor may be a processing module, a microprocessor, or an integrated circuit integrated on the chip. In the above method embodiments, the transmission operation of the first communication device can be understood as an output of the chip, and the reception operation of the first communication device or the second communication device in the above method embodiments can be understood as an input of the chip.

[0300] When the device 1500 is a chip, in another possible design, the transceiver of the chip may be an input / output interface, and the memory is external to the chip.

[0301] The present application also provides an apparatus 1600, which can be a network device or a chip. Apparatus 1600 can be used to perform the operations performed by the first communication device in FIG. 6 (for example, the first communication device is a base station) or the operations performed by the second communication device in FIG. 8 (for example, the second communication device is a base station).

[0302] When the apparatus 1600 is a network device, for example, a base station, FIG16 shows a simplified schematic diagram of a base station structure. The base station includes a portion 1610, a portion 1620, and a portion 1630.

[0303] Part 1610 is mainly used for baseband processing, base station control, etc.; Part 1610 is usually the control center of the base station, which can usually be called a processor, used to control the base station to perform the processing operations on the first communication device or the second communication device side in the above method embodiment.

[0304] The 1620 section is primarily used to store computer program code and data.

[0305] Section 1630 is primarily used for receiving and transmitting RF signals and converting RF signals to baseband signals. Section 1630 can generally be referred to as a transceiver module, transceiver, transceiver circuit, or transceiver. The transceiver module in section 1630, which can also be referred to as a transceiver or transceiver, includes an antenna 1633 and a RF circuit (not shown in FIG16 ), where the RF circuit is primarily used for RF processing. Alternatively, the device used to implement the receiving function in section 1630 can be considered a receiver, and the device used to implement the transmitting function can be considered a transmitter. That is, section 1630 includes a receiver 1632 and a transmitter 1631. A receiver can also be referred to as a receiving module, receiver, or receiving circuit, and a transmitter can be referred to as a transmitting module, transmitter, or transmitting circuit.

[0306] Sections 1610 and 1620 may include one or more boards, each of which may include one or more processors and one or more memories. The processor is used to read and execute programs in the memory to implement baseband processing functions and control the base station. If multiple boards are present, the boards may be interconnected to enhance processing capabilities. As an optional implementation, multiple boards may share one or more processors, multiple boards may share one or more memories, or multiple boards may simultaneously share one or more processors.

[0307] For example, in one implementation, the transceiver module in section 1630 is used to execute the transceiver-related processes performed by the first communication device or the second communication device in the above-mentioned embodiment. The processor in section 1610 is used to execute the processing-related processes performed by the first communication device or the second communication device in the above-mentioned embodiment.

[0308] It should be understood that FIG16 is merely an example and not a limitation, and the network device including the processor, memory, and transceiver may not rely on the structures shown in FIG13 , FIG14 , or FIG16 .

[0309] When device 1600 is a chip, in one possible design, the chip includes a transceiver, a memory, and a processor. The transceiver may be an input / output circuit (or input / output interface) or a communication interface; the processor may be a processor, microprocessor, or integrated circuit integrated on the chip. The sending operation of the first communication device or the second communication device in the above method embodiment can be understood as the output of the chip, and the receiving operation of the first communication device or the second communication device in the above method embodiment can be understood as the input of the chip.

[0310] When the device 1500 is a chip, in another possible design, the transceiver of the chip may be an input / output interface, and the memory is external to the chip.

[0311] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program (also referred to as code, or instructions). When the computer-readable storage medium is run on a computer, the computer executes the method performed by the first communication device or the second communication device in the above method embodiment.

[0312] An embodiment of the present application further provides a computer program product comprising a computer program or instructions, which, when executed by a computer, enables the computer to implement the method performed by the first communication device or the second communication device in the above method embodiment.

[0313] The present application also provides a communication system including a first communication device and a second communication device in the above embodiment. The first communication device is configured to perform some or all of the operations performed by the first communication device in the above method embodiment, and the second communication device is configured to perform some or all of the operations performed by the second communication device in the above method embodiment.

[0314] An embodiment of the present application further provides a chip device, comprising a processor, configured to call a computer program or instruction stored in the memory so that the processor executes the method provided in the above embodiment.

[0315] In a possible implementation, the input of the chip device corresponds to the receiving operation in any one of the foregoing embodiments, and the output of the chip device corresponds to the sending operation in any one of the foregoing embodiments.

[0316] Optionally, the processor is coupled to the memory via an interface.

[0317] Optionally, the chip device further includes a memory, in which computer programs or instructions are stored.

[0318] The processor mentioned in any of the above may be a general-purpose central processing unit, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the method provided in any of the above embodiments. The memory mentioned in any of the above may be a read-only memory (ROM) or other type of static storage device that can store static information and computer programs or instructions, a random access memory (RAM), etc.

[0319] Those skilled in the art can clearly understand that, for the sake of convenience and brevity of description, the explanation of the relevant contents and beneficial effects of any of the above-mentioned devices can refer to the corresponding method embodiments provided above, and will not be repeated here.

[0320] In the several embodiments provided in this 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 merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, 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 an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0321] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0322] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0323] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the part that essentially contributes to the technical solution of the present application or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several computer programs or instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method 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 ROM, a RAM, a magnetic disk, or an optical disk.

[0324] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A model switching method, characterized in that, Including: Determine an artificial intelligence (AI) model, where the AI model includes multiple levels of sub-models, and each level of sub-model in the multiple levels of sub-models applies to at least one object; Obtain the identifier of the target object; Based on the identifier of the target object, determine the target sub-model of the level corresponding to the target object, where the target sub-model applies to the target object; Switch the sub-model of the level corresponding to the target object in the multiple levels of sub-models to the target sub-model.

2. The method according to claim 1, characterized in that Before determining the AI model, the method further includes: Receive a first message, where the first message indicates the parameters for constructing the AI model, the hierarchical structure of the AI model, and at least one object to which each level of sub-model applies, and the multiple levels of sub-models are determined from the AI model based on the hierarchical structure of the AI model.

3. The method according to claim 2, wherein The hierarchical structure of the AI model indicates the starting neural network layer and the ending neural network layer of each level of sub-model in the AI model.

4. The method according to any one of claims 1 to 3, characterized in that The determining, based on the identifier of the target object, the target sub-model of the level corresponding to the target object includes: Based on the identifier of the target object, determine whether at least one object to which the sub-model of the level corresponding to the target object in the multiple levels of sub-models applies includes the target object; In the case where at least one object to which the sub-model of the level corresponding to the target object in the multiple levels of sub-models applies does not include the target object, obtain the parameters for constructing the target sub-model; Based on the parameters for constructing the target sub-model, construct the target sub-model.

5. The method according to claim 4, wherein The parameters for constructing the target sub-model are carried by the following message: Radio Resource Control (RRC) message, Medium Access Control Control Element (MAC CE), Downlink Control Information (DCI), or Uplink Control Information (UCI).

6. The method according to any one of claims 1 to 5, characterized in that, The identifier of the target object is carried in a handover message, and the handover message is used to indicate handover to the target object.

7. The method according to any one of claims 1 to 6, characterized in that, The multiple levels of sub-models include multiple levels of sub-models obtained by dividing the AI model based on network coverage areas of multiple different granularities.

8. The method according to claim 7, wherein The object is a network coverage area, and the multiple different granularities of network coverage areas are: the network coverage area of a cell group, the network coverage area of a cell, and the network coverage area of a beam; The multiple levels of sub-models include: a first-level sub-model, a second-level sub-model, and a third-level sub-model. The network coverage area to which the first-level sub-model applies is the network coverage area of the cell group, the network coverage area to which the second-level sub-model applies is the network coverage area of at least one cell in the cell group, and the network coverage area to which the third-level sub-model applies is the network coverage area of at least one beam in the at least one cell.

9. The method according to claim 8, wherein The method further includes: Store the first-level sub-model and / or the second-level sub-model.

10. The method according to claim 8 or 9, characterized in that The identifier of the target object includes a cell group identifier, a cell identifier, and a beam identifier.

11. The method according to any one of claims 7 to 10, characterized in that, The method further includes: Based on the data collected in the first network coverage area, train or evaluate the sub-model of the level corresponding to the first network coverage area in the multiple levels of sub-models.

12. The method according to claim 11, wherein Before training / evaluating the sub-model of the level corresponding to the first network coverage area in the multi-level sub-model based on the data collected in the first network coverage area, the method further includes: Marking the data collected in the first network coverage area with the identifier of the first network coverage area.

13. The method according to any one of claims 1 to 6, characterized in that, The multi-level sub-model includes a multi-level sub-model obtained by dividing the AI model based on multiple data processing functions, and the multiple data processing functions are the functions provided by the AI model based on the input data.

14. The method according to claim 13, wherein The object is a task, and the multiple data processing functions are: a feature extraction function related to a task group corresponding to the input data, a feature extraction function related to at least one task in the task group, and an output function related to one task in the at least one task; The multi-level sub-model includes: a first-level sub-model, a second-level sub-model, and a third-level sub-model. The first-level sub-model is applicable to the task group, the second-level sub-model is applicable to the at least one task, and the third-level sub-model is applicable to one task in the at least one task.

15. The method according to claim 14, characterized in that, The method further includes: Storing the first-level sub-model and / or the second-level sub-model.

16. The method according to claim 14 or 15, characterized in that, The identifier of the target object includes the identifier of the task group to which the target object belongs.

17. A communication device, characterized in that, Includes a module for implementing the method according to any one of claims 1 to 16.

18. A communication device, characterized in that, Includes a processor for executing a computer program or instruction in a memory to implement the method according to any one of claims 1 to 16.

19. A computer-readable storage medium, characterized in that, For storing a computer program or instruction, when the computer program or instruction runs on a computer, the method according to any one of claims 1 to 16 is executed.

Images