WO2024094157A1 - Ai模型更新方法及通信装置 - Google Patents

Ai模型更新方法及通信装置 Download PDF

Info

Publication number
WO2024094157A1
WO2024094157A1 PCT/CN2023/129535 CN2023129535W WO2024094157A1 WO 2024094157 A1 WO2024094157 A1 WO 2024094157A1 CN 2023129535 W CN2023129535 W CN 2023129535W WO 2024094157 A1 WO2024094157 A1 WO 2024094157A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
information
terminal device
association
target
Prior art date
Application number
PCT/CN2023/129535
Other languages
English (en)
French (fr)
Inventor
陈晓宇
Original Assignee
展讯通信（上海）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 展讯通信（上海）有限公司 filed Critical 展讯通信（上海）有限公司
Publication of WO2024094157A1 publication Critical patent/WO2024094157A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • H04L41/0813—Configuration setting characterised by the conditions triggering a change of settings
    • H04L41/082—Configuration setting characterised by the conditions triggering a change of settings the condition being updates or upgrades of network functionality

Definitions

  • the present application relates to the field of communication technology, and in particular to a method and a communication device for updating an artificial intelligence (AI) model.
  • AI artificial intelligence
  • AI technology can be widely used in various fields. Applying AI technology in the field of wireless communication technology can improve the throughput of wireless communication systems, effectively shorten processing delays, and increase user capacity. This is an important task for future wireless communication technology.
  • the AI function can be realized through the deployed AI model.
  • 3GPP 3rd Generation Partnership Project
  • 3GPP 3rd Generation Partnership Project
  • An embodiment of the present application provides an AI model updating method, which can quickly update the AI model of a terminal device, thereby helping to improve communication efficiency.
  • an embodiment of the present application provides an AI model update method, which can be executed by a terminal device, or by a device matching the terminal device, such as a chip, a chip module, or a processor.
  • the method may include: receiving a first signaling from a network device, the first signaling is used to indicate first model association information, the first model association information is associated with the attribute information of the first AI model; the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information is associated with the attribute information of the second AI model.
  • the terminal device when the terminal device receives the first signaling indicating the first model association information, it can update the second AI model currently used to the first AI model that has an association relationship with the first model association information, so that the terminal device can quickly update the AI model, thereby helping to improve communication efficiency.
  • the second AI model currently in use when the second AI model currently in use is updated to the first AI model, specifically, when the first AI model is different from the second AI model and the terminal device supports the first AI model, the second AI model currently in use is updated to the first AI model.
  • the second AI model currently in use by the terminal device can be updated to the first AI model that has an association relationship with the first model association information, which helps to improve communication quality and communication efficiency.
  • the terminal device supports the first AI model, and the state of the first AI model is not deactivated, the second AI model currently in use is updated to the first AI model. That is to say, when the state of the first AI model is deactivated, the second AI model currently in use will not be updated to the first AI model.
  • a target AI model is selected from the AI models supported by the terminal device, and the target AI model satisfies the similarity condition; the first AI model or the second AI model is updated to the target AI model.
  • a target AI model with attribute information closest to the first AI model can be selected, which can improve communication efficiency to a certain extent.
  • a first request message is sent to the network device, and the first request message is used to indicate the attribute information of the third AI model. That is, when the terminal device does not support the first AI model, the terminal device requests the network device to use the third AI model.
  • a first response message is received from the network device, and when the first response message indicates confirmation information, the first AI model or the second AI model is updated to the third AI model.
  • the AI model update can be stopped, or the first request message can be sent to the network device again, at which time the first request message can indicate that the terminal device requests to use the attribute information of another AI model (that is, the other AI model is different from the third AI model), and waits for confirmation information or rejection information from the network device.
  • the association relationship between the current model association information and the attribute information of the second AI model is updated to the association relationship between the first model association information and the attribute information of the second AI model.
  • the association relationship between the model association information and the attribute information of the AI model can be updated in real time.
  • the second AI model may be deactivated to avoid using two or more AI models at the same time and reduce the processing load of the terminal device. Updating the second AI model to the first AI model implicitly means activating and using the first AI model.
  • the method further includes: sending a second request message to the network device, the second request message is used to determine the first information; and receiving a second response message from the network device, the second response message includes the first information.
  • the first information includes attribute information of a fourth AI model; the method further includes: updating the first AI model or the second AI model to the fourth AI model.
  • the first information includes first reference signal information; the method further includes: obtaining channel quality information according to the first reference signal information; updating the first AI model or the second AI model to an AI model associated with the channel quality information according to the channel quality information, and/or updating the attribute information of the AI model associated with the channel quality information.
  • the terminal device can retrain the AI model to ensure that the updated AI model matches the channel quality information.
  • the first information also includes a reporting threshold; the method further includes: selecting target channel quality information greater than or equal to the reporting threshold from the channel quality information, and determining reference signal information corresponding to the target channel quality information.
  • the method further includes: sending target channel quality information and reference signal information corresponding to the target channel quality information to the network device, so that the network device can obtain the reference signal information used by the terminal device to retrain the AI model.
  • the method further includes: updating the attribute information of the AI model associated with the channel quality information according to the target channel quality information and the reference signal information corresponding to the target channel quality information.
  • the terminal device can retrain the AI model to ensure that the updated AI model matches the channel quality information.
  • the method further includes: if the output result of the first AI model does not meet the preset condition, based on the first model association information, selecting a target AI model from the AI models supported by the terminal device, and updating the first AI model to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the output result of the first AI model does not meet the performance requirements, the first AI model can be updated to the target AI model.
  • the method further includes: when the state of the first AI model is a deactivated state, based on the first model association information, selecting a target AI model from the AI models supported by the terminal device, and updating the first AI model or the second AI model to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the first AI model is deactivated, the first AI model or the second AI model can be updated to the target AI model.
  • the method further includes: the first AI model is different from the AI model indicated by the AI model indication information, and based on the first model association information, a target AI model is selected from the AI models supported by the terminal device, and the first AI model or the second AI model is updated to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the first AI model is different from the AI model indicated by the network device, the first AI model or the second AI model can be updated to the target AI model.
  • the target AI model satisfies a similarity condition, including at least one of the following:
  • the difference between the model identifier of the first AI model and the model identifier of the target AI model is less than or equal to the first threshold;
  • the first model association information is identical to the model association information associated with the target AI model or the similarity is greater than a second threshold;
  • the target AI model is an AI model associated with the preset cell information; the preset cell information is included in the configured or activated cell information; or, the preset cell information is the cell information associated with the target transmission configuration indication TCI state, and the target TCI state is indicated or configured or activated by the second signaling; Optionally, the time difference between the reception time of the second signaling and the current system time is less than the third threshold;
  • the target AI model is the AI model that is used most times in the second preset time period before the second AI model is updated or the AI model that was used last time.
  • the method further includes: receiving a third signaling from the network device, the third signaling being used to indicate the association relationship between the first model association information and the attribute information of the first AI model.
  • the network device may indicate the association relationship between each model association information and the attribute information of each AI model, so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the method further includes: determining an association relationship between the first model association information and the attribute information of the first AI model.
  • the terminal device maintains the association relationship between each model association information and the attribute information of each AI model, which can improve the initiative of the terminal device.
  • the association relationship between the first model association information and the attribute information of the first AI model is predefined by the protocol. That is, the protocol predefines the association relationship between each model association information and the attribute information of each AI model, so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the first model association information may include but is not limited to at least one of the following information: first operator identification information, first public land mobile network (PLMN) information, first network equipment vendor identification information, first tracking area identifier (TAI) information, first roaming guidance (SOR) information, first terminal routing selection strategy (user equipment route selection policy, URSP) information, first operating frequency information, first cell information, reference signal information associated with the first cell information, a first neighboring cell list, first timestamp information, and first terminal device information;
  • PLMN public land mobile network
  • TAI tracking area identifier
  • SOR roaming guidance
  • URSP user equipment route selection policy
  • the first cell information includes but is not limited to at least one of the following information: first cell group information, first service cell identification information, first physical cell identification information, first transmission and reception point (transmission and reception point, TRP) identification information, and first bandwidth part (bandwidth part, BWP) identification.
  • the attribute information of the first AI model includes but is not limited to at least one of the following information: a model identifier of the first AI model, model structure information of the first AI model, and model parameters of the first AI model.
  • an embodiment of the present application provides an AI model update method, which can be executed by a network device, or by a device matching the network device, such as a chip, a chip module, or a processor.
  • the method may include: sending a first signaling to a terminal device, the first signaling is used to indicate first model association information, the first model association information has an association relationship with the attribute information of the first AI model, and the first model association information is different from the second model association information currently used.
  • the method further includes: receiving a first request message from a terminal device, the first request message being used to indicate attribute information of a third AI model; and sending a first response message to the terminal device, the first response message being used to indicate confirmation information.
  • the method further includes: receiving a second request message from the terminal device, and determining the first information according to the second request message; sending a second response message to the terminal device, wherein the second response message includes the first information;
  • the first information includes attribute information of the fourth AI model.
  • the first information includes first reference signal information, where the first reference signal information is reference signal information of a reference signal configured or sent by the network device, or reference signal information that matches the purpose of the AI model recommended by the terminal device, or reference signal information that is associated with attribute information of the AI model expected by the network device.
  • first reference signal information is reference signal information of a reference signal configured or sent by the network device, or reference signal information that matches the purpose of the AI model recommended by the terminal device, or reference signal information that is associated with attribute information of the AI model expected by the network device.
  • the first information also includes a reporting threshold; the above method also includes: receiving target channel quality information from the terminal device and reference signal information corresponding to the target channel quality information, so that the network device can obtain the reference signal information used by the terminal device to retrain the AI model.
  • the attribute information of the AI model associated with the channel quality information is updated to enable the network device to retrain the AI model to ensure that the updated AI model matches the channel quality information.
  • the method further includes: sending a third signaling to the terminal device, the third signaling being used to indicate the association relationship between the first model association information and the attribute information of the first AI model.
  • the network device can indicate the association relationship between each model association information and the attribute information of each AI model, so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the association relationship between the first model association information and the attribute information of the first AI model is predefined by the protocol. That is, the protocol predefines the association relationship between each model association information and the attribute information of each AI model, so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the first model association information may include but is not limited to at least one of the following information: first operator identification information, first PLMN information, first network equipment vendor identification information, first TAI information, first SOR information, first URSP information, first operating frequency information, first cell information, reference signal information associated with the first cell information, first neighboring cell list, first timestamp information, and first terminal device information;
  • the first cell information includes but is not limited to at least one of the following information: first cell group information, first service cell identification information, first physical cell identification information, first TRP identification information, and first BWP identification.
  • the attribute information of the first AI model includes but is not limited to at least one of the following information: a model identifier of the first AI model, model structure information of the first AI model, and model parameters of the first AI model.
  • an embodiment of the present application provides an AI model update method, which can be executed by a terminal device, or by a device matching the terminal device, such as a chip, a chip module, or a processor.
  • the method may include: receiving a first signaling from a network device, the first signaling is used to indicate attribute information of a first AI model; the attribute information of the first AI model is different from the attribute information of a second AI model currently in use, and updating the second AI model to the first AI model.
  • the terminal device when the terminal device receives the first signaling indicating the attribute information of the first AI model, it can update the currently used second AI model to the first AI model, so that the terminal device can quickly update the AI model, thereby helping to improve communication efficiency.
  • the attribute information of the first AI model is the attribute information of the AI model expected by the network device.
  • the attribute information of the first AI model is different from the attribute information of the second AI model currently in use.
  • the attribute information of the first AI model is associated with the terminal device information of the terminal device.
  • the second AI model currently in use may be updated to the first AI model if the terminal device supports the first AI model.
  • the second AI model currently in use by the terminal device may be updated to the first AI model indicated by the first signaling, which helps to improve communication quality and communication efficiency.
  • the second AI model is updated to the first AI model. That is, when the state of the first AI model is deactivated, the second AI model will not be updated to the first AI model.
  • a target AI model is selected from the AI models supported by the terminal device, and the target AI model satisfies the similarity condition; the first AI model or the second AI model is updated to the target AI model.
  • a target AI model with the closest attribute information to the first AI model can be selected, which can improve communication efficiency to a certain extent.
  • a first request message is sent to the network device, and the first request message is used to indicate the attribute information of the third AI model. That is, when the terminal device does not support the first AI model, the terminal device requests the network device to use the third AI model.
  • a first response message is received from the network device, and when the first response message indicates confirmation information, the first AI model or the second AI model is updated to the third AI model.
  • the AI model update can be stopped, or the first request message can be sent to the network device again.
  • the first request message can indicate that the terminal device requests to use the attribute information of another AI model (that is, the other AI model is different from the third AI model), and waits for confirmation information or rejection information from the network device.
  • the second AI model may be deactivated to avoid using two or more AI models at the same time and reduce the processing load of the terminal device. Updating the second AI model to the first AI model implicitly means activating and using the first AI model.
  • the method further includes: sending a second request message to the network device, the second request message is used to determine the first information; and receiving a second response message from the network device, the second response message includes the first information.
  • the first information includes attribute information of a fourth AI model; the method further includes: updating the first AI model or the second AI model to the fourth AI model.
  • the first information includes first reference signal information; the method further includes: obtaining channel quality information according to the first reference signal information; updating the first AI model or the second AI model to an AI model associated with the channel quality information according to the channel quality information, and/or updating the attribute information of the AI model associated with the channel quality information.
  • the terminal device can retrain the AI model to ensure that the updated AI model matches the channel quality information.
  • the first information also includes a reporting threshold; the method further includes: selecting target channel quality information greater than or equal to the reporting threshold from the channel quality information, and determining reference signal information corresponding to the target channel quality information.
  • the method further includes: sending target channel quality information and reference signal information corresponding to the target channel quality information to the network device, so that the network device can obtain the reference signal information used by the terminal device to retrain the AI model.
  • the method further includes: updating the attribute information of the AI model associated with the channel quality information according to the target channel quality information and the reference signal information corresponding to the target channel quality information.
  • the terminal device retrains the AI model to ensure that the updated AI model matches the channel quality information.
  • the method further includes: if the output result of the first AI model does not meet the preset condition, based on the first model association information, selecting a target AI model from the AI models supported by the terminal device, and updating the first AI model to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the output result of the first AI model does not meet the performance requirements, the first AI model can be updated to the target AI model.
  • the method further includes: when the state of the first AI model is a deactivated state, based on the first model association information, selecting a target AI model from the AI models supported by the terminal device, and updating the first AI model or the second AI model to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the first AI model is deactivated, the first AI model or the second AI model can be updated to the target AI model.
  • the method further includes: the first AI model is different from the AI model indicated by the AI model indication information, and based on the first model association information, a target AI model is selected from the AI models supported by the terminal device, and the first AI model or the second AI model is updated to the target AI model.
  • the target AI model satisfies the similarity condition. That is, when the first AI model is different from the AI model indicated by the network device, the first AI model or the second AI model can be updated to the target AI model.
  • the target AI model satisfies a similarity condition, including at least one of the following:
  • the difference between the model identifier of the first AI model and the model identifier of the target AI model is less than or equal to the first threshold;
  • the target AI model is the AI model associated with the preset cell information; the preset cell information is included in the configured or activated cell information; or, the preset cell information is the cell information associated with the target transmission configuration indicating the TCI state, and the target TCI state is indicated or configured or activated by the second signaling; optionally, the time difference between the reception time of the second signaling and the current system time is less than the third threshold;
  • the target AI model is the AI model that is used most times in the second preset time period before the second AI model is updated or the AI model that was used last time.
  • an embodiment of the present application provides an AI model update method, which can be executed by a network device, or by a device matching the network device, such as a chip, a chip module, or a processor.
  • the method may include: sending a first signaling to a terminal device, the first signaling being used to indicate attribute information of a first AI model.
  • an embodiment of the present application provides a communication device, which includes a communication unit and a processing unit.
  • the communication unit is used to receive a first signaling from a network device, the first signaling is used to indicate first model association information, and the first model association information has an association relationship with the attribute information of the first AI model; the processing unit, the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information has an association relationship with the attribute information of the second AI model.
  • the communication unit is used to send a first signaling to a terminal device, the first signaling is used to indicate the first model association information, the first model association information has an association relationship with the attribute information of the first AI model, and the first model association information is different from the second model association information currently in use.
  • the present application provides a communication device, comprising a processor, a memory, and a computer program or instructions stored in the memory, characterized in that the processor executes the computer program or instructions to implement a method provided in any one of the first to fourth aspects.
  • the present application provides a chip.
  • the chip is used to receive a first signaling from a network device, the first signaling is used to indicate first model association information, and the first model association information is associated with the attribute information of the first AI model; the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information is associated with the attribute information of the second AI model.
  • the chip is used to send a first signaling to a terminal device, the first signaling is used to indicate the first model association information, the first model association information is associated with the attribute information of the first AI model, and the first model association information is different from the second model association information currently in use.
  • the present application provides a computer-readable storage medium, in which computer-readable instructions are stored.
  • the communication device executes the method provided in any one of the first to fourth aspects above.
  • the present application provides a computer program or a computer program product, comprising codes or instructions, which, when executed on a computer, enable the computer to execute the method provided in any one of the first to fourth aspects.
  • the present application provides a chip module, which includes a communication module, a power module, a storage module and a chip, wherein: the power module is used to provide power to the chip module; the storage module is used to store data and instructions; the communication module is used for internal communication within the chip module, or for the chip module to communicate with external devices; the chip is used to execute the method provided in any one of the first to fourth aspects.
  • FIG1 is a schematic diagram of the architecture of a communication system to which the present application is applied;
  • FIG2 is a flow chart of an AI model updating method provided in an embodiment of the present application.
  • FIG3 is a flow chart of another AI model updating method provided in an embodiment of the present application.
  • FIG4 is a flow chart of another AI model updating method provided in an embodiment of the present application.
  • FIG5 is a schematic diagram of the structure of a communication device provided by the present application.
  • FIG6 is a schematic diagram of the structure of another communication device provided by the present application.
  • FIG. 7 is a schematic diagram of the structure of a chip module provided in an embodiment of the present application.
  • words such as “first” and “second” are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that words such as “first” and “second” do not limit the quantity and execution order, and words such as “first” and “second” do not necessarily limit differences.
  • “And/or” describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and/or B may represent: A exists alone, A and B exist at the same time, and B exists alone. The character “/" generally indicates that the previously associated objects are in an "or” relationship.
  • the present application can be applied to a fifth generation (5G) system, also known as a new radio (NR) system; or to a sixth generation (6G) system, or a seventh generation (7G) system, or other future communication systems; or can also be used in a device to device (D2D) system, a machine to machine (M2M) system, a vehicle to everything (V2X) system, and the like.
  • 5G fifth generation
  • NR new radio
  • 6G sixth generation
  • 7G seventh generation
  • D2D device to device
  • M2M machine to machine
  • V2X vehicle to everything
  • the present application may be applied to the system architecture shown in FIG1.
  • the communication system 10 shown in FIG1 may include, but is not limited to, a network device 110 and a terminal device 120.
  • the number and form of the devices in FIG1 are for example only and do not constitute a limitation on the embodiments of the present application. For example, multiple terminal devices may be included in an actual application.
  • Terminal equipment also known as user equipment (UE), mobile station (MS), mobile terminal (MT), etc.
  • UE user equipment
  • MS mobile station
  • MT mobile terminal
  • terminal equipment refers to equipment that provides voice and/or data connectivity to users.
  • handheld devices with wireless connection functions vehicle-mounted devices, etc.
  • some examples of terminal equipment are: mobile phones, tablet computers, laptops, PDAs, mobile internet devices (MID), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc.
  • MID mobile internet devices
  • VR virtual reality
  • AR augmented reality
  • the device for realizing the function of the terminal device may be a terminal device; or it may be a device that can support the terminal device to realize the function, such as a chip or a chip module, etc.
  • the device can be installed in the terminal device or used in combination with the terminal device.
  • the technical solution provided in the present application is described by taking the terminal device as an example of the device for realizing the function of the terminal device.
  • the terminal device involved in the embodiment of the present application has an AI model, or is described as having an AI model deployed.
  • the number of AI models supported by the terminal device is not limited in the present application, and depends on the terminal device. It depends on the factory settings or the actual situation.
  • the network device may be an access network device, which refers to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network, and may also be referred to as a base station.
  • RAN nodes are: a gNB, a transmission reception point (TRP), an evolved Node B (eNB), a radio network controller (RNC), a Node B (NB), a base station controller (BSC), a base transceiver station (BTS), a home base station (e.g., home evolved Node B, or home Node B, HNB), a base band unit (BBU), or a wireless fidelity (Wifi) access point (AP), etc.
  • TRP transmission reception point
  • eNB evolved Node B
  • RNC radio network controller
  • NB Node B
  • BSC base station controller
  • BTS base transceiver station
  • home base station e.g., home evolved Node B, or home Node B, HNB
  • BBU wireless
  • the network device may include a centralized unit (CU) node, or a distributed unit (DU) node, or a RAN device including a CU node and a DU node.
  • CU centralized unit
  • DU distributed unit
  • RAN device including a CU node and a DU node.
  • the network device may be a core network device.
  • the core network device is mainly responsible for managing and controlling the entire network, connecting the upper and lower parts, sorting the received data, and transmitting it to the corresponding terminal device or data network (data network, DN).
  • the core network device may also be referred to as a core network node.
  • the core network device may be, for example, a network data analysis function (network data analytics function, NWDAF) node or a location management function (location management function, LMF) node.
  • NWDAF node used to provide network analysis services based on the request data of the network service.
  • LMF node used for overall coordination and scheduling of resources required for the location of the terminal device, and also used to provide positioning services.
  • the network device may be a neural network processing node, which refers to a node that provides services to an AI model, such as providing model parameters for the AI model.
  • the device for realizing the function of the network device may be a network device; or it may be a device that can support the network device to realize the function, such as a chip or a chip module, etc.
  • the device may be installed in the network device or used in combination with the network device.
  • the technical solution provided in the present application is described by taking the device for realizing the function of the network device as an example, that is, the network device.
  • the communication system described in the embodiment of the present application is for more clearly illustrating the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided by the embodiment of the present application.
  • Those skilled in the art can know that with the evolution of the system architecture and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.
  • AI models can be used to implement AI functions and can also be described as machine learning models.
  • AI models can be implemented in at least one of the following ways, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc.
  • neural network is composed of neurons, including: input parameters, multiplicative coefficients, additive coefficients, activation functions, etc.
  • Common activation functions include: Sigmoid function, hyperbolic function (tanh function), rectified linear unit (ReLU) function, etc.
  • the parameters of the neural network are optimized using a gradient optimization algorithm, which is a type of minimization or maximization algorithm.
  • the objective function (also called loss function) is an algorithm that is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we build a neural network model f. With the model, we can predict the output information based on the input, and calculate the difference between the predicted value and the true value. This is the loss function.
  • the purpose of building a neural network model is to minimize the value of the loss function as much as possible. The smaller the loss value, the closer the neural network model is to the actual situation.
  • the error back propagation (BP) algorithm is a common optimization algorithm used to optimize AI models.
  • the basic idea of the BP algorithm is that the learning process consists of two processes: forward propagation of signals and back propagation of errors.
  • forward propagation the input sample is passed from the input layer, processed layer by layer by each hidden layer, and then passed to the output layer. If the actual output of the output layer does not match the expected output, it enters the back propagation stage of the error.
  • the back propagation stage of the error is referred to as error back propagation.
  • Error back propagation is to propagate the output error layer by layer through the hidden layer to the input layer in some form, and distribute the error to all units in each layer, so as to obtain the error signal of each layer unit, and this error signal is used as the basis for correcting the weights of each unit.
  • This process of adjusting the weights of each layer of the signal forward propagation and error back propagation is repeated.
  • the process of continuous adjustment of weights is the learning and training process of the AI model. This process continues until the error output by the AI model is reduced to an acceptable level, or until the preset number of learning times is reached.
  • Optimization algorithms for optimizing AI models can also be Gradient Descent algorithm, Stochastic Gradient Descent (SGD) algorithm, mini-batch Gradient Descent algorithm, Momentum method (Momentum), Stochastic Gradient Descent with momentum (Nesteroy) algorithm, adaptive Gradient Descent (Adagrad) algorithm, Root Mean Square Prop (Root Mean Square Prop) algorithm, Adaptive Moment Estimation (Adaptive Moment Estimation, Adam) algorithm, etc.
  • the derivative/partial derivative of the current neuron is calculated based on the error obtained by the loss function, and the influence of the learning rate, previous gradient/derivative/partial derivative, etc. is added to obtain the gradient, which is then passed to the previous layer.
  • the AI model involved in the embodiments of the present application refers to an AI model applied in the field of wireless communication technology, which can be used to improve the throughput of wireless communication systems, shorten processing delays, and increase user capacity.
  • the AI model involved in the embodiments of the present application refers to an AI model of a terminal device, thereby improving the communication capabilities of the terminal device.
  • the attribute information of the AI model is used to describe at least one of the identification information of the AI model, the model structure information of the AI model, and the model parameters of the AI model.
  • the attribute information of the AI model may include at least one of the following: the identification information of the AI model, the model structure information of the AI model, and the model parameters of the AI model.
  • the identification information is used to identify the AI model, for example, it can be a model identifier (identifier, ID), different model IDs identify different AI models, and different AI models have different model IDs.
  • the model structure information is used to describe the model structure, for example, it can be a neural network, a decision tree, or structural information such as neurons and layers in the model. Different model structures can be considered different AI models.
  • Model parameters may include at least one parameter, such as an activation function, a multiplicative coefficient, an additive coefficient, etc. Different model parameters can be considered different AI models. For example, if two AI models have the same model parameters except for the activation function, the two AI models are also considered to be different AI models.
  • the name of the attribute information of the AI model is used for example, and can also be described as the model information of the AI model, the feature information of the AI model, or the basic information of the AI model, etc.
  • the model-related information is used to describe information associated with the AI model, or to describe Information that has an association relationship with the attribute information of the AI model. Changes in model association information may lead to changes in the AI model.
  • the association relationship between model association information and the AI model can be a one-to-one relationship, that is, one type of model association information is associated with one AI model; it can also be a many-to-one relationship, that is, multiple types of model association information can be associated with one AI model; it can also be a one-to-many relationship, that is, one type of model association information can be associated with multiple AI models.
  • Model-associated information may include, but is not limited to, at least one of the following information: PLMN information, operator identification information, network equipment vendor identification information, TAI information, SOR information, URSP information, operating frequency information, cell information, reference signal information associated with cell information, neighbor cell list, timestamp information, and terminal device information.
  • PLMN information is used to describe a cellular mobile communication network of a certain standard of a certain operator in a certain area.
  • PLMN information is used to describe the 5G standard network of operator 1 in the home area of the terminal device.
  • PLMN information is used to describe the 5G standard network of operator 2 in the visited area of the terminal device.
  • the operator identification information is used to identify the operator, for example, it may be the operator that the terminal device has signed a contract with, or the operator to which the terminal device currently belongs.
  • the network equipment vendor identification information is used to identify the network equipment vendor, for example, it may be a base station manufacturer or a core network equipment manufacturer.
  • TAI PLMN + tracking area code (TAC).
  • Tracking area (TA) is a concept established for the location management of terminal devices.
  • the core network device can know the TA where the terminal device is located.
  • paging the terminal device in an idle state paging is performed in all cells of the TA registered by the terminal device.
  • TA is a cell-level configuration. Multiple cells can be configured with the same TA, and one cell belongs to one TA.
  • SOR information is used by the home PLMN to guide the terminal device from one network to another.
  • SOR is a technology that encourages roaming terminal devices to roam to a preferred roaming network through the home PLMN. For example, the terminal device is registered on one PLMN, and for some reason, the home PLMN of the terminal device wants the terminal device to register on another PLMN.
  • URSP information is used to describe the routing selection strategy of the terminal device.
  • the strategy describes the correspondence between the application and the network slice on the terminal device, and the terminal device can select the network slice for the application according to the URSP.
  • the operating frequency information is used to describe the frequency that the terminal device will use, is currently using, or can use, and may include but is not limited to at least one of the frequency number, frequency band number, etc.
  • the cell information is used to describe the serving cell information of the terminal device, and may include but is not limited to at least one of the following information: cell group information, serving cell identification information, physical cell identifier (PCI), TRP identification information, BWP identification.
  • the cell group information may include a cell group ID, which is used to identify the cell group to which the serving cell belongs.
  • the serving cell identification information may include a serving cell ID, which is used to identify the serving cell.
  • PCI is used to identify the logical cell of the serving cell.
  • the TRP identification information may include but is not limited to at least one of the following: a control-resource set (CORESET) identifier, a control-resource set pool (CORESET pool) identifier, a control-resource set group identifier, a transmission configuration indication (TCI) status pool identifier, a TCI status identifier, a reference signal resource identifier, and a reference signal resource set identifier.
  • CORESET control-resource set
  • CORESET pool control-resource set pool
  • TCI transmission configuration indication
  • BWP identifier is used to identify the BWP.
  • Reference signal (RS) information associated with the cell information which is used to describe the reference signal associated with the cell information, and may include but is not limited to the number of reference signal resources, the set of reference signal resources, the number ...
  • the index of the reference signal, the pattern of the reference signal, and the repetition parameter value of the reference signal for example, repetition off or repetition on, repetition off means not turning on the repeated transmission of the reference signal, and repetition on means turning on the repeated transmission of the reference signal).
  • the reference signal refers to a downlink reference signal, for example, a synchronization signal block (SSB) or a channel state information-reference signal (CSI-RS).
  • SSB synchronization signal block
  • CSI-RS channel state information-reference signal
  • Neighbor cell list used to describe the neighbor cells of the serving cell.
  • the neighbor cell list can be configured by the network device through high-level signaling.
  • Timestamp information is used to describe the time interval of the AI model, such as the training time interval or the application time interval. It can be understood that the timestamp information is the effective time of the AI model.
  • the network device can broadcast the timestamp information of the AI model.
  • the terminal device information is used to describe the identification information of the terminal device, and may include but is not limited to at least one of the terminal device vendor identification information, the terminal device identification information, etc.
  • the terminal device vendor identification information is used to identify the manufacturer of the terminal device, such as the brand of the mobile phone, etc.
  • the identification information of the terminal device is used to uniquely identify the terminal device, and may include but is not limited to the international mobile subscriber identity (IMSI), the subscription permanent identifier (SUPI), the subscription concealed identifier (SUCI), etc.
  • IMSI international mobile subscriber identity
  • SUPI subscription permanent identifier
  • SUCI subscription concealed identifier
  • the device vendor identification information of the terminal device and the identification information of the terminal device are associated with the AI model.
  • AI model 1 is applicable to mobile phones of brand A
  • AI model 2 is applicable to mobile phones of brand B.
  • the network device can broadcast the device vendor identification information of the terminal device to which the AI model is applicable, or broadcast the identification information of the terminal device to which the AI model is applicable.
  • the network device can broadcast that AI model 1 is applicable to mobile phones of brand A.
  • model association information listed above is for example only and does not constitute a limitation on the embodiments of the present application.
  • the model association information may also include other types of information.
  • the information included in the model association information may change with the change of the application scenario, such as adding or reducing certain information.
  • association relationship between the attribute information of the AI model and the model-related information which actually means that there is an association relationship between the AI model and the model-related information. Specifically, it refers to the association relationship between the attribute information of each AI model in at least one AI model and the model-related information under different environments or conditions.
  • the association relationship can be a one-to-one relationship, a many-to-one relationship, or a one-to-many relationship.
  • the model association information is the operator identification information.
  • the association relationship between the attribute information of the two AI models and the two operator identification information can be seen in Table 1 below.
  • model IDs of the two AI models are 1 and 2 respectively.
  • Model ID 1 and model parameter 1 are associated with operator ID A
  • model ID 2 and model parameter 2 are associated with operator ID B.
  • the association relationship between the attribute information of the AI model and the model association information is indicated by the network device, for example, the network device may indicate the association relationship through high-level signaling or downlink control information (DCI) or
  • DCI downlink control information
  • the broadcast message indicates the association relationship between the attribute information of the AI model and the model association information.
  • the terminal device supports 2 AI models, and the network device can indicate the association relationship between each of the 2 AI models and the model association information associated with each of them. It can be understood that when indicating the association relationship, the network device can implicitly indicate the attribute information and model association information of the AI model with the association relationship. That is, the network device can indicate the attribute information of the AI model, as well as the model association information associated with the attribute information.
  • the network device indicates the association relationship so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the association relationship between the attribute information of the AI model and the model association information is determined by the terminal device.
  • the terminal device maintains the association relationship.
  • the terminal device may report its determined association relationship to the network device so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the association relationship between the attribute information of the AI model and the model association information is predefined by a protocol.
  • the protocol predefines the association relationship so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • updating an AI model can also be understood as at least one of the following: updating the AI model to (or switching to) another AI model; deactivating the AI model and activating a new AI model; updating the attribute information of the AI model; training the AI model; fine-tuning the AI model; adjusting the AI model; testing the AI model; supervising the AI model; verifying/validating the AI model, etc.
  • Figure 2 is a flowchart of an AI model update method provided in an embodiment of the present application. Taking the terminal device executing the process shown in Figure 2 as an example, the process shown in Figure 2 may specifically include the following steps:
  • the network device sends a first signaling to the terminal device.
  • the terminal device receives the first signaling from the network device.
  • the first signaling is used to indicate first model association information, and the first model association information is associated with attribute information of the first AI model.
  • the first model-associated information may include, but is not limited to, at least one of the following information: first PLMN information, first operator identification information, first network equipment vendor identification information, first TAI information, first SOR information, first URSP information, first operating frequency information, first cell information, reference signal information associated with the first cell information, first neighboring cell list, first timestamp information, and first terminal device information.
  • the first cell information includes, but is not limited to, at least one of the following information: first cell group information, first service cell identification information, first PCI, first TRP identification information, and first BWP identification.
  • the information included in the first model-associated information can refer to the aforementioned specific description of the information included in the model-associated information, which will not be repeated here.
  • the first signaling is different according to the different contents included in the first model association information.
  • the first model association information is a first neighbor cell list
  • the first signaling may be a radio resource control (RRC) signaling
  • RRC radio resource control
  • the RRC signaling is used to configure the first neighbor cell list.
  • the first signaling may be an RRC signaling used to configure the first neighbor cell list.
  • the first model association information is the first PCI
  • the first signaling may be a media access control-control element (MAC-CE) signaling
  • the MAC-CE signaling is used to indicate activation of the first PCI. That is, the first signaling may be a MAC-CE signaling indicating activation of the first PCI.
  • MAC-CE media access control-control element
  • the first model association information is the first cell information
  • the first signaling may be an RRC signaling, which is used to indicate cell switching, specifically indicating switching to the first cell.
  • the RRC signaling may directly indicate or implicitly indicate the first cell information.
  • the RRC signaling may, for example, be an RRC reconfiguration (RRCReconfiguration) signaling. That is, the first signaling may be an RRC signaling indicating a cell switching.
  • the first signaling may be a MAC-CE signaling, which is used to indicate activation of the first cell.
  • the MAC-CE signaling may directly indicate or implicitly indicate the first cell information. That is, the first signaling may be a MAC-CE signaling indicating activation of the first cell.
  • the first model association information is a first BWP identifier
  • the first signaling may be a DCI, which is used to indicate BWP switching, specifically indicating switching to the first BWP.
  • the DCI may directly indicate or implicitly indicate the first BWP identifier. That is, the first signaling may be a DCI indicating BWP switching.
  • the first model association information is the first SOR information or the first URSP information
  • the first signaling may be signaling sent directly by the core network device to the terminal device, or may be signaling sent by the core network device to the terminal device through the base station.
  • the first model association information has an association relationship with the attribute information of the first AI model.
  • the network device sends a third signaling to the terminal device, and accordingly, the terminal device receives the third signaling from the network device.
  • the third signaling is used to indicate the association relationship between the first model association information and the attribute information of the first AI model. This allows the network device and the terminal device to have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the terminal device determines or maintains the association relationship between the first model association information and the attribute information of the first AI model.
  • the association relationship between the first model association information and the attribute information of the first AI model is predefined by the protocol, so that the network device and the terminal device have a consistent understanding of the association relationship between the AI model of the terminal device and the model association information.
  • the first model association information is different from the currently used second model association information, and the currently used second AI model is updated to the first AI model.
  • the second model association information and the attribute information of the second AI model are associated.
  • the second model association information currently in use is the current model association information, which can be understood as the model association information currently in use.
  • the URSP information currently in use is the second URSP information
  • the operating frequency information currently in use is the second operating frequency information, etc.
  • the information included in the second model association information can refer to the specific description of the information included in the model association information, which will not be repeated here.
  • the terminal device can determine whether the first model association information indicated by the first signaling is the same as the second model association information currently in use. If at least one type of model association information in the first model association information and the second model association information is different, it can be determined that the first model association information is different from the second model association information. For example, if the first PCI in the first model association information is different from the second PCI in the second model association information, it can be determined that the first model association information is different from the second model association information.
  • the terminal device may update the second AI model currently in use to the first AI model.
  • the terminal device updates the second AI model currently in use to the first AI model that has an association relationship with the first model association information.
  • the update can be performed in the following two situations.
  • the second AI model is updated to the first AI model. That is to say, when the terminal device updates the AI model, it not only considers Different AI models require consideration of whether the terminal device supports the new AI model to avoid a situation where the updated AI model is unavailable.
  • the second AI model is updated to the first AI model. That is to say, when the terminal device considers whether the AI models are different and whether the terminal device supports the new AI model, it also considers whether the new AI model is in the deactivated state. If it is not in the deactivated state, it can be updated to the new AI model; if it is in the deactivated state, it cannot be updated to the new AI model. Whether the state of the first AI model is in the deactivated state can be indicated by the network device or determined autonomously by the terminal device.
  • the terminal device updates from AI model 1 to AI model 2, it deactivates AI model 1.
  • AI model 2 can be updated to AI model 1, but the state of AI model 1 is in the deactivated state, so AI model 2 will not be updated to AI model 1.
  • the change of the state of the AI model can be determined by the terminal device or indicated by the network device.
  • the above situation 1 and situation 2 are for the terminal device to support the first AI model. If the terminal device does not support the first AI model, the following method 1 or method 2 can be used.
  • Mode 1 When the first AI model is different from the second AI model and the terminal device does not support the first AI model, based on the first model association information, a target AI model is selected from the AI models supported by the terminal device, and the first AI model or the second AI model is updated to the target AI model. That is, when the terminal device does not support the first AI model, the terminal device selects a suboptimal AI model, and the suboptimal AI model is the target AI model. That is, when the terminal device finds that the first AI model is not supported, the first AI model or the second AI model can be updated to the target AI model.
  • the terminal device supports AI model 1 and AI model 2.
  • the terminal device moves from the network of operator 1 to the network of operator 3 (that is, the AI model associated with the current model association information is model 1), and the AI model associated with the first model association information is AI model 3, but the terminal device does not support AI model 3.
  • the terminal device selects AI model 2 as the target AI model. It’s just that AI model 2 is associated with the network of operator 2 instead of the network of operator 3.
  • the terminal device updates AI model 1 to AI model 2, it can establish an association relationship between the attribute information of AI model 2 and the first model association information.
  • the target AI model satisfies the similarity condition, which may include at least one of the following:
  • the difference between the model identification of the first AI model and the model identification of the target AI model is less than or equal to the first threshold. That is, the model ID of the first AI model is closest to the model ID of the target AI model.
  • the first threshold is 1
  • the model ID of the first AI model is 1
  • the value range of the model ID of the AI model supported by the terminal device is ⁇ 2,3 ⁇
  • the AI model with the model ID of the AI model being 2 is the target AI model.
  • the model ID of the second AI model is 2
  • the second AI model may not be updated, or the attribute information of the second AI model may be updated to better adapt to changes in the communication environment.
  • the first model association information is the same as the model association information associated with the target AI model or the similarity is greater than the second threshold.
  • the second threshold may be different for different types of model association information. For example, if the model association information is TAI information, the second threshold may be 80%; if the model association information is USRP information, the second threshold may be 90%.
  • model association information including multiple types of information if the similarity of one model association information is greater than its corresponding second threshold, it can be considered that the overall similarity is greater than the second threshold.
  • the first model association information includes first TAI information and first USRP information. If the first If the similarity between the TAI information and the TAI information associated with the target AI model is greater than the second threshold, it is considered that the similarity between the first model-associated information and the model-associated information associated with the target AI model is greater than the second threshold.
  • the target AI model is an AI model associated with the preset cell information.
  • the preset cell information is included in the configured or activated cell information, for example, the target AI model is the AI model associated with the default BWP in the BWP activated by the network device.
  • the preset cell information is the cell information associated with the target TCI state, and the target TCI state is indicated or configured or activated by the second signaling. Among them, the time difference between the reception time of the second signaling and the current system time is less than the third threshold.
  • the target TCI state is the TCI state indicated, configured or activated by the second signaling recently sent by the network device.
  • the second signaling is a high-level signaling
  • the target TCI state is the TCI state configured by the high-level signaling.
  • the second signaling is a MAC-CE signaling
  • the target TCI state is the TCI state activated by the MAC-CE signaling.
  • the target TCI state can also be understood as the most recently used TCI state.
  • the cell information associated with the target AI model can be understood as the cell information associated with the target TCI state.
  • the preset cell information includes a default cell identifier in a neighbor cell list, so that the target AI model is an AI model associated with the default cell identifier in the neighbor cell list configured by the network device for the terminal device.
  • the target AI model is the AI model that is used most times in the second preset time period before the second AI model is updated or the AI model that was used last.
  • the specific value range of the second preset time period is not limited in the embodiment of the present application.
  • the last used AI model refers to the AI model used before the second AI model is used.
  • Mode 2 When the first AI model is different from the second AI model and the terminal device does not support the first AI model, a first request message is sent to the network device, and the first request message is used to indicate the attribute information of the third AI model. That is, when the terminal device does not support the first AI model, the terminal device sends a first request message to the network device, indicating the attribute information of the third AI model, indicating that the terminal device requests to use the third AI model, and the third AI model may be different from the first AI model.
  • the third AI model may be a target AI model or another AI model.
  • the terminal device receives a first response message from the network device, and when the first response message indicates confirmation information, the first AI model or the second AI model is updated to the third AI model.
  • the AI model update may be stopped, or the first request message may be sent to the network device again, at which time the first request message may indicate that the terminal device requests to use the attribute information of another AI model (that is, the other AI model is different from the third AI model), and waits for confirmation information or rejection information from the network device.
  • the first request message does not specify the attribute information of any AI model, and the first request message is used to request the network device to indicate the attribute information of an AI model.
  • the network device Upon receiving the first request message, the network device sends a first response message to the terminal device, and the first response message is used to indicate the attribute information of the new AI model.
  • the terminal device may update the second AI model to the new AI model. Furthermore, the terminal device may establish an association relationship between the new AI model and the first model association information.
  • the terminal device sends a first request message to the network device to request the use of a third AI model or a target AI model.
  • the terminal device receives a first response message indicating confirmation information
  • the first AI model or the second AI model is updated to a third AI model or a target AI model.
  • the terminal device receives a first response message indicating confirmation information
  • the second AI model is updated to the third AI model or the target AI model; after the terminal device is updated to the first AI model, it is found that the first AI model is not supported.
  • the terminal device receives a first response message indicating confirmation information
  • the first AI model is updated to the third AI model. Type or target AI model.
  • the above-mentioned cases 1 and 2, as well as methods 1 and 2, are for the case where the first AI model is different from the second AI model. If the first AI model is the same as the second AI model, the terminal device updates the association relationship between the current model association information and the attribute information of the second AI model to the association relationship between the first model association information and the attribute information of the second AI model. Thus, the association relationship between the model association information and the attribute information of the AI model can be updated in real time.
  • the above-mentioned methods 1 and 2 are for the terminal device not supporting the first AI model. If the terminal device finds at least one of the following defects 1 to 3 before or after updating the second AI model to the first AI model, the target AI model can be selected from the AI models supported by the terminal device based on the first model association information.
  • Defect 1 The output result of the first AI model does not meet the preset conditions, that is, the output result of the first AI model does not meet the performance requirements. In other words, when the output result of the first AI model does not meet the performance requirements, the first AI model can be updated to the target AI model.
  • the similarity conditions satisfied by the target AI model can be referred to the specific description in method 1, which will not be repeated here.
  • Defect 2 The state of the first AI model is deactivated, that is, the first AI model is indicated as deactivated by the network device or the terminal device determines that the state of the first AI model is deactivated, and the first AI model or the second AI model can be updated to the target AI model. If the terminal device has found that the state of the first AI model is deactivated before the second AI model is updated to the first AI model, the second AI model is updated to the target AI model. If the terminal device finds that the state of the first AI model is deactivated after the second AI model is updated to the first AI model, the first AI model is updated to the target AI model.
  • the first AI model is different from the AI model indicated by the AI model indication information.
  • the AI model indication information is sent by the network device and is different from the first signaling and the second signaling.
  • the network device may send AI model indication information to the terminal device when the terminal device accesses. If before updating the second AI model to the first AI model, the terminal device finds that the AI model indicated by the network device is not the first AI model, then the second AI model may be updated to the target AI model. If after updating the second AI model to the first AI model, the terminal device finds that the AI model indicated by the network device is not the first AI model, then the first AI model may be updated to the target AI model.
  • the terminal device can also adopt method 2 to send a first request message to the network device, the first request message is used to indicate the attribute information of the third AI model, and when receiving the first response message of the indication confirmation information from the network device, the first AI model or the second AI model is updated to the third AI model according to the confirmation information.
  • the first request message is used to request the network device to indicate the attribute information of an AI model, and when receiving the first response message, the first response message indicates the attribute information of the new AI model, the first AI model or the second AI model is updated to the new AI model.
  • the first AI model can be updated to the third AI model or the target AI model; or the second AI model can be updated to the third AI model or the target AI model.
  • the situation where the first AI model is unavailable may include but is not limited to at least one of the following situations:
  • the terminal device does not support the first AI model
  • the first AI model is different from the AI model indicated by the AI model indication information.
  • the terminal device may also update the second AI model to The AI model determined by the reference signal sent by the network device.
  • the terminal device obtains a measurement result (for example, including channel quality information) by measuring the reference signal sent by the network device in the first cell, determines an AI model according to the measurement result, and updates the second AI model to the AI model.
  • the terminal device can also update the first AI model to an AI model determined by measuring the reference signal sent by the network device.
  • the terminal device may deactivate the second AI model to avoid using two or more AI models at the same time, thereby reducing the processing load of the terminal device.
  • the terminal device when the terminal device receives the first signaling indicating the first model association information, it can update the second AI model currently used to the first AI model that has an association relationship with the first model association information, so that the terminal device can quickly update the AI model, thereby helping to improve communication efficiency.
  • FIG3 is a flowchart of another AI model updating method provided in an embodiment of the present application.
  • the method may specifically include the following steps:
  • the network device sends a first signaling to the terminal device.
  • the terminal device receives the first signaling from the network device.
  • the first signaling is used to indicate first model association information, and the first model association information is associated with attribute information of the first AI model.
  • step 301 may refer to the specific description of step 201 in the embodiment shown in FIG2 , which will not be repeated here.
  • the terminal device may directly update the second AI model to the first AI model, or execute steps 302 to 304 when the first AI model is not available, or execute steps 302 to 304 when the terminal device does not accept the use of the first AI model.
  • the terminal device sends a second request message to the network device.
  • the network device receives the second request message from the terminal device.
  • the second request message may include but is not limited to at least one of the following information:
  • the purpose of the AI model recommended by the terminal device can be used by the network device to provide corresponding reference signal information to the terminal device based on the purpose, such as providing corresponding reference signal resources.
  • the uses of the AI model may include but are not limited to: training, reasoning, updating, switching, activation, deactivation, monitoring, etc. Training can be divided into initial training or retraining, and training can be model parameters of the training AI model, etc.
  • the attribute information of the AI model recommended by the terminal device can be used by the network device to determine whether to approve the terminal device to use the AI model.
  • the attribute information of the AI model recommended by the terminal device for example, at least one of the model ID, recommended model structure information, or recommended model parameters of the AI model recommended by the terminal device.
  • the AI model update duration recommended by the terminal device is used to indicate the duration required for the AI model update determined by the terminal device.
  • the recommended AI model update duration may include at least one of the network switching duration, cell switching duration, BWP switching duration, and beam switching duration.
  • the reference signal information recommended by the terminal device is used to indicate how the terminal device recommends the network device to send the reference signal.
  • the reference signal information recommended by the terminal device may include, but is not limited to: the number of reference signal resources, the pattern of the reference signal, the repetition parameter value of the reference signal (e.g., repetition off or repetition on, repetition off means not turning on the repeated transmission of the reference signal, repetition on means turning on the repeated transmission of the reference signal), and at least one of the reporting thresholds of the reference signal.
  • the number of receiving beams expected by the terminal device is used to indicate how many transmitting beams the terminal device expects the network device to use to send reference signals, which is used to match the number of receiving beams of the terminal device.
  • the uplink resource corresponding to the second request message may include but is not limited to at least one of the following resources:
  • Preconfigured periodic physical uplink channel resources such as preconfigured periodic physical uplink control channel (PUCCH) resources.
  • PUCCH physical uplink control channel
  • the uplink authorized resource that is used the most times in the first preset time period before sending the second request message The specific value range of the first preset time period is not limited in the embodiment of the present application.
  • the first preset time period and the second preset time period may be the same or different.
  • next available uplink grant resource (next available uplink grant) can also be described as the most recently available uplink grant resource or the uplink grant resource to be used, and "available" means that the most recently arrived uplink grant resource can carry the second request message.
  • the next available uplink grant resource can be a dynamic grant resource (dynamic grant) or a configured grant resource (configured grant).
  • a random access channel (RACH) resource associated with the first model association information A random access channel (RACH) resource associated with the first model association information.
  • the network device preconfigures an association relationship between the random access resource and the model association information, and the UE determines to use the random access resource associated with the first model association information based on the detected first model association information, and sends a second request message.
  • RACH random access channel
  • Random access resources corresponding to the reference signal information associated with the first model association information For example, RACH resources corresponding to the SSB associated with the first cell.
  • the terminal device determines to use the RACH resources corresponding to the SSB of the first cell based on detecting the SSB associated with the first cell, and sends a second request message.
  • the network device sends a second response message to the terminal device.
  • the terminal device receives the second response message from the network device.
  • the second response message is used to respond to the second request message.
  • the network device determines the first information according to the second request message.
  • the second response message may include the first information, and the first information may be at least one of the following information:
  • the A1 information corresponds to the A information in the above second request message.
  • the network device provides the corresponding reference signal information to the terminal device for the AI model of the purpose requested by the terminal device.
  • the B1 information corresponds to the B information in the above-mentioned second request message.
  • the B information is the attribute information of the AI model recommended by the terminal device.
  • the terminal device recommends the use of the fourth AI model, that is, the second request message includes the attribute information of the fourth AI model recommended by the terminal device. If the network device agrees that the terminal device uses the fourth AI model, then the first information is confirmation information, and the confirmation information indicates that the network device agrees that the terminal device uses the requested AI model; if the network device does not agree that the terminal device uses the fourth AI model, then the first information is rejection information, and the rejection information indicates that the network device does not agree that the terminal device uses the requested AI model. When the terminal device receives the rejection information, it can stop updating the AI model, or send the second request message to the network device again.
  • B1 attribute information of the AI model recommended by the terminal device. That is, the first information includes the attribute information of the AI model recommended by the terminal device in the second request message. It can be understood that the attribute information of the AI model in the second request message is the same as the attribute information of the AI model in the first information, that is, the network device agrees that the terminal device uses the requested AI model. For example, the terminal device recommends using the fourth AI model, that is, the second request message includes the attribute information of the fourth AI model recommended by the terminal device.
  • the first information Including attribute information of the fourth AI model; if the network device does not agree that the terminal device uses the fourth AI model, then the first information includes attribute information of another AI model, or the first information includes D1 information.
  • the fourth AI model may be the same as or different from the first AI model.
  • B1 attribute information of the AI model expected by the network device.
  • the attribute information of the AI model expected by the network device is independent of the B information in the above-mentioned second request message.
  • the network device directly feeds back to the terminal device the attribute information of the AI model that the network device expects the terminal device to use, regardless of which AI model the terminal device requests to use.
  • the terminal device receives the attribute information of the AI model expected by the network device, if the terminal device agrees to use the AI model, the second AI model can be updated to the AI model; if the terminal device does not agree to use the AI model, the AI model update can be stopped, or the second request message can be sent to the network device again.
  • the specified AI model update duration that is, the duration required for the AI model update specified by the network device.
  • the specified AI model update duration may include at least one of the network switching duration, cell switching duration, BWP switching duration, and beam switching duration.
  • the network device determines the specified AI model update duration based on the second request message, which may be greater than or equal to the duration required for the terminal device to complete the switch to the first operator network.
  • the C1 information may or may not correspond to the C information in the above-mentioned second request message.
  • the network device may determine the specified AI model update duration based on the AI model update duration recommended by the terminal device; the network device may also determine the specified AI model update duration based on other information in the second request message.
  • reference signal information of the reference signal configured or sent by the network device may match the reference signal information recommended by the terminal device, that is, the terminal device recommends how the network device should be configured or sent, and the network device configures or sends according to the terminal device's suggestion; or it may not match, that is, the network device does not configure or send the reference signal according to the terminal device's suggestion.
  • the number of transmit beams supported by the network device to match the number of receive beams of the terminal device.
  • the E1 information matches the E information.
  • the terminal device updates the second AI model according to the first information.
  • the terminal device may update the second AI model to the AI model requested for use.
  • the terminal device may update the second AI model to the AI model recommended by the terminal device.
  • the terminal device may update the second AI model to the fourth AI model.
  • the terminal device may update the second AI model to the AI model expected by the network device.
  • the terminal device may update the second AI model to the fourth AI model.
  • the first information includes first reference signal information, which may be reference signal information of a reference signal configured or sent by a network device, or may be reference signal information that matches the purpose of an AI model recommended by a terminal device, or may be reference signal information that is associated with attribute information of an AI model expected by a network device.
  • the terminal device may measure the received reference signal based on the first reference signal information to obtain channel quality information.
  • the channel quality information may include, but is not limited to, at least one of the following: signal to interference plus noise ratio (SINR), reference signal receiving power (RSRP), received signal strength indicator (RSSI ... Quality (reference signal receiving quality, RSRQ), channel quality indication (channel quality indication, CQI).
  • the terminal device further updates the second AI model to an AI model associated with the channel quality information based on the channel quality information.
  • the terminal device updates the attribute information of the AI model associated with the channel quality information based on the channel quality information. It can be understood that the terminal device adjusts or fine-tunes the attribute information of the AI model associated with the channel quality information based on the channel quality information. Thereby, the terminal device retrains the AI model to ensure that the updated AI model matches the channel quality information.
  • the terminal device further selects target channel quality information that is greater than or equal to the reporting threshold from the channel quality information, and determines the reference signal information corresponding to the target channel quality information.
  • the terminal device may update the attribute information of the AI model that is associated with the channel quality information based on the target channel quality information.
  • the terminal device may report the target channel quality information and the reference signal information corresponding to the target channel quality information to the network device, so that the network device is aware of the reference signal information used by the terminal device to retrain the AI model.
  • the network device may update the attribute information of the AI model that is associated with the channel quality information based on the target channel quality information, and implement retraining of the AI model by the network device to ensure that the updated AI model matches the channel quality information.
  • the terminal device when the terminal device receives the first signaling indicating the first model association information, it can negotiate with the network device to update the AI model of the terminal device, thereby helping to improve communication efficiency.
  • the terminal device may also perform steps 302 and 303, and update the first AI model according to the first information.
  • the process of the terminal device updating the first AI model according to the first information is similar to the process of updating the second AI model according to the first information, and will not be repeated here.
  • FIG. 4 is a flowchart of another AI model updating method provided in an embodiment of the present application.
  • the method may specifically include the following steps:
  • the network device sends a first signaling to the terminal device.
  • the terminal device receives the first signaling from the network device.
  • the first signaling is used to indicate attribute information of the first AI model.
  • the attribute information of the first AI model indicated by the first signaling may be the attribute information of the AI model expected by the network device, that is, the attribute information of the AI model that the network device expects the terminal device to use.
  • the attribute information of the first AI model may include all model parameters of the AI model expected by the network device, or some model parameters of the AI model expected by the network device.
  • model parameters of the first AI model include model parameter values arranged in a preset order, such as neuron parameters arranged in the order of neurons in each layer.
  • the attribute information of the first AI model indicated by the first signaling is different from the attribute information of the second AI model currently used. That is, the network device sends the first signaling to the terminal device when learning the attribute information of the second AI model currently used by the terminal.
  • the model structure of the first AI model is different from the model structure of the second AI model.
  • the attribute information of the first AI model indicated by the first signaling is associated with the terminal device information of the terminal device.
  • the first signaling is a broadcast message
  • the broadcast message includes the attribute information of the AI model associated with the device vendor identification information of the terminal device.
  • step 401 indicates the attribute information of the first AI model; while the first signaling in step 301 indicates the first model association information.
  • the attribute information of the first AI model is different from the attribute information of the second AI model currently in use, and the second AI model is updated to the first AI model.
  • the terminal device can determine whether the attribute information of the first AI model is the same as the attribute information of the second AI model currently in use. If at least one type of attribute information in the attribute information of the first AI model and the attribute information of the second AI model is different, it can be determined that the attribute information of the first AI model is different from the attribute information of the second AI model.
  • the process of updating the second AI model to the first AI model in step 402 may refer to the detailed description in step 202, which will not be repeated here.
  • the terminal device when the terminal device receives the first signaling indicating the attribute information of the first AI model, it can update the currently used second AI model to the first AI model, so that the terminal device can quickly update the AI model, thereby helping to improve communication efficiency.
  • this embodiment A may include:
  • the network device sends a first signaling to the terminal device.
  • the terminal device receives the first signaling from the network device.
  • the first signaling is used to indicate attribute information of the first AI model.
  • the terminal device sends a second request message to the network device.
  • the network device receives the second request message from the terminal device.
  • the network device sends a second response message to the terminal device.
  • the terminal device receives the second response message from the network device.
  • the first signaling is used to indicate the attribute information of the first AI model, while in the embodiment shown in FIG3 , the first signaling is used to indicate the first model association information.
  • the network device determines the first information according to the second request message.
  • the network device determines the first information according to the capability information of the terminal device.
  • the capability information of the terminal device includes the attribute information of each AI model in the multiple AI models supported by the terminal device, which can be reported to the network device when accessing the network device.
  • the first information may include at least one of the following information: attribute information of the AI model supported by the terminal device (for example, it may be the attribute information of a certain AI model in multiple AI models), attribute information of the AI model expected by the network device, the specified AI model update duration, the reference signal information of the reference signal configured or sent by the network device, and the number of transmit beams supported by the network device.
  • the attribute information of the AI model supported by the terminal device is similar to the attribute information of the AI model recommended by the above-mentioned terminal device.
  • the rest of the information in the first information can refer to the specific description of this information in step 303, which will not be repeated here.
  • the terminal device sends a second request message to the network device after receiving the first signaling.
  • the terminal device sends a third request message to the network device before receiving the first signaling, and receives a third response message from the network device.
  • the third request message is similar to the second request message, and the third response message is similar to the second response message.
  • the second request message is sent after receiving the first signaling, and the third request message is sent before receiving the first signaling.
  • the terminal device can predict in advance that it will move to the first cell, then the terminal device can send a third request message to the network device before receiving the first signaling to switch to the first cell.
  • FIG 5 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application.
  • the communication device 50 can be a terminal device or a device matching a terminal device.
  • the communication device 50 includes a processing unit 501 and a communication unit 502.
  • the communication unit 502 is used to receive a first signaling from a network device, where the first signaling is used to indicate first model association information, and the first model association information is associated with attribute information of the first AI model; the processing unit 501, the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information is associated with attribute information of the second AI model.
  • the communication unit 502 is used to send a first signaling to the terminal device, where the first signaling is used to indicate first model association information, the first model association information is associated with the attribute information of the first AI model, and the first model association information is different from the second model association information currently used.
  • FIG. 6 is a schematic diagram of the structure of another communication device provided in an embodiment of the present application.
  • the communication device 60 can be a terminal device or a device matched with a terminal device.
  • the communication device may also include a memory 603.
  • the transceiver 601, the processor 602, and the memory 603 may be connected via a bus 604 or other methods.
  • the bus is represented by a bold line in Figure 6, and the connection method between other components is only for schematic illustration and is not limited.
  • the bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one bold line is used in Figure 6, but it does not mean that there is only one bus or one type of bus.
  • the coupling in the embodiment of the present application is an indirect coupling or communication connection between devices, units or modules, which can be electrical, mechanical or other forms, and is used for information exchange between devices, units or modules.
  • the specific connection medium between the above-mentioned transceiver 601, processor 602, and memory 603 is not limited in the embodiment of the present application.
  • the memory 603 may include a read-only memory and a random access memory, and provides instructions and data to the processor 602. A portion of the memory 603 may also include a nonvolatile random access memory.
  • the processor 602 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor, and optionally, the processor 602 may also be any conventional processor, etc.
  • the memory 603 is used to store program instructions; the processor 602 is used to call the program instructions stored in the memory 603 to execute the steps executed by the terminal device in the corresponding embodiments of Figures 2 to 4.
  • the method provided in the embodiment of the present application can be implemented by running a computer program (including program code) capable of executing each step involved in the above method on a general computing device such as a computer including a CPU, a random access memory (RAM), a read-only memory (ROM) and other processing elements and storage elements.
  • the computer program can be recorded on a computer-readable recording medium, for example, and loaded into the above computing device through the computer-readable recording medium and run therein.
  • the communication device 60 provided in the embodiment of the present application solves the problem and has the same beneficial effects as The principles and beneficial effects of solving the problems in the embodiments shown in Figures 2 to 4 of the present application are similar. Please refer to the principles and beneficial effects of the implementation of the method. For the sake of concise description, they will not be repeated here.
  • the aforementioned communication device may be, for example, a chip or a chip module.
  • An embodiment of the present application also provides a chip, which includes a processor, and the processor can execute the relevant steps of the terminal device in the aforementioned method embodiment.
  • the chip is used to: receive a first signaling from a network device, the first signaling is used to indicate first model association information, the first model association information is associated with attribute information of the first AI model; the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information is associated with attribute information of the second AI model.
  • the chip is specifically used when the first AI model is different from the second AI model, and the terminal device supports the first AI model, and the second AI model is updated to the first AI model.
  • the chip is also used for the first AI model being different from the second AI model and the terminal device not supporting the first AI model.
  • a target AI model is selected from the AI models supported by the terminal device; the target AI model satisfies a similarity condition; and the first AI model or the second AI model is updated to the target AI model.
  • the chip is also used for sending a first request message to a network device when the first AI model is different from the second AI model and the terminal device does not support the first AI model, the first request message is used to indicate attribute information of a third AI model; receiving a first response message from the network device; and updating the first AI model or the second AI model to a third AI model when the first response message indicates confirmation information.
  • the chip is also used for the first AI model being the same as the second AI model, updating the association relationship between the second model association information and the attribute information of the second AI model to the association relationship between the first model association information and the attribute information of the second AI model.
  • the chip is also used to deactivate the second AI model.
  • the chip is further used to send a second request message to the network device, where the second request message is used to determine the first information
  • a second response message is received from the network device, where the second response message includes the first information.
  • the chip is also used for the first information including attribute information of a fourth AI model; and updating the first AI model or the second AI model to the fourth AI model.
  • the chip is also used for the first information including first reference signal information; obtaining channel quality information based on the first reference signal information; updating the first AI model or the second AI model to an AI model associated with the channel quality information based on the channel quality information; and/or updating attribute information of the AI model associated with the channel quality information.
  • the chip is also used for the first information also including a reporting threshold; selecting target channel quality information greater than or equal to the reporting threshold from the channel quality information, and determining reference signal information corresponding to the target channel quality information.
  • the chip is also used to send target channel quality information and reference signal information corresponding to the target channel quality information to the network device.
  • the chip is also used to update attribute information of an AI model associated with the channel quality information based on the target channel quality information and reference signal information corresponding to the target channel quality information.
  • the chip is also used for, if the output result of the first AI model does not meet the preset condition, information, selecting a target AI model from the AI models supported by the terminal device, and updating the first AI model to the target AI model;
  • the state of the first AI model is a deactivated state, based on the first model association information, a target AI model is selected from the AI models supported by the terminal device, and the first AI model or the second AI model is updated to the target AI model;
  • the first AI model is different from the AI model indicated by the AI model indication information, and based on the first model association information, a target AI model is selected from the AI models supported by the terminal device, and the first AI model or the second AI model is updated to the target AI model;
  • the target AI model meets the similarity conditions.
  • the target AI model satisfies similarity conditions, including at least one of the following:
  • the difference between the model identifier of the first AI model and the model identifier of the target AI model is less than or equal to the first threshold;
  • the first model association information is identical to the model association information associated with the target AI model or the similarity is greater than a second threshold;
  • the target AI model is an AI model associated with the preset cell information; the preset cell information is included in the configured or activated cell information; or, the preset cell information is the cell information associated with the target TCI state, and the target TCI state is indicated or configured or activated by the second signaling;
  • the target AI model is the AI model that is used most times in the second preset time period before the second AI model is updated or the AI model that was used last time.
  • the chip is also used to receive a third signaling from a network device, and the second signaling is used to indicate an association relationship between the first model association information and the attribute information of the first AI model.
  • the chip is also used to determine the association relationship between the first model association information and the attribute information of the first AI model.
  • the association relationship between the first model association information and the attribute information of the first AI model is predefined by the protocol.
  • the first model association information includes at least one of the following: first PLMN information, first operator information, first network equipment vendor identification information, first TAI information, first SOR information, first URSP information, first operating frequency information, first cell information, reference signal information associated with the first cell information, first neighboring cell list, first timestamp information, and first terminal device information;
  • the first cell information includes at least one of the following: first cell group information, first service cell identification information, first physical cell identification, first TRP identification information, and first BWP identification.
  • the attribute information of the first AI model includes at least one of the following: a model identifier of the first AI model, model structure information of the first AI model, and a model parameter of the first AI model.
  • the chip is used to: send a first signaling to a terminal device, the first signaling is used to indicate first model association information, the first model association information is associated with attribute information of the first AI model, and the first model association information is different from the second model association information currently used.
  • the chip is also used to receive a first request message from a terminal device, the first request message is used to indicate attribute information of a third AI model; and a first response message is sent to the terminal device, the first response message is used to indicate confirmation information.
  • the chip is also used to receive a second request message from a terminal device, determine the first information based on the second request message; send a second response message to the terminal device, the second response message including the first information; or determine the first information based on the capability information of the terminal device; and send the first information to the terminal device.
  • the first information includes attribute information of the fourth AI model.
  • the first information includes first reference signal information, where the first reference signal information is reference signal information of a reference signal configured or sent by the network device, or reference signal information matching the purpose of the AI model recommended by the terminal device. Or reference signal information that is associated with attribute information of the AI model expected by the network device.
  • the chip is also used for the first information also including a reporting threshold; receiving target channel quality information and reference signal information corresponding to the target channel quality information from a terminal device; and updating attribute information of an AI model that has an association with the channel quality information based on the target channel quality information and the reference signal information corresponding to the target channel quality information.
  • the chip is also used to send a third signaling to the terminal device, where the third signaling is used to indicate the association relationship between the first model association information and the attribute information of the first AI model.
  • the association relationship between the first model association information and the attribute information of the first AI model is predefined by the protocol.
  • the first model association information includes at least one of the following: first public land mobile network PLMN information, first operator information, first network equipment vendor identification information, first tracking area identification TAI information, first roaming guide SOR information, first terminal routing selection strategy URSP information, first operating frequency information, first cell information, reference signal information associated with the first cell information, first neighboring cell list, first timestamp information, and first terminal device information;
  • the first cell information includes at least one of the following: first cell group information, first service cell identification information, first physical cell identification, first sending and receiving point TRP identification information, and first bandwidth part BWP identification.
  • the attribute information of the first AI model includes at least one of the following: a model identifier of the first AI model, model structure information of the first AI model, and a model parameter of the first AI model.
  • FIG7 is a schematic diagram of the structure of a chip module provided in an embodiment of the present application.
  • the chip module 70 can execute the relevant steps of the terminal device in the aforementioned method embodiment, and the chip module 70 includes: a communication interface 701 and a chip 702 .
  • the communication interface is used for internal communication of the chip module, or for the chip module to communicate with an external device.
  • the communication interface can also be described as a communication module.
  • the chip 702 is used to implement the functions of the terminal device in the embodiment of the present application.
  • chip 702 is used to receive a first signaling from a network device, where the first signaling is used to indicate first model association information, where the first model association information is associated with attribute information of the first AI model; the first model association information is different from the second model association information currently in use, and the second AI model currently in use is updated to the first AI model; wherein the second model association information is associated with attribute information of the second AI model.
  • chip 702 is used to receive a first signaling from a network device, where the first signaling is used to indicate attribute information of a first AI model; the attribute information of the first AI model is different from the attribute information of a second AI model currently in use, and the second AI model is updated to the first AI model.
  • the chip module 70 may further include a storage module 703 and a power module 704.
  • the storage module 703 is used to store data and instructions.
  • the power module 704 is used to provide power to the chip module.
  • each module contained therein can be implemented by hardware such as circuits, and different modules can be located in the same component of the chip module (such as a chip, circuit module, etc.) or in different components.
  • the modules can be implemented by a software program that runs on a processor integrated inside the chip module, and the remaining (if any) modules can be implemented by hardware such as circuits.
  • An embodiment of the present application also provides a computer-readable storage medium, in which one or more instructions are stored, and the one or more instructions are suitable for being loaded by a processor and executing the method provided by the above method embodiment.
  • the embodiment of the present application also provides a computer program product including a computer program or instructions.
  • the computer program or instructions When the computer program or instructions are executed on a computer, the computer executes the method provided by the above method embodiment.
  • the steps of the method or algorithm described in the embodiments of the present application can be implemented in hardware or by executing software instructions by a processor.
  • the software instructions can be composed of corresponding software modules, and the software modules can be stored in RAM, flash memory, ROM, erasable programmable ROM (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art.
  • An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium.
  • the storage medium can also be a component of the processor.
  • the processor and the storage medium can be located in an ASIC.
  • the ASIC can be located in a terminal device or a management device.
  • the processor and the storage medium can also be present in a terminal device or a management device as discrete components.
  • the functions described in the embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof.
  • software When implemented using software, it can be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.
  • the computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the computer instructions may be transmitted from a website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.
  • the computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated.
  • the available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a digital video disc (DVD)), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.
  • a magnetic medium e.g., a floppy disk, a hard disk, a magnetic tape
  • an optical medium e.g., a digital video disc (DVD)
  • DVD digital video disc
  • SSD solid state disk
  • the various modules/units included in the various devices and products described in the above embodiments can be software modules/units, or hardware modules/units, or they can be partially software modules/units and partially hardware modules/units.
  • the various modules/units included therein can all be implemented in the form of hardware such as circuits, or at least some of the modules/units can be implemented in the form of software programs, which run on a processor integrated inside the chip, and the remaining (if any) modules/units can be implemented in the form of hardware such as circuits;
  • the various modules/units included therein can all be implemented in the form of hardware such as circuits, and different modules/units can be located in the same component of the chip module (such as a chip, circuit module, etc.) or in different components, or at least some of the modules/units can be implemented in the form of software programs, which run on a processor integrated inside the chip module
  • Circuits and other hardware means for each device or product applied to or integrated in the terminal, each module/unit contained therein can be implemented by circuits and other hardware means, different modules/units can be located in the same component (for example, a chip, circuit module, etc.) or in different components in the terminal, or, at least some modules/units can be implemented by software programs, which run on a processor integrated in the terminal, and the remaining (if any) modules/units can be implemented by circuits and other hardware means.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请提供一种AI模型更新方法及通信装置，该方法可包括：接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型。采用本申请，网络设备可通过信令触发终端设备更新AI模型，使得终端设备可以快速地更新AI模型，从而有助于提高通信效率。

Description

AI模型更新方法及通信装置
本申请要求于2022年11月3日提交中国专利局、申请号为202211372492.8、申请名称为“AI模型更新方法及通信装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及通信技术领域，尤其涉及一种人工智能(artificial intelligence，AI)模型更新的方法及通信装置。
背景技术
随着科学技术的发展，AI技术应运而生。AI技术可被广泛地应用在各个领域。将AI技术应用在无线通信技术领域中，可达到提升无线通信系统的吞吐量，有效缩短处理时延，以及增强用户容量等优势，这是未来无线通信技术的重要任务。
对一个设备，其AI功能可通过部署的AI模型实现。目前，在第三代合作伙伴计划(3rd-generation partnership project，3GPP)标准化中，支持在终端设备上部署AI模型，以实现AI模型应用在无线通信系统中。如何触发终端设备的AI模型更新是亟待解决的技术问题。
发明内容
本申请实施例提供一种AI模型更新方法，可以快速地更新终端设备的AI模型，从而有助于提高通信效率。
第一方面，本申请实施例提供一种AI模型更新方法，该方法可由终端设备执行，或由与终端设备匹配的装置执行，例如芯片、芯片模组或处理器等。该方法可包括：接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。
可见，终端设备在接收到指示第一模型关联信息的第一信令时，可将当前使用的第二AI模型更新为与第一模型关联信息具有关联关系的第一AI模型，使得终端设备可以快速地更新AI模型，从而有助于提高通信效率。
在一种可能的实现方式中，在将当前使用的第二AI模型更新为第一AI模型时，具体可在第一AI模型不同于第二AI模型，且终端设备支持第一AI模型的情况下，将当前使用的第二AI模型更新为第一AI模型。从而在移动场景或网络切换场景下，可将终端设备当前使用的第二AI模型更新为，与第一模型关联信息具有关联关系的第一AI模型，有助于提高通信质量和通信效率。
进一步的，在第一AI模型不同于第二AI模型，且终端设备支持第一AI模型，且第一AI模型的状态不为去激活状态的情况下，将当前使用的第二AI模型更新为第一AI模型。也就是说，在第一AI模型的状态为去激活状态的情况下，不会将当前使用的第二AI模型更新为第一AI模型。
在一种可能的实现方式中，在第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型的情况下，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，目标AI模型满足相似度条件；将第一AI模型或第二AI模型更新为目标AI模型。在终端设备不支持第一AI模型的情况下，可以选择与第一AI模型的属性信息最接近的目标AI模型，在一定程度上可以提高通信效率。
在一种可能的实现方式中，在第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型的情况下，向网络设备发送第一请求消息，第一请求消息用于指示第三AI模型的属性信息。也就是说，在终端设备不支持第一AI模型的情况下，终端设备向网络设备请求使用第三AI模型。接收来自网络设备的第一响应消息，在第一响应消息指示确认信息的情况下，将第一AI模型或第二AI模型更新为第三AI模型。反之，在第一响应消息指示拒绝信息的情况下，可停止AI模型更新，或者可再次向网络设备发送第一请求消息，此时第一请求消息可指示终端设备请求使用另外的AI模型的属性信息(也就是说另外的AI模型不同于第三AI模型)，等待接收来自网络设备的确认信息或拒绝信息。
在一种可能的实现方式中，在第一AI模型与第二AI模型相同的情况下，将当前模型关联信息与第二AI模型的属性信息之间的关联关系，更新为第一模型关联信息与第二AI模型的属性信息之间的关联关系。从而可以实时更新模型关联信息与AI模型的属性信息之间的关联关系。
在一种可能的实现方式中，在将当前使用的第二AI模型更新为第一AI模型之后，可去激活第二AI模型，以避免同时使用2个或2个以上的AI模型，减少终端设备的处理负荷。将第二AI模型更新为第一AI模型，隐含表示激活并使用第一AI模型。
在一种可能的实现方式中，上述方法还包括：向网络设备发送第二请求消息，第二请求消息用于确定第一信息；接收来自网络设备的第二响应消息，第二响应消息包括第一信息。
在一种可能的实现方式中，第一信息包括第四AI模型的属性信息；上述方法还包括：将第一AI模型或第二AI模型更新为第四AI模型。
在一种可能的实现方式中，第一信息包括第一参考信号信息；上述方法还包括：根据第一参考信号信息，获得信道质量信息；根据信道质量信息，将第一AI模型或第二AI模型更新为与信道质量信息具有关联关系的AI模型，和/或，更新与信道质量信息具有关联关系的AI模型的属性信息。从而实现终端设备通过对AI模型进行重训练，保证更新后的AI模型与信道质量信息相匹配。
在一种可能的实现方式中，第一信息还包括上报阈值；上述方法还包括：从信道质量信息中选择大于或等于上报阈值的目标信道质量信息，并确定目标信道质量信息对应的参考信号信息。
在一种可能的实现方式中，上述方法还包括：向网络设备发送目标信道质量信息和目标信道质量信息对应的参考信号信息。以便网络设备获知终端设备对AI模型进行重训练所使用的参考信号信息。
在一种可能的实现方式中，上述方法还包括：根据目标信道质量信息和目标信道质量信息对应的参考信号信息，更新与信道质量信息具有关联关系的AI模型的属性信息。从而 实现终端设备通过对AI模型进行重训练，保证更新后的AI模型与信道质量信息相匹配。
在一种可能的实现方式中，上述方法还包括：若第一AI模型的输出结果不满足预设条件，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，在第一AI模型的输出结果不满足性能需求时，可将第一AI模型更新为目标AI模型。
在一种可能的实现方式中，上述方法还包括：在第一AI模型的状态为去激活状态的情况下，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，第一AI模型被去激活时，可将第一AI模型或第二AI模型更新为目标AI模型。
在一种可能的实现方式中，上述方法还包括：第一AI模型与AI模型指示信息指示的AI模型不同，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，第一AI模型与网络设备指示的AI模型不同时，可将第一AI模型或第二AI模型更新为目标AI模型。
在一种可能的实现方式中，上述目标AI模型满足相似度条件，包括以下至少一项：
第一AI模型的模型标识与目标AI模型的模型标识之差小于或等于第一阈值；
第一模型关联信息与目标AI模型关联的模型关联信息相同或相似度大于第二阈值；
目标AI模型为与预设小区信息关联的AI模型；预设小区信息包含于配置的或激活的小区信息中；或，预设小区信息为目标传输配置指示TCI状态关联的小区信息，目标TCI状态由第二信令指示或配置或激活；可选的，第二信令的接收时间与当前系统时间之间的时间差小于第三阈值；
目标AI模型为更新第二AI模型之前的第二预设时间段内使用次数最多的AI模型或上一次使用的AI模型。
在一种可能的实现方式中，上述方法还包括：接收来自网络设备的第三信令，第三信令用于指示第一模型关联信息与第一AI模型的属性信息之间的关联关系。也就是说，网络设备可指示各个模型关联信息与各个AI模型的属性信息之间的关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在一种可能的实现方式中，上述方法还包括：确定第一模型关联信息与第一AI模型的属性信息之间的关联关系。也就是说，终端设备维护各个模型关联信息与各个AI模型的属性信息之间的关联关系，可提高终端设备的能动性。
在一种可能的实现方式中，第一模型关联信息与第一AI模型的属性信息之间的关联关系是协议预定义的。也就是说，协议预定义各个模型关联信息与各个AI模型的属性信息之间的关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在一种可能的实现方式中，第一模型关联信息可以包括但不限于如下至少一项信息：第一运营商标识信息、第一公共陆地移动网络(public land mobile network，PLMN)信息、第一网络设备商标识信息、第一跟踪区标识(tracking area identifier，TAI)信息、第一漫游引导(steering of roaming，SOR)信息、第一终端路由选择策略(user  equipment route selection policy，URSP)信息、第一工作频率信息、第一小区信息、第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
其中，第一小区信息包括但不限于如下至少一项信息：第一小区组信息、第一服务小区标识信息、第一物理小区标识信息、第一发送接收点(transmission and reception point，TRP)标识信息、第一带宽部分(bandwidth part，BWP)标识。
在一种可能的实现方式中，第一AI模型的属性信息包括但不限于如下至少一项信息：第一AI模型的模型标识、第一AI模型的模型结构信息、第一AI模型的模型参数。
第二方面，本申请实施例提供一种AI模型更新方法，该方法可由网络设备执行，或由与网络设备匹配的装置执行，例如芯片、芯片模组或处理器等。该方法可包括：向终端设备发送第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系，第一模型关联信息和当前使用的第二模型关联信息不同。
在一种可能的实现方式中，上述方法还包括：接收来自终端设备的第一请求消息，第一请求消息用于指示第三AI模型的属性信息；向终端设备发送的第一响应消息，第一响应消息用于指示确认信息。
在一种可能的实现方式中，上述方法还包括：接收来自终端设备的第二请求消息，根据第二请求消息确定第一信息；向终端设备发送第二响应消息，第二响应消息包括第一信息；
或根据终端设备的能力信息，确定第一信息；向终端设备发送第一信息。
在一种可能的实现方式中，第一信息包括第四AI模型的属性信息。
在一种可能的实现方式中，第一信息包括第一参考信号信息，第一参考信号信息为网络设备配置的或发送的参考信号的参考信号信息，或与终端设备建议的AI模型的用途匹配的参考信号信息，或与网络设备期望的AI模型的属性信息具有关联关系的参考信号信息。
在一种可能的实现方式中，第一信息还包括上报阈值；上述方法还包括：接收来自终端设备的目标信道质量信息和目标信道质量信息对应的参考信号信息，从而网络设备可以获知终端设备对AI模型进行重训练所使用的参考信号信息。可选的，根据目标信道质量信息和目标信道质量信息对应的参考信号信息，更新与信道质量信息具有关联关系的AI模型的属性信息，实现网络设备对AI模型的重训练，以保证更新后的AI模型与信道质量信息相匹配。
在一种可能的实现方式中，上述方法还包括：向终端设备发送第三信令，第三信令用于指示第一模型关联信息与第一AI模型的属性信息的关联关系。也就是说，网络设备可指示各个模型关联信息与各个AI模型的属性信息之间的关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在一种可能的实现方式中，第一模型关联信息与第一AI模型的属性信息的关联关系是协议预定义的。也就是说，协议预定义各个模型关联信息与各个AI模型的属性信息之间的关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在一种可能的实现方式中，第一模型关联信息可以包括但不限于如下至少一项信息：第一运营商标识信息、第一PLMN信息、第一网络设备商标识信息、第一TAI信息、第一 SOR信息、第一URSP信息、第一工作频率信息、第一小区信息、第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
其中，第一小区信息包括但不限于如下至少一项信息：第一小区组信息、第一服务小区标识信息、第一物理小区标识信息、第一TRP标识信息、第一BWP标识。
在一种可能的实现方式中，第一AI模型的属性信息包括但不限于如下至少一项信息：第一AI模型的模型标识、第一AI模型的模型结构信息、第一AI模型的模型参数。
第三方面，本申请实施例提供一种AI模型更新方法，该方法可由终端设备执行，或由与终端设备匹配的装置执行，例如芯片、芯片模组或处理器等。该方法可包括：接收来自网络设备的第一信令，第一信令用于指示第一AI模型的属性信息；第一AI模型的属性信息与当前使用的第二AI模型的属性信息不同，将第二AI模型更新为第一AI模型。
可见，终端设备在接收到指示第一AI模型的属性信息的第一信令时，可将当前使用的第二AI模型更新为与第一AI模型，使得终端设备可以快速地更新AI模型，从而有助于提高通信效率。
在一种可能的实现方式中，第一AI模型的属性信息为网络设备期望的AI模型的属性信息。
在一种可能的实现方式中，第一AI模型的属性信息，不同于当前使用的第二AI模型的属性信息。
在一种可能的实现方式中，第一AI模型的属性信息，与终端设备的终端设备信息具有关联关系。
在一种可能的实现方式中，在将第二AI模型更新为第一AI模型时，具体可在终端设备支持第一AI模型的情况下，将当前使用的第二AI模型更新为第一AI模型。从而在移动场景或网络切换场景下，可将终端设备当前使用的第二AI模型更新为，第一信令指示的第一AI模型，有助于提高通信质量和通信效率。
进一步的，在终端设备支持第一AI模型，且第一AI模型的状态不为去激活状态的情况下，将第二AI模型更新为第一AI模型。也就是说，在第一AI模型的状态为去激活状态的情况下，不会将第二AI模型更新为第一AI模型。
在一种可能的实现方式中，在终端设备不支持第一AI模型的情况下，从终端设备支持的AI模型中选择目标AI模型，目标AI模型满足相似度条件；将第一AI模型或第二AI模型更新为目标AI模型。在终端设备不支持第一AI模型的情况下，可以选择与第一AI模型的属性信息最接近的目标AI模型，在一定程度上可以提高通信效率。
在一种可能的实现方式中，在终端设备不支持第一AI模型的情况下，向网络设备发送第一请求消息，第一请求消息用于指示第三AI模型的属性信息。也就是说，在终端设备不支持第一AI模型的情况下，终端设备向网络设备请求使用第三AI模型。接收来自网络设备的第一响应消息，在第一响应消息指示确认信息的情况下，将第一AI模型或第二AI模型更新为第三AI模型。反之，在第一响应消息指示拒绝信息的情况下，可停止AI模型更新，或者可再次向网络设备发送第一请求消息，此时第一请求消息可指示终端设备请求使用另外的AI模型的属性信息(也就是说另外的AI模型不同于第三AI模型)，等待接收来自网络设备的确认信息或拒绝信息。
在一种可能的实现方式中，在将第二AI模型更新为第一AI模型之后，可去激活第二AI模型，以避免同时使用2个或2个以上的AI模型，减少终端设备的处理负荷。将第二AI模型更新为第一AI模型，隐含表示激活并使用第一AI模型。
在一种可能的实现方式中，上述方法还包括：向网络设备发送第二请求消息，第二请求消息用于确定第一信息；接收来自网络设备的第二响应消息，第二响应消息包括第一信息。
在一种可能的实现方式中，第一信息包括第四AI模型的属性信息；上述方法还包括：将第一AI模型或第二AI模型更新为第四AI模型。
在一种可能的实现方式中，第一信息包括第一参考信号信息；上述方法还包括：根据第一参考信号信息，获得信道质量信息；根据信道质量信息，将第一AI模型或第二AI模型更新为与信道质量信息具有关联关系的AI模型，和/或，更新与信道质量信息具有关联关系的AI模型的属性信息。从而实现终端设备通过对AI模型进行重训练，保证更新后的AI模型与信道质量信息相匹配。
在一种可能的实现方式中，第一信息还包括上报阈值；上述方法还包括：从信道质量信息中选择大于或等于上报阈值的目标信道质量信息，并确定目标信道质量信息对应的参考信号信息。
在一种可能的实现方式中，上述方法还包括：向网络设备发送目标信道质量信息和目标信道质量信息对应的参考信号信息。以便网络设备获知终端设备对AI模型进行重训练所使用的参考信号信息。
在一种可能的实现方式中，上述方法还包括：根据目标信道质量信息和目标信道质量信息对应的参考信号信息，更新与信道质量信息具有关联关系的AI模型的属性信息。从而实现终端设备通过对AI模型进行重训练，保证更新后的AI模型与信道质量信息相匹配。
在一种可能的实现方式中，上述方法还包括：若第一AI模型的输出结果不满足预设条件，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，在第一AI模型的输出结果不满足性能需求时，可将第一AI模型更新为目标AI模型。
在一种可能的实现方式中，上述方法还包括：在第一AI模型的状态为去激活状态的情况下，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，第一AI模型被去激活时，可将第一AI模型或第二AI模型更新为目标AI模型。
在一种可能的实现方式中，上述方法还包括：第一AI模型与AI模型指示信息指示的AI模型不同，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型。其中，目标AI模型满足相似度条件。也就是说，第一AI模型与网络设备指示的AI模型不同时，可将第一AI模型或第二AI模型更新为目标AI模型。
在一种可能的实现方式中，上述目标AI模型满足相似度条件，包括以下至少一项：
第一AI模型的模型标识与目标AI模型的模型标识之差小于或等于第一阈值；
目标AI模型为与预设小区信息关联的AI模型；预设小区信息包含于配置的或激活的 小区信息中；或，预设小区信息为目标传输配置指示TCI状态关联的小区信息，目标TCI状态由第二信令指示或配置或激活；可选的，第二信令的接收时间与当前系统时间之间的时间差小于第三阈值；
目标AI模型为更新第二AI模型之前的第二预设时间段内使用次数最多的AI模型或上一次使用的AI模型。
第四方面，本申请实施例提供一种AI模型更新方法，该方法可由网络设备执行，或由与网络设备匹配的装置执行，例如芯片、芯片模组或处理器等。该方法可包括：向终端设备发送第一信令，第一信令用于指示第一AI模型的属性信息。
第五方面，本申请实施例提供一种通信装置，该通信装置包括通信单元和处理单元。在一种方式中，通信单元，用于接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；处理单元，第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。在另一种方式中，通信单元，用于向终端设备发送第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系，第一模型关联信息和当前使用的第二模型关联信息不同。
第六方面，本申请提供一种通信装置，该装置包括处理器、存储器及存储在存储器上的计算机程序或指令，其特征在于，处理器执行计算机程序或指令以实现如第一方面至第四方面中任一方面提供的方法。
第七方面，本申请提供一种芯片。在一种实现方式中，该芯片用于接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。在另一种实现方式中，该芯片用于向终端设备发送第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系，第一模型关联信息和当前使用的第二模型关联信息不同。
第八方面，本申请提供一种计算机可读存储介质，该计算机存储介质中存储有计算机可读指令，当该计算机可读指令在计算机上运行时，使得该通信装置执行上述第一方面至第四方面中任一方面提供的方法。
第九方面，本申请提供一种计算机程序或计算机程序产品，包括代码或指令，当代码或指令在计算机上运行时，使得计算机执行如第一方面至第四方面中任一方面提供的方法。
第十方面，本申请提供一种芯片模组，芯片模组包括通信模组、电源模组、存储模组以及芯片，其中：所述电源模组用于为所述芯片模组提供电能；所述存储模组用于存储数据和指令；所述通信模组用于进行芯片模组内部通信，或者用于所述芯片模组与外部设备进行通信；所述芯片用于执行如第一方面至第四方面中任一方面提供的方法。
附图说明
图1是应用本申请的一种通信系统的架构示意图；
图2是本申请实施例提供的一种AI模型更新方法的流程示意图；
图3是本申请实施例提供的另一种AI模型更新方法的流程示意图；
图4是本申请实施例提供的又一种AI模型更新方法的流程示意图；
图5是本申请提供的一种通信装置的结构示意图；
图6是本申请提供的另一种通信装置的结构示意图；
图7是本申请实施例提供的一种芯片模组的结构示意图。
具体实施方式
在本申请中，“第一”、“第二”等字样用于对功能和作用基本相同的相同项或相似项进行区分。本领域技术人员可以理解“第一”、“第二”等字样并不对数量和执行次序进行限定，并且“第一”、“第二”等字样也并不限定一定不同。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。字符“/”一般表示前后关联对象是一种“或”的关系。
应当理解，本申请中，“至少一个”指的是一个或多个；“多个”是指两个或两个以上。此外，本申请的“等于”可以与“大于”连用，也可以与“小于”连用。在“等于”与“大于”连用的情况下，采用“大于”的技术方案；在“等于”与“小于”连用的情况下，采用“小于”的技术方案。
首先，对本申请涉及的系统架构进行阐述。
本申请可应用于第五代(5th generation，5G)系统，也可以称为新空口(new radio，NR)系统；或者可应用于第六代(6th generation，6G)系统，或者第七代(7th generation，7G)系统，或未来的其他通信系统；或者还可用于设备到设备(device to device，D2D)系统，机器到机器(machine to machine，M2M)系统、车联网(vehicle to everything，V2X)等等。
本申请可应用于图1所示的系统架构中。图1所示的通信系统10可包括但不限于：网络设备110和终端设备120。图1中设备的数量和形态用于举例，并不构成对本申请实施例的限定，例如实际应用中可以包括多个终端设备。
终端设备，又称之为用户设备(user equipment，UE)、移动台(mobile station，MS)、移动终端(mobile terminal，MT)等，是指向用户提供语音和/或数据连通性的设备。例如，具有无线连接功能的手持式设备、车载设备等。目前，一些终端设备的举例为：手机(mobile phone)、平板电脑、笔记本电脑、掌上电脑、移动互联网设备(mobile internet device，MID)、可穿戴设备，虚拟现实(virtual reality，VR)设备、增强现实(augmented reality，AR)设备、工业控制(industrial control)中的无线终端、无人驾驶(self driving)中的无线终端、远程手术(remote medical surgery)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端、智慧家庭(smart home)中的无线终端等。
在本申请中，用于实现终端设备的功能的装置可以是终端设备；也可以是能够支持终端设备实现该功能的装置，例如芯片或芯片模组等，该装置可以被安装在终端设备中或者和终端设备匹配使用。在本申请提供的技术方案中，以用于实现终端设备的功能的装置是终端设备为例，描述本申请提供的技术方案。本申请实施例涉及的终端设备具有AI模型，或描述为部署有AI模型。终端设备支持的AI模型数量在本申请不作限定，视终端设备出 厂设置而定或视实际情况而定。
在一种实现方式中，网络设备可以是接入网设备，是指将终端设备接入到无线网络的无线接入网(radio access network，RAN)节点(或设备)，又可以称为基站。目前，一些RAN节点的举例为：继续演进的节点B(gNB)、发送接收点(transmission reception point，TRP)、演进型节点B(evolved Node B，eNB)、无线网络控制器(radio network controller，RNC)、节点B(Node B，NB)、基站控制器(base station controller，BSC)、基站收发台(base transceiver station，BTS)、家庭基站(例如，home evolved NodeB，或home Node B，HNB)、基带单元(base band unit，BBU)，或无线保真(wireless fidelity，Wifi)接入点(access point，AP)等。另外，在一种网络结构中，网络设备可以包括集中单元(centralized unit，CU)节点、或分布单元(distributed unit，DU)节点、或包括CU节点和DU节点的RAN设备。需要说明的是，集中单元节点、分布单元节点还可能采用其他名称，本申请并不限定。
在另一种实现方式中，网络设备可以是核心网设备。核心网设备主要负责对整个网络进行管理和控制，负责承上启下，对收到的数据进行分拣，并将其传输到相应的终端设备或数据网络(data network，DN)。核心网设备也可以称为核心网节点。在本申请中，核心网设备例如可以是网络数据分析功能(network data analytics function，NWDAF)节点或定位管理功能(location management function，LMF)节点等。NWDAF节点，用于根据网络服务的请求数据提供网络分析服务。LMF节点，用于对终端设备的位置所需的资源进行总体协调和调度，还用于提供定位服务。
在又一种实现方式中，网络设备可以是神经网络处理节点，是指对AI模型提供服务的节点，例如为AI模型提供模型参数。
在本申请中，用于实现网络设备的功能的装置可以是网络设备；也可以是能够支持网络设备实现该功能的装置，例如芯片或芯片模组等，该装置可以被安装在网络设备中或者和网络设备匹配使用。在本申请提供的技术方案中，以用于实现网络设备的功能的装置是网络设备为例，描述本申请提供的技术方案。
可以理解的是，本申请实施例描述的通信系统是为了更加清楚的说明本申请实施例的技术方案，并不构成对本申请实施例提供的技术方案的限定，本领域技术人员可知，随着系统架构的演变和新业务场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。
其次，对本申请涉及的相关名称或术语进行阐述，以便于本领域技术人员理解。
1、AI模型
AI模型可用于实现AI功能，也可以描述为机器学习模型。AI模型可通过如下至少一种方式实现，例如神经网络、决策树、支持向量机、贝叶斯分类器等。
以神经网络为例，神经网络由神经元组成，其中包括：输入参数、乘性系数、加性系数、激活函数等。常见的激活函数包括：Sigmoid函数、双曲线函数(tanh函数)、线性整流单元(rectified linear unit，ReLU)函数等等。
神经网络的参数通过梯度优化算法进行优化，梯度优化算法是一类最小化或者最大化 目标函数(也可称为损失函数)的算法，而目标函数往往是模型参数和数据的数学组合。例如给定数据X和其对应的标签Y，我们构建一个神经网络模型f，有了模型后，根据输入就可以预测出输出信息，并且可以计算出预测值和真实值之间的差距，这就是损失函数。构建神经网络模型的目的是使损失函数的值尽可能达到最小。损失值越小，则说明神经网络模型越接近于真实情况。
误差反向传播(errorBackPropagation，BP)算法是一种常见的优化算法，用于对AI模型进行优化。BP算法的基本思想是，学习过程由信号的正向传播与误差的反向传播两个过程组成。正向传播时，输入样本从输入层传入，经各个隐层逐层处理后传向输出层。若输出层的实际输出与期望的输出不符，则转入误差的反向传播阶段。为了描述方便，将误差的反向传播阶段简称为误差反传。误差反传是将输出误差以某种形式通过隐层向输入层逐层反传，并将误差分摊给各层的所有单元，从而获得各层单元的误差信号，此误差信号即作为修正各单元权值的依据。这种信号正向传播与误差反传的各层权值调整过程，是周而复始的进行着。权值不断调整的过程，也就是AI模型的学习训练过程。此过程一直进行到AI模型输出的误差减少到可接受的程度，或进行到预先设定的学习次数为止。
对AI模型进行优化的优化算法还可以是梯度下降(Gradient Descent)算法、随机梯度下降(Stochatic Gradient Descent，SGD)算法、小批量梯度下降(mini-batch Gradient Descent)算法、动量法(Momentum)、带动量的随机梯度下降(Nesteroy)算法、自适应梯度下降(adaptive Gradient Descent，Adagrad)算法、均方根误差降速(Root Mean Square Prop)算法、自适应动量估计(Adaptive Moment Estimation，Adam)算法等。
上述优化算法在误差反传时，根据损失函数得到的误差，对当前神经元求导数/偏导，加上学习速率、之前的梯度/导数/偏导等影响，得到梯度，将梯度传给上一层。
本申请实施例涉及的AI模型，指的是应用在无线通信技术领域中的AI模型，可用于提升无线通信系统的吞吐量、缩短处理时延、以及增强用户容量等。本申请实施例涉及的AI模型，指的是终端设备的AI模型，从而可以提升终端设备的通信能力。
(1)AI模型的属性信息
AI模型的属性信息用于描述AI模型的标识信息、AI模型的模型结构信息、AI模型的模型参数中的至少一项。换言之，AI模型的属性信息可以包括如下至少一项：AI模型的标识信息、AI模型的模型结构信息、AI模型的模型参数。
其中，标识信息用于标识AI模型，例如可以是模型标识(identifier，ID)，不同的模型ID标识不同的AI模型，不同的AI模型具有不同的模型ID。模型结构信息用于描述模型结构，例如可以是神经网络、决策树、或是模型中的神经元和层等结构信息。不同的模型结构，可以认为不同的AI模型。模型参数可以包括至少一种参数，例如包括激活函数、乘性系数、加性系数等。不同的模型参数，可以认为不同的AI模型。例如两个AI模型除激活函数不同，其他模型参数均相同，也认为这两个AI模型为不同的AI模型。
需要说明的是，在本申请实施例中，AI模型的属性信息这个名称用于举例，也可以描述为AI模型的模型信息、AI模型的特征信息或AI模型的基本信息等。
2、模型关联信息
在本申请实施例中，模型关联信息用于描述与AI模型具有关联关系的信息，或用于描 述与AI模型的属性信息具有关联关系的信息。模型关联信息的改变，可能会导致AI模型的变化。模型关联信息与AI模型之间的关联关系可以是一对一的关系，即一种模型关联信息关联一种AI模型；也可以是多对一的关系，即多种模型关联信息可以关联一个AI模型；也可以是一对多的关系，即一种模型关联信息可以关联多种AI模型。
模型关联信息可以包括但不限于如下至少一项信息：PLMN信息、运营商标识信息、网络设备商标识信息、TAI信息、SOR信息、URSP信息、工作频率信息、小区信息、小区信息关联的参考信号信息、邻区列表、时间戳信息、终端设备信息。
1)PLMN信息用于描述某个地区的某个运营商的某种制式的蜂窝移动通信网络。例如PLMN信息用于描述终端设备的归属地的运营商1的5G制式网络。再例如，PLMN信息用于描述终端设备的拜访地的运营商2的5G制式网络。
2)运营商标识信息用于标识运营商，例如可以是终端设备签约的运营商，或终端设备当前所属的运营商等。
3)网络设备商标识信息用于标识网络设备商，例如可以是基站的制造商或核心网设备制造商等。
4)TAI信息用于标识TAI，可表示为TAI＝PLMN+跟踪区编码(tracking area code，TAC)。跟踪区(tracking area，TA)是为终端设备的位置管理设立的概念。当终端设备处于空闲状态时，核心网设备能够知道终端设备所在的TA，同时寻呼处于空闲状态的终端设备时，在终端设备所注册的TA的所有小区进行寻呼。TA是小区级的配置，多个小区可以配置相同的TA，且一个小区属于一个TA。
5)SOR信息用于归属地PLMN能够将终端设备从一个网络引导到另一个网络。SOR是一种通过归属地PLMN来鼓励漫游终端设备漫游到偏好漫游网络的技术。例如，终端设备在一个PLMN上被注册，并且由于某种原因，终端设备的归属地PLMN希望终端设备在另一个PLMN上注册。
6)URSP信息用于描述终端设备的路由选择策略。该策略描述终端设备上的应用与网络切片之间的对应关系，终端设备根据URSP可以为应用选择网络切片。
7)工作频率信息用于描述终端设备将使用的或正在使用的或可使用的频率，可以包括但不限于频率编号、频段编号等中的至少一项。
8)小区信息用于描述终端设备的服务小区信息，可以包括但不限于如下至少一项信息：小区组信息、服务小区标识信息、物理小区标识(physical cell identifier，PCI)、TRP标识信息、BWP标识。小区组信息可以包括小区组标识(cell group ID)，用于标识服务小区所属的小区组。服务小区标识信息可以包括服务小区标识(serving cell ID)，用于标识服务小区。PCI，用于标识服务小区的逻辑小区。TRP标识信息可以包括但不限于如下至少一项：控制资源集(control-resource set，CORESET)标识、控制资源集池(CORESET pool)标识、控制资源集的组标识、传输配置指示(transmission configuration indication，TCI)状态池标识、TCI状态标识、参考信号资源标识、参考信号资源集合标识。BWP标识，用于标识BWP。
9)小区信息关联的参考信号(reference signal，RS)信息，用于描述与小区信息关联的参考信号，可以包括但不限于参考信号资源的数量、参考信号资源的集合、参考信号资源 的索引、参考信号的样式(pattern)、参考信号的重复参数值(例如repetition off或repetition on，repetition off表示不开启参考信号的重复传输，repetition on表示开启参考信号的重复传输)。其中，参考信号指的是下行参考信号，例如可以是同步信号块(synchronization signal block，SSB)或信道状态信息参考信号(channel state information-reference signal，CSI-RS)。
10)邻区列表，用于描述服务小区的邻区。邻区列表可以是网络设备通过高层信令配置的。
11)时间戳信息用于描述AI模型的时间区间，例如训练的时间区间或应用的时间区间等。可以理解的是，时间戳信息是AI模型的有效时间。网络设备可广播AI模型的时间戳信息。
12)终端设备信息用于描述终端设备的标识信息，可以包括但不限于终端设备的设备商标识信息、终端设备的标识信息等中的至少一项。其中，终端设备的设备商标识信息用于标识终端设备的制造商，例如手机的品牌等。终端设备的标识信息用于唯一标识终端设备，可以包括但不限于国际移动用户识别码(international mobile subscriber identity，IMSI)、用户永久标识符(subscription permanent identifier，SUPI)、用户隐藏标识符(subscription concealed identifier，SUCI)等。
进一步的，终端设备的设备商标识信息以及终端设备的标识信息与AI模型具有关联关系。例如，AI模型1适用于手机品牌为A的手机，AI模型2适用于手机品牌为B的手机。网络设备可广播AI模型适用的终端设备的设备商标识信息，或广播AI模型适用的终端设备的标识信息。例如，网络设备可广播AI模型1适用于手机品牌为A的手机。
上述列举的模型关联信息用于举例，并不构成对本申请实施例的限定，模型关联信息可能还包括其他类型的信息。模型关联信息包括的信息可随着应用场景的变化而变化，例如增加或减少某些信息。
3、AI模型的属性信息与模型关联信息之间具有关联关系
AI模型的属性信息与模型关联信息之间具有关联关系，实际上是AI模型与模型关联信息之间具有关联关系。具体指的是，至少一个AI模型中每个AI模型的属性信息与不同环境下或不同条件下的模型关联信息之间的关联关系。该关联关系可以是一对一的关系，也可以是多对一的关系，还可以是一对多的关系。
示例的，模型关联信息为运营商标识信息，2个AI模型的属性信息与2个运营商标识信息之间的关联关系可参见下表1所示。
表1
表1中，2个AI模型的模型ID分别为1和2，模型ID 1和模型参数1与运营商标识A具有关联关系，模型ID 2和模型参数2与运营商标识B具有关联关系。
在一种实现方式中，AI模型的属性信息与模型关联信息之间的关联关系由网络设备指示，例如网络设备可通过高层信令或下行控制信息(downlink control information，DCI)或 广播消息指示AI模型的属性信息与模型关联信息之间的关联关系。例如，终端设备支持2个AI模型，网络设备可指示2个AI模型中每个AI模型与各自关联的模型关联信息之间的关联关系。可以理解的是，网络设备在指示关联关系时，可以隐含指示具有关联关系的AI模型的属性信息和模型关联信息。即网络设备可指示AI模型的属性信息，以及属性信息关联的模型关联信息。网络设备指示关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在另一种实现方式中，AI模型的属性信息与模型关联信息之间的关联关系由终端设备确定。也就是说，终端设备维护关联关系。可选的，终端设备可将其确定的关联关系上报至网络设备，以使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在又一种实现方式中，AI模型的属性信息与模型关联信息之间的关联关系由协议预定义。协议预定义关联关系，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
在本申请实施例中，更新某个AI模型，也可以理解为以下至少一项：将该AI模型更新为(或切换为)其他AI模型；去激活该AI模型，并激活新的AI模型；对该AI模型的属性信息进行更新；训练该AI模型；微调该AI模型；调整该AI模型；测试该AI模型；监督该AI模型；校验/验证该AI模型等。
下面对本申请实施例提供的AI模型更新方法进行详细阐述。
请参见图2，是本申请实施例提供的一种AI模型更新方法的流程示意图。以终端设备执行图2所示的流程为例，图2所示的流程具体可包括以下步骤：
201，网络设备向终端设备发送第一信令。相应的，终端设备接收来自网络设备的第一信令。其中，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系。
第一模型关联信息可以包括但不限于如下至少一项信息：第一PLMN信息、第一运营商标识信息、第一网络设备商标识信息、第一TAI信息、第一SOR信息、第一URSP信息、第一工作频率信息、第一小区信息、第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息。其中，第一小区信息包括但不限于如下至少一项信息：第一小区组信息、第一服务小区标识信息、第一PCI、第一TRP标识信息、第一BWP标识。第一模型关联信息包括的信息，可参考前述对模型关联信息包括的信息的具体描述，在此不再赘述。
根据第一模型关联信息包括的不同内容，第一信令有所不同。
示例的，第一模型关联信息为第一邻区列表，第一信令可以是无线资源控制(radio resource control，RRC)信令，该RRC信令用于配置第一邻区列表。也就是说，第一信令可以是用于配置第一邻区列表的RRC信令。
示例的，第一模型关联信息为第一PCI，第一信令可以是介质接入控制-控制元素(media access control-control element，MAC-CE)信令，该MAC-CE信令用于指示激活第一PCI。也就是说，第一信令可以是指示激活第一PCI的MAC-CE信令。
示例的，第一模型关联信息为第一小区信息，第一信令可以是RRC信令，该RRC信令用于指示小区切换，具体指示切换至第一小区。该RRC信令可直接指示或隐含指示第一小区信息。该RRC信令例如可以是RRC重配置(RRCReconfiguration)信令。也就是说，第一信令可以是指示小区切换的RRC信令。或者，第一信令可以是MAC-CE信令，该MAC-CE信令用于指示激活第一小区。该MAC-CE信令可直接指示或隐含指示第一小区信息。也就是说，第一信令可以是指示激活第一小区的MAC-CE信令。
示例性的，第一模型关联信息为第一BWP标识，第一信令可以是DCI，该DCI用于指示BWP切换，具体指示切换至第一BWP。该DCI可直接指示或隐含指示第一BWP标识。也就是说，第一信令可以是指示BWP切换的DCI。
示例性的，第一模型关联信息为第一SOR信息或第一URSP信息，第一信令可以是核心网设备直接向终端设备发送的信令，也可以是核心网设备通过基站向终端设备发送的信令。
第一模型关联信息与第一AI模型的属性信息具有关联关系。在一种实现方式中，网络设备向终端设备发送第三信令，相应的，终端设备接收来自网络设备的第三信令。其中，第三信令用于指示第一模型关联信息与第一AI模型的属性信息的关联关系。使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。在另一种实现方式中，终端设备确定或维护第一模型关联信息与第一AI模型的属性信息之间的关联关系。在又一种实现方式中，第一模型关联信息与第一AI模型的属性信息之间的关联关系是协议预定义的，使得网络设备和终端设备对于终端设备的AI模型与模型关联信息之间关联关系，理解一致。
202，第一模型关联信息与当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型。其中，第二模型关联信息和第二AI模型的属性信息具有关联关系。
其中，当前使用的第二模型关联信息即当前模型关联信息，可以理解为当前正在使用的模型关联信息。例如当前正在使用的URSP信息即为第二URSP信息，或当前正在使用的工作频率信息即为第二工作频率信息等。第二模型关联信息包括的信息，可参考前述对模型关联信息包括的信息的具体描述，在此不再赘述。
终端设备在接收到第一信令时，可判断第一信令指示的第一模型关联信息与当前使用的第二模型关联信息是否相同。第一模型关联信息与第二模型关联信息中至少一种类型的模型关联信息不同，即可判断出第一模型关联信息与第二模型关联信息不同。例如，第一模型关联信息中的第一PCI不同于第二模型关联信息中的第二PCI，即可判断出第一模型关联信息与第二模型关联信息不同。
在判断出第一模型关联信息与第二模型关联信息不同时，终端设备可将当前使用的第二AI模型更新为第一AI模型。也就是说，终端设备将当前使用的第二AI模型更新为与第一模型关联信息具有关联关系的第一AI模型。
终端设备将当前使用的第二AI模型更新为第一AI模型时，可在如下两种情况下更新。
情况1，在第一AI模型不同于第二AI模型，且终端设备支持第一AI模型的情况下，将第二AI模型更新为第一AI模型。也就是说，终端设备在进行AI模型更新时，不仅考虑 AI模型不同，还需要考虑终端设备是否支持新的AI模型，以避免出现更新后的AI模型不可用的情况。
情况2，在第一AI模型不同于第二AI模型，且终端设备支持第一AI模型，且第一AI模型的状态不为去激活状态的情况下，将第二AI模型更新为第一AI模型。也就是说，终端设备在考虑AI模型不同、终端设备是否支持新的AI模型时，还考虑新的AI模型是否为去激活状态。若不为去激活状态，则可更新为新的AI模型；若为去激活状态，则不能更新为新的AI模型。第一AI模型的状态是否为去激活状态，可由网络设备指示或由终端设备自主确定。例如终端设备在从AI模型1更新为AI模型2后，去激活AI模型1，在检测到第一模型关联信息时，确定可将AI模型2更新为AI模型1，但是AI模型1的状态为去激活状态，因此不会将AI模型2更新为AI模型1。AI模型的状态的改变，可由终端设备确定或由网络设备指示。
上述情况1和情况2针对终端设备支持第一AI模型，若终端设备不支持第一AI模型，那么可采用如下方式1或方式2。
方式1，在第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型的情况下，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型。也就是说，在终端设备不支持第一AI模型的情况下，终端设备选择次优的AI模型，次优的AI模型即为目标AI模型。也就是说，终端设备在发现不支持第一AI模型时，可将第一AI模型或第二AI模型更新为目标AI模型。
示例的，终端设备支持AI模型1和AI模型2，终端设备从运营商1的网络移动至运营商3的网络(即与当前模型关联信息具有关联关系的AI模型为模型1)，与第一模型关联信息具有关联关系的AI模型为AI模型3，但是终端设备不支持AI模型3。进而终端设备选择AI模型2作为目标AI模型。只不过AI模型2关联的是运营商2的网络而不是运营商3的网络。进而，终端设备在将AI模型1更新为AI模型2之后，可建立AI模型2的属性信息与第一模型关联信息之间的关联关系。
其中，目标AI模型满足相似度条件，可以包括以下至少一项：
1)第一AI模型的模型标识与目标AI模型的模型标识之差小于或等于第一阈值。也就是说，第一AI模型的模型ID与目标AI模型的模型ID最接近。例如，第一阈值为1，第一AI模型的模型ID为1，终端设备支持的AI模型的模型ID的取值范围为{2,3}，那么AI模型的模型ID为2的AI模型即为目标AI模型。进一步的，若第二AI模型的模型ID为2，那么可不对第二AI模型进行更新，或对第二AI模型的属性信息进行更新，以更好地适应通信环境的改变。
2)第一模型关联信息与目标AI模型关联的模型关联信息相同或相似度大于第二阈值。第一模型关联信息与目标AI模型关联的模型关联信息的相似度越大，表示第一AI模型与目标AI模型的匹配度越高，两个AI模型越接近。需要说明的是，对于不同类型的模型关联信息，第二阈值可能有所不同。例如，模型关联信息为TAI信息，第二阈值可以为80％；模型关联信息为USRP信息，第二阈值可以为90％。对于模型关联信息包括多种类型的信息而言，比较一种模型关联信息的相似度大于其对应的第二阈值，便可以认为整体的相似度大于第二阈值。例如，第一模型关联信息包括第一TAI信息和第一USRP信息，若第一 TAI信息与目标AI模型关联的TAI信息的相似度大于第二阈值，则认为第一模型关联信息与目标AI模型关联的模型关联信息相似度大于第二阈值。
3)目标AI模型为与预设小区信息关联的AI模型。可选的，预设小区信息包含于配置的或激活的小区信息中，例如目标AI模型为网络设备激活的BWP中默认BWP关联的AI模型。可选的，预设小区信息为目标TCI状态关联的小区信息，目标TCI状态由第二信令指示或配置或激活。其中，第二信令的接收时间与当前系统时间之间的时间差小于第三阈值。也就是说，目标TCI状态为最近网络设备发送的第二信令所指示的或配置的或激活的TCI状态。例如，第二信令为高层信令，目标TCI状态为高层信令配置的TCI状态。再例如，第二信令为MAC-CE信令，目标TCI状态为该MAC-CE信令激活的TCI状态。目标TCI状态也可以理解为最近使用的TCI状态。从而目标AI模型关联的小区信息可以理解为目标TCI状态关联的小区信息。可选的，预设小区信息包括邻区列表中默认小区标识，从而目标AI模型为网络设备为终端设备配置的邻区列表中默认小区标识关联的AI模型。
4)目标AI模型为更新第二AI模型之前的第二预设时间段内使用次数最多的AI模型或上一次使用的AI模型。其中，第二预设时间段的具体取值范围在本申请实施例不作限定。在对第二AI模型更新之前，上一次使用的AI模型指的是在使用第二AI模型之前，使用的AI模型。
方式2，在第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型的情况下，向网络设备发送第一请求消息，第一请求消息用于指示第三AI模型的属性信息。也就是说，在终端设备不支持第一AI模型的情况下，终端设备向网络设备发送第一请求消息，指示第三AI模型的属性信息，表示终端设备请求使用第三AI模型，第三AI模型可以不同于第一AI模型。可选的，第三AI模型可以是目标AI模型，也可以是其他AI模型。终端设备接收来自网络设备的第一响应消息，在第一响应消息指示确认信息的情况下，将第一AI模型或第二AI模型更新为第三AI模型。反之，在第一响应消息指示拒绝信息的情况下，可停止AI模型更新，或者可再次向网络设备发送第一请求消息，此时第一请求消息可指示终端设备请求使用另外的AI模型的属性信息(也就是说另外的AI模型不同于第三AI模型)，等待接收来自网络设备的确认信息或拒绝信息。
可选的，第一请求消息不指定任何AI模型的属性信息，第一请求消息用于请求网络设备指示一个AI模型的属性信息。网络设备在接收到该第一请求消息的情况下，向终端设备发送第一响应消息，第一响应消息用于指示新的AI模型的属性信息。终端设备在接收到第一响应消息的情况下，可将第二AI模型更新为新的AI模型。进一步的，终端设备可建立新的AI模型与第一模型关联信息之间的关联关系。
可以理解的是，终端设备在更新至第一AI模型之前或之后，发现不支持第一AI模型，进而终端设备向网络设备发送第一请求消息，用于请求使用第三AI模型或目标AI模型。终端设备在接收到指示确认信息的第一响应消息的情况下，将第一AI模型或第二AI模型更新为第三AI模型或目标AI模型。对于终端设备在更新至第一AI模型之前，发现不支持第一AI模型，在接收到指示确认信息的第一响应消息的情况下，将第二AI模型更新为第三AI模型或目标AI模型；对于终端设备在更新至第一AI模型之后，发现不支持第一AI模型，在接收到指示确认信息的第一响应消息的情况下，将第一AI模型更新为第三AI模 型或目标AI模型。
上述情况1和情况2，以及方式1和方式2，针对第一AI模型与第二AI模型不同而言。若第一AI模型与第二AI模型相同，那么终端设备将当前模型关联信息与第二AI模型的属性信息之间的关联关系，更新为第一模型关联信息与第二AI模型的属性信息之间的关联关系。从而可以实时更新模型关联信息与AI模型的属性信息之间的关联关系。
上述方式1和方式2针对终端设备不支持第一AI模型而言。若终端设备在将第二AI模型更新为第一AI模型之前或之后，发现存在以下缺陷1至缺陷3中的至少一种，可基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型。
缺陷1，第一AI模型的输出结果不满足预设条件，即第一AI模型的输出结果不满足性能需求，也就是说，在第一AI模型的输出结果不满足性能需求时，可将第一AI模型更新为目标AI模型。目标AI模型满足的相似度条件，可参考方式1中对其的具体描述，在此不再赘述。
缺陷2，第一AI模型的状态为去激活状态，即第一AI模型被网络设备指示为去激活状态或终端设备确定出第一AI模型的状态为去激活状态，可将第一AI模型或第二AI模型更新为目标AI模型。若在将第二AI模型更新为第一AI模型之前，终端设备已经发现第一AI模型的状态为去激活状态，则将第二AI模型更新为目标AI模型。若在将第二AI模型更新为第一AI模型之后，终端设备发现第一AI模型的状态为去激活状态，则将第一AI模型更新为目标AI模型。
缺陷3，第一AI模型与AI模型指示信息指示的AI模型不同。其中，AI模型指示信息由网络设备发送，不同于第一信令和第二信令。可选的，网络设备可在终端设备接入时，向终端设备发送AI模型指示信息。若在将第二AI模型更新为第一AI模型之前，终端设备发现网络设备指示的AI模型不是第一AI模型，那么可将第二AI模型更新为目标AI模型。若在将第二AI模型更新为第一AI模型之后，终端设备发现网络设备指示的AI模型不是第一AI模型，那么可将第一AI模型更新为目标AI模型。
对于上述缺陷1至缺陷3而言，终端设备除了可以选择目标AI模型之外，还可以采用方式2，向网络设备发送第一请求消息，第一请求消息用于指示第三AI模型的属性信息，在接收到来自网络设备的指示确认信息第一响应消息的情况下，根据确认信息，将第一AI模型或第二AI模型更新为第三AI模型。或者，第一请求消息用于请求网络设备指示一个AI模型的属性信息，在接收到第一响应消息，第一响应消息指示新的AI模型的属性信息的情况下，将第一AI模型或第二AI模型更新为新的AI模型。
由上述方式1、方式2以及缺陷1至缺陷3可知，在第一AI模型不可用的情况下，可将第一AI模型更新为第三AI模型或目标AI模型；或可将第二AI模型更新为第三AI模型或目标AI模型。第一AI模型不可用的情况可以包括但不限于如下至少一种情况：
(1)终端设备不支持第一AI模型；
(2)第一AI模型的输出结果不满足预设条件；
(3)第一AI模型的状态为去激活状态；
(4)第一AI模型与AI模型指示信息指示的AI模型不同。
可选的，对于情况(1)至(4)而言，终端设备还可以将第二AI模型更新为通过测量 网络设备发送的参考信号而确定的AI模型。例如，终端设备通过测量网络设备在第一小区发送的参考信号，得到测量结果(例如包括信道质量信息)，根据测量结果确定一个AI模型，将第二AI模型更新为该AI模型。或者，终端设备还可以将第一AI模型更新为通过测量网络设备发送的参考信号而确定的AI模型。
可选的，终端设备在对第二AI模型进行更新之后，可去激活第二AI模型，以避免同时使用2个或2个以上的AI模型，减少终端设备的处理负荷。
在图2所示的实施例中，终端设备在接收到指示第一模型关联信息的第一信令时，可将当前使用的第二AI模型更新为与第一模型关联信息具有关联关系的第一AI模型，使得终端设备可以快速地更新AI模型，从而有助于提高通信效率。
请参见图3，是本申请实施例提供的另一种AI模型更新方法的流程示意图，该方法具体可包括以下步骤：
301，网络设备向终端设备发送第一信令。相应的，终端设备接收来自网络设备的第一信令。其中，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系。
步骤301的实现过程，可参考图2所示实施例中步骤201的具体描述，在此不再赘述。终端设备在接收到第一信令时，可直接将第二AI模型更新为第一AI模型，或在第一AI模型不可用的情况下执行步骤302至步骤304，或在终端设备不接受使用第一AI模型的情况下执行步骤302至步骤304。
302，终端设备向网络设备发送第二请求消息。相应的，网络设备接收来自终端设备的第二请求消息。
其中，第二请求消息可以包括但不限于如下至少一项信息：
A，终端设备建议的AI模型的用途。终端设备建议的AI模型的用途，可用于网络设备基于该用途向终端设备提供相应的参考信号信息，例如提供相应的参考信号资源。AI模型的用途可以包括但不限于：训练、推理、更新、切换、激活、去激活、监视等。训练可分为首次训练或重训练，训练可以是训练AI模型的模型参数等。
B，终端设备建议的AI模型的属性信息。终端设备建议的AI模型的属性信息，可用于网络设备判断是否同意终端设备使用该AI模型。终端设备建议的AI模型的属性信息，例如，终端设备建议的AI模型的模型ID、建议的模型结构信息或建议的模型参数中的至少一项。
C，终端设备建议的AI模型更新时长，用于指示终端设备确定的AI模型更新所需的时长。可选的，建议的AI模型更新时长可以包括网络切换时长、小区切换时长、BWP切换时长、波束切换时长中至少一项时长。
D，终端设备建议的参考信号信息，用于指示终端设备建议网络设备如何发送参考信号。终端设备建议的参考信号信息可以包括但不限于：参考信号资源的数量、参考信号的样式(pattern)、参考信号的重复参数值(例如repetition off或repetition on，repetition off表示不开启参考信号的重复传输，repetition on表示开启参考信号的重复传输)、参考信号的上报阈值中的至少一项。
E，终端设备期望的接收波束数量，用于指示终端设备期望网络设备使用多少个发送波束以发送参考信号，用来匹配终端设备的接收波束数量。
进一步的，第二请求消息对应的上行资源可以包括但不限于如下至少一项资源：
1)预配置的周期性的物理上行信道资源，例如预配置的周期性的物理上行控制信道(physical uplink control channel，PUCCH)资源。
2)在发送第二请求消息之前的第一预设时间段内使用次数最多的上行授权资源。其中，第一预设时间段的具体取值范围在本申请实施例不作限定。第一预设时间段与第二预设时间段可以相同或不同。
3)上一次使用的上行授权资源。
4)下一个可用的上行授权资源。其中，下一个可用的上行授权资源(next available uplink grant)，也可以描述为最近可用的上行授权资源或即将使用的上行授权资源，“可用”指的是最近到达的上行授权资源能够承载所述第二请求消息。下一个可用的上行授权资源可以是动态授权资源(dynamic grant)或配置的授权资源(configured grant)。
5)与第一模型关联信息关联的随机接入(random access channel，RACH)资源。可选的，网络设备预配置随机接入资源与模型关联信息之间的关联关系，UE根据检测到的第一模型关联信息，确定使用第一模型关联信息关联的随机接入资源，发送第二请求消息。
6)与第一模型关联信息关联的参考信号信息对应的随机接入资源。例如，第一小区关联的SSB对应的RACH资源，终端设备根据检测到第一小区关联的SSB，确定使用第一小区的SSB对应的RACH资源，发送第二请求消息。
303，网络设备向终端设备发送第二响应消息。相应的，终端设备接收来自网络设备的第二响应消息。
其中，第二响应消息用于响应第二请求消息。网络设备根据第二请求消息，确定第一信息。第二响应消息可以包括第一信息，第一信息可以是如下至少一项信息：
A1，与终端设备建议的AI模型的用途匹配的参考信号信息。A1信息与上述第二请求消息中的A信息对应。也就是说，终端设备请求哪种用途的AI模型，网络设备向终端设备提供相应的参考信号信息。
B1，确认信息或拒绝信息。B1信息与上述第二请求消息中的B信息对应。B信息为终端设备建议的AI模型的属性信息。例如，终端设备建议使用第四AI模型，即第二请求消息包括终端设备建议的第四AI模型的属性信息，若网络设备同意终端设备使用第四AI模型，那么第一信息为确认信息，确认信息表示网络设备同意终端设备使用所请求的AI模型；若网络设备不同意终端设备使用第四AI模型，那么第一信息为拒绝信息，拒绝信息表示网络设备不同意终端设备使用所请求的AI模型。终端设备接收到拒绝信息时，可停止AI模型更新，或再次向网络设备发送第二请求消息。
可选的，B1，终端设备建议的AI模型的属性信息。也就是说，第一信息包括第二请求消息中，终端设备建议的AI模型的属性信息。可以理解的是，第二请求消息中的AI模型的属性信息，与第一信息中的AI模型的属性信息相同，即网络设备同意终端设备使用所请求的AI模型。例如，终端设备建议使用第四AI模型，即第二请求消息包括终端设备建议的第四AI模型的属性信息，若网络设备同意终端设备使用第四AI模型，那么第一信息 包括第四AI模型的属性信息；若网络设备不同意终端设备使用第四AI模型，那么第一信息包括另一个AI模型的属性信息，或者第一信息包括D1信息。其中，第四AI模型与第一AI模型可能相同或不同。
可选的，B1，网络设备期望的AI模型的属性信息。网络设备期望的AI模型的属性信息，与上述第二请求消息中的B信息无关。也就是说，网络设备不管终端设备请求使用哪个AI模型，直接向终端设备反馈，网络设备期望终端设备使用的AI模型的属性信息。终端设备在接收到网络设备期望的AI模型的属性信息时，若终端设备同意使用该AI模型，那么可将第二AI模型更新为该AI模型；若终端设备不同意使用该AI模型，那么可停止AI模型更新，或再次向网络设备发送第二请求消息。
C1，指定的AI模型更新时长，即网络设备指定的AI模型更新所需的时长。可选的，指定的AI模型更新时长可以包括网络切换时长、小区切换时长、BWP切换时长、波束切换时长中至少一项时长。例如，当终端设备检测到第一运营商网络，并且在第二请求消息中包括了建议的对应于第一运营商网络的AI模型的属性信息，网络设备根据该第二请求消息来确定指定的AI模型更新时长，该时长可以大于或等于终端设备完成切换至第一运营商网络所需时长。C1信息与上述第二请求消息中的C信息可以对应，也可以不对应。也就是说，网络设备可根据终端设备建议的AI模型更新时长，确定指定的AI模型更新时长；网络设备也可以根据第二请求消息中的其他信息，确定指定的AI模型更新时长。
D1，网络设备配置的或发送的参考信号的参考信号信息。网络设备配置的或发送的参考信号可以与终端设备建议的参考信号信息匹配，即终端设备建议网络设备如何配置或发送，网络设备就按照终端设备的建议配置或发送；也可以不匹配，即网络设备不按照终端设备的建议配置或发送参考信号。
E1，网络设备支持的发送波束数量，以匹配终端设备的接收波束数量。也就是说，E1信息与E信息匹配。
304，终端设备根据第一信息，对第二AI模型进行更新。
在一种实现方式中，在第一信息包括确认信息的情况下，终端设备可将第二AI模型更新为请求使用的AI模型。在第一信息包括终端设备建议的AI模型的属性信息的情况下，终端设备可将第二AI模型更新为终端设备建议的AI模型。例如，第一信息包括第四AI模型的属性信息，那么终端设备可将第二AI模型更新为第四AI模型。在第一信息包括网络设备期望的AI模型的属性信息的情况下，终端设备可将第二AI模型更新为网络设备期望的AI模型。例如，第一信息包括第四AI模型的属性信息，那么终端设备可将第二AI模型更新为第四AI模型。
在另一种实现方式中，第一信息包括第一参考信号信息，第一参考信号信息可以是网络设备配置的或发送的参考信号的参考信号信息，也可以是与终端设备建议的AI模型的用途匹配的参考信号信息，还可以是与网络设备期望的AI模型的属性信息具有关联关系的参考信号信息。进而，终端设备可基于第一参考信号信息对接收到的参考信号进行测量，获得信道质量信息。其中，信道质量信息可以包括但不限于如下至少一种：信干噪比(signal to interference plus noise ratio，SINR)、参考信号接收功率(reference signal receiving power，RSRP)、接收信号的强度指示(received signal strength indicator，RSSI)、参考信号接收 质量(reference signal receiving quality，RSRQ)、信道质量指示(channel quality indication，CQI)。
可选的，终端设备进一步根据信道质量信息，将第二AI模型为更新为与信道质量信息具有关联关系的AI模型。或者，终端设备根据信道质量信息，更新与信道质量信息具有关联关系的AI模型的属性信息。可以理解为，终端设备根据信道质量信息，调整或微调与信道质量信息具有关联关系的AI模型的属性信息。从而实现终端设备通过对AI模型进行重训练，保证更新后的AI模型与信道质量信息相匹配。
可选的，在第一信息进一步包括参考信号的上报阈值的情况下，终端设备进一步从信道质量信息中选择大于或等于上报阈值的目标信道质量信息，并确定目标信道质量信息对应的参考信号信息。终端设备可根据目标信道质量信息，更新与信道质量信息具有关联关系的AI模型的属性信息。或者，终端设备可将目标信道质量信息和目标信道质量信息对应的参考信号信息上报至网络设备，以便网络设备获知终端设备对AI模型进行重训练所使用的参考信号信息。可选的，网络设备可根据目标信道质量信息，更新与信道质量信息具有关联关系的AI模型的属性信息，实现网络设备对AI模型的重训练，以保证更新后的AI模型与信道质量信息相匹配。
在图3所示的实施例中，终端设备在接收到指示第一模型关联信息的第一信令时，可与网络设备协商，以更新终端设备的AI模型，从而有助于提高通信效率。
作为一种可选的实施例，终端设备在更新至第一AI模型之后，也可以执行步骤302和步骤303，并根据第一信息，对第一AI模型进行更新。终端设备根据第一信息对第一AI模型进行更新的过程，与根据第一信息对第二AI模型进行更新的过程类似，在此不再赘述。
请参见图4，是本申请实施例提供的又一种AI模型更新方法的流程示意图，该方法具体可包括以下步骤：
401，网络设备向终端设备发送第一信令。相应的，终端设备接收来自网络设备的第一信令。其中，第一信令用于指示第一AI模型的属性信息。
在一种实现方式中，第一信令指示的第一AI模型的属性信息，可以是网络设备期望的AI模型的属性信息，即网络设备期望终端设备使用的AI模型的属性信息。例如，第一AI模型的属性信息可以包括网络设备期望的AI模型的全部模型参数，或网络设备期望的AI模型的部分模型参数。
可选的，第一AI模型的模型参数包括按预设顺序排列的模型参数取值，例如按照各层神经元顺序排列的神经元参数。
在另一种实现方式中，第一信令指示的第一AI模型的属性信息，不同于当前使用的第二AI模型的属性信息。也就是说，网络设备在获知终端当前使用的第二AI模型的属性信息的情况下，向终端设备发送第一信令。例如，第一AI模型的模型结构与第二AI模型的模型结构不同。
在又一种实现方式中，第一信令指示的第一AI模型的属性信息，与终端设备的终端设备信息具有关联关系。例如，第一信令为广播消息，该广播消息包括与终端设备的设备商标识信息具有关联关系的AI模型的属性信息。
步骤401与步骤301的区别在于，步骤401中的第一信令指示第一AI模型的属性信息；而步骤301中的第一信令指示第一模型关联信息。
402，第一AI模型的属性信息与当前使用的第二AI模型的属性信息不同，将第二AI模型更新为第一AI模型。
终端设备在接收到第一信令时，可判断第一AI模型的属性信息与当前使用的第二AI模型的属性信息是否相同。第一AI模型的属性信息与第二AI模型的属性信息中至少一种类型的属性信息不同，即可判断出第一AI模型的属性信息与第二AI模型的属性信息不同。
步骤402中将第二AI模型更新为第一AI模型的过程，可参考步骤202中对其的具体描述，在此不再赘述。
在图4所示的实施例中，终端设备在接收到指示第一AI模型的属性信息的第一信令时，可将当前使用的第二AI模型更新为与第一AI模型，使得终端设备可以快速地更新AI模型，从而有助于提高通信效率。
作为一种可选的实施例A，该实施例A可包括：
(1)网络设备向终端设备发送第一信令。相应的，终端设备接收来自网络设备的第一信令。其中，第一信令用于指示第一AI模型的属性信息。
(2)终端设备向网络设备发送第二请求消息。相应的，网络设备接收来自终端设备的第二请求消息。
(3)网络设备向终端设备发送第二响应消息。相应的，终端设备接收来自网络设备的第二响应消息。
该实施例A与图3所示实施例的不同之处在于，该实施例中第一信令用于指示第一AI模型的属性信息，而图3所示实施例中第一信令用于指示第一模型关联信息。
图3所示的实施例和实施例A中，网络设备根据第二请求消息，确定第一信息。作为一种可选的实施例，网络设备根据终端设备的能力信息，确定第一信息。其中，终端设备的能力信息包括终端设备支持的多个AI模型中每个AI模型的属性信息，可在接入网络设备时上报至网络设备。该实施例中，第一信息可以包括以下至少一项信息：终端设备支持的AI模型的属性信息(例如可以是多个AI模型中某个AI模型的属性信息)、网络设备期望的AI模型的属性信息、指定的AI模型更新时长、网络设备配置的或发送的参考信号的参考信号信息、网络设备支持的发送波束数量。其中，终端设备支持的AI模型的属性信息与上述终端设备建议的AI模型的属性信息类似。第一信息中的其余信息可参考步骤303中对这些信息的具体描述，在此不再赘述。
图3所示的实施例和实施例A中，终端设备在接收到第一信令之后，向网络设备发送第二请求消息。作为一种可选的实施例，终端设备在接收到第一信令之前，向网络设备发送第三请求消息，接收来自网络设备的第三响应消息。第三请求消息与第二请求消息类似，第三响应消息与第二响应消息类似。该实施例与图3所示实施例的不同之处在于，第二请求消息在接收到第一信令之后发送，第三请求消息在接收到第一信令之前发送。例如，终端设备可以提前预测将移动至第一小区，那么终端设备可在接收到切换至第一小区的第一信令之前，向网络设备发送第三请求消息。
请参见图5，图5是本申请实施例提供的一种通信装置的结构示意图。该通信装置50可以是终端设备，也可以是与终端设备匹配的装置。如图5所示，该通信装置50包括处理单元501和通信单元502。
在一种实现方式中，通信单元502，用于接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；处理单元501，第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。
在另一种实现方式中，通信单元502，用于向终端设备发送第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系，第一模型关联信息和当前使用的第二模型关联信息不同。
基于同一发明构思，本申请实施例中提供的通信装置50解决问题的原理与有益效果与本申请图2至图4所示实施例中解决问题的原理和有益效果相似，可以参见方法的实施的原理和有益效果，为简洁描述，在这里不再赘述。
请参见图6，图6是本申请实施例提供的另一种通信装置的结构示意图。该通信装置60可以是终端设备，也可以是与终端设备匹配的装置。可选的，该通信装置还可以包括存储器603。其中，收发器601、处理器602、存储器603可以通过总线604或其他方式连接。总线在图6中以粗线表示，其它部件之间的连接方式，仅是进行示意性说明，并不引以为限。总线可以分为地址总线、数据总线、控制总线等。为便于表示，图6中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。
本申请实施例中的耦合是装置、单元或模块之间的间接耦合或通信连接，可以是电性，机械或其它的形式，用于装置、单元或模块之间的信息交互。本申请实施例中不限定上述收发器601、处理器602、存储器603之间的具体连接介质。
存储器603可以包括只读存储器和随机存取存储器，并向处理器602提供指令和数据。存储器603的一部分还可以包括非易失性随机存取存储器。
处理器602可以是中央处理单元(Central Processing Unit，CPU)，该处理器602还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器，可选的，该处理器602也可以是任何常规的处理器等。
在一种可选的实施方式中，存储器603，用于存储程序指令；处理器602，用于调用存储器603中存储的程序指令，以用于执行图2至图4对应实施例中终端设备所执行的步骤。
在本申请实施例中，可以通过在包括CPU、随机存取存储介质(Random Access Memory，RAM)、只读存储介质(Read-Only Memory，ROM)等处理元件和存储元件的例如计算机的通用计算装置上运行能够执行上述方法所涉及的各步骤的计算机程序(包括程序代码)，以及来实现本申请实施例所提供的方法。计算机程序可以记载于例如计算机可读记录介质上，并通过计算机可读记录介质装载于上述计算装置中，并在其中运行。
基于同一发明构思，本申请实施例中提供的通信装置60解决问题的原理与有益效果与 本申请图2至图4所示实施例中解决问题的原理和有益效果相似，可以参见方法的实施的原理和有益效果，为简洁描述，在这里不再赘述。
前述通信装置，例如可以是：芯片、或者芯片模组。
本申请实施例还提供一种芯片，该芯片包括处理器，处理器可以执行前述方法实施例中终端设备的相关步骤。
在一种实现方式中，该芯片用于：接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。
可选的，该芯片具体用于第一AI模型不同于第二AI模型，且终端设备支持第一AI模型，将第二AI模型更新为第一AI模型。
可选的，该芯片还用于第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型；目标AI模型满足相似度条件；将第一AI模型或第二AI模型更新为目标AI模型。
可选的，该芯片还用于第一AI模型不同于第二AI模型，且终端设备不支持第一AI模型，向网络设备发送第一请求消息，第一请求消息用于指示第三AI模型的属性信息；接收来自网络设备的第一响应消息；在第一响应消息指示确认信息的情况下，将第一AI模型或第二AI模型更新为第三AI模型。
可选的，该芯片还用于第一AI模型与第二AI模型相同，将第二模型关联信息与第二AI模型的属性信息之间的关联关系，更新为第一模型关联信息与第二AI模型的属性信息之间的关联关系。
可选的，该芯片还用于去激活第二AI模型。
可选的，该芯片还用于向网络设备发送第二请求消息，第二请求消息用于确定第一信息；
接收来自网络设备的第二响应消息，第二响应消息包括第一信息。
可选的，该芯片还用于第一信息包括第四AI模型的属性信息；将第一AI模型或第二AI模型更新为第四AI模型。
可选的，该芯片还用于第一信息包括第一参考信号信息；根据第一参考信号信息，获得信道质量信息；根据信道质量信息，将第一AI模型或第二AI模型更新为与信道质量信息具有关联关系的AI模型；和/或，更新与信道质量信息具有关联关系的AI模型的属性信息。
可选的，该芯片还用于第一信息还包括上报阈值；从信道质量信息中选择大于或等于上报阈值的目标信道质量信息，并确定目标信道质量信息对应的参考信号信息。
可选的，该芯片还用于向网络设备发送目标信道质量信息和目标信道质量信息对应的参考信号信息。
可选的，该芯片还用于根据目标信道质量信息和目标信道质量信息对应的参考信号信息，更新与信道质量信息具有关联关系的AI模型的属性信息。
可选的，该芯片还用于第一AI模型的输出结果不满足预设条件，基于第一模型关联信 息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型更新为目标AI模型；
或，第一AI模型的状态为去激活状态，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型；
或，第一AI模型与AI模型指示信息指示的AI模型不同，基于第一模型关联信息，从终端设备支持的AI模型中选择目标AI模型，将第一AI模型或第二AI模型更新为目标AI模型；
其中，目标AI模型满足相似度条件。
可选的，目标AI模型满足相似度条件，包括以下至少一项：
第一AI模型的模型标识与目标AI模型的模型标识之差小于或等于第一阈值；
第一模型关联信息与目标AI模型关联的模型关联信息相同或相似度大于第二阈值；
目标AI模型为与预设小区信息关联的AI模型；预设小区信息包含于配置的或激活的小区信息中；或，预设小区信息为目标TCI状态关联的小区信息，目标TCI状态由第二信令指示或配置或激活；
目标AI模型为更新第二AI模型之前的第二预设时间段内使用次数最多的AI模型或上一次使用的AI模型。
可选的，该芯片还用于接收来自网络设备的第三信令，第二信令用于指示第一模型关联信息与第一AI模型的属性信息的关联关系。
可选的，该芯片还用于确定第一模型关联信息与第一AI模型的属性信息的关联关系。
可选的，第一模型关联信息与第一AI模型的属性信息的关联关系是协议预定义的。
可选的，第一模型关联信息包括以下至少一项：第一PLMN信息、第一运营商信息、第一网络设备商标识信息、第一TAI信息、第一SOR信息、第一URSP信息、第一工作频率信息、第一小区信息、第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
其中，第一小区信息包括以下至少一项：第一小区组信息、第一服务小区标识信息、第一物理小区标识、第一TRP标识信息、第一BWP标识。
可选的，第一AI模型的属性信息包括以下至少一项：第一AI模型的模型标识、第一AI模型的模型结构信息、第一AI模型的模型参数。
在一种实现方式中，该芯片用于：向终端设备发送第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系，第一模型关联信息和当前使用的第二模型关联信息不同。
可选的，该芯片还用于接收来自终端设备的第一请求消息，第一请求消息用于指示第三AI模型的属性信息；向终端设备发送的第一响应消息，第一响应消息用于指示确认信息。
可选的，该芯片还用于接收来自终端设备的第二请求消息，根据第二请求消息确定第一信息；向终端设备发送第二响应消息，第二响应消息包括第一信息；或根据终端设备的能力信息，确定第一信息；向终端设备发送第一信息。
可选的，第一信息包括第四AI模型的属性信息。
可选的，第一信息包括第一参考信号信息，第一参考信号信息为网络设备配置的或发送的参考信号的参考信号信息，或与终端设备建议的AI模型的用途匹配的参考信号信息， 或与网络设备期望的AI模型的属性信息具有关联关系的参考信号信息。
可选的，该芯片还用于第一信息还包括上报阈值；接收来自终端设备的目标信道质量信息和目标信道质量信息对应的参考信号信息；根据目标信道质量信息和目标信道质量信息对应的参考信号信息，更新与信道质量信息具有关联关系的AI模型的属性信息。
可选的，该芯片还用于向终端设备发送第三信令，第三信令用于指示第一模型关联信息与第一AI模型的属性信息的关联关系。
可选的，第一模型关联信息与第一AI模型的属性信息的关联关系是协议预定义的。
可选的，第一模型关联信息包括以下至少一项：第一公共陆地移动网络PLMN信息、第一运营商信息、第一网络设备商标识信息、第一跟踪区标识TAI信息、第一漫游引导SOR信息、第一终端路由选择策略URSP信息、第一工作频率信息、第一小区信息、第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
其中，第一小区信息包括以下至少一项：第一小区组信息、第一服务小区标识信息、第一物理小区标识、第一发送接收点TRP标识信息、第一带宽部分BWP标识。
可选的，第一AI模型的属性信息包括以下至少一项：第一AI模型的模型标识、第一AI模型的模型结构信息、第一AI模型的模型参数。
请参阅图7，图7是本申请实施例提供的一种芯片模组的结构示意图。该芯片模组70可以执行前述方法实施例中终端设备的相关步骤，该芯片模组70包括：通信接口701和芯片702。
其中，通信接口用于进行芯片模组内部通信，或者用于该芯片模组与外部设备进行通信。通信接口也可以描述为通信模组。芯片702用于实现本申请实施例中终端设备的功能。
例如，芯片702，用于接收来自网络设备的第一信令，第一信令用于指示第一模型关联信息，第一模型关联信息与第一AI模型的属性信息具有关联关系；第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为第一AI模型；其中，第二模型关联信息与第二AI模型的属性信息具有关联关系。
再例如，芯片702，用于接收来自网络设备的第一信令，第一信令用于指示第一AI模型的属性信息；第一AI模型的属性信息与当前使用的第二AI模型的属性信息不同，将第二AI模型更新为第一AI模型。
可选的，芯片模组70还可以包括存储模组703、电源模组704。存储模组703用于存储数据和指令。电源模组704用于为芯片模组提供电能。
对于应用于或集成于芯片模组的各个装置、产品，其包含的各个模块可以都采用电路等硬件的方式实现，不同的模块可以位于芯片模组的同一组件(例如芯片、电路模块等)或者不同组件中，或者，至少部分模块可以采用软件程序的方式实现，该软件程序运行于芯片模组内部集成的处理器，剩余的(如果有)部分模块可以采用电路等硬件方式实现。
本申请实施例还提供一种计算机可读存储介质，计算机可读存储介质中存储有一条或多条指令，一条或多条指令适于由处理器加载并执行上述方法实施例所提供的方法。
本申请实施例还提供一种包含计算机程序或指令的计算机程序产品，当计算机程序或指令在计算机上运行时，使得计算机执行上述方法实施例所提供的方法。
需要说明的是，对于上述的各个实施例，为了简单描述，将其都表述为一系列的动作组合。本领域技术人员应该知悉，本申请不受所描述的动作顺序的限制，因为本申请实施例中的某些步骤可以采用其他顺序或者同时进行。另外，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作、步骤、模块或单元等并不一定是本申请实施例所必须的。
在上述实施例中，本申请实施例对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
本申请实施例所描述的方法或者算法的步骤可以以硬件的方式来实现，也可以是由处理器执行软件指令的方式来实现。软件指令可以由相应的软件模块组成，软件模块可以被存放于RAM、闪存、ROM、可擦除可编程只读存储器(erasable programmable ROM，EPROM)、电可擦可编程只读存储器(electrically EPROM，EEPROM)、寄存器、硬盘、移动硬盘、只读光盘(CD-ROM)或者本领域熟知的任何其它形式的存储介质中。一种示例性的存储介质耦合至处理器，从而使处理器能够从该存储介质读取信息，且可向该存储介质写入信息。当然，存储介质也可以是处理器的组成部分。处理器和存储介质可以位于ASIC中。另外，该ASIC可以位于终端设备或管理设备中。当然，处理器和存储介质也可以作为分立组件存在于终端设备或管理设备中。
本领域技术人员应该可以意识到，在上述一个或多个示例中，本申请实施例所描述的功能可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行该计算机程序指令时，全部或部分地产生按照本申请实施例所述的流程或功能。该计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。该计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输。例如，该计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。该计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。该可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如，数字视频光盘(digital video disc，DVD))、或者半导体介质(例如，固态硬盘(solid state disk，SSD))等。
关于上述实施例中描述的各个装置、产品包含的各个模块/单元，其可以是软件模块/单元，也可以是硬件模块/单元，或者也可以部分是软件模块/单元，部分是硬件模块/单元。例如，对于应用于或集成于芯片的各个装置、产品，其包含的各个模块/单元可以都采用电路等硬件的方式实现，或者，至少部分模块/单元可以采用软件程序的方式实现，该软件程序运行于芯片内部集成的处理器，剩余的(如果有)部分模块/单元可以采用电路等硬件方式实现；对于应用于或集成于芯片模组的各个装置、产品，其包含的各个模块/单元可以都采用电路等硬件的方式实现，不同的模块/单元可以位于芯片模组的同一组件(例如芯片、电路模块等)或者不同组件中，或者，至少部分模块/单元可以采用软件程序的方式实现，该软件程序运行于芯片模组内部集成的处理器，剩余的(如果有)部分模块/单元可以采用 电路等硬件方式实现；对于应用于或集成于终端的各个装置、产品，其包含的各个模块/单元可以都采用电路等硬件的方式实现，不同的模块/单元可以位于终端内同一组件(例如，芯片、电路模块等)或者不同组件中，或者，至少部分模块/单元可以采用软件程序的方式实现，该软件程序运行于终端内部集成的处理器，剩余的(如果有)部分模块/单元可以采用电路等硬件方式实现。
以上所述的具体实施方式，对本申请实施例的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本申请实施例的具体实施方式而已，并不用于限定本申请实施例的保护范围，凡在本申请实施例的技术方案的基础之上，所做的任何修改、等同替换、改进等，均应包括在本申请实施例的保护范围之内。

Claims (42)

  1. 一种人工智能AI模型更新方法，其特征在于，包括：
    接收来自网络设备的第一信令，所述第一信令用于指示第一模型关联信息，所述第一模型关联信息与第一AI模型的属性信息具有关联关系；
    所述第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为所述第一AI模型；所述第二模型关联信息和所述第二AI模型的属性信息具有关联关系。
  2. 根据权利要求1所述的方法，其特征在于，所述将当前使用的第二AI模型更新为所述第一AI模型，包括：
    所述第一AI模型不同于所述第二AI模型，且终端设备支持所述第一AI模型，将所述第二AI模型更新为所述第一AI模型。
  3. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    所述第一AI模型不同于所述第二AI模型，且所述终端设备不支持所述第一AI模型，基于所述第一模型关联信息，从所述终端设备支持的AI模型中选择目标AI模型；所述目标AI模型满足相似度条件；
    将所述第一AI模型或所述第二AI模型更新为所述目标AI模型。
  4. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    所述第一AI模型不同于所述第二AI模型，且终端设备不支持所述第一AI模型，向所述网络设备发送第一请求消息，所述第一请求消息用于指示第三AI模型的属性信息；
    接收来自所述网络设备的第一响应消息；
    在所述第一响应消息指示确认信息的情况下，将所述第一AI模型或所述第二AI模型更新为所述第三AI模型。
  5. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    所述第一AI模型与所述第二AI模型相同，将所述第二模型关联信息与所述第二AI模型的属性信息之间的关联关系，更新为所述第一模型关联信息与所述第二AI模型的属性信息之间的关联关系。
  6. 根据权利要求1至5任一项所述的方法，其特征在于，所述方法还包括：
    去激活所述第二AI模型。
  7. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    向所述网络设备发送第二请求消息，所述第二请求消息用于确定第一信息；
    接收来自所述网络设备的第二响应消息，所述第二响应消息包括第一信息。
  8. 根据权利要求7所述的方法，其特征在于，所述第一信息包括第四AI模型的属性信息；
    所述方法还包括：
    将所述第一AI模型或所述第二AI模型更新为所述第四AI模型。
  9. 根据权利要求7所述的方法，其特征在于，所述第一信息包括第一参考信号信息；
    所述方法还包括：
    根据所述第一参考信号信息，获得信道质量信息；
    根据所述信道质量信息，将所述第一AI模型或所述第二AI模型更新为与所述信道质量信息具有关联关系的AI模型；和/或，更新与所述信道质量信息具有关联关系的AI模型的属性信息。
  10. 根据权利要求9所述的方法，其特征在于，所述第一信息还包括上报阈值；
    所述方法还包括：
    从所述信道质量信息中选择大于或等于所述上报阈值的目标信道质量信息，并确定所述目标信道质量信息对应的参考信号信息。
  11. 根据权利要求10所述的方法，其特征在于，所述方法还包括：
    向所述网络设备发送所述目标信道质量信息和所述目标信道质量信息对应的参考信号信息。
  12. 根据权利要求10所述的方法，其特征在于，所述方法还包括：
    根据所述目标信道质量信息和所述目标信道质量信息对应的参考信号信息，更新与所述信道质量信息具有关联关系的AI模型的属性信息。
  13. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    所述第一AI模型的输出结果不满足预设条件，基于所述第一模型关联信息，从所述终端设备支持的AI模型中选择目标AI模型，将所述第一AI模型更新为目标AI模型；
    或，所述第一AI模型的状态为去激活状态，基于所述第一模型关联信息，从所述终端设备支持的AI模型中选择目标AI模型，将所述第一AI模型或所述第二AI模型更新为目标AI模型；
    或，所述第一AI模型与AI模型指示信息指示的AI模型不同，基于所述第一模型关联信息，从所述终端设备支持的AI模型中选择目标AI模型，将所述第一AI模型或所述第二AI模型更新为目标AI模型；
    其中，所述目标AI模型满足相似度条件。
  14. 根据权利要求3或13所述的方法，其特征在于，所述目标AI模型满足所述相似度 条件，包括以下至少一项：
    所述第一AI模型的模型标识与所述目标AI模型的模型标识之差小于或等于第一阈值；
    所述第一模型关联信息与所述目标AI模型关联的模型关联信息相同或相似度大于第二阈值；
    所述目标AI模型为与预设小区信息关联的AI模型；所述预设小区信息包含于配置的或激活的小区信息中；或，所述预设小区信息为目标传输配置指示TCI状态关联的小区信息，所述目标TCI状态由第二信令指示或配置或激活；
    所述目标AI模型为更新所述第二AI模型之前的第二预设时间段内使用次数最多的AI模型或上一次使用的AI模型。
  15. 根据权利要求1至5任一项所述的方法，其特征在于，所述方法还包括：
    接收来自所述网络设备的第三信令，所述第三信令用于指示所述第一模型关联信息与所述第一AI模型的属性信息的关联关系。
  16. 根据权利要求1至5任一项所述的方法，其特征在于，所述方法还包括：
    确定所述第一模型关联信息与所述第一AI模型的属性信息的关联关系。
  17. 根据权利要求1至5任一项所述的方法，其特征在于，所述第一模型关联信息与所述第一AI模型的属性信息的关联关系是协议预定义的。
  18. 根据权利要求1至5任一项所述的方法，其特征在于，所述第一模型关联信息包括以下至少一项：第一公共陆地移动网络PLMN信息、第一运营商信息、第一网络设备商标识信息、第一跟踪区标识TAI信息、第一漫游引导SOR信息、第一终端路由选择策略URSP信息、第一工作频率信息、第一小区信息、所述第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
    其中，所述第一小区信息包括以下至少一项：第一小区组信息、第一服务小区标识信息、第一物理小区标识、第一发送接收点TRP标识信息、第一带宽部分BWP标识。
  19. 根据权利要求1至5任一项所述的方法，其特征在于，所述第一AI模型的属性信息包括以下至少一项：所述第一AI模型的模型标识、所述第一AI模型的模型结构信息、所述第一AI模型的模型参数。
  20. 一种人工智能AI模型更新方法，其特征在于，包括：
    接收来自网络设备的第一信令，所述第一信令用于指示第一AI模型的属性信息；
    所述第一AI模型的属性信息与当前使用的第二AI模型的属性信息不同，将所述第二AI模型更新为所述第一AI模型。
  21. 根据权利要求20所述的方法，其特征在于，所述将所述第二AI模型更新为所述第 一AI模型，包括：
    终端设备支持所述第一AI模型，将所述第二AI模型更新为所述第一AI模型。
  22. 根据权利要求21所述的方法，其特征在于，所述方法还包括：
    所述终端设备不支持所述第一AI模型，从所述终端设备支持的AI模型中选择目标AI模型；所述目标AI模型满足相似度条件；
    将所述第一AI模型或所述第二AI模型更新为所述目标AI模型。
  23. 根据权利要求21所述的方法，其特征在于，所述方法还包括：
    所述终端设备不支持所述第一AI模型，向所述网络设备发送第一请求消息，所述第一请求消息用于指示第三AI模型的属性信息；
    接收来自所述网络设备的第一响应消息；
    在所述第一响应消息指示确认信息的情况下，将所述第一AI模型或所述第二AI模型更新为所述第三AI模型。
  24. 根据权利要求20所述的方法，其特征在于，所述方法还包括：
    向所述网络设备发送第二请求消息，所述第二请求消息用于确定第一信息；
    接收来自所述网络设备的第二响应消息，所述第二响应消息包括第一信息。
  25. 根据权利要求24所述的方法，其特征在于，所述第一信息包括第四AI模型的属性信息；
    所述方法还包括：
    将所述第一AI模型或所述第二AI模型更新为所述第四AI模型。
  26. 根据权利要求24所述的方法，其特征在于，所述第一信息包括第一参考信号信息；
    所述方法还包括：
    根据所述第一参考信号信息，获得信道质量信息；
    根据所述信道质量信息，将所述第一AI模型或所述第二AI模型更新为与所述信道质量信息具有关联关系的AI模型；和/或，更新与所述信道质量信息具有关联关系的AI模型的属性信息。
  27. 一种人工智能AI模型更新方法，其特征在于，所述方法包括：
    向终端设备发送第一信令，所述第一信令用于指示第一模型关联信息，所述第一模型关联信息与第一AI模型的属性信息具有关联关系，所述第一模型关联信息和当前使用的第二模型关联信息不同。
  28. 根据权利要求27所述的方法，其特征在于，所述方法还包括：
    接收来自所述终端设备的第一请求消息，所述第一请求消息用于指示第三AI模型的属 性信息；
    向所述终端设备发送的第一响应消息，所述第一响应消息用于指示确认信息。
  29. 根据权利要求27所述的方法，其特征在于，所述方法还包括：
    接收来自所述终端设备的第二请求消息，根据所述第二请求消息确定第一信息；向所述终端设备发送第二响应消息，所述第二响应消息包括第一信息；
    或根据所述终端设备的能力信息，确定第一信息；向所述终端设备发送所述第一信息。
  30. 根据权利要求29所述的方法，其特征在于，所述第一信息包括第四AI模型的属性信息。
  31. 根据权利要求30所述的方法，其特征在于，所述第一信息包括第一参考信号信息，所述第一参考信号信息为所述网络设备配置的或发送的参考信号的参考信号信息，或与所述终端设备建议的AI模型的用途匹配的参考信号信息，或与所述网络设备期望的AI模型的属性信息具有关联关系的参考信号信息。
  32. 根据权利要求31所述的方法，其特征在于，所述第一信息还包括上报阈值；
    所述方法还包括：
    接收来自所述终端设备的目标信道质量信息和所述目标信道质量信息对应的参考信号信息。
  33. 根据权利要求27至32任一项所述的方法，其特征在于，所述方法还包括：
    向所述终端设备发送第三信令，所述第三信令用于指示所述第一模型关联信息与所述第一AI模型的属性信息的关联关系。
  34. 根据权利要求27至32任一项所述的方法，其特征在于，所述第一模型关联信息与所述第一AI模型的属性信息的关联关系是协议预定义的。
  35. 根据权利要求27至32任一项所述的方法，其特征在于，所述第一模型关联信息包括以下至少一项：第一公共陆地移动网络PLMN信息、第一运营商信息、第一网络设备商标识信息、第一跟踪区标识TAI信息、第一漫游引导SOR信息、第一终端路由选择策略URSP信息、第一工作频率信息、第一小区信息、所述第一小区信息关联的参考信号信息、第一邻区列表、第一时间戳信息、第一终端设备信息；
    其中，所述第一小区信息包括以下至少一项：第一小区组信息、第一服务小区标识信息、第一物理小区标识、第一发送接收点TRP标识信息、第一带宽部分BWP标识。
  36. 根据权利要求27至32任一项所述的方法，其特征在于，所述第一AI模型的属性信息包括以下至少一项：所述第一AI模型的模型标识、所述第一AI模型的模型结构信息、 所述第一AI模型的模型参数。
  37. 一种通信装置，其特征在于，所述装置包括：
    通信单元，用于接收来自网络设备的第一信令，所述第一信令用于指示第一模型关联信息，所述第一模型关联信息与第一AI模型的属性信息具有关联关系；
    处理单元，用于所述第一模型关联信息和当前使用的第二模型关联信息不同，将当前使用的第二AI模型更新为所述第一AI模型。
  38. 一种通信装置，其特征在于，所述装置包括：
    通信单元，用于向终端设备发送第一信令，所述第一信令用于指示第一模型关联信息，所述第一模型关联信息与第一AI模型的属性信息具有关联关系，所述第一模型关联信息和当前使用的第二模型关联信息不同，所述第一模型关联信息用于触发将当前使用的第二AI模型更新为所述第一AI模型。
  39. 一种通信装置，其特征在于，包括处理器、存储器及存储在所述存储器上的计算机程序或指令，其特征在于，所述处理器执行所述计算机程序或指令以实现权利要求1-26中任一项所述方法的步骤，或实现权利要求27-36中任一项所述方法的步骤。
  40. 一种芯片，包括处理器，其特征在于，所述处理器执行权利要求1-26中任一项所述方法的步骤，或执行权利要求27-36中任一项所述方法的步骤。
  41. 一种芯片模组，其特征在于，所述芯片模组包括通信模组、电源模组、存储模组以及芯片，其中：所述电源模组用于为所述芯片模组提供电能；所述存储模组用于存储数据和指令；所述通信模组用于进行所述芯片模组内部通信，或者用于所述芯片模组与外部设备进行通信；所述芯片用于执行权利要求1-26中任一项所述方法的步骤，或执行权利要求27-36中任一项所述方法的步骤。
  42. 一种计算机可读存储介质，其特征在于，其存储有计算机程序或指令，所述计算机程序或指令被执行时实现权利要求1-26中任一项所述方法的步骤，或实现权利要求27-36中任一项所述方法的步骤。
PCT/CN2023/129535 2022-11-03 2023-11-03 Ai模型更新方法及通信装置 WO2024094157A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211372492.8A CN117997738A (zh) 2022-11-03 2022-11-03 Ai模型更新方法及通信装置
CN202211372492.8 2022-11-03

Publications (1)

Publication Number Publication Date
WO2024094157A1 true WO2024094157A1 (zh) 2024-05-10

Family

ID=90895992

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/129535 WO2024094157A1 (zh) 2022-11-03 2023-11-03 Ai模型更新方法及通信装置

Country Status (2)

Country Link
CN (1) CN117997738A (zh)
WO (1) WO2024094157A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114091679A (zh) * 2020-08-24 2022-02-25 华为技术有限公司 一种更新机器学习模型的方法及通信装置
CN114143799A (zh) * 2020-09-03 2022-03-04 华为技术有限公司 通信方法及装置
WO2022122997A1 (en) * 2020-12-11 2022-06-16 Telefonaktiebolaget Lm Ericsson (Publ) Predicting random access procedure performance based on ai/ml models
CN114915983A (zh) * 2021-02-07 2022-08-16 展讯通信（上海）有限公司 一种数据获取方法及装置
CN115150287A (zh) * 2021-03-30 2022-10-04 华为技术有限公司 网络模型管理方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114091679A (zh) * 2020-08-24 2022-02-25 华为技术有限公司 一种更新机器学习模型的方法及通信装置
CN114143799A (zh) * 2020-09-03 2022-03-04 华为技术有限公司 通信方法及装置
WO2022122997A1 (en) * 2020-12-11 2022-06-16 Telefonaktiebolaget Lm Ericsson (Publ) Predicting random access procedure performance based on ai/ml models
CN114915983A (zh) * 2021-02-07 2022-08-16 展讯通信（上海）有限公司 一种数据获取方法及装置
CN115150287A (zh) * 2021-03-30 2022-10-04 华为技术有限公司 网络模型管理方法和装置

Also Published As

Publication number Publication date
CN117997738A (zh) 2024-05-07

Similar Documents

Publication Publication Date Title
US11997722B2 (en) Random access procedure reporting and improvement for wireless networks
US11979754B2 (en) Connection behavior identification for wireless networks
US20230090022A1 (en) Method and device for selecting service in wireless communication system
WO2021238277A1 (zh) 网络优化方法、服务器、网络侧设备、系统和存储介质
CN113661727A (zh) 用于无线网络的无线电接入网(ran)节点的神经网络的配置
WO2020253613A1 (zh) 通信的方法和通信装置
US12096249B2 (en) Systems and methods for machine learning model augmentation using target distributions of key performance indicators in a wireless network
WO2023082878A1 (zh) 一种通信方法及装置
US11576055B2 (en) Method, apparatus and computer readable media for network optimization
CN113965474A (zh) 网络质量评估的方法、电子设备及存储介质
US20230112127A1 (en) Electronic device for deploying application and operation method thereof
US9264960B1 (en) Systems and methods for determinng access node candidates for handover of wireless devices
WO2021077372A1 (en) Method and access network node for beam management
CN115843054A (zh) 参数选择方法、参数配置方法、终端及网络侧设备
WO2024094157A1 (zh) Ai模型更新方法及通信装置
WO2024094103A1 (zh) Ai模型更新方法及通信装置
CN115150851A (zh) 用于小区选择的方法、设备、存储介质和计算机程序产品
US20230344717A1 (en) Policy conflict management method, apparatus, and system
WO2024139923A1 (zh) 信息传输方法及通信装置
WO2024197505A1 (zh) 用于无线通信的方法及设备
WO2023240546A1 (zh) 模型监测方法、装置、设备及介质
WO2024169622A1 (zh) 配置参考信号资源的方法以及通信装置
WO2024067248A1 (zh) 一种获取训练数据集的方法和装置
WO2024169462A1 (zh) 模型监控方法、装置、终端及网络侧设备
CN114978256B (zh) 一种无蜂窝大规模mimo系统、调整方法及调整装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23885082

Country of ref document: EP

Kind code of ref document: A1