Communication method and apparatus

By using the target identifier with a smaller number of bits than the global unique model identification for the life cycle management of the artificial intelligence model, the signaling overhead problem caused by the global unique model identification is solved, and communication efficiency and system flexibility are improved.

WO2024234989A1PCT designated stage expired Publication Date: 2024-11-21HUAWEI TECH CO LTD

Patent Information

Application Number
PCT/CN2024/090419
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-12
Filing Date
2024-04-28
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

In the life cycle management process of artificial intelligence models, the use of global unique model identification leads to a large signaling overhead between access network devices and terminal devices, especially due to the huge number of global AI models and the number of bits consumed, resulting in Communication is inefficient.

Method used

A target identifier with a smaller number of bits than the global unique model identifier (such as a combination of bitmap, bit sequence, model information and index) is used to replace the global unique model identifier for life cycle management and reduce signaling overhead.

Benefits of technology

It effectively reduces signaling overhead during the life cycle management process, improves communication efficiency, reduces unnecessary equipment work, and improves the dynamic flexibility and configuration capabilities of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024090419_21112024_PF_FP_ABST
    Figure CN2024090419_21112024_PF_FP_ABST
Patent Text Reader

Abstract

A communication method and apparatus, for use in solving the problem of large signaling overhead in the lifecycle management process of an existing artificial intelligence (AI) model. An access network device synchronizes, with a terminal device, a globally unique model identifier of a first AI model that requires lifecycle management, and then generates a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier. The access network device performs lifecycle management on the first AI model according to the target identifier of the first AI model.
Need to check novelty before this filing date? Find Prior Art

Description

Communication method and device

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of China on May 12, 2023, with application number 202310538880.7 and application name “A Communication Method and Device”, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of communication technology, and in particular to a communication method and device. Background Art

[0004] Artificial intelligence (AI) is a technology that simulates the human brain to perform complex calculations. With advances in data storage and computing power, AI is increasingly being used. The Third Generation Partnership Project (3GPP) proposes incorporating AI into fifth-generation (5G) communication systems to improve network performance and user experience through intelligent data collection and analysis.

[0005] Currently, AI models are mainly identified by globally unique model IDs (GMIs). Since the number of global AI models may be very large, the GMIs occupy a large number of bits. For example, if the number of global AI models is 2 60 , then the globally unique model identifier needs to use 60 bits.

[0006] In this way, in the lifecycle management (LCM) process of the AI ​​model, the use of a globally unique model identifier to identify the AI ​​model between the access network device and the terminal device will result in a large signaling overhead.

[0007] Summary of the Invention

[0008] The present application provides a communication method and apparatus for reducing the signaling overhead between access network equipment and terminal equipment in the lifecycle management process of an AI model.

[0009] In a first aspect, the present application provides a communication method, which can be executed by an access network device or a module (such as a chip) in the access network device. The following description takes the execution of the access network device as an example.

[0010] The access network device synchronizes the globally unique model identifier of a first AI model requiring lifecycle management with the terminal device, and then generates a target identifier for the first AI model corresponding to the globally unique model identifier of the first AI model, where the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier. The access network device performs lifecycle management on the first AI model based on the target identifier of the first AI model. Examples of lifecycle management include activating or deactivating the first AI model and updating parameters in the first AI model.

[0011] In the above technical solution, after the access network device and the terminal device synchronize the globally unique model identifier of the first AI model that needs to undergo lifecycle management, a target identifier of the first AI model is generated. During the lifecycle management process, the target identifier of the first AI model replaces the globally unique model identifier of the first AI model. In this way, the bits occupied by the identifier of the first AI model during the lifecycle management process are effectively reduced, thereby reducing the signaling overhead during the lifecycle management process.

[0012] In one possible implementation, the target identifier is one of the following:

[0013] (1) Bitmap: The bitmap includes N bits, each of which corresponds to N AI models. The N AI models are AI models that have synchronized a globally unique model identifier with the terminal device and require lifecycle management. The N AI models include the first AI model.

[0014] (2) A first bit sequence, wherein the first bit sequence includes m bits, and the m bits may have M decimal values, and the M values ​​correspond to M AI models respectively.

[0015] (3) A second bit sequence, where the second bit sequence is composed of a plurality of bits at preset positions in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different.

[0016] (4) A combination of model information of an AI model and an index of the AI ​​model under the model information, wherein the corresponding combination of any two AI models in the M AI models is different. The model information of the AI ​​model includes one or more of the radio resource control (RRC) parameters of the AI ​​model, the scenario ID of the AI ​​model, and the applicable condition ID of the AI ​​model.

[0017] Among them, M AI models are AI models that access network devices need to perform lifecycle management with terminal devices, and N AI models are included in M ​​AI models, where m, M, and N are all positive integers.

[0018] In the above technical solution, multiple representation modes of target identifiers are provided, which occupy fewer bits than the globally unique model identifier.

[0019] In a first possible implementation, the access network device generates a target identifier for the first AI model by itself and sends the target identifier of the first AI model to the terminal device. In this way, both the access network device and the terminal device can perform lifecycle management of the first AI model based on the target identifier of the first AI model.

[0020] In one possible example, when the target identifier is a bitmap, when the access network device generates the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the access network device may specifically sort N AI models (or the globally unique model identifiers of N AI models) according to the globally unique model identifiers of the N AI models to obtain a sorting result, wherein the first AI model is located at the nth position in the sorting result, where n and N are both positive integers and n is less than or equal to N; the access network device generates a bitmap for the first AI model based on the sorting result, and the bit at the nth position in the bitmap of the first AI model corresponds to the first AI model. Optionally, the value of the bit at the nth position in the bitmap of the first AI model can be used to indicate the activation status of the first AI model.

[0021] In the above technical solution, the access network device generates a bitmap for the first AI model based on the sorting results of N AI models, and then replaces the first AI model's globally unique model identifier with the bitmap of the first AI model, helping to reduce signaling overhead during lifecycle management. Furthermore, in the bitmap of the first AI model, each bit can correspond to an AI model with a synchronized globally unique model identifier that requires lifecycle management. The value of the bit can indicate the activation status of the AI ​​model, facilitating dynamic and flexible configuration.

[0022] In one possible example, when the target identifier is a bitmap, the access network device may further generate an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and requires lifecycle management. Furthermore, when generating the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the access network device may specifically sort the N AI models (or the globally unique model identifiers of the N AI models) according to the indexes of the N AI models to obtain a sorting result, wherein the first AI model is located at the Nth position in the sorting result; and based on the sorting result, the access network device generates a bitmap for the first AI model, where the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model. Optionally, the value of the bit at the Nth position in the bitmap of the first AI model can be used to indicate the activation status of the first AI model.

[0023] In the above technical solution, the access network device generates an index N of the first AI model for the first AI model. The index N of the first AI model is used to indicate that the Nth bit in the bitmap of the first AI model corresponds to the first AI model, thereby avoiding the need for the access network device and the terminal device to synchronize the globally unique model identifier of the AI ​​model that requires lifecycle management each time, that is, reordering multiple AI models with synchronized globally unique model identifiers, and refreshing the correspondence between the AI ​​model and the bits in the bitmap, which helps to avoid unnecessary work for the access network device.

[0024] Optionally, after generating the index N of the first AI model, the access network device also sends the index N of the first AI model to the terminal device. The index N is used by the terminal device to determine that the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model.

[0025] In the above technical solution, the terminal device can determine the bit corresponding to the first AI model from the bitmap of the first AI model according to the index N, avoiding the terminal device from reordering multiple AI models that have synchronized globally unique model identifiers, and refreshing the correspondence between the AI ​​model and the bits in the bitmap, which helps to avoid unnecessary work for the terminal device.

[0026] In one possible example, when an access network device performs lifecycle management on a first AI model based on a target identifier of the first AI model, the access network device may specifically send the target identifier of the first AI model to a terminal device. The target identifier of the first AI model is used to instruct the terminal device to activate the first AI model. Furthermore, after sending the target identifier of the first AI model to the terminal device, the access network device may synchronize with the terminal device the globally unique model identifier of a second AI model requiring lifecycle management, and generate a target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model. In other words, the AI ​​models requiring lifecycle management for which the globally unique model identifiers have been synchronized with the terminal device include not only the first AI model but also the second AI model. Subsequently, the access network device sends the target identifier of the second AI model to the terminal device. The target identifier of the second AI model is used to instruct the terminal device to activate the second AI model and deactivate the first AI model. The target identifier of the first AI model may be a bitmap of the first AI model, and the target identifier of the second AI model may be a bitmap of the second AI model.

[0027] In the above technical solution, if the terminal device moves, causing the performance of the current first AI model to degrade and the first AI model to no longer be applicable to the current channel state between the access network device and the terminal device, the access network device sends a bitmap of a second AI model to the terminal device to activate the second AI model and deactivate the first AI model, thereby implementing a switch from the first AI model to the second AI model during lifecycle management. Furthermore, the access network device sends the bitmap of the second AI model to activate the second AI model and deactivate the first AI model, helping to reduce signaling overhead.

[0028] When the access network equipment includes a centralized unit (CU) and a distributed unit (DU), that is, when the access network equipment adopts a CU-DU separation architecture:

[0029] In one possible implementation, when the access network device generates the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the DU may specifically generate the target identifier of the first AI model. Furthermore, the DU may also send the target identifier of the first AI model to the terminal device; or, the DU may send the target identifier of the first AI model to the CU, and the CU may send the target identifier of the first AI model to the terminal device.

[0030] In the above technical solution, the DU can be responsible for the functions of the medium access control (MAC) layer, and the MAC layer generates the target identifier of the first AI model, which helps to improve the processing speed.

[0031] In one possible implementation, the first AI model is located in a DU. The DU generates an index N for the first AI model and sends the index N to the terminal device. Alternatively, the DU generates an index N for the first AI model and sends the index N to the CU, which then sends the index N to the terminal device.

[0032] In one possible implementation, the first AI model is located in a CU. The DU sends the globally unique model identifier of the first AI model to the CU, and the CU generates an index N of the first AI model based on the globally unique model identifier of the first AI model. Subsequently, the CU sends the index N of the first AI model to the DU, and the DU sends the index N of the first AI model to the terminal device. Alternatively, the CU directly sends the index N of the first AI model to the terminal device.

[0033] In a second possible implementation, the access network device sends the target rule to the terminal device. Accordingly, both the access network device and the terminal device can generate a target identifier for the first AI model based on the target rule. In this way, both the access network device and the terminal device can manage the lifecycle of the first AI model based on the target identifier of the first AI model.

[0034] In one possible example, the access network device directly sends a rule (i.e., a target rule) to the terminal device, and instructs the terminal device to generate a target identifier of the first AI model according to the target rule.

[0035] In the above technical solution, the access network device directly sends the target rules to the terminal device, which reduces the interaction between the access network device and the terminal device and helps to reduce signaling consumption.

[0036] In another possible example, the access network device sends a rule set to the terminal device. Accordingly, after receiving the rule set from the access network device, the terminal device selects a rule from the rule set as a target rule and sends the target rule to the access network device.

[0037] In the above technical solution, the terminal device selects the target rule by itself, and can select the target rule that is more suitable for the capabilities of the terminal device, which helps to improve the stability of communication.

[0038] When the access network device includes a CU and a DU, that is, when the access network device has a CU-DU separation architecture: when the access network device sends a target rule to the terminal device, specifically, the DU sends the target rule to the terminal device; or, alternatively, the DU sends the target rule to the CU, and the CU sends the target rule to the terminal device. When the access network device generates a target identifier for the first AI model based on the target rule, specifically, the DU generates the target identifier for the first AI model based on the target rule.

[0039] In the above technical solution, since the model reasoning of the first AI model is usually located in the DU, generating the target rules of the first AI model through the DU helps to improve the processing speed.

[0040] In a second aspect, the present application provides a communication method, which can be executed by a terminal device or a module (such as a chip) in the terminal device. The following description takes the execution of the terminal device as an example.

[0041] The terminal device and the access network device synchronize a globally unique model identifier of a first AI model requiring lifecycle management. The terminal device performs lifecycle management on the first AI model based on a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model. The number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier.

[0042] In one possible implementation, the target identifier is one of the following:

[0043] (1) Bitmap, the bitmap includes N bits, the N bits correspond to N AI models respectively, the N AI models are AI models that have synchronized the globally unique model identifier with the access network device and require lifecycle management, and the N AI models include the first AI model.

[0044] (2) A first bit sequence, wherein the first bit sequence includes m bits, and the m bits have a total of M decimal values, and the M values ​​correspond to M AI models respectively.

[0045] (3) A second bit sequence, where the second bit sequence is composed of a plurality of bits at preset positions in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different.

[0046] (4) A combination of the model information of the AI ​​model and the index of the AI ​​model under the model information, and the corresponding combination of any two AI models in the M AI models is different, and the model information includes one or more of the RRC parameters, the scenario identifier, and the applicable condition identifier.

[0047] Among them, M AI models are AI models that need to be managed with the terminal device for life cycle, N AI models are included in the M AI models, and m, M, and N are all positive integers.

[0048] In one possible implementation, when a terminal device performs lifecycle management on a first AI model based on the target identifier of the first AI model, the terminal device may specifically activate the first AI model based on the target identifier of the first AI model. Furthermore, after activating the first AI model, the terminal device may synchronize with the access network device the globally unique model identifier of a second AI model requiring lifecycle management. That is, the AI ​​models requiring lifecycle management for which the globally unique model identifier has been synchronized with the access network device include not only the first AI model but also the second AI model. Subsequently, the terminal device may activate the second AI model and deactivate the first AI model based on the target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model.

[0049] In one possible implementation, before performing lifecycle management on the first AI model based on the target identifier of the first AI model, the terminal device may also receive the target identifier of the first AI model from the access network device, where the target identifier of the first AI model is generated by the access network device.

[0050] In one possible implementation, before performing lifecycle management on the first AI model based on the target identifier of the first AI model, the terminal device may also receive a target rule from the access network device, where the target rule is a rule for generating the target identifier of the first AI model; then, the terminal device generates the target identifier of the first AI model based on the target rule.

[0051] In a third aspect, an embodiment of the present application provides a communication device.

[0052] The communication device has the function of implementing the access network device in the above-mentioned first aspect or any possible implementation of the first aspect. The device can be an access network device or a chip included in the access network device.

[0053] The communication device may also have the function of implementing the terminal device in the above-mentioned second aspect or any possible implementation of the second aspect. The device may be a terminal device or a chip included in the terminal device.

[0054] The functions of the above-mentioned communication device can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules, units or means corresponding to the above-mentioned functions.

[0055] In one possible implementation, the structure of the device includes a processing module and a transceiver module, wherein the processing module is configured to support the device to perform the corresponding functions of the access network device in the above-mentioned first aspect or any one of the implementations of the first aspect, or to perform the corresponding functions of the terminal device in the above-mentioned second aspect or any one of the implementations of the second aspect. The transceiver module is used to support communication between the device and other communication devices. For example, when the device is an access network device, it can be used to synchronize the globally unique model identifier of the first AI model that requires lifecycle management with the terminal device. The communication device may also include a storage module, which is coupled to the processing module and stores the necessary program instructions and data for the device. As an example, the processing module may be a processor, the communication module may be a transceiver, and the storage module may be a memory. The memory may be integrated with the processor or may be set separately from the processor.

[0056] In another possible implementation, the structure of the device includes a processor and may also include a memory. The processor is coupled to the memory and can be used to execute computer program instructions stored in the memory, so that the device performs the method in the above-mentioned first aspect or any possible implementation of the first aspect, or performs the method in the above-mentioned second aspect or any possible implementation of the second aspect. Optionally, the device also includes a communication interface, and the processor is coupled to the communication interface. When the device is an access network device or a terminal device, the communication interface can be a transceiver or an input / output interface; when the device is a chip included in the access network device or a chip included in the terminal device, the communication interface can be the input / output interface of the chip. Optionally, the transceiver can be a transceiver circuit, and the input / output interface can be an input / output circuit.

[0057] In a fourth aspect, an embodiment of the present application provides a chip system, comprising: a processor, the processor being coupled to a memory, the memory being used to store programs or instructions, and when the programs or instructions are executed by the processor, the chip system implements the method of the above-mentioned first aspect or any possible implementation of the first aspect, or implements the method of the above-mentioned second aspect or any possible implementation of the second aspect.

[0058] Optionally, the chip system further includes an interface circuit for transmitting interactive code instructions to the processor.

[0059] Optionally, there may be one or more processors in the chip system, and the processor may be implemented in hardware or software. When implemented in hardware, the processor may be a logic circuit, an integrated circuit, etc. When implemented in software, the processor may be a general-purpose processor implemented by reading software code stored in a memory.

[0060] Optionally, the chip system may include one or more memories. The memory may be integrated with the processor or provided separately from the processor. For example, the memory may be a non-transient processor, such as a read-only memory (ROM), which may be integrated with the processor on the same chip or provided on separate chips.

[0061] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program or instruction stored thereon. When the computer program or instruction is executed, the computer executes the method in the above-mentioned first aspect or any possible implementation of the first aspect, or the method in the above-mentioned second aspect or any possible implementation of the second aspect.

[0062] In a sixth aspect, an embodiment of the present application provides a computer program product. When a computer reads and executes the computer program product, the computer executes the method in the above-mentioned first aspect or any possible implementation of the first aspect, or executes the method in the above-mentioned second aspect or any possible implementation of the second aspect.

[0063] In a seventh aspect, embodiments of the present application provide a communications system, comprising an access network device and at least one terminal device. The access network device is configured to execute the method of the first aspect or any possible implementation of the first aspect, and the terminal device is configured to execute the method of the second aspect or any possible implementation of the second aspect.

[0064] The technical effects that can be achieved in any of the second to seventh aspects mentioned above can refer to the description of the beneficial effects in the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] FIG1 is a schematic diagram of a communication system architecture provided by this application;

[0066] FIG2 is a schematic diagram of an AI model deployment provided by this application;

[0067] FIG3 is a schematic diagram of another AI model deployment provided by this application;

[0068] FIG4 is a schematic diagram of an application architecture of an AI model provided in this application;

[0069] FIG5 is a schematic diagram of a channel state information compression model provided by the present application;

[0070] FIG6 is a schematic diagram of a channel state information prediction model provided by the present application;

[0071] FIG7 is a lifecycle management process for an AI model provided by this application;

[0072] FIG8 is a flow chart of a communication method provided by the present application;

[0073] FIG9 is a flow chart of a first specific communication method provided in this application;

[0074] FIG10 is a flow chart of a second specific communication method provided in this application;

[0075] FIG11 is a flow chart of a third specific communication method provided in this application;

[0076] FIG12 is a flow chart of a fourth specific communication method provided in this application;

[0077] FIG13 is a flow chart of a communication method under a CU-DU separation architecture provided by the present application;

[0078] FIG14 is a flow chart of a fifth specific communication method provided in this application;

[0079] FIG15 is a flow chart of a sixth specific communication method provided in this application;

[0080] FIG16 is a schematic structural diagram of a communication device provided by the present application;

[0081] FIG17 is a schematic structural diagram of another communication device provided in this application. DETAILED DESCRIPTION

[0082] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0083] Figure 1 is a schematic diagram of the architecture of a communication system 1000 to which this application can be applied. As shown in Figure 1 , the communication system includes a radio access network 100 and a core network 200. Optionally, the communication system 1000 may further include the Internet 300.

[0084] The wireless access network 100 may include at least one access network device (such as 110a and 110b in FIG. 1 ) and may also include at least one terminal device (such as 120a-120j in FIG. 1 ). The terminal device is wirelessly connected to the access network device, and the access network device is wirelessly or wiredly connected to the core network. The core network device and the access network device may be independent and distinct physical devices, or the functions of the core network device and the logical functions of the access network device may be integrated into the same physical device, or a single physical device may integrate some of the functions of the core network device and some of the functions of the access network device. Terminal devices and access network devices may be interconnected via wired or wireless connections. FIG. 1 is merely a schematic diagram; the communication system may also include other network devices, such as wireless relay devices and wireless backhaul devices, which are not shown in FIG.

[0085] An access network device may be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next-generation NodeB (gNB) in a fifth-generation (5G) mobile communication system, an access network device in an open radio access network (O-RAN), a next-generation base station in a sixth-generation (6G) mobile communication system, a base station in a future mobile communication system, or an access node in a wireless fidelity (WiFi) system. Alternatively, the access network device may be a module or unit that performs some of the functions of a base station, such as a central unit (CU), a distributed unit (DU), a CU control plane (CU-CP) module, or a CU user plane (CU-UP) module. The access network device may be a macro base station (such as 110a in FIG. 1 ), a micro base station, an indoor station (such as 110b in FIG. 1 ), a relay node, or a donor node. This application does not limit the specific technology and specific device form used by the access network equipment.

[0086] In this application, a device for implementing the functions of an access network device may be an access network device; it may also be a device capable of supporting the access network device in implementing the functions, such as a chip system, a hardware circuit, a software module, or a hardware circuit and a software module, which may be installed in the access network device or used in conjunction with the access network device. In this application, a chip system may consist of a chip or may include a chip and other discrete components.

[0087] Furthermore, the communication between the access network device and the terminal device follows a certain protocol layer structure. The protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure. For example, the control plane protocol layer structure may include the functions of protocol layers such as the radio resource control (RRC) layer, the packet data convergence protocol (PDCP) layer, the radio link control (RLC) layer, the media access control (MAC) layer and the physical layer (PHY). For example, the user plane protocol layer structure may include the functions of protocol layers such as the PDCP layer, the RLC layer, the MAC layer and the physical layer. In one possible implementation, a service data adaptation protocol (SDAP) layer may also be included above the PDCP layer. Optionally, the protocol layer structure between the access network device and the terminal device may also include an artificial intelligence (AI) layer for transmitting data related to the AI ​​function.

[0088] The access device may include a CU and a DU. Multiple DUs may be centrally controlled by one CU. As an example, the interface between the CU and the DU may be referred to as an F1 interface. Among them, the control plane (CP) interface may be F1-C, and the user plane (UP) interface may be F1-U. This application does not limit the specific names of the interfaces. The CU and DU may be divided according to the protocol layers of the wireless network: for example, the functions of the PDCP layer and above protocol layers are set in the CU, and the functions of the protocol layers below the PDCP layer (such as the RLC layer and the MAC layer, etc.) are set in the DU; for another example, the functions of the protocol layers above the PDCP layer are set in the CU, and the functions of the protocol layers below the PDCP layer are set in the DU, without limitation.

[0089] The above division of the processing functions of CU and DU according to the protocol layer is only an example, and can also be divided in other ways. For example, the CU or DU can be divided into functions with more protocol layers, and for another example, the CU or DU can be divided into partial processing functions with the protocol layer. In one design, some functions of the RLC layer and the functions of the protocol layers above the RLC layer are set in the CU, and the remaining functions of the RLC layer and the functions of the protocol layers below the RLC layer are set in the DU. In another design, the functions of the CU or DU can also be divided according to the service type or other system requirements, for example, by delay, the functions whose processing time needs to meet the delay requirements are set in the DU, and the functions that do not need to meet the delay requirements are set in the CU. In another design, the CU can also have one or more functions of the core network. For example, the CU can be set on the network side to facilitate centralized management. In another design, the radio unit (RU) of the DU is set remotely. Optionally, the RU can have radio frequency functions.

[0090] Optionally, the DU and RU can be divided at the physical layer. For example, the DU can implement higher-level functions in the physical layer, and the RU can implement lower-level functions in the physical layer. When used for transmission, the physical layer functions may include at least one of the following: adding a cyclic redundancy check (CRC) code, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, resource mapping, physical antenna mapping, or radio frequency transmission functions. When used for reception, the physical layer functions may include at least one of the following: CRC checking, channel decoding, rate matching, descrambling, demodulation, layer demapping, channel detection, resource demapping, physical antenna demapping, or radio frequency reception functions. The higher-level functions in the physical layer may include a portion of the physical layer functions, such as those closer to the MAC layer, and the lower-level functions in the physical layer may include another portion of the physical layer functions, such as those closer to the radio frequency functions. For example, the higher-level functions in the physical layer may include adding a CRC code, channel coding, rate matching, scrambling, modulation, and layer mapping, while the lower-level functions in the physical layer may include precoding, resource mapping, physical antenna mapping, and radio frequency transmission functions.

[0091] Exemplarily, the functions of the CU can be implemented by one entity, or can be implemented by different entities. For example, the functions of the CU can be further divided, that is, the control plane and the user plane are separated and implemented through different entities, namely the control plane CU entity (i.e., CU-CP entity) and the user plane CU entity (i.e., CU-UP entity), and the CU-CP entity and the CU-UP entity can be connected through the E1 interface. The CU-CP entity and the CU-UP entity can be coupled with the DU to jointly complete the functions of the access network device. The control plane CU-CP of the CU also includes a further divided architecture, that is, the existing CU-CP is further divided into CU-CP1 and CU-CP2, wherein CU-CP1 includes various wireless resource management functions, and CU-CP2 only includes RRC functions and PDCP-C functions (i.e., the basic functions of the control plane signaling at the PDCP layer).

[0092] Optionally, any of the above-mentioned DU, CU, CU-CP, CU-UP, and RU can be a software module, a hardware structure, or a software module + hardware structure, without limitation. The existence forms of different entities can be different and are not limited. For example, DU, CU, CU-CP, and CU-UP are software modules, and RU is a hardware structure. These modules and their execution methods are also within the scope of protection of this application.

[0093] Terminal devices may also be referred to as terminals, user equipment (UE), mobile stations, mobile terminal devices, etc. Terminal devices can be widely used for communications in various scenarios, for example, including but not limited to at least one of the following scenarios: device-to-device (D2D), vehicle to everything (V2X), machine-type communication (MTC), Internet of Things (IOT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, or smart city, etc. The terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver function, a wearable device, a vehicle, a drone, a helicopter, an airplane, a ship, a robot, a robotic arm, or a smart home device, etc. This application does not limit the specific technology and specific device form adopted by the terminal device.

[0094] In the present application, the device for realizing the function of the terminal device can be the terminal device; it can also be a device that can support the terminal device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. The device can be installed in the terminal device or can be used in combination with the terminal device.

[0095] The access network equipment and terminal equipment can be fixed or movable. The access network equipment and / or terminal equipment can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on the water surface; they can also be deployed on aircraft, balloons and artificial satellites in the air. This application does not limit the application scenarios of the access network equipment and terminal equipment. The access network equipment and terminal equipment can be deployed in the same scenario or different scenarios. For example, the access network equipment and terminal equipment are deployed on land at the same time; or, the access network equipment is deployed on land and the terminal equipment is deployed on the water surface, etc., and no further examples are given.

[0096] The roles of access network devices and terminal devices can be relative. For example, the helicopter or drone 120i in Figure 1 can be configured as a mobile access network device. To terminal devices 120j that access the wireless access network 100 via 120i, 120i is an access network device. However, to access network device 110a, 120i is a terminal device, meaning that communication between 110a and 120i occurs via a wireless air interface protocol. Communication between 110a and 120i can also occur via an interface protocol between access network devices. In this case, relative to 110a, 120i is also an access network device. Therefore, both access network devices and terminal devices can be collectively referred to as communication devices. 110a and 110b in Figure 1 can be referred to as communication devices with access network device functionality, while 120a-120j in Figure 1 can be referred to as communication devices with terminal device functionality.

[0097] In the present application, an independent network element (e.g., an AI network element, or AI node) may be introduced into the communication system shown in FIG. 1 to implement AI-related operations. The AI ​​network element may be directly connected to the access network equipment in the communication system, or may be indirectly connected to the access network equipment through a third-party network element. The third-party network element may be a core network element such as an authentication management function (AMF) or a user plane function (UPF). Alternatively, an AI module (also referred to as an AI function or AI entity) may be configured in other network elements in the communication system to implement AI-related operations. For example, the other network element may be an access network device (e.g., a gNB), a core network device, or an operation, administration, and maintenance (OAM) system. In this case, the network element that performs AI-related operations is a network element with a built-in AI module. The OAM is used to operate, manage, and maintain access network devices and / or core network devices. For example, when the AI ​​module is located in the OAM, its communication with the access network device can reuse the current northbound interface; when the AI ​​module is located in the access network device, it can reuse the current F1, Xn, Uu and other interfaces; when the AI ​​module becomes an independent network entity, it is necessary to re-establish the communication link from the network entity to the OAM and the access network device, and the communication link can be a wired link or a wireless link.

[0098] In the present application, as shown in Figure 2 or Figure 3, at least one of the core network equipment, access network equipment, terminal equipment or OAM equipment can be deployed with an AI model, and the corresponding functions can be implemented using the AI ​​model. Among them, the AI ​​model represents the mapping relationship between the input and output of the AI ​​model. The AI ​​model can be a neural network, a linear regression AI model, a decision tree AI model, a support vector machine (SVM), a Bayesian network, a Q learning AI model or other machine learning AI models. The AI ​​model may include algorithms or parameters for implementing AI. In the present application, the AI ​​model can be used for load prediction, terminal device trajectory prediction, channel state information prediction, optimal beam prediction, positioning prediction, etc. In addition, the AI ​​model can also perform strategy reasoning from the perspectives of network energy saving, mobility optimization, etc. based on the prediction results of the network performance of the access network device by the trained AI model, so as to obtain reasonable and efficient energy saving strategies, mobility optimization strategies, etc. In the present application, the AI ​​model can also be referred to as a model.

[0099] In this application, the AI ​​models deployed in different nodes can be the same or different. Differences in AI models include at least one of the following: differences in AI model structural parameters, such as the number of AI model layers and / or weights; differences in AI model input parameters; or differences in AI model output parameters. Differences in AI model input parameters and / or AI model output parameters can be described as differences in AI model functionality. Furthermore, differences in AI models can also be distinguished by their applicable scenarios or the datasets used for training. Unlike Figure 2 above, Figure 3 separates the functionality of the access network device into a CU and a DU. Optionally, the CU and DU may be those within an O-RAN architecture. One or more AI models may be deployed in the CU, and / or one or more AI models may be deployed in the DU. Optionally, the CU in Figure 3 can be further separated into a CU-CP and a CU-UP. Optionally, one or more AI models may be deployed in the CU-CP, and / or one or more AI models may be deployed in the CU-UP. Optionally, in FIG. 2 or FIG. 3 , the OAM of the access network device and the OAM of the core network device may be deployed independently.

[0100] As shown in Figure 4, it is a schematic diagram of an application architecture of an AI model. The data source is used to store training data and inference data. The AI ​​model training host obtains an AI model by analyzing or training the training data provided by the data source, and deploys the AI ​​model in the AI ​​model inference host. Optionally, the AI ​​model training node can also update the AI ​​model deployed on the AI ​​model inference node. The AI ​​model inference node can also feedback relevant information of the deployed AI model to the AI ​​model training node, so that the AI ​​model training node can optimize or update the deployed AI model.

[0101] The AI ​​model training node learns the AI ​​model, which is equivalent to the AI ​​model training node using the training data to learn the mapping relationship between the input and output of the AI ​​model. The AI ​​model inference node uses the AI ​​model to perform inference based on the inference data provided by the data source to obtain the inference result. This method can also be described as follows: the AI ​​model inference node inputs the inference data into the AI ​​model, and obtains the output through the AI ​​model, which is the inference result.

[0102] The reasoning result may indicate: the configuration parameters used (executed) by the execution object, and / or the operations performed by the execution object. The reasoning result may be uniformly planned by the execution (actor) entity and sent to one or more execution objects (e.g., network entities) for execution. Optionally, the execution entity or execution object may feed back the parameters or measurements it collects to the data source. This process may be referred to as performance feedback, and the fed-back parameters may be used as training data or reasoning data. Optionally, the execution entity or execution object may also determine feedback information related to the performance of the AI ​​model based on the reasoning result output by the AI ​​model reasoning node, and feed the feedback information back to the AI ​​model reasoning node. The AI ​​model reasoning node may feed back the performance information of the AI ​​model to the AI ​​model training node based on the feedback information, so that the AI ​​model training node can optimize or update the deployed AI model, etc. This process may be referred to as AI model feedback.

[0103] The 3rd Generation Partnership Project (3GPP) is currently discussing AI use cases, which mainly include the following three types:

[0104] 1) Channel State Information Feedback Enhancement (CSI feedback enhancement) scenario

[0105] Channel state information (CSI) refers to the channel properties of a communication link. Terminal devices report CSI to access network equipment so the equipment can select a more appropriate modulation and coding scheme (MCS) for the terminal. This MCS is used for data transmission between the access network and the terminal. Scenarios for enhanced CSI feedback include CSI compression and CSI prediction.

[0106] The channel state information compression scenario is implemented based on a two-sided AI model (two-sided (AI / ML) model). See the AI ​​model schematic diagram shown in Figure 5. The bilateral AI model (i.e., the channel state information compression model) includes an encoder and quantizer on the terminal device side, and a decoder and dequantizer on the access network device side. The input of the encoder is the actual channel state information obtained by the terminal device measuring the reference signal. The channel state information can be expressed in the form of the original channel matrix / precoded channel matrix / eigenvector after the channel is decomposed, etc. The output of the encoder is a floating-point vector carrying compressed channel state information; further, the quantizer maps the floating-point vector carrying the compressed channel state information into a quantized bit sequence (possible quantization methods include scalar quantization, vector quantization, etc.). The functions of the decoder and dequantizer are opposite to those of the encoder and quantizer. The channel state information output by the decoder can be considered as inferred channel state information. For example, the encoder or decoder can be implemented by AI models such as convolutional neural network (CNN) and transformer.

[0107] The channel state information prediction scenario is based on a one-sided AI model (one-sided (AI / ML) model) deployed on the terminal device side. Referring to the AI ​​model schematic diagram shown in Figure 6, the input of the one-sided AI model (i.e., the channel state information prediction model) is the channel state information of multiple historical moments (shown in Figure 6 as the time from t1-t2 to t1, where t1 and t2 are both positive integers and t1 is greater than t2). The output of the one-sided AI model is the channel state information (i.e., the inferred channel state information) at the target moment (shown in Figure 6 as the time t1+1). The one-sided AI model can specifically be a multi-layer perceptron (MLP).

[0108] 2) Beam Management (BM) Scenario

[0109] BM scenarios are implemented based on a unilateral AI model (i.e., a beam management model). This model is primarily used to discover the strongest transmit / receive beam pair. This beam management model can be deployed on the network side (e.g., in access network equipment) or in terminal devices.

[0110] When the beam management model is deployed on the terminal device side, it can be used to predict the optimal beam on the access network device side, wherein the beam management model can be built into the terminal device or sent to the terminal device by the network side. During the use of the AI ​​model, the terminal device measures the downlink synchronization signal block (SSB) and channel state information-reference signal (CSI-RS) to obtain the angle domain information of the channel, and inputs the angle domain information of the channel into the beam management model to infer the top-k beams among multiple beams (such as the top k beams in the reference signal received power (RSRP) ranking, where k is a positive integer). The identifiers of the inferred top-k beams are fed back to the access network device (for the second round of scanning and subsequent determination of the optimal beam).

[0111] When the beam management model is deployed on the network side, it relies on the channel angle domain information measured by the terminal device for beam prediction, such as using the beam RSRP measured by the terminal device for prediction: the access network device transmits some beams in the full codebook, the terminal device measures the sparse beams to obtain RSRP, and feeds the RSRP back to the access network device. The access network device inputs the obtained RSRP into the beam management model, determines the top-k beams (such as the first k beams in the RSRP sorting, where k is a positive integer), and performs a second round of scanning on the top-k beams. After measurement, the terminal device reports the optimal beam in the second round.

[0112] 3) Positioning accuracy enhancements scenarios

[0113] The positioning accuracy enhancement scenario is implemented based on a unilateral AI model (i.e., the positioning accuracy enhancement model). This AI can identify line of sight (LOS) and non-light of sight (NLOS) conditions, thereby improving positioning accuracy with a smaller number of transmission reception point (TRP) antennas. For example, in indoor scenarios (also known as heavy NLOS scenarios), where there may not be a sufficient number of LOS paths, the positioning accuracy enhancement model can improve positioning accuracy in these scenarios.

[0114] AI model lifecycle management (LCM) is the management of AI models from start to finish. It can be understood as the steps, processes, and strategies that organizations and individuals should follow when leveraging AI to achieve real business value. A possible AI model lifecycle management process might include: AI-related capability interaction, data collection for AI model training, AI model training, AI model operations (including AI model registration and configuration), data collection for AI model inference, AI model inference, AI model performance feedback and / or monitoring, and AI model switching / updates / deactivation / fallback to traditional methods. See Figure 7 for details.

[0115] In the lifecycle management process of Figure 7, a model ID is required to uniquely identify the AI ​​model. For example, in the "AI model operation" step, AI model registration involves the model ID; in the "AI model switching" step, for example, when the access network device instructs the terminal device to switch from AI model 1 to AI model 2, the model ID of AI model 1 and the model ID of AI model 2 are involved.

[0116] Currently, AI models primarily use globally unique model identifiers (GUIs). However, since there may be tens of thousands of AI models globally, GUIs occupy a large number of bits. Consequently, using GUIs to identify AI models between two devices (e.g., access network devices and terminal devices, or access network devices and core network devices) during the AI ​​model lifecycle management process results in high signaling overhead.

[0117] To address the above issues, this application provides a communication method that uses an identifier (denoted as a target identifier) ​​with a smaller number of bits than a globally unique model identifier as the identifier of the AI ​​model during lifecycle management. The following describes the solution using access network devices and terminal devices as examples, but this solution can also be implemented by other devices.

[0118] For details, please refer to the flowchart of the communication method shown in FIG8 :

[0119] In step 801, the access network device and the terminal device synchronize the globally unique model identifier of the first AI model that requires lifecycle management.

[0120] In one possible approach, the OAM stores globally unique model identifiers of multiple AI models. The OAM can determine the globally unique model identifier of the first AI model that requires lifecycle management from the globally unique model identifiers of the multiple AI models, and then synchronize the globally unique model identifier of the first AI model to the access network device and the terminal device respectively. Optionally, the OAM also stores multiple AI models. When the first AI model is a bilateral AI model, the OAM can also send the sub-AI model inferred on the access network device side in the first AI model to the access network device, and send the sub-AI model inferred on the terminal device side in the first AI model to the terminal device; when the first AI model is a unilateral AI model and the inference is on the terminal device side, the OAM can also send the first AI model to the terminal device; when the first AI model is a unilateral AI model and the inference is on the access network device side, the OAM can also send the first AI model to the access network device.

[0121] In another possible approach, the access network device stores the globally unique model identifiers of multiple AI models. The access network device can determine the globally unique model identifier of the first AI model that requires lifecycle management from the globally unique model identifiers of the multiple AI models, and then synchronize the globally unique model identifier of the first AI model to the terminal device. Optionally, the access network device also stores multiple AI models. When the first AI model is a bilateral AI model, the access network device can also synchronize the sub-AI model inferred on the terminal device side of the first AI model to the terminal device; when the first AI model is a unilateral AI model and the inference is on the terminal device side, the access network device can also synchronize the first AI model to the terminal device.

[0122] In another possible approach, the terminal device stores the globally unique model identifiers of multiple AI models. The terminal device can determine the globally unique model identifier of a first AI model requiring lifecycle management from among the globally unique model identifiers of the multiple AI models, and then synchronize the globally unique model identifier of the first AI model to the access network device. Optionally, the terminal device also stores multiple AI models. If the first AI model is a bilateral AI model, the terminal device can also send the sub-AI model of the first AI model that is inferred on the access network device side to the access network device. If the first AI model is a unilateral AI model and inference is performed on the access network device side, the terminal device can also send the first AI model to the access network device.

[0123] In addition, the access network device and the terminal device also synchronize at least one or more of the following model information of the first AI model: the radio resource control (RRC) parameters of the first AI model, the scenario ID of the first AI model, and the applicable condition ID of the first AI model. The specific method is similar to the above-mentioned method of synchronizing the first AI model. The model information of the first AI model can be synchronized with the first AI model in one message, or it can be synchronized in multiple messages.

[0124] Optionally, the access network device further synchronizes with the terminal device the total number M of AI models requiring lifecycle management, where M is a positive integer. It is understood that the access network device and the terminal device may synchronize the globally unique model identifiers of the M AI models in one or more batches, with the globally unique model identifiers of one or more AI models being synchronized each time.

[0125] For example, the number M of AI models that require lifecycle management for access network devices and terminal devices is equal to 4, that is, the AI ​​models that require lifecycle management for access network devices and terminal devices are AI model 1 to AI model 4, and the globally unique model identifiers of AI model 1 to AI model 4 are respectively represented as GMI-001 to GMI-004. When the access network device and the terminal device synchronize for the first time, GMI-001 and the total number of AI models that require lifecycle management, 4, are synchronized; when the access network device and the terminal device synchronize for the second time, GMI-002 is synchronized; when the access network device and the terminal device synchronize for the third time, GMI-003 is synchronized; and when the access network device and the terminal device synchronize for the fourth time, GMI-004 is synchronized. Furthermore, the first AI model can be any one of the M AI models.

[0126] Optionally, OAM determines the quantity M and synchronizes the quantity M to the access network device and the terminal device. Alternatively, the access network device determines the quantity M and synchronizes the quantity M to the terminal device. Alternatively, the terminal device determines the quantity M and synchronizes the quantity M to the access network device.

[0127] In step 802 , the access network device generates a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier.

[0128] The target identifier may be one or more of the following implementations: (1) bitmap, (2) first bit sequence, (3) second bit sequence, (4) combination of model information and index.

[0129] The following explains how to implement different target identifiers.

[0130] Target identification implementation method: (1) bitmap

[0131] In the first possible approach, the bitmap includes M bits, where M is the total number of AI models that require lifecycle management for the access network device and the terminal device, and the M AI models include the first AI model. For example, the access network device and the terminal device need to perform lifecycle management for AI models 1 to 4, that is, M = 4, and the bitmap includes 4 bits. In one example, AI model 1 corresponds to bitmap: 1000, AI model 2 corresponds to bitmap: 0100, AI model 3 corresponds to bitmap: 0010, and AI model 4 corresponds to bitmap: 0001.

[0132] In the above example, for a bitmap, the N models whose globally unique model identifiers have been synchronized between the access network device and the terminal device correspond to N bits in the bitmap respectively. The N models are included in the M models, and the N AI models include the first AI model. The value of each bit in the N bits can be used to indicate the activation state of the corresponding AI model (for example, activation or deactivation). Exemplarily, when the value of the bit is 1, it indicates that the activation state of the corresponding AI model is activation; when the value of the bit is 0, it indicates that the activation state of the corresponding AI model is deactivation.

[0133] For example, the access network device and the terminal device have synchronized the globally unique model identifiers of AI model 1 to AI model 4, that is, the four bits from left to right in the bitmap correspond to AI model 1 to AI model 4 respectively. For example, the bitmap is 1000. In this bitmap: 1000, the value of the first bit is 1, which is used to indicate that the activation state of AI model 1 is activated; the value of the second bit is 0, which is used to indicate that the activation state of AI model 2 is deactivated; the value of the third bit is 0, which is used to indicate that the activation state of AI model 3 is deactivated; the value of the fourth bit is 0, which is used to indicate that the activation state of AI model 4 is deactivated.

[0134] In this application, the first bit of the bitmap is specifically the first bit from left to right in the bitmap, and the second bit of the bitmap is specifically the second bit from left to right in the bitmap. The others are similar and will not be repeated here.

[0135] When the access network device generates the bitmap of the first AI model, it can specifically be that the access network device generates the bitmap of the first AI model based on the number M of AI models that the access network device and the terminal device need to perform lifecycle management, where the bitmap includes M bits.

[0136] In a specific example, an access network device sorts N AI models based on their globally unique model identifiers to obtain a sorting result. The access network device then generates a bitmap for a first AI model based on the sorting result. Since the globally unique model identifier of the first AI model is at the nth position in the sorting result, the bit at the nth position in the bitmap corresponds to the first AI model, and the value of the bit at the nth position indicates the activation status of the first AI model. It will be understood that the values ​​of other bits indicate the activation status of the AI ​​model corresponding to the bit. n is a positive integer less than or equal to N.

[0137] For example, M=4, and the four AI models are AI model 1 to AI model 4. The access network device and the terminal device have synchronized the globally unique model identifiers of AI model 1 and AI model 2. AI model 2 is the first AI model, wherein the globally unique model identifier of AI model 1 is less than the globally unique model identifier of AI model 2, that is, AI model 1 is before AI model 2. Since other AI models have not yet been synchronized between the access network device and the terminal device, the other AI models are all located after AI model 2. The access network device generates a bitmap of 0100 for AI model 2. The second bit in the bitmap corresponds to AI model 2, and the value of the second bit is 1, which is used to indicate that the activation state of AI model 2 is activated.

[0138] Optionally, the access network device or the terminal device may store a mapping relationship between a synchronized globally unique model identifier (or a model corresponding to a synchronized globally unique model identifier) ​​and a bit in a bitmap. In combination with the above example, the access network device and the terminal device have synchronized AI models with globally unique model identifiers, namely AI model 1 and AI model 2. The globally unique model identifiers of AI model 1 and AI model 2 are GMI-001 and GMI-002 respectively. Then the mapping relationship between the access network device and the terminal device can be: GMI-001 is mapped to the first bit in the bitmap; GMI-002 is mapped to the second bit in the bitmap.

[0139] In the second possible approach, the bitmap includes N bits, where N is the number of AI models whose globally unique model identifiers have been synchronized between the access network device and the terminal device. These N AI models include the first AI model. For example, if the access network device has synchronized the globally unique model identifiers of AI model 1 and AI model 2 with the terminal device, that is, N = 2, then AI model 1 corresponds to bitmap: 10, and AI model 2 corresponds to bitmap: 01.

[0140] Optionally, the N bits correspond to N AI models respectively, and the value of each bit is used to indicate the activation state of the corresponding AI model (such as activation or deactivation). Exemplarily, when the value of the bit is 1, it indicates that the activation state of the corresponding AI model is activation; when the value of the bit is 0, it indicates that the activation state of the corresponding AI model is deactivation. Taking bitmap:10 of AI model 1 as an example, the two bits in bitmap:10 correspond to AI model 1 and AI model 2 respectively, where the value of the first bit is 1, used to indicate that the activation state of AI model 1 is activation; the value of the second bit is 0, used to indicate that the activation state of AI model 2 is deactivation.

[0141] When the access network device generates the bitmap of the first AI model, it can specifically be that the access network device generates a bitmap based on the number N of AI models with globally unique model identifiers that have been synchronized with the terminal device, where the bitmap includes N bits.

[0142] In a specific example, the access network device sorts N AI models according to their globally unique model identifiers to obtain a sorting result, and then generates a bitmap of the first AI model based on the sorting result. Since the globally unique model identifier of the first AI model is located at the nth position in the sorting result, the bit at the nth position in the bitmap corresponds to the first AI model, where n is a positive integer less than or equal to N.

[0143] For example, the access network device and the terminal device have synchronized the globally unique model identifiers of AI models 1 and AI model 2 (i.e., n=2), where AI model 2 is the first AI model, and the globally unique model identifier of AI model 1 is less than the globally unique model identifier of AI model 2, i.e., AI model 1 is before AI model 2. Furthermore, the bitmap of AI model 2 generated by the access network device is 01, and the second bit in the bitmap corresponds to AI model 2.

[0144] Optionally, the access network device or terminal device may store the synchronized globally unique model identifier (or the model corresponding to the synchronized globally unique model identifier) ​​and the bit mapping relationship in the bitmap. For specific instructions, please refer to the description in the first possible method above and will not be repeated here.

[0145] It can be understood that the number of bits in the bitmap in the second possible manner is less than the number of bits in the bitmap in the first possible manner, that is, the second possible manner helps to further reduce signaling overhead.

[0146] In the first implementation method mentioned above, each time the access network device and the terminal device synchronize the globally unique model identifier of a new AI model, it is necessary to reorder the multiple AI models (or the globally unique model identifiers of multiple AI models) that have synchronized the globally unique model identifier, and generate a bitmap corresponding to the newly synchronized globally unique model identifier based on the sorting result obtained by the reordering, and may refresh the mapping relationship between the bits and models in the bitmap.

[0147] For example, the access network device and the terminal device synchronize AI model 1 to AI model 3 in sequence, where the order according to the globally unique model identifier is AI model 1, AI model 3, and AI model 2. When the access network device and the terminal device synchronize the globally unique model identifier of AI model 1, a bitmap:1000 of AI model 1 is generated, where the first bit corresponds to AI model 1. When the access network device and the terminal device synchronize the globally unique model identifier of AI model 2, the sorting result of AI model 1 and AI model 2 is: AI model 1, AI model 2, thereby generating a bitmap:0100 of AI model 2, where the first bit corresponds to AI model 1, and the second bit corresponds to AI model 2. When the access network device and the terminal device synchronize the globally unique model identifier of AI model 3, the sorting result of AI model 1, AI model 2, and AI model 3 is: AI model 1, AI model 3, AI model 2, thereby generating a bitmap:0100 of AI model 3, where the first bit corresponds to AI model 1, the second bit corresponds to AI model 3, and the third bit corresponds to AI model 2. It can be seen that the second bit in the bitmap originally corresponds to AI model 2, but after the third synchronization of the globally unique model identifier of the model, the second bit in the bitmap needs to correspond to AI model 3, and the third bit in the bitmap corresponds to AI model 2, that is, the mapping relationship between the bits and models in the bitmap is refreshed.

[0148] In order to avoid the access network device reordering multiple AI models (or the globally unique model identifiers of multiple AI models) and refreshing the mapping relationship between bits and models in the bitmap after each synchronization of the globally unique model identifier of a new AI model, the present application provides the following third implementation method:

[0149] The access network device can assign a corresponding index to the AI ​​model (or the globally unique model identifier of the AI ​​model) each time the globally unique model identifier of a new AI model is synchronized. The index is used to indicate the number of times the globally unique model identifier of the AI ​​model is synchronized, and the access network device generates a bitmap based on the index. Specifically, when the access network device synchronizes the globally unique model identifier of the first AI model with the terminal device, it determines that the first AI model is the globally unique model identifier of the Nth synchronized AI model, and then assigns index N to the first AI model, and then generates a bitmap for the first AI model, wherein the Nth bit in the bitmap of the first AI model corresponds to the first AI model. Furthermore, the value of the Nth bit is used to indicate the activation status of the first AI model.

[0150] For example, the access network device and the terminal device synchronize AI model 1 to AI model 3 in sequence. When the access network device and the terminal device synchronize the globally unique model identifier of AI model 1, index 1 is assigned to the globally unique model identifier of AI model 1, and then bitmap:1000 of AI model 1 is generated, wherein the first bit corresponds to index 1 and AI model 1; when the access network device and the terminal device synchronize the globally unique model identifier of AI model 2, index 2 is assigned to the globally unique model identifier of AI model 2, and then bitmap:0100 of AI model 2 is generated, wherein the first bit corresponds to index 1 and AI model 1, and the second bit corresponds to index 2 and AI model 2; when the access network device and the terminal device synchronize the globally unique model identifier of AI model 3, index 3 is assigned to the globally unique model identifier of AI model 3, and then bitmap:0010 of AI model 1 is generated, wherein the first bit corresponds to index 1 and AI model 1, the second bit corresponds to index 2 and AI model 2, and the third bit corresponds to index 3 and AI model 3.

[0151] In this way, the bitmap is generated based on the index order, without the need to re-sort multiple AI models (or the globally unique model identifiers of multiple AI models) and refresh the mapping relationship between bits and models in the bitmap, which helps avoid unnecessary work for access network equipment.

[0152] Similarly, based on the second implementation above, this application provides a fourth implementation. The difference between the fourth implementation and the third implementation is that the bitmap of the first AI model only includes the first N bits. No further details will be given.

[0153] It should also be added that, when the present application sorts the globally unique model identifiers, it can be sorted from small to large, that is, each bit from left to right in the bitmap corresponds to the globally unique model identifiers from small to large (or the AI ​​model corresponding to the globally unique model identifier); or, the present application sorts the globally unique model identifiers, it can be sorted from large to small, that is, each bit from left to right in the bitmap corresponds to the globally unique model identifiers from large to small (or the AI ​​model corresponding to the globally unique model identifier), which is not limited in the present application. Similarly, when the present application sorts the indexes corresponding to the globally unique model identifiers, it can be sorted from small to large or from large to small. For the convenience of description, the present application takes the sorting of the globally unique model identifiers (or the indexes corresponding to the globally unique model identifiers) from small to large as an example.

[0154] It is worth mentioning that when the number M of models that need lifecycle management is small, the gain brought by the bitmap approach is more obvious, because the bitmap can be configured and processed at the MAC layer, and the processing latency is low.

[0155] It is also worth mentioning that each bit in the above bitmap corresponds to a model. In another possible manner, multiple bits in the bitmap may correspond to a model. Furthermore, multiple states of the model can be indicated based on the values ​​of the multiple bits. For example, two bits in the bitmap correspond to a model. When the values ​​of the two bits are 00, it indicates that the model is in state 0; when the values ​​of the two bits are 10, it indicates that the model is in state 1; when the values ​​of the two bits are 01, it indicates that the model is in state 2; when the values ​​of the two bits are 11, it indicates that the model is in state 3; of course, there can be other examples, which are not limited in this application.

[0156] Target identification implementation method: (2) First bit sequence

[0157] Specifically, the first bit sequence includes m bits, and the m bits have M decimal values, each of which corresponds to M AI models. The M AI models are AI models for which lifecycle management is required for access network devices and terminal devices. For example, M = 4, and the four AI models are AI models 1 to 4, meaning that the access network devices and terminal devices need to perform lifecycle management for AI models 1 to 4. Furthermore, the first bit sequence includes 2 bits: the first bit sequence of AI model 1 is 00, the first bit sequence of AI model 2 is 10, the first bit sequence of AI model 2 is 10, and the first bit sequence of AI model 4 is 11.

[0158] When the access network device generates the first bit sequence of the first AI model, it can be that the access network device determines that the first bit sequence of each AI model includes m bits based on the number M of AI models that the access network device and the terminal device need to perform lifecycle management, and then determines the value of each bit in the first bit sequence of each AI model.

[0159] Target identification implementation method: (3) Second bit sequence

[0160] Specifically, the second bit sequence is composed of multiple bits at preset positions in the globally unique model identifier. It can be understood that the second bit sequences corresponding to any two AI models in the M AI models are different. For example, the access network device and the terminal device need to perform lifecycle management on AI models 1 to 4. The globally unique model identifiers of AI models 1 to 3 are 00...0001, 00...0010, 00...0011, respectively. Then the second bit sequences of AI models 1 to 3 can be the last four bits of their respective globally unique model identifiers, i.e., 0001, 0010, 0011.

[0161] When the access network device generates the second bit sequence of the first AI model, the access network device may specifically predetermine a preset position (for example, the last K bits of the globally unique model identifier, where K is a positive integer), and then select one or more bits at the preset position from the globally unique model identifier of the first AI model to form the second bit sequence.

[0162] Target identification implementation method: (4) model information of the AI ​​model, or a combination of the model information of the AI ​​model and the index of the AI ​​model under the model information.

[0163] In one example, the model information corresponding to any two of the M AI models is different. For example, the access network device and the terminal device need to perform lifecycle management on AI models 1 to 4. The model information of AI models 1 to 4 is model information 1 to model information 4, respectively, where any two of model information 1 to model information 4 are different.

[0164] When the access network device generates the target identifier of the first AI model, the access network device may specifically determine the model information of the first AI model and use the model information of the first AI model as the target identifier of the first AI model.

[0165] In another example, multiple AI models among the M AI models correspond to the same model information, and the AI ​​models are further identified by a combination of the model information and the different indexes of each AI model under the model information. For example, the access network device and the terminal device need to perform lifecycle management for AI models 1 to 4. The model information of AI models 1 and 2 is model information 1, and the model information of AI models 3 and 4 is model information 2. In this case, AI model 1 is identified by model information 1 + index 1, AI model 2 is identified by model information 1 + index 2, AI model 3 is identified by model information 2 + index 1, and AI model 4 is identified by model information 2 + index 2.

[0166] The model information may be one or more of the following: radio resource control (RRC) parameters, a scenario ID, and an applicable condition ID.

[0167] When the access network device generates the target identifier of the first AI model, the access network device may specifically determine the model information of the first AI model and the index of the first AI model under the model information to obtain a combination of the two, and use the combination of the first AI model as the target identifier of the first AI model.

[0168] It can be understood that the access network device pre-stores a rule for generating the target identifier of the first AI model (referred to as the target rule), and then the access network device can generate the target identifier of the first AI model according to the target rule. For the convenience of description, the rule for generating the bitmap of the first AI model is referred to as the first rule, the rule for generating the first bit sequence of the first AI model is referred to as the second rule, the rule for generating the second bit sequence of the first AI model is referred to as the third rule, and the rule for generating the model information / combination corresponding to the first AI model is referred to as the fourth rule. The target rule can be any one of the first to fourth rules.

[0169] In one possible embodiment, after the access network device generates a target identifier of the first AI model according to the target rule, it can send the target identifier of the first AI model to the terminal device. In this way, both the access network device and the terminal device can perform lifecycle management of the first AI model based on the target identifier of the first AI model (i.e., step 803).

[0170] Furthermore, when the access network device generates a bitmap based on the index, the access network device can also send the index N assigned by the access network device to the first AI model to the terminal device. Accordingly, after receiving the index N, the terminal device can determine that the Nth bit in the bitmap corresponds to the first AI model. Alternatively, the access network device and the terminal device each generate an index corresponding to each globally unique model identifier based on the synchronization order of the globally unique model identifier. That is, when the access network device and the terminal device synchronize the globally unique model identifier of the first AI model, each generates an index N corresponding to the globally unique model identifier of the first AI model. The access network device sends the bitmap of the first AI model to the terminal device, and the terminal device determines that the Nth bit in the bitmap corresponds to the first AI model based on the locally generated index N. Furthermore, the terminal device can also determine the activation state of the first AI model based on the value of the Nth bit in the bitmap.

[0171] In this way, the terminal device does not need to re-sort multiple AI models (or the globally unique model identifiers of multiple AI models) and refresh the mapping relationship between bits and models in the bitmap, which helps avoid unnecessary work for the terminal device.

[0172] In another possible embodiment, the access network device can also send the target rule to the terminal device. Accordingly, the terminal device also generates a target identifier of the first AI model according to the target rule (the specific generation method can refer to the above-mentioned method for the access network device to generate the target identifier of the first AI model, and "access network device" can be replaced with "terminal device" for understanding). In this way, both the access network device and the terminal device can perform lifecycle management of the first AI model based on the target identifier of the first AI model (i.e., step 803).

[0173] Furthermore, when the access network device sends the target rule to the terminal device:

[0174] In one example, the access network device directly sends a rule (i.e., a target rule) to the terminal device, and instructs the terminal device to generate a target identifier for the first AI model according to the target rule. Optionally, the access network device first obtains the capability information of the terminal device. For example, the access network device sends a reporting instruction to the terminal device, and the reporting instruction is used to instruct the terminal device to report the capability information of the terminal device to the access network device. Accordingly, the terminal device reports the capability information of the terminal device to the access network device in response to the reporting instruction. Furthermore, the access network device determines a target rule that matches the capability information of the terminal device based on the capability information of the terminal device, and then sends the target rule to the terminal device.

[0175] In another example, an access network device sends a rule set to a terminal device, the rule set including one or more of the first to fourth rules. Accordingly, after receiving the rule set from the access network device, the terminal device selects a rule from the rule set as a target rule and sends the target rule to the access network device. Alternatively, the terminal device may select a rule from the rule set as a target rule based on the terminal device's capability information.

[0176] Among them, the capability information of the terminal device includes at least one or more of the following: whether the terminal device supports dynamic capability reporting, data set transmission capability, data collection capability, model training capability, model reasoning capability, model monitoring capability, model switching capability, and model update capability.

[0177] Furthermore, when the access network device generates a bitmap based on the index, the target rule sent by the access network device to the terminal device may also include instruction information, which is used to instruct the access network device and the terminal device to each generate an index for the AI ​​model corresponding to the globally unique model identifier according to the synchronization order of the globally unique model identifiers. Accordingly, when the access network device generates a bitmap based on the index N corresponding to the globally unique model identifier of the first AI model, the terminal device may also generate a bitmap based on the index N corresponding to the globally unique model identifier of the first AI model.

[0178] In addition, in the present application, the way in which the access network device and the terminal device negotiate the target rules can also be that the access network device and the terminal device are both pre-configured with a rule set, the access network device selects the target rule from the locally pre-configured rule set, and sends the identifier of the target rule to the terminal device; or, the terminal device selects the target rule from the locally pre-configured rule set and sends the identifier of the target rule to the access network device.

[0179] It should be added that the above only exemplifies the first to fourth rules. This application may also have other rules (such as the fifth rule, the sixth rule, etc.). Accordingly, the target rule can also be other rules, or the rule set sent by the access network device to the terminal device also includes other rules, which is not limited in this application.

[0180] In step 803, the access network device performs lifecycle management on the first AI model based on the target identifier of the first AI model. Correspondingly, the terminal device performs lifecycle management on the first AI model based on the target identifier of the first AI model. Alternatively, it can be understood that the access network device and the terminal device use the target identifier of the first AI model during interaction to implement lifecycle management of the first AI model.

[0181] In the above technical solution, the access network device includes a correspondence between the target identifier of the first AI model and the globally unique model identifier of the first AI model, and the terminal device also includes a correspondence between the target identifier of the first AI model and the globally unique model identifier of the first AI model. During the lifecycle management of the first AI model, both the access network device and the terminal device may no longer use the globally unique model identifier of the first AI model, but instead use the target identifier of the first AI model. This helps to reduce signaling overhead.

[0182] Among them, lifecycle management may include at least one or more of the following: activating or deactivating the first AI model, updating parameters in the first AI model, monitoring the performance of the first AI model, and returning the first AI model to the traditional mode.

[0183] The following takes activating the first AI model as an example:

[0184] In one possible manner, the access network device sends a target identifier and an activation indication of the first AI model to the terminal device. Correspondingly, the terminal device activates the first AI model according to the target identifier and the activation indication of the first AI model.

[0185] In another possible approach, the target identifier of the first AI model may specifically be a bitmap of the first AI model, and the value of the bit corresponding to the first AI model in the bitmap of the first AI model is used to indicate the activation status of the first AI model. Specifically, the access network device sends the bitmap of the first AI model to the terminal device, and the terminal device activates the first AI model based on the bitmap of the first AI model. Exemplarily, the terminal device may sort N AI models (or N AI models' globally unique model identifiers) based on the globally unique model identifiers of the N AI models synchronized with the access network device to obtain a sorting result, and then determine, based on the sorting result, a mapping relationship between the N AI models and the bits in the bitmap of the first AI model. For example, if the first AI model corresponds to bitmap:0100, the terminal device determines that the second bit in the bitmap corresponds to the first AI model, and the value of the second bit in the bitmap is 1, the terminal device may activate the first AI model based on the value of the second bit in the bitmap.

[0186] Optionally, after step 803, the access network device may also synchronize the globally unique model identifier of the second AI model that requires lifecycle management with the terminal device, and determine the target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model (the specific implementation method can be found in the description of steps 801 and 802 above, replacing "first AI model" with "second AI model" for understanding).

[0187] Furthermore, in an AI model switching scenario, the access network device may send a target identifier and deactivation instruction for the first AI model, as well as a target identifier and activation instruction for the second AI model, to the terminal device. Accordingly, the terminal device deactivates the first AI model based on the target identifier and deactivation instruction for the first AI model, and activates the second AI model based on the target identifier and activation instruction for the second AI model. In this manner, the terminal device switches the AI ​​model from the first AI model to the second AI model. Specifically, the AI ​​model switching scenario may involve performance degradation of the first AI model due to movement of the terminal device, making the first AI model no longer suitable for the current channel state between the access network device and the terminal device, necessitating a switch to a more suitable second AI model.

[0188] Furthermore, when the target identifier of the second AI model is the bitmap of the second AI model, the first AI model can be deactivated by taking the value of the bit corresponding to the first AI model in the bitmap of the second AI model, and the second AI model can be activated by taking the value of the bit corresponding to the second AI model in the bitmap of the second AI model.

[0189] For example, the second AI model corresponds to bitmap:0010. The second bit in the bitmap of the second AI model corresponds to the first AI model, and the value of the second bit is 0. The terminal device can activate the first AI model based on the value of the second bit; the third bit in the bitmap of the second AI model corresponds to the second AI model, and the value of the third bit is 1. The terminal device can activate the second AI model based on the value of the third bit.

[0190] It can be understood that the first bit in the bitmap of the second AI model also corresponds to the third AI model. The globally unique model identifier of the third AI model is synchronized before the globally unique model identifier of the second AI model, and the order based on the globally unique model identifier (or AI model index) is the third AI model, the first AI model, and the second AI model. When the value of the first bit is 0, the terminal device can deactivate the third AI model based on the value of the first bit; when the value of the first bit is 1, the terminal device can activate the third AI model based on the value of the first bit.

[0191] It should be added that the terminal device in this application can be one or more. That is, when there are multiple terminal devices, the access network device can synchronize the globally unique model identifier of the first AI model that needs to be managed for lifecycle with the multiple terminal devices. Furthermore, in one example, the access network device sends the target identifier of the first AI model generated by itself to the multiple terminal devices respectively, and then the access network device and each terminal device perform lifecycle management through the target identifier of the first AI model. In another example, the access network device sends a target rule to each of the multiple terminal devices. The access network device and the multiple terminal devices all generate the target identifier of the first AI model based on the target rule, and then the access network device and each terminal device perform lifecycle management through the target identifier of the first AI model. In another example, the access network device sends a rule set to each of the multiple terminal devices. It can be considered that each terminal device selects a target rule from the rule set and reports the target rule to the access network device. That is, the target rules negotiated by the access network device with different terminal devices can be the same or different. Accordingly, the target identifiers of the first AI model used by the access network device when performing lifecycle management with different terminal devices can be the same or different.

[0192] To better explain the embodiments of the present application, four specific examples are provided below:

[0193] As shown in Figure 9, which is a flowchart of the first specific example provided in this application, there are 4 AI models (i.e., M=4) that require lifecycle management for access network devices and terminal devices. The access network device generates a bitmap based on the synchronized global unique model identifier, and the bitmap includes 4 bits.

[0194] In step 901, the access network device and the terminal device synchronize GMI-003, where GMI-003 is the globally unique model identifier of the first AI model. The first AI model can be considered to be AI model GMI-003.

[0195] In step 902, the access network device determines the bitmap:1000 corresponding to GMI-003 based on the synchronized globally unique model identifier. Specifically, the currently synchronized globally unique model identifier is only GMI-003, that is, GMI-003 is ranked first in the order. Therefore, in the generated bitmap:1000, the first bit corresponds to the AI ​​model GMI-003 (or corresponds to GMI-003). Exemplarily, the access network device maintains a mapping relationship between the synchronized globally unique model identifier and the bits in bitmap:1000 (see Table 1 in Figure 9).

[0196] In step 903, the access network device sends bitmap: 1000 to the terminal device. Accordingly, the terminal device receives bitmap: 1000 from the access network device. For example, the terminal device maintains a synchronized mapping relationship between a globally unique model identifier and bits in bitmap: 1000 (see Table 1 in FIG9 ).

[0197] In step 904, the terminal device activates AI model GMI-003 (ie, the first AI model).

[0198] In step 905, the access network device and the terminal device synchronize GMI-001, where GMI-001 is the globally unique model identifier of the second AI model. The second AI model can be considered to be the AI ​​model GMI-001.

[0199] In step 906 , the access network device sorts the two synchronized globally unique model identifiers according to the two synchronized globally unique model identifiers to obtain a sorting result, and further determines the bitmap: 1000 corresponding to GMI-001.

[0200] Specifically, the currently synchronized globally unique model identifiers are GMI-003 and GMI-001. Furthermore, GMI-001 is smaller than GMI-003, that is, GMI-001 is ranked first in the order and GMI-003 is ranked second in the order. Therefore, in the generated bitmap:1000, the first bit corresponds to the AI ​​model GMI-001 (or corresponds to GMI-001), and the second bit in bitmap:1000 corresponds to the AI ​​model GMI-003 (or corresponds to GMI-003). Exemplarily, the access network device maintains a mapping relationship between the two synchronized globally unique model identifiers and the bits in bitmap:1000 (see Table 2 in Figure 9).

[0201] In step 907, the access network device sends bitmap: 1000 to the terminal device. Accordingly, the terminal device receives bitmap: 1000 from the access network device. For example, the terminal device maintains a mapping relationship between two synchronized globally unique model identifiers and bits in bitmap: 1000 (see Table 2 in FIG9 ).

[0202] In step 908, the terminal device activates AI model GMI-001 and deactivates AI model GMI-003.

[0203] The embodiment related to FIG9 is a specific implementation of the embodiment related to FIG8 , and the specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not fully described in step 901 can be referred to the description of step 801; any content not fully described in steps 902 and 903 can be referred to the description of step 802; any content not fully described in steps 904 to 908 can be referred to the description of step 803.

[0204] As shown in Figure 10, which is a flowchart of the second specific example provided in this application, there are 4 AI models (i.e., M=4) that require lifecycle management for access network devices and terminal devices. The access network device generates a bitmap based on the index corresponding to the synchronized global unique model identifier, and the bitmap includes 4 bits.

[0205] Step 1001: The access network device and the terminal device synchronize GMI-003 and index 1.

[0206] GMI-003 is the globally unique model identifier of the first AI model, and can be considered to be AI model GMI-003. Index 1 is used to indicate that GMI-003 is the first globally unique model identifier synchronized between the access network device and the terminal device. Furthermore, the access network device can associate GMI-003 with Index 1, and the terminal device can also associate GMI-003 with Index 1.

[0207] In step 1002, the access network device generates a bitmap:1000 corresponding to GMI-003 based on index 1 corresponding to GMI-003, where the first bit corresponds to index 1 and AI model GMI-003 (or corresponds to GMI-003). Exemplarily, the access network device maintains a synchronized mapping between a globally unique model identifier, index, and bits in bitmap:1000 (see Table 1 in FIG10 ).

[0208] In step 1003, the access network device sends bitmap:1000 to the terminal device. Accordingly, the terminal device receives bitmap:1000 from the access network device. Specifically, the terminal device determines that the first bit in bitmap:1000 corresponds to index 1, i.e., GMI-003. Exemplarily, the terminal device maintains a mapping relationship between a synchronized globally unique model identifier, the index corresponding to the globally unique model identifier, and the bits in bitmap:1000 (see Table 1 in FIG10 ).

[0209] Step 1004: The terminal device activates AI model GMI-003 (i.e., the first AI model).

[0210] Step 1005: The access network device and the terminal device synchronize GMI-001 and index 2.

[0211] GMI-001 is the globally unique model identifier of the second AI model, and can be considered to be AI model GMI-001. Index 2 is used to indicate that GMI-001 is the second globally unique model identifier synchronized between the access network device and the terminal device. Furthermore, the access network device can associate GMI-001 with Index 2, and the terminal device can also associate GMI-001 with Index 2.

[0212] In step 1006, the access network device generates a bitmap:0100 corresponding to GMI-001 based on index 2 corresponding to GMI-001. The first bit in bitmap:0100 corresponds to index 1, AI model GMI-003 (or corresponds to GMI-003), and the second bit in bitmap:0100 corresponds to index 2, AI model GMI-001 (or corresponds to GMI-001). Exemplarily, the access network device maintains a mapping relationship between two synchronized globally unique model identifiers, the indexes corresponding to the two globally unique model identifiers, and the bits in bitmap:1000 (see Table 2 in FIG10 ).

[0213] In step 1007, the access network device sends bitmap:0100 to the terminal device. Accordingly, the terminal device receives bitmap:0100 from the access network device. Specifically, the terminal device determines that the first bit in bitmap:0100 corresponds to index 1, that is, corresponds to GMI-003. Also, the terminal device determines that the second bit in bitmap:0100 corresponds to index 2, that is, corresponds to GMI-001. Exemplarily, the terminal device maintains a mapping relationship between two synchronized globally unique model identifiers, the indexes corresponding to the two globally unique model identifiers, and the bits in bitmap:1000 (see Table 2 in Figure 10).

[0214] In step 1008, the terminal device activates AI model GMI-001 and deactivates AI model GMI-003.

[0215] The embodiment of FIG10 is a specific implementation of the embodiment of FIG8 . The specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not described in detail in step 1001 can be referred to the description of step 801; any content not described in detail in steps 1002 and 1003 can be referred to the description of step 802; any content not described in detail in steps 1004 to 1008 can be referred to the description of step 803.

[0216] As shown in Figure 11, which is a flowchart of the third specific example provided in this application, there are 4 AI models (i.e., M=4) that need to be managed for lifecycle by the access network device and the terminal device. The access network device sends target rules to the terminal device, and then the access network device and the terminal device synchronize the globally unique model identifier of the first AI model, and then each generates a target identifier of the first AI model according to the target rules.

[0217] In step 1101, the access network device and the terminal device synchronize GMI-003, where GMI-003 is the globally unique model identifier of the first AI model. The first AI model can be considered to be AI model GMI-003.

[0218] Step 1102: The access network device sends a target rule to the terminal device. Correspondingly, the terminal device receives the target rule from the access network device.

[0219] The target rule may also be referred to as a mapping rule configuration, and the target rule may be one of the first to fourth rules. Optionally, before the access network device sends the target rule to the terminal device, it obtains capability information of the terminal device and determines the target rule based on the capability information of the terminal device.

[0220] Optionally, after receiving the target rule from the access network device, the terminal device may also send a first confirmation message to the access network device, where the first confirmation message is used to indicate that the terminal device can use the target rule to determine the target identifier of the first AI model.

[0221] In this application, the order between step 1101 and step 1102 is not limited.

[0222] In step 1103, the access network device determines the target identifier of the first AI model according to the target rule.

[0223] In step 1104, the terminal device determines the target identifier of the first AI model according to the target rule.

[0224] In this application, the order between step 1103 and step 1104 is not limited.

[0225] Optionally, after determining the target identifier of the first AI model, the terminal device may further send a second confirmation message to the access network device, where the second confirmation message is used to indicate that the terminal device is able to determine the target identifier of the first AI model using the target rule. Exemplarily, the second confirmation message includes the target identifier of the first AI model.

[0226] Furthermore, after receiving the second confirmation message from the terminal device, the access network device obtains the target identifier of the first AI model from the second confirmation message and compares it with the target identifier of the first AI model determined by itself to determine whether the two are consistent. If they are consistent, it is determined that the terminal device can perform lifecycle management of the first AI model based on the target identifier of the first AI model. In other words, the access network device and the terminal device execute the following step 1105.

[0227] It can be understood that in this application, the terminal device can send a first confirmation message or a second confirmation message to the access network device.

[0228] In step 1105, the access network device and the terminal device perform lifecycle management on the first AI model according to the target identifier of the first AI model.

[0229] The embodiment of FIG11 is a specific implementation of the embodiment of FIG8 . The specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not fully described in step 1101 can be referred to the description of step 801; any content not fully described in steps 1102 to 1104 can be referred to the description of step 802; and any content not fully described in step 1105 can be referred to the description of step 803.

[0230] As shown in Figure 12, which is a flowchart of the fourth specific example provided in this application, there are 4 AI models (i.e., M=4) that need to be managed for lifecycle in the access network device and the terminal device. The access network device sends a rule set to the terminal device, and the terminal device selects the target rule from the rule set and reports the selected target rule to the access network device. Then, after synchronizing the globally unique model identifier of the first AI model, the access network device and the terminal device each generate a target identifier for the first AI model according to the target rule.

[0231] In step 1201, the access network device and the terminal device synchronize GMI-003, where GMI-003 is the globally unique model identifier of the first AI model. The first AI model can be considered to be AI model GMI-003.

[0232] Step 1202: The access network device sends a rule set to the terminal device. Correspondingly, the terminal device receives the rule set from the access network device.

[0233] The rule set may also be referred to as a mapping rule configuration set, and the rule set includes at least one or more of the first to fourth rules.

[0234] In this application, the order between step 1201 and step 1202 is not limited.

[0235] Step 1203: The terminal device selects a target rule from the rule set.

[0236] Exemplarily, the terminal device selects a target rule from the rule set according to its own capability information. For example, the terminal device selects the first rule in the rule set as the target rule.

[0237] Step 1204: The terminal device sends the target rule (or the identifier of the target rule) to the access network device. Correspondingly, the access network device receives the target rule (or the identifier of the target rule) from the terminal device.

[0238] In step 1205, the access network device determines the target identifier of the first AI model according to the target rule.

[0239] In step 1206, the terminal device determines the target identifier of the first AI model according to the target rule.

[0240] In this application, the order between step 1205 and step 1206 is not limited.

[0241] Step 1207: The access network device and the terminal device perform lifecycle management based on the target identifier of the first AI model.

[0242] Optionally, after determining the target rule and determining the target identifier of the first AI model based on the target rule, the terminal device may also send the target identifier of the first AI model to the access network device. Accordingly, the access network device performs lifecycle management with the terminal device based on the target identifier of the first AI model. In this manner, the terminal device may not send the target rule (or the identifier of the target rule) to the access network device. The access network device may determine the target rule (or the identifier of the target rule) selected by the terminal device based on the target identifier of the first AI model from the terminal device.

[0243] The embodiment of FIG12 is a specific implementation of the embodiment of FIG8 . The specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not described in detail in step 1201 can be referred to the description of step 801; any content not described in detail in steps 1202 to 1206 can be referred to the description of step 802; and any content not described in detail in step 1207 can be referred to the description of step 803.

[0244] In the above technical solution, the access network device and the terminal device use the target identifier of the first AI model to manage the lifecycle of the first AI model. Compared with using the first AI model's globally unique model identifier to manage the lifecycle of the first AI model, this can effectively reduce air interface signaling overhead during lifecycle management. Furthermore, the access network device can activate or deactivate the AI ​​model using a bitmap, achieving dynamic configuration.

[0245] In this application, the access network device includes a CU and a DU, and the CU and the DU communicate through the F1 interface, that is, the access network device has a CU-DU separation architecture. In the separation architecture, the DU generates a target identifier for the first AI model and then sends the target identifier of the first AI model to the terminal device, or the DU sends the target identifier of the first AI model to the terminal device through the CU. For the specific process, see the communication method exemplified in Figure 13.

[0246] In step 1301, the DU synchronizes with the terminal device the globally unique model identifier of the first AI model that requires lifecycle management.

[0247] In one possible way, the DU stores the globally unique model identifiers of multiple AI models. The DU can determine the globally unique model identifier of the first AI model that needs to be lifecycle managed from the globally unique model identifiers of multiple AI models, and then synchronize the globally unique model identifier of the first AI model to the terminal device. In one example, the DU sends the globally unique model identifier of the first AI model to the CU, and then the CU sends the globally unique model identifier of the first AI model to the terminal device. Optionally, after receiving the globally unique model identifier of the first AI model from the CU, the terminal device sends an ACK (acknowledgement, ACK) to the CU, and accordingly, after receiving the ACK from the terminal device, the CU sends an ACK to the DU. In another example, the DU sends the globally unique model identifier of the first AI model to the terminal device. Optionally, after receiving the globally unique model identifier of the first AI model from the DU, the terminal device sends an ACK to the DU.

[0248] Optionally, the DU also stores multiple AI models. When the first AI model is a bilateral AI model, the DU can also synchronize the sub-AI model inferred on the terminal device side in the first AI model to the terminal device; when the first AI model is a unilateral AI model and the inference is on the terminal device side, the DU can also synchronize the first AI model to the terminal device. The specific synchronization method is similar to the way in which the DU synchronizes the globally unique model identifier of the first AI model to the terminal device.

[0249] Furthermore, if the first AI model is stored in the CU, the DU may first request the first AI model from the CU. For example, the DU sends a request message to the CU, including the globally unique model identifier of the first AI model. In response to the request message, the CU sends the first AI model to the DU.

[0250] In another possible embodiment, the terminal device stores the globally unique model identifiers of multiple AI models. The terminal device can determine the globally unique model identifier of the first AI model that needs to be lifecycle managed from the globally unique model identifiers of the multiple AI models, and then synchronize the globally unique model identifier of the first AI model to the DU. In one example, the terminal device can send the globally unique model identifier of the first AI model to the CU, and then the CU sends the globally unique model identifier of the first AI model to the DU. Optionally, after receiving the globally unique model identifier of the first AI model from the CU, the DU sends an ACK to the CU, and accordingly, after receiving the ACK from the DU, the CU sends an ACK to the terminal device. In another example, the terminal device sends the globally unique model identifier of the first AI model to the DU. Optionally, after receiving the globally unique model identifier of the first AI model from the terminal device, the DU sends an ACK to the terminal device.

[0251] Optionally, the terminal device also stores multiple AI models. When the first AI model is a bilateral AI model, the terminal device can also synchronize the sub-AI model inferred on the DU side in the first AI model to the DU; when the first AI model is a unilateral AI model and the inference is on the DU side, the terminal device can also synchronize the first AI model to the DU. The specific synchronization method is similar to the method in which the terminal device synchronizes the globally unique model identifier of the first AI model to the DU.

[0252] In addition, the model information of the first AI model and the number M of AI models that require lifecycle management are also synchronized between the DU and the terminal device. For specific implementation methods, please refer to the description of synchronizing the AI ​​model and the globally unique model identifier of the AI ​​model between the DU and the terminal device, as well as the description in the above step 801.

[0253] In step 1302 , the DU determines a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model.

[0254] Step 1303: The DU sends the target identifier of the first AI model to the terminal device.

[0255] In one example, the DU sends the target identifier of the first AI model to the CU, and then the CU sends the target identifier of the first AI model to the terminal device. Optionally, after receiving the target identifier of the first AI model from the CU, the terminal device sends an ACK to the CU. Correspondingly, after receiving the ACK from the terminal device, the CU sends an ACK to the DU. This example is applicable to the case where the lifecycle management process is closed between the CU and the terminal device.

[0256] In another example, the DU directly sends the target identifier of the first AI model to the terminal device. Optionally, after receiving the target identifier of the first AI model from the DU, the terminal device sends an ACK to the DU. This example is applicable to the case where the lifecycle management process is closed only between the DU and the terminal device.

[0257] The embodiment of FIG13 is a specific implementation of the embodiment of FIG8 , and the specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not fully described in step 1301 can be referred to the description of step 801; any content not fully described in steps 1302 and 1303 can be referred to the description of step 802.

[0258] To better explain the embodiments of the present application, two specific examples are provided below:

[0259] As shown in Figure 14, which is a flowchart of the first specific example provided in this application, the first AI model and the globally unique model identifier of the first AI model are stored in the DU, and the lifecycle management process is closed between the CU and the terminal device.

[0260] In step 1401, the DU sends GMI-003 to the CU. In response, the CU receives GMI-003. GMI-003 is the globally unique model identifier of the first AI model, and the first AI model can be considered to be AI model GMI-003.

[0261] Step 1402: The CU sends GMI-003 to the terminal device, and the terminal device receives GMI-003 accordingly.

[0262] Step 1403: DU generates bitmap 1000 corresponding to GMI-003.

[0263] In step 1404, the DU sends bitmap: 1000 to the CU. Correspondingly, the CU receives bitmap: 1000.

[0264] In step 1405, the CU sends bitmap: 1000 to the terminal device. Correspondingly, the terminal device receives bitmap: 1000.

[0265] The embodiment related to FIG14 is a specific implementation of the embodiment related to FIG13 , and the specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not fully described in steps 1401 and 1402 can be referred to the description of step 1301; any content not fully described in step 1403 can be referred to the description of step 1302; and any content not fully described in steps 1404 and 1405 can be referred to the description of step 1303.

[0266] For the description of GMI-003 and bitmap in FIG14 , reference can be made to the description of the relevant embodiments in FIG9 to FIG12 .

[0267] As shown in Figure 15, which is a flowchart of the second specific example provided in this application, the first AI model and the globally unique model identifier of the first AI model are stored in the terminal device, and the lifecycle management process is closed between the CU and the terminal device.

[0268] In step 1501, the terminal device sends GMI-003 to the CU. The CU receives GMI-003. GMI-003 is the globally unique model identifier of the first AI model, and the first AI model can be considered to be AI model GMI-003.

[0269] Step 1502: CU sends GMI-003 to DU, and DU receives GMI-003 accordingly.

[0270] Step 1503: DU generates bitmap 1000 corresponding to GMI-003.

[0271] Step 1504: DU sends bitmap: 1000 to CU. Correspondingly, CU receives bitmap: 1000.

[0272] In step 1505, the CU sends bitmap: 1000 to the terminal device. Correspondingly, the terminal device receives bitmap: 1000.

[0273] The embodiment related to Figure 15 is a specific implementation of the embodiment related to Figure 13 , and the specific implementation methods of the two embodiments can be referenced to each other. Specifically, any content not fully described in steps 1501 and 1502 can be referred to the description of step 1301; any content not fully described in step 1503 can be referred to the description of step 1302; and any content not fully described in steps 1504 and 1505 can be referred to the description of step 1303.

[0274] For the description of GMI-003 and bitmap in FIG15 , please refer to the description in the related embodiments of FIG9 to FIG12 .

[0275] It can be understood that after step 1303, step 1405 or step 1505, the lifecycle management of the AI ​​model is also performed between the DU and the terminal device, or the lifecycle management of the AI ​​model is also performed between the DU, CU and the terminal device. For specific instructions, please refer to the description in the above step 803.

[0276] It should be noted that in Figure 13 above, the DU can also synchronize the index N of the first AI model with the terminal device. For specific synchronization methods, see the method for synchronizing the globally unique model identifier of the first AI model between the DU and the terminal device. For example, the DU and the terminal device can synchronize the globally unique model identifier and index N of the first AI model together.

[0277] In one possible approach, the DU stores multiple AI models and their globally unique model identifiers. After the DU determines a first AI model requiring lifecycle management from among the multiple AI models, it assigns the first AI model an index N and then synchronizes the first AI model's globally unique model identifier and index N to the terminal device. In one specific implementation, the DU sends the first AI model's globally unique model identifier and index N to the CU, which then sends the first AI model's globally unique model identifier and index N to the terminal device. Optionally, the DU also sends the first AI model to the CU, and in response, the CU also sends the first AI model to the terminal device.

[0278] In another possible implementation, a CU stores multiple AI models and their globally unique model identifiers, and the multiple AI models include a first AI model. The DU determines the globally unique model identifier of the first AI model requiring lifecycle management and sends a request message to the CU, including the globally unique model identifier of the first AI model. In response to the request message, the CU assigns the first AI model index N to the first AI model and sends the first AI model index N to the DU. Optionally, in response to the request message, the CU also sends the globally unique model identifier and index N of the first AI model to the terminal device. Optionally, the CU also sends the first AI model to the terminal device.

[0279] In another possible implementation, a CU stores multiple AI models and their globally unique model identifiers, including a first AI model. The DU assigns the first AI model an index N, and instructs the CU to send the first AI model's globally unique model identifier and index N to a terminal device. Exemplarily, the DU determines the globally unique model identifier of the first AI model requiring lifecycle management, assigns the index N to the first AI model, and sends a request message to the CU, the request message including the globally unique model identifier and index N of the first AI model. In response, the CU sends the globally unique model identifier and index N of the first AI model to the terminal device. Optionally, the CU further sends the first AI model to the terminal device in response to the request message.

[0280] In addition, in the scenario where the access network device directly sends a rule (i.e., target rule) to the terminal device:

[0281] In one possible approach, the DU generates a target rule and sends it to the terminal device; or, the DU generates a target rule and sends it to the CU; correspondingly, the CU receives the target rule from the DU and sends it to the terminal device. In another possible approach, the CU generates a target rule and sends it to the DU and the terminal device respectively. Furthermore, the DU and the terminal device each generate a target identifier for the first AI model based on the target rule.

[0282] In the scenario where the access network device directly sends a rule set (including target rules) to the terminal device:

[0283] In one possible approach, the DU generates a rule set and sends the rule set to the terminal device. Correspondingly, the terminal device receives the rule set from the DU, selects a target rule from the rule set, and sends the target rule to the DU. Furthermore, the DU and the terminal device each generate a target identifier for the first AI model based on the target rule.

[0284] Alternatively, the DU generates a rule set and sends it to the CU; the CU receives the rule set from the DU and sends it to the terminal device; correspondingly, the terminal device receives the rule set from the CU, selects a target rule from the rule set, and sends it to the CU; the CU receives the target rule from the terminal device and sends it to the DU. Furthermore, the DU and the terminal device each generate a target identifier for the first AI model based on the target rule.

[0285] In another possible approach, the CU generates a rule set and sends the rule set to the terminal device. The terminal device receives the rule set from the CU, selects a target rule from the rule set, and sends the target rule to the CU. The CU receives the target rule from the terminal device and sends the target rule to the DU. Furthermore, the DU and the terminal device each generate a target identifier for the first AI model based on the target rule.

[0286] Optionally, the target rule / rule set sent by the DU to the terminal device may be carried in the control unit (MAC CE) of the media access control layer; the target rule / rule set sent by the CU to the terminal device may be carried in the RRC message.

[0287] It should be noted that in 5G, access network equipment includes CU and DU. In future 6G or other communication systems, CU and DU may continue to be referred to as CU and DU, or by other names; accordingly, the interface between CU and DU may continue to be the F1 interface, or may be another interface. This application does not limit this.

[0288] The above description focuses on the differences between the relevant method embodiments of Figures 8 to 15. Except for the differences, the relevant method embodiments of Figures 8 to 15 can refer to each other; in addition, in the same embodiment, different implementation methods or different examples can also refer to each other.

[0289] It can be understood that in the above-mentioned method embodiments, the methods and operations implemented by the terminal device can also be implemented by components that can be used for the terminal device (such as chips or circuits), and the methods and operations implemented by the access network device can also be implemented by components that can be used for the access network device (such as chips or circuits).

[0290] In the embodiments provided above, the methods provided in the embodiments of the present application are described from the perspective of interaction between various devices. In order to implement the various functions in the methods provided in the embodiments of the present application, the terminal device and the access network device may include a hardware structure and / or a software module, and implement the above functions in the form of a hardware structure, a software module, or a hardware structure plus a software module. Whether a function of the above functions is executed in the form of a hardware structure, a software module, or a hardware structure plus a software module depends on the specific application and design constraints of the technical solution.

[0291] The division of modules in the embodiments of the present application is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the present application may be integrated into a single processor, or may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0292] Corresponding to the aforementioned implementation scheme, Figures 16 and 17 are schematic diagrams of the structures of possible communication devices provided by this application. These communication devices can be used to implement the functions of the access network device or terminal device in the above-mentioned method embodiments related to Figures 8 to 15, and thus can also achieve the beneficial effects of the above-mentioned method embodiments related to Figures 8 to 15.

[0293] In the present application, the communication device may be an access network device as shown in Figures 1 to 3, or a terminal device as shown in Figures 1 to 3, or a module (such as a chip) applied to an access network device or a terminal device.

[0294] As shown in Figure 16, the communication device 1600 includes a transceiver module 1601 and a processing module 1602. The communication device 1600 is used to implement the functions of the access network device or terminal device in the above-mentioned method embodiments related to Figures 8 to 15.

[0295] When the communication device 1600 implements the functions of the access network device in the above-mentioned method embodiments of FIG. 8 to FIG. 15 :

[0296] The transceiver module 1601 is used to synchronize the globally unique model identifier of a first AI model that requires lifecycle management with a terminal device; the processing module 1602 is used to generate a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier; and, based on the target identifier of the first AI model, perform lifecycle management on the first AI model.

[0297] In one possible implementation, the target identifier is one of the following: a bitmap, where the bitmap includes N bits, each of the N bits corresponding to N AI models, where the N AI models are AI models that have synchronized a globally unique model identifier with the terminal device and require lifecycle management, and the N AI models include a first AI model; a first bit sequence, where the first bit sequence includes m bits, and the m bits have a total of M decimal values, and the M values ​​correspond to the M AI models; a second bit sequence, where the second bit sequence is composed of multiple bits at preset positions in the globally unique model identifier, and the second bit sequences corresponding to any two AI models in the M AI models are different; a combination of model information of the AI ​​model and an index of the AI ​​model under the model information, and the combination corresponding to any two AI models in the M AI models is different, and the model information includes one or more of RRC parameters, scenarios, and applicable conditions; wherein the M AI models are AI models that require lifecycle management with the terminal device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

[0298] In one possible implementation, the target identifier is a bitmap; when generating the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module 1602 is specifically configured to: sort the N AI models according to their globally unique model identifiers, wherein the N AI models are AI models whose globally unique model identifiers have been synchronized with the terminal device and require lifecycle management, and the first AI model is located at the nth position in the sorting, where n and N are both positive integers, and n is less than or equal to N; and generate a bitmap for the first AI model, where the bit at the nth position in the bitmap of the first AI model corresponds to the first AI model.

[0299] In one possible implementation, the target identifier is a bitmap; the processing module 1602 is further used to: generate an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and requires lifecycle management; when generating the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module 1602 is specifically used to: generate a bitmap of the first AI model, where the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model; wherein N is a positive integer.

[0300] In one possible implementation, after the processing module 1602 generates the index N of the first AI model, the transceiver module 1601 is further used to: send the index N of the first AI model to the terminal device, where the index N is used by the terminal device to determine that the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model.

[0301] In one possible implementation, when performing lifecycle management on the first AI model based on the target identifier of the first AI model, the processing module 1602 is specifically configured to: control the transceiver module 1601 to send the target identifier of the first AI model to the terminal device, where the target identifier of the first AI model is used to instruct the terminal device to activate the first AI model; after controlling the transceiver module 1601 to send the target identifier of the first AI model to the terminal device, the processing module 1602 is further configured to: control the transceiver module 1601 to send the target identifier of the second AI model to the terminal device; wherein the AI ​​models that have synchronized the globally unique model identifier with the terminal device and require lifecycle management also include the second AI model, and the target identifier of the second AI model is used to instruct the terminal device to activate the second AI model and deactivate the first AI model.

[0302] In one possible implementation, when generating the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module 1602 is specifically used to: control the transceiver module 1601 to send a target rule to the terminal device, where the target rule is a rule for generating the target identifier of the first AI model; and generate the target identifier of the first AI model according to the target rule.

[0303] When the communication device 1600 implements the functions of the terminal device in the above-mentioned method embodiments of FIG. 8 to FIG. 15 :

[0304] Transceiver module 1601 is used to synchronize the globally unique model identifier of a first AI model that requires lifecycle management with an access network device; processing module 1602 is used to perform lifecycle management on the first AI model based on the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model; wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier.

[0305] In one possible implementation, the target identifier is one of the following: a bitmap, where the bitmap includes N bits, each of the N bits corresponding to N AI models, where the N AI models are AI models that have synchronized a globally unique model identifier with an access network device and require lifecycle management, and the N AI models include a first AI model; a first bit sequence, where the first bit sequence includes m bits, where the m bits have a total of M decimal values, and the M values ​​correspond to the M AI models; a second bit sequence, where the second bit sequence is composed of multiple bits at preset positions in the globally unique model identifier, and the second bit sequences corresponding to any two of the M AI models are different; a combination of model information of the AI ​​model and an index of the AI ​​model under the model information, where the combination corresponding to any two of the M AI models is different, and the model information includes one or more of RRC parameters, scenarios, and applicable conditions; wherein the M AI models are AI models that require lifecycle management with the access network device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

[0306] In one possible implementation, when performing lifecycle management on the first AI model based on the target identifier of the first AI model, the processing module 1602 is specifically configured to: activate the first AI model based on the target identifier of the first AI model; after activating the first AI model based on the target identifier of the first AI model, the processing module 1602 is further configured to: activate the second AI model and deactivate the first AI model based on the target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model; wherein the AI ​​models that have synchronized the globally unique model identifier with the access network device and require lifecycle management also include the second AI model.

[0307] In one possible implementation, before performing lifecycle management on the first AI model based on the target identifier of the first AI model, the processing module 1602 is also used to: control the transceiver module 1601 to receive the target identifier of the first AI model from the access network device, where the target identifier of the first AI model is generated by the access network device.

[0308] In one possible implementation, before performing lifecycle management on the first AI model based on the target identifier of the first AI model, the processing module 1602 is further used to: control the transceiver module 1601 to receive a target rule from the access network device, where the target rule is a rule for generating the target identifier of the first AI model; and generate the target identifier of the first AI model based on the target rule.

[0309] FIG17 shows an apparatus 1700 provided in an embodiment of the present application. The apparatus shown in FIG17 may be a hardware circuit implementation of the apparatus shown in FIG16. The apparatus may be applied to the flowcharts shown above to perform the functions of the access network device or terminal device in the method embodiments related to FIG8 to FIG15.

[0310] For ease of explanation, FIG17 shows only the main components of the device.

[0311] The device 1700 shown in FIG17 includes a communication interface 1710, a processor 1720, and a memory 1730, wherein the memory 1730 is used to store program instructions and / or data. The processor 1720 may operate in conjunction with the memory 1730. The processor 1720 may execute program instructions stored in the memory 1730. When the instructions or program stored in the memory 1730 are executed, the processor 1720 is used to perform the operations performed by the processing module 1602 in the above embodiment, and the communication interface 1710 is used to perform the operations performed by the transceiver module 1601 in the above embodiment.

[0312] Memory 1730 is coupled to processor 1720. In the embodiments of the present application, coupling refers to an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, for information exchange between the devices, units, or modules. At least one of memory 1730 may be included in processor 1720.

[0313] In the embodiments of the present application, the communication interface may be a transceiver, circuit, bus, module, or other type of communication interface. In the embodiments of the present application, when the communication interface is a transceiver, the transceiver may include an independent receiver or an independent transmitter; or a transceiver or communication interface that integrates transceiver functions.

[0314] Device 1700 may also include a communication line 1740. Communication interface 1710, processor 1720, and memory 1730 may be interconnected via communication line 1740; communication line 1740 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, for example. Communication line 1740 may be classified as an address bus, a data bus, a control bus, and the like. For ease of illustration, FIG17 shows only one thick line, but this does not imply that there is only one bus or only one type of bus.

[0315] Corresponding to the aforementioned implementation scheme, an embodiment of the present application provides a chip system, including: a processor, the processor being coupled to a memory, the memory being used to store programs or instructions, and when the programs or instructions are executed by the processor, the chip system implements the method of the access network device in the method embodiments related to Figures 8 to 15 above, or implements the method of the terminal device in the method embodiments related to Figures 8 to 15 above. Optionally, the chip system also includes an interface circuit, which is used to exchange code instructions to the processor.

[0316] Optionally, there may be one or more processors in the chip system, and the processor may be implemented in hardware or software. When implemented in hardware, the processor may be a logic circuit, an integrated circuit, etc. When implemented in software, the processor may be a general-purpose processor implemented by reading software code stored in a memory.

[0317] Optionally, the chip system may include one or more memories. The memory may be integrated with the processor or provided separately from the processor. For example, the memory may be a non-transient processor, such as a read-only memory (ROM), which may be integrated with the processor on the same chip or provided on separate chips.

[0318] Corresponding to the implementation scheme described above, an embodiment of the present application provides a chip comprising at least one processor and an interface; the interface is used to provide program instructions or data to at least one processor; and at least one processor is used to execute program line instructions to implement the method of the access network device in the above-mentioned method embodiments related to Figures 8 to 15, or to implement the method of the terminal device in the above-mentioned method embodiments related to Figures 8 to 15.

[0319] Corresponding to the above implementation scheme, an embodiment of the present application provides a computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed, the computer implements the method of the access network device in the above-mentioned method embodiments related to Figures 8 to 15, or implements the method of the terminal device in the above-mentioned method embodiments related to Figures 8 to 15.

[0320] Corresponding to the implementation scheme described above, an embodiment of the present application provides a computer program product. When a computer reads and executes the computer program product, the computer implements the method of the access network device in the above-mentioned method embodiments related to Figures 8 to 15, or implements the method of the terminal device in the above-mentioned method embodiments related to Figures 8 to 15.

[0321] Corresponding to the above implementation scheme, an embodiment of the present application provides a communication system, which includes the access network device in the relevant method embodiments of Figures 8 to 15 above and at least one terminal device.

[0322] The terms "system" and "network" in the embodiments of the present application can be used interchangeably. "At least one" refers to one or more, and "plurality" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in the embodiments of the present application are used to distinguish multiple objects, and are not used to limit the order, timing, priority or importance of multiple objects.

[0323] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0324] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each flow and / or box in the flow chart and / or block diagram, as well as the combination of the flow chart and / or box in the flow chart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one or more flow charts and / or one or more boxes in the block diagram.

[0325] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0326] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

Claims

1. A communication method, characterized in that: include: A globally unique model identifier of a first artificial intelligence AI model that needs lifecycle management synchronized with the terminal device; Generate a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier; The lifecycle management of the first AI model is performed according to the target identifier of the first AI model.

2. The method according to claim 1, characterized in that The target identifier is one of the following: A bitmap bitmap, wherein the bitmap includes N bits, the N bits respectively correspond to N AI models, the N AI models are AI models that have synchronized a globally unique model identifier with the terminal device and require lifecycle management, and the N AI models include the first AI model; A first bit sequence, wherein the first bit sequence includes m bits, the m bits have a total of M decimal values, and the M values ​​correspond to M AI models respectively; a second bit sequence, where the second bit sequence is composed of a plurality of bits at a preset position in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different; A combination of model information of an AI model and an index of the AI ​​model under the model information, wherein the combinations corresponding to any two AI models in the M AI models are different, and the model information includes one or more of a radio resource control RRC parameter, a scenario, and an applicable condition; Among them, the M AI models are AI models that need to perform the lifecycle management with the terminal device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

3. The method according to claim 1 or 2, characterized in that The target identifier is a bitmap; The generating a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model includes: Sorting the N AI models according to their globally unique model identifiers, where the N AI models are AI models that have synchronized their globally unique model identifiers with the terminal device and require lifecycle management, and the first AI model is located at the nth position in the sorting, where n and N are both positive integers, and n is less than or equal to N; A bitmap of the first AI model is generated, wherein the bit at the nth position in the bitmap of the first AI model corresponds to the first AI model.

4. The method according to claim 1 or 2, characterized in that: The target identifier is a bitmap; The method further comprises: Generate an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and needs to perform lifecycle management; The generating a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model includes: Generate a bitmap of the first AI model, where the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model; Wherein, N is a positive integer.

5. The method according to claim 4, characterized in that After generating the index N of the first AI model, the method further includes: Send an index N of the first AI model to the terminal device, where the index N is used by the terminal device to determine that the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model.

6. The method according to any one of claims 1 to 5, characterized in that The performing the lifecycle management on the first AI model according to the target identifier of the first AI model includes: Sending a target identifier of the first AI model to the terminal device, where the target identifier of the first AI model is used to instruct the terminal device to activate the first AI model; After sending the target identifier of the first AI model to the terminal device, the method further includes: Sending a target identifier of a second AI model to the terminal device; Among them, the AI ​​model that has synchronized the globally unique model identifier with the terminal device and requires lifecycle management also includes the second AI model, and the target identifier of the second AI model is used to instruct the terminal device to activate the second AI model and deactivate the first AI model.

7. The method according to any one of claims 1 to 6, characterized in that The method is applicable to an access network device, wherein the access network device includes a centralized unit CU and a distributed unit DU; The generating a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model includes: The DU generates a target identifier of the first AI model; The method further comprises: The DU sends a target identifier of the first AI model to the terminal device; or, The DU sends a target identifier of the first AI model to the CU, and the CU sends the target identifier of the first AI model to the terminal device.

8. The method according to any one of claims 1 to 7, characterized in that The method is applicable to access network equipment, wherein the access network equipment includes a CU and a DU; The method further comprises: The DU generates an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and needs lifecycle management; The DU sends the index N of the first AI model to the terminal device; or, the DU sends the index N of the first AI model to the CU, and the CU sends the index N of the first AI model to the terminal device; Wherein, N is a positive integer.

9. The method according to any one of claims 1 to 7, characterized in that The method is applicable to access network equipment, wherein the access network equipment includes a CU and a DU; The method further comprises: The DU sends the globally unique model identifier of the first AI model to the CU; The CU generates an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and needs to perform lifecycle management; The CU sends the index N of the first AI model to the DU, and the DU sends the index N of the first AI model to the terminal device; or, the CU sends the index N of the first AI model to the terminal device; Wherein, N is a positive integer.

10. The method according to any one of claims 1 to 5, characterized in that The generating a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model includes: Sending a target rule to the terminal device, where the target rule is a rule for generating a target identifier of the first AI model; According to the target rule, a target identifier of the first AI model is generated.

11. The method according to claim 10, characterized in that The method is applicable to access network equipment, wherein the access network equipment includes a CU and a DU; The sending the target rule to the terminal device includes: The DU sends the target rule to the terminal device; or, the DU sends the target rule to the CU, and the CU sends the target rule to the terminal device; Generating a target identifier of the first AI model according to the target rule includes: The DU generates a target identifier of the first AI model according to the target rule.

12. A communication method, characterized in that: include: A globally unique model identifier of a first artificial intelligence AI model that needs lifecycle management synchronized with the access network device; Performing the lifecycle management on the first AI model according to the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model; The number of bits occupied by the target identifier is smaller than the number of bits occupied by the globally unique model identifier.

13. The method according to claim 12, characterized in that The target identifier is one of the following: A bitmap bitmap, wherein the bitmap includes N bits, the N bits respectively correspond to N AI models, the N AI models are AI models that have synchronized a globally unique model identifier with the access network device and require lifecycle management, and the N AI models include the first AI model; A first bit sequence, wherein the first bit sequence includes m bits, the m bits have a total of M decimal values, and the M values ​​correspond to M AI models respectively; a second bit sequence, where the second bit sequence is composed of a plurality of bits at a preset position in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different; A combination of model information of an AI model and an index of the AI ​​model under the model information, wherein the combinations corresponding to any two AI models in the M AI models are different, and the model information includes one or more of a radio resource control RRC parameter, a scenario, and an applicable condition; Among them, the M AI models are AI models that need to perform the lifecycle management with the access network device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

14. The method according to claim 12 or 13, characterized in that The performing the lifecycle management on the first AI model according to the target identifier of the first AI model includes: activating the first AI model according to the target identifier of the first AI model; After activating the first AI model according to the target identifier of the first AI model, the method further includes: According to the target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model, the second AI model is activated and the first AI model is deactivated; wherein, the AI ​​models that have synchronized the globally unique model identifier with the access network device and need to perform lifecycle management also include the second AI model.

15. The method according to any one of claims 12 to 14, characterized in that Before performing the lifecycle management on the first AI model according to the target identifier of the first AI model, the method further includes: A target identifier of the first AI model is received from the access network device, where the target identifier of the first AI model is generated by the access network device.

16. The method according to any one of claims 12 to 14, characterized in that Before performing the lifecycle management on the first AI model according to the target identifier of the first AI model, the method further includes: Receiving a target rule from the access network device, where the target rule is a rule for generating a target identifier of the first AI model; According to the target rule, a target identifier of the first AI model is generated.

17. A communication device, characterized in that: include: A transceiver module, used for synchronizing with the terminal device a globally unique model identifier of a first artificial intelligence AI model that needs lifecycle management; A processing module is used to generate a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, wherein the number of bits occupied by the target identifier is less than the number of bits occupied by the globally unique model identifier; and, according to the target identifier of the first AI model, perform the lifecycle management on the first AI model.

18. The device according to claim 17, characterized in that The target identifier is one of the following: A bitmap bitmap, wherein the bitmap includes N bits, the N bits respectively correspond to N AI models, the N AI models are AI models that have synchronized a globally unique model identifier with the terminal device and require lifecycle management, and the N AI models include the first AI model; A first bit sequence, wherein the first bit sequence includes m bits, the m bits have a total of M decimal values, and the M values ​​correspond to M AI models respectively; a second bit sequence, where the second bit sequence is composed of a plurality of bits at a preset position in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different; A combination of model information of an AI model and an index of the AI ​​model under the model information, wherein the combinations corresponding to any two AI models in the M AI models are different, and the model information includes one or more of a radio resource control RRC parameter, a scenario, and an applicable condition; Among them, the M AI models are AI models that need to perform the lifecycle management with the terminal device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

19. The device according to claim 17 or 18, characterized in that The target identifier is a bitmap; When the processing module generates the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module is specifically configured to: Sorting the N AI models according to their globally unique model identifiers, where the N AI models are AI models that have synchronized their globally unique model identifiers with the terminal device and require lifecycle management, and the first AI model is located at the nth position in the sorting, where n and N are both positive integers, and n is less than or equal to N; A bitmap of the first AI model is generated, wherein the bit at the nth position in the bitmap of the first AI model corresponds to the first AI model.

20. The device according to claim 17 or 18, characterized in that The target identifier is a bitmap; The processing module is also used for: Generate an index N of the first AI model, where the index N is used to indicate that the first AI model is the Nth AI model that synchronizes a globally unique model identifier with the terminal device and needs to perform lifecycle management; When the processing module generates the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module is specifically configured to: Generate a bitmap of the first AI model, where the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model; Wherein, N is a positive integer.

21. The device according to claim 20, characterized in that After the processing module generates the index N of the first AI model, the transceiver module is further used to: Send an index N of the first AI model to the terminal device, where the index N is used by the terminal device to determine that the bit at the Nth position in the bitmap of the first AI model corresponds to the first AI model.

22. The device according to any one of claims 17 to 21, characterized in that When the processing module performs the lifecycle management on the first AI model according to the target identifier of the first AI model, the processing module is specifically used to: Control the transceiver module to send a target identifier of the first AI model to the terminal device, where the target identifier of the first AI model is used to instruct the terminal device to activate the first AI model; After controlling the transceiver module to send the target identifier of the first AI model to the terminal device, the processing module is further used to: Control the transceiver module to send a target identifier of a second AI model to the terminal device; Among them, the AI ​​model that has synchronized the globally unique model identifier with the terminal device and requires lifecycle management also includes the second AI model, and the target identifier of the second AI model is used to instruct the terminal device to activate the second AI model and deactivate the first AI model.

23. The device according to any one of claims 17 to 21, characterized in that When the processing module generates the target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model, the processing module is specifically configured to: Control the transceiver module to send a target rule to the terminal device, where the target rule is a rule for generating a target identifier of the first AI model; According to the target rule, a target identifier of the first AI model is generated.

24. A communication device, characterized in that: include: A transceiver module, used to synchronize with the access network device a globally unique model identifier of a first artificial intelligence AI model that needs lifecycle management; a processing module, configured to perform the lifecycle management on the first AI model according to a target identifier of the first AI model corresponding to the globally unique model identifier of the first AI model; The number of bits occupied by the target identifier is smaller than the number of bits occupied by the globally unique model identifier.

25. The device according to claim 24, characterized in that The target identifier is one of the following: A bitmap bitmap, wherein the bitmap includes N bits, the N bits respectively correspond to N AI models, the N AI models are AI models that have synchronized a globally unique model identifier with the access network device and require lifecycle management, and the N AI models include the first AI model; A first bit sequence, wherein the first bit sequence includes m bits, the m bits have a total of M decimal values, and the M values ​​correspond to M AI models respectively; a second bit sequence, where the second bit sequence is composed of a plurality of bits at a preset position in the globally unique model identifier, and the second bit sequences corresponding to any two AI models among the M AI models are different; A combination of model information of an AI model and an index of the AI ​​model under the model information, wherein the combinations corresponding to any two AI models in the M AI models are different, and the model information includes one or more of a radio resource control RRC parameter, a scenario, and an applicable condition; Among them, the M AI models are AI models that need to perform the lifecycle management with the access network device, the N AI models are included in the M AI models, and m, M, and N are all positive integers.

26. The device according to claim 24 or 25, characterized in that When the processing module performs the lifecycle management on the first AI model according to the target identifier of the first AI model, the processing module is specifically used to: activating the first AI model according to the target identifier of the first AI model; After activating the first AI model according to the target identifier of the first AI model, the processing module is further configured to: According to the target identifier of the second AI model corresponding to the globally unique model identifier of the second AI model, the second AI model is activated and the first AI model is deactivated; wherein, the AI ​​models that have synchronized the globally unique model identifier with the access network device and need to perform lifecycle management also include the second AI model.

27. The device according to any one of claims 24 to 26, characterized in that Before performing the lifecycle management on the first AI model according to the target identifier of the first AI model, the processing module is further configured to: Control the transceiver module to receive a target identifier of the first AI model from the access network device, where the target identifier of the first AI model is generated by the access network device.

28. The device according to any one of claims 24 to 26, characterized in that Before performing the lifecycle management on the first AI model according to the target identifier of the first AI model, the processing module is further configured to: Control the transceiver module to receive a target rule from the access network device, where the target rule is a rule for generating a target identifier of the first AI model; According to the target rule, a target identifier of the first AI model is generated.

29. A communication device, characterized in that: It includes a processor and a communication interface, wherein the communication interface is used to receive signals from other communication devices outside the communication device and transmit them to the processor or send signals from the processor to other communication devices outside the communication device, and the processor is used to implement the method described in any one of claims 1 to 11 or any one of claims 12 to 16 through logic circuits or execution code instructions.

30. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program or instruction. When the computer program or instruction is executed by the communication device, the method according to any one of claims 1 to 11 or any one of claims 12 to 16 is implemented.

31. A computer program product, characterized in that The computer program product comprises a computer program or instructions, and when the computer program or instructions are executed by a communication device, the method according to any one of claims 1 to 11, or the method according to any one of claims 12 to 16 is implemented.

32. A communication system, characterized in that: The invention comprises the communication device according to any one of claims 17 to 23 and the communication device according to any one of claims 24 to 28.

Citation Information

Patent Citations

  • Generic Method and System for Lifecycle Management

    US20130268911A1

  • Systems, methods, and apparatus for artificial intelligence and machine learning for a physical layer of communication system

    US20230131694A1

  • Transmitting and receiving method and device

    WO2012171196A1

  • Wireless device, first and second radio network nodes, and methods performed therein for determining global id of the second radio network node

    WO2020091678A1

  • Resource deployment and application data routing methods and apparatus, and storage medium

    WO2023025295A1

Cited By

  • Model management method and wireless communication device

    WO2026193766A1