Communication method, apparatus, and system

By defining the constraint relationships of AI function groups in the wireless communication system and managing AI functions at the group level, the problem of low efficiency in managing individual models in the existing technology is solved, and more efficient model management and system performance optimization are achieved.

WO2026233245A1PCT designated stage Publication Date: 2026-11-12HUAWEI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2026/094341
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-09
Filing Date
2026-04-30
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

In existing technologies, management nodes manage AI functions at the granularity of individual models, which is inefficient and cannot effectively optimize the performance of wireless communication systems.

Method used

By defining and indicating the constraints between AI functions, and managing them at the granular level using AI function groups, including operations such as activation, deactivation, monitoring, and inference configuration, the efficiency of model management is improved.

Benefits of technology

It improved the efficiency of model management, optimized the performance of the wireless communication system, and enhanced the efficiency of model operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2026094341_12112026_PF_FP_ABST
    Figure CN2026094341_12112026_PF_FP_ABST
Patent Text Reader

Abstract

The present application provides a communication method, an apparatus, and a system, aiming to improve the efficiency of model management. The method comprises: a third apparatus determining that one or more AI functions satisfy a first constraint relationship, wherein the one or more AI functions are AI functions supported by a first apparatus; and sending first information, wherein the first information is used for indicating that the one or more AI functions satisfy the first constraint relationship. In this way, a second apparatus can manage at least one AI function among the one or more AI functions on the basis of the first constraint relationship, wherein the management includes, for example, activation, deactivation, inference configuration, monitoring, etc. Therefore, the efficiency of model management can be improved, thereby improving the operation efficiency of a model, and optimizing the performance of a wireless communication system.
Need to check novelty before this filing date? Find Prior Art

Description

Communication methods, devices and systems

[0001] This application claims priority to Chinese Patent Application No. 202510603678.7, filed on May 9, 2025, entitled "Communication Method, Apparatus and System", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of communications, and more particularly to a communication method, apparatus and system. Background Technology

[0003] With the development of artificial intelligence (AI) and machine learning (ML), AI / ML technologies have been introduced into wireless air interfaces. For example, AI models can be applied to transmitters for data transmission processing, such as encoding, modulation, and precoding; AI models can also be applied to receivers for channel estimation and data reception processing, such as decoding, demodulation, and equalization; AI models can also be applied to both transmitting and receiving ends for channel compression and reconstruction; and AI models are used for beam management.

[0004] The model can be managed by a unified management node, which can perform operations such as model activation, deactivation, monitoring, and inference configuration. As AI / ML technologies become increasingly versatile in wireless air interfaces, the functions that models can perform are also expanding. Therefore, multiple models can be deployed in transmitters and / or receivers to achieve different functionalities.

[0005] Currently, management nodes manage models at the granularity of individual models (or functions), which is inefficient. Summary of the Invention

[0006] This application provides a communication method, apparatus, and system to improve the efficiency of model management.

[0007] Firstly, a communication method is provided, which can be applied to a third device, which can be a terminal side. The terminal side can be a terminal-side device, or a chip or circuit of the terminal-side device. The terminal-side device can be replaced by a device on a terminal device side, which can include at least one of a terminal device or an AI entity on the terminal device side. The AI ​​entity on the terminal device side can be the terminal device itself, or an AI entity serving the terminal device, such as a server, such as an over-the-top (OTT) server or a cloud server. Components deployed in the terminal device can be, for example, circuits or chips inside the terminal device (such as modem chips, also known as baseband chips, or system-on-chip (SoC) chips or system-in-package (SIP) chips containing modem cores, or chips for AI or SoC or SIP chips containing chips for AI, etc.); or, they can be applied to logic modules or software capable of implementing all or part of the functions of the communication device, etc.

[0008] For example, the method includes: determining that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a first device; and sending first information to indicate that the one or more AI functions satisfy the first constraint relationship.

[0009] Based on the above scheme, the third device indicates to the second device, through the first information, a first constraint relationship satisfied between one or more AI functions. This allows the second device to conveniently manage the one or more AI functions based on this first constraint relationship. Therefore, it can improve the efficiency of model management, thereby improving the model's operating efficiency and optimizing the performance of the wireless communication system.

[0010] In conjunction with the first aspect, in some possible implementations of the first aspect, one or more AI functions satisfy the first constraint relationship, including: the activation of one or more AI functions satisfies the first constraint relationship, and / or, the inference configuration of one or more AI functions satisfies the first constraint relationship.

[0011] Defining the first constraint relationship from the perspective of activation and / or inference configuration can conveniently provide a reference for the subsequent activation and inference configuration of one or more AI functions.

[0012] In conjunction with the first aspect, in some possible implementations of the first aspect, the method further includes: receiving second information for managing at least one AI function among one or more AI functions that satisfy the first constraint relationship, the management including one or more of the following: activation, deactivation, monitoring or inference configuration.

[0013] The second device can perform one or more of the following actions on the one or more AI functions based on the first constraint: activation, deactivation, monitoring, or inference configuration. For example, the second device can determine whether the one or more AI functions can be activated simultaneously, or determine the inference configuration for the one or more AI functions, based on the first constraint. This can improve the efficiency of model management, enhance model operating efficiency, and optimize the performance of the wireless communication system.

[0014] In conjunction with the first aspect, in some possible implementations of the first aspect, the second information is used to perform inference configuration on at least one of one or more AI functions that satisfy the first constraint relationship, and the inference configuration of the at least one AI function has a corresponding relationship.

[0015] Among multiple AI functions that satisfy the first constraint relationship, the activation or deactivation of some AI functions may affect the inference configuration of other AI functions. Therefore, the inference configuration of one or more AI functions can be designed so that the inference configuration for one or more AI functions can be adapted to the first constraint relationship they satisfy.

[0016] Optionally, the second information may also indicate a correspondence.

[0017] By indicating this correspondence, it is easier for a third device to use the corresponding inference configuration when using one or more AI functions, thereby improving the efficiency of model operation.

[0018] In conjunction with the first aspect, in some possible implementations of the first aspect, one or more AI functions are one of one or more AI function groups, each of the one or more AI function groups includes one or more AI functions supported by the first device, and one or more AI functions in each of the one or more AI function groups satisfy one of the following multiple constraint relationships: independence, dependency, substitution, or condition switching; an AI function satisfying the independence relationship is one AI function, and an AI function satisfying the dependency, selection, substitution, or condition switching relationship is multiple AI functions; wherein, among the multiple AI functions satisfying the dependency relationship, the activation of one AI function depends on the activation of another AI function; multiple AI functions satisfying the substitution relationship are not activated simultaneously; among the multiple AI functions satisfying the condition switching relationship, the inference configuration of one or more AI functions is determined based on whether another one or more AI functions are activated.

[0019] One or more AI functions that satisfy a constraint relationship are defined as an AI function group, which makes it convenient for the second device to manage AI function groups at the granularity.

[0020] In conjunction with the first aspect, in some possible implementations of the first aspect, the first information is used to indicate that one or more AI functions satisfy a first constraint relationship, including: the first information is used to indicate one or more AI functions included in each of the one or more AI function groups, and the constraint relationships they satisfy.

[0021] When the third device sends the first information, it can indicate one or more AI functions that satisfy the first constraint relationship to the second device in the form of an AI function group, so that the second device can manage them in the form of AI function groups in the future.

[0022] Optionally, the first information is also used to indicate the identifier of each AI function group in one or more AI function groups.

[0023] When the third device manages AI function groups at the granularity, it can indicate one or more AI functions that need to be managed by the identifier of the AI ​​function group. Therefore, indicating the identifier of the AI ​​function group can save instruction overhead compared to indicating each AI function separately.

[0024] In conjunction with the first aspect, in some possible implementations of the first aspect, one or more AI function groups are determined based on one or more AI functions supported by the first device.

[0025] The third device can instruct the second device on all the AI ​​functions supported by the first device in the form of AI function groups, so that the second device can understand the capabilities of the first device and the relationship between the various AI functions, and facilitate the subsequent management of the AI ​​functions supported by the first device.

[0026] In conjunction with the first aspect, in some possible implementations of the first aspect, one or more AI function groups are determined based on attribute information related to each AI function among one or more AI functions supported by the first device.

[0027] The third device can also determine the AI ​​function group to be indicated to the second device based on the attribute information related to each AI function supported by the first device, so that the currently available AI functions can be indicated to the second device for the second device to manage them.

[0028] Optionally, the method further includes: receiving status information from the second device; one or more AI function groups are determined based on attribute information related to each AI function among one or more AI functions supported by the first device, including: one or more AI function groups are determined based on the status information and attribute information related to each AI function among one or more AI functions supported by the first device.

[0029] Optionally, the status information of the second device is used to indicate a status related to one or more of the baseband processing, radio frequency, or antenna.

[0030] The third device can combine the status information of the second device and the attribute information related to each AI function supported by the first device to determine the AI ​​function group to be indicated to the second device. At this time, the AI ​​function group indicated to the second device is the AI ​​function group currently available to the first device. That is, the third device can avoid indicating the AI ​​function group that is currently unavailable to the second device, which can reduce the indication overhead and facilitate the third device's management of the currently available AI function groups deployed in the first device.

[0031] Secondly, a communication method is provided, which can be applied to a second device, which can be a network side. The network side can be a network-side device, or a chip or circuit of a network-side device. The network-side device can be replaced by a device on the network device side, which can include at least one of a network device or an AI entity on the network device side. The AI ​​entity on the network device side can be the network device itself or an AI entity serving the network device, such as an intelligent network element on the network device side. Components deployed in the network device can be, for example, circuits or chips inside the network device (such as a modem chip, also known as a baseband chip, or a SoC chip or SIP chip containing a modem core, or a chip for AI or a SoC or SIP chip containing a chip for AI, etc.); or, they can be applied to logic modules or software capable of implementing all or part of the functions of the second device, etc.

[0032] It should be understood that the technical solution of the second aspect corresponds to the technical solution of the first aspect. Therefore, the relevant descriptions and technical effects of the first aspect are equally applicable to the second aspect. Thus, the content that is the same as or corresponding to the first aspect can be referred to the relevant descriptions of the first aspect, and will not be repeated here.

[0033] For example, the method includes: receiving first information, the first information indicating that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a first device; sending second information, the second information being used to manage at least one of the one or more AI functions satisfying the first constraint relationship, or in other words, the second information being used to indicate the management of at least one of the one or more AI functions satisfying the first constraint relationship, the management including one or more of the following: activation, deactivation, inference configuration, or monitoring.

[0034] For example, the method includes: receiving first information, the first information being used to indicate that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a first device, the first information being used to determine management of at least one of the one or more AI functions satisfying the first constraint relationship, the management including one or more of the following: activation, deactivation, inference configuration, or monitoring.

[0035] Optionally, the first information is used to determine the second information, which is used to instruct the management of at least one AI function among one or more AI functions that satisfy the first constraint relationship, or in other words, the second information is used to manage at least one AI function among one or more AI functions that satisfy the first constraint relationship.

[0036] In conjunction with the second aspect, in some possible implementations of the second aspect, one or more AI functions satisfy the first constraint relationship, including: the activation of one or more AI functions satisfies the first constraint relationship, and / or, the inference configuration of one or more AI functions satisfies the first constraint relationship.

[0037] In conjunction with the second aspect, in some possible implementations of the second aspect, the second information is used to perform inference configuration on at least one of one or more AI functions that satisfy the first constraint relationship, and the inference configuration of the at least one AI function has a corresponding relationship.

[0038] Optionally, the second information may also indicate a correspondence.

[0039] In conjunction with the second aspect, in some possible implementations of the second aspect, one or more AI functions are one of one or more AI function groups, each of the one or more AI function groups includes one or more AI functions supported by the first device, and one or more AI functions in each of the one or more AI function groups satisfy one of the following multiple constraint relationships: independence, dependency, substitution, or condition switching; an AI function that satisfies the independence relationship is one AI function, and an AI function that satisfies the dependency, selection, substitution, or condition switching relationship is multiple AI functions; wherein, among the multiple AI functions that satisfy the dependency relationship, the activation of one AI function depends on the activation of another AI function; the multiple AI functions that satisfy the substitution relationship are not activated simultaneously; among the multiple AI functions that satisfy the condition switching relationship, the inference configuration of one or more AI functions is determined based on whether the other one or more AI functions are activated.

[0040] In conjunction with the second aspect, in some possible implementations of the second aspect, the first information is used to indicate that one or more AI functions satisfy a first constraint relationship, including: the first information is used to indicate the AI ​​functions included in each of the one or more AI function groups, and the constraint relationships they satisfy, wherein one of the one or more AI function groups includes one or more AI functions.

[0041] Optionally, the first information is also used to indicate the identifier of each AI function group in one or more AI function groups.

[0042] In conjunction with the second aspect, in some possible implementations of the second aspect, before receiving the first information, the method further includes: sending status information, the first information being determined based on the status information and attribute information related to each of one or more AI functions supported by the first device.

[0043] Optionally, the status information of the second device is used to indicate a status related to one or more of the baseband processing, radio frequency, or antenna.

[0044] In conjunction with the first or second aspect, in some possible implementations, one or more AI functions include one or more of the following: channel estimation, channel prediction, channel compression and reconstruction, coding, modulation, precoding, equalization, demodulation, or decoding.

[0045] Applying AI to various operations in wireless communication allows neural networks to be used to optimize the performance of transmitters and / or receivers, thereby improving the performance of wireless communication systems.

[0046] The following provides an example of one or more AI functions and the first constraint they satisfy.

[0047] Optionally, one or more AI functions are for channel estimation, and the first information is used to indicate that the first constraint relationship satisfied by the channel estimation is an independence relationship.

[0048] Optionally, one or more AI functions include channel estimation and equalization, and the first information is used to indicate that a first constraint relationship between channel estimation and equalization is a dependency relationship.

[0049] Optionally, one or more AI functions include channel estimation and channel prediction, and the first information is used to indicate that a first constraint relationship between the channel estimation and channel prediction is a dependency relationship.

[0050] Optionally, one or more AI functions include channel estimation and channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between channel estimation and channel compression and reconstruction is a conditional switching relationship.

[0051] Optionally, one or more AI functions include channel prediction and channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between channel prediction and channel compression and reconstruction is a conditional switching relationship.

[0052] Optionally, one or more AI functions include equalization and decoding, and the first information is used to indicate that the first constraint relationship satisfied between equalization and decoding is a substitution relationship.

[0053] Optionally, one or more AI functions include demodulation and decoding, and the first information is used to indicate that a first constraint relationship satisfied between demodulation and decoding is a substitution relationship.

[0054] Thirdly, an apparatus is provided. This apparatus may include functional modules corresponding to the methods / operations / steps / actions described in the first aspect and any possible implementation thereof, or may include functional modules corresponding to the methods / operations / steps / actions described in the second aspect and any possible implementation thereof. The functional modules may be hardware circuits, software, or a combination of hardware circuits and software.

[0055] In one design, the device may include a processing module and a communication module. The communication module is used to perform the sending and receiving actions performed by the third device as described in the first aspect and any possible implementation thereof, while the processing module is used to perform the processing-related actions performed by the third device in the methods described in the first aspect and any possible implementation thereof.

[0056] In one design, the device can be a terminal device, or a device, module, circuit, or chip configured in the terminal device.

[0057] In one design, the device may include a processing module and a communication module. The communication module is used to perform the sending and receiving actions performed by the second device as described in the second aspect and any possible implementation thereof, while the processing module is used to perform the processing-related actions performed by the second device in the methods described in the second aspect and any possible implementation thereof.

[0058] In one design, the device can be a network device, or a device, module, circuit, or chip configured in the network device.

[0059] Fourthly, an apparatus is provided, including a processor and a storage medium storing instructions that, when executed by the processor, cause the method as described in the first aspect and any possible implementation thereof to be implemented, or cause the method as described in the second aspect and any possible implementation thereof to be implemented.

[0060] Fifthly, an apparatus is provided, comprising a processing circuit for processing data and / or information such that a method as described in the first aspect and any possible implementation thereof is implemented, or a method as described in the second aspect and any possible implementation thereof is implemented.

[0061] The processing circuit may include one or more processors, or all or part of the circuitry in one or more processors used for control or processing functions.

[0062] Optionally, the apparatus may further include a memory for storing programs or instructions, and the processor for running the programs or instructions to implement methods such as any of the first to fourth aspects, and any possible implementation of any of the aspects.

[0063] Optionally, the device may also include the transceiver circuit, or an input / output interface.

[0064] In a sixth aspect, a chip is provided, including processing circuitry for running a program or instructions to implement a method as described in the first aspect and any possible implementation thereof, or to implement a method as described in the second aspect and any possible implementation thereof.

[0065] Optionally, the chip may further include a memory for storing programs or instructions.

[0066] Optionally, the chip may also include transceiver circuitry, or input / output interfaces.

[0067] A seventh aspect provides a computer-readable storage medium comprising instructions that, when executed by a processor, cause the method as described in the first aspect and any possible implementation thereof to be implemented, or cause the method as described in the second aspect and any possible implementation thereof to be implemented.

[0068] Eighthly, a computer program product is provided, the computer program product comprising computer program code or instructions, which, when executed, cause the method as described in the first aspect and any possible implementation thereof to be implemented, or cause the method as described in the second aspect and any possible implementation thereof to be implemented.

[0069] Ninth aspect, a communication system is provided, the communication system including a third device and a second device, and optionally, a first device.

[0070] Optionally, the third device is used to perform the method in the first aspect and any possible implementation of the first aspect, and the second device is used to perform the method in the second aspect and any possible implementation of the second aspect.

[0071] It should be understood that the third to ninth aspects of this application correspond to the technical solutions of the first to second aspects of this application, and the beneficial effects obtained by each aspect and the corresponding feasible implementation are similar, and will not be repeated here. Attached Figure Description

[0072] Figure 1 is a schematic diagram of a communication system applicable to the communication method of this application embodiment;

[0073] Figure 2 is a schematic diagram of another communication system applicable to the communication method of this application embodiment;

[0074] Figure 3 is a schematic diagram of a possible application framework in a communication system;

[0075] Figure 4 is a schematic diagram of another possible application framework in a communication system;

[0076] Figure 5 is a schematic diagram of the AI / ML functional framework for wireless air interface;

[0077] Figure 6 is a schematic diagram of the AI ​​model used to implement channel prediction function;

[0078] Figure 7 is a schematic diagram of the AI ​​model used to implement channel compression and reconstruction functions;

[0079] Figure 8 is a schematic diagram of the AI ​​model used to implement the channel estimation function;

[0080] Figure 9 is a schematic diagram of an AI model used to implement multiple input multiple output (MIMO) equalization.

[0081] Figure 10 is a schematic diagram of the AI ​​model used to implement the demodulation function;

[0082] Figure 11 is a schematic diagram of the AI ​​model used to implement the decoding function;

[0083] Figure 12 is a schematic diagram of an wireless communication system including a transceiver based on a neural network, provided in an embodiment of this application.

[0084] Figure 13 is an example of a transceiver based on a neural network provided in an embodiment of this application;

[0085] Figure 14 is another example of a transceiver based on a neural network provided in the embodiments of this application;

[0086] Figure 15 is a schematic diagram of the training process of the intelligent air interface provided in the embodiment of this application;

[0087] Figure 16 is a schematic flowchart of the communication method provided in an embodiment of this application;

[0088] Figure 17 is a schematic diagram of various constraint relationships provided in the embodiments of this application;

[0089] Figure 18 is an example of multiple AI functions provided in the embodiments of this application satisfying various constraint relationships;

[0090] Figure 19 is another example of how multiple AI functions provided in the embodiments of this application satisfy multiple constraint relationships;

[0091] Figure 20 is a schematic diagram of multiple AI functions deployed in a first device and their respective AI function groups provided in an embodiment of this application;

[0092] Figure 21 is another schematic flowchart of the communication method provided in an embodiment of this application;

[0093] Figure 22 is another schematic flowchart of the communication method provided in the embodiments of this application;

[0094] Figure 23 is a schematic diagram of the communication device provided in an embodiment of this application;

[0095] Figure 24 is a schematic diagram of the structure of a communication device provided in another embodiment of this application;

[0096] Figure 25 is a schematic diagram of the structure of the AI ​​processor provided in an embodiment of this application. Detailed Implementation

[0097] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0098] To facilitate understanding of the embodiments of this application, the following points will be explained first:

[0099] First, in this application, the indication includes explicit indication (also known as direct indication) and implicit indication (also known as indirect indication). Explicit indication information A means including information A; implicit indication information A means indicating information A through the correspondence between information A and information B, and direct indication information B. The correspondence between information A and information B can be predefined, pre-stored, pre-burned, or pre-configured; or it can refer to indicating information A through information B and preset rules.

[0100] Second, in this application, information C is used to determine information D, which includes both determining information D based solely on information C and determining it based on information C and other information. Furthermore, information C can also be used to determine information D indirectly, for example, in the case where information D is determined based on information E, and information E is determined based on information C.

[0101] Third, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0102] Fourth, in this application, the use of prefixes such as "first" and "second" is merely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first information" and "second information" are simply different pieces of information, and there is no temporal sequence, size, or priority relationship between them.

[0103] Fifth, in this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to a third device" can be understood as the destination of the information being the third device, which may include direct transmission via the air interface or indirect transmission via the air interface by other units or modules. "Receive information from a second device" can be understood as the source of the information being the second device, which may include direct reception from the second device via the air interface or indirect reception from the second device via the air interface by other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface.

[0104] In other words, sending and receiving can occur between devices, such as between terminal devices and network devices; or they can occur within a device, such as between components, modules, chips, software modules, or hardware modules within a device via a bus, wiring, or interface.

[0105] Sixth, in the embodiments of this application, "when," "if," and "if" all refer to the device making corresponding processing under certain objective circumstances, and are not limited to a time, nor do they require the device to make a judgment action when it is implemented, nor do they mean that there are other limitations.

[0106] Seventh, in this application, the words "example," "exemplarily," "for example," or "such as" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "example," "exemplarily," "for example," or "such as" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "example," "exemplarily," "for example," or "such as" is intended to present the relevant concepts in a specific manner.

[0107] Eighth, in this application, the terms "model," "AI model," and "neural network model" are used interchangeably, and they all convey the same meaning.

[0108] The technical solutions provided in this application can be applied to various communication systems, such as 5th generation (5G) or new radio (NR) systems, frequency division duplex (FDD) systems, time division duplex (TDD) systems, wireless local area network (WLAN) systems, satellite communication systems, future communication systems, or integrated systems of multiple systems. The technical solutions provided in this application can also be applied to device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), and Internet of Things (IoT) communication systems or other communication systems.

[0109] In a communication system, a device can send signals to or receive signals from another device. These signals can include information, signaling, or data. The device can also be replaced by an entity, network entity, communication equipment, communication module, node, communication node, etc.; this disclosure uses a device as an example. For instance, a communication system can include at least one terminal device and at least one network device. The network device can send downlink signals to the terminal device, and / or the terminal device can send uplink signals to the network device. It is understood that the terminal device in this disclosure can be replaced by a first communication device, and the network device can be replaced by a second communication device, both performing the corresponding communication methods described in this disclosure.

[0110] In wireless communication networks, such as mobile communication networks, the services supported by the networks are becoming increasingly diverse, thus requiring increasingly diverse demands. For example, networks need to support ultra-high speeds, ultra-low latency, and / or massive connectivity. This characteristic makes network planning, network configuration, and / or resource scheduling increasingly complex. Furthermore, as network functions become more powerful, such as supporting higher spectrum levels, supporting higher-order multiple-input multiple-output (MIMO) technologies, supporting beamforming (BF), and / or supporting beam management, network energy efficiency has become a hot research topic. These new demands, new scenarios, and new characteristics bring unprecedented challenges to network planning, operation, and efficient operation. To meet these challenges, artificial intelligence (AI) technology can be introduced into wireless communication networks to achieve network intelligence. To support AI technology in wireless networks, AI nodes may also be introduced. AI nodes can be the AI ​​network elements shown in Figure 2 below or the AI ​​modules shown in Figure 3.

[0111] Figure 1 is a schematic diagram of a communication system applicable to the communication method of this application embodiment. As shown in Figure 1, the communication system 100A may include at least one access network device, such as access network device 110 shown in Figure 1; the communication system 100A may also include at least one terminal device, such as terminal device 120 and terminal device 130 shown in Figure 1. Access network device 110 and terminal devices (such as terminal device 120 and terminal device 130) can communicate via a wireless link. The communication devices in this communication system, for example, access network device 110 and terminal device 120, can communicate via multi-antenna technology.

[0112] Figure 2 is a schematic diagram of another communication system applicable to the communication method of this application embodiment. Compared with the communication system 100A shown in Figure 1, the communication system 100B shown in Figure 2 further includes an AI network element 140. The AI ​​network element 140 is used to perform AI-related operations, such as building datasets or AI models. The AI ​​network element can also be simply referred to as an intelligent network element. In this disclosure, the AI ​​model can be simply referred to as a model.

[0113] In one possible implementation, access network device 110 can send data related to the training of the AI ​​model to AI network element 140, whereby AI network element 140 constructs a dataset and trains the AI ​​model. For example, the data related to the training of the AI ​​model may include data reported by terminal devices. AI network element 140 can send the results of operations related to the AI ​​model to access network device 110, and then forward them to terminal devices via access network device 110. For example, the results of operations related to the AI ​​model may include at least one of the following: a trained AI model, model evaluation results, or test results, etc. Exemplarily, a portion of the trained AI model may be deployed on access network device 110, and another portion on terminal devices 120 and / or 130. Alternatively, the trained AI model may be deployed on access network device 110. Or, the trained AI model may be deployed on terminal devices 120 and / or 130.

[0114] It should be understood that Figure 2 is only used as an example of the AI ​​network element 140 being directly connected to the access network device 110. In other scenarios, the AI ​​network element 140 can also be connected to a terminal device (such as terminal device 120 and / or terminal device 130); or, the AI ​​network element 140 can be connected to both the access network device 110 and the terminal device (such as terminal device 120 and / or terminal device 130) simultaneously; or, the AI ​​network element 140 can also be connected to the access network device 110 through a third-party network element; the AI ​​network element 140 can also be set as a module in the access network device 110 or the terminal device (such as terminal device 120 and / or terminal device 130) shown in Figure 1. This application embodiment does not limit the connection relationship between the AI ​​network element and other network elements.

[0115] It should be noted that Figures 1 and 2 are simplified schematic diagrams for ease of understanding. For example, the communication system may also include other devices, such as wireless relay devices and / or wireless backhaul devices, which are not shown in Figures 1 and 2. In practical applications, the communication system may include multiple access network devices and multiple terminal devices. The embodiments of this application do not limit the number of access network devices and terminal devices included in the communication system.

[0116] In the embodiments of this application, the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user apparatus.

[0117] Terminal devices can be devices that provide voice / data, such as handheld devices with wireless connectivity, in-vehicle devices, etc. Currently, examples of terminals include: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, wearable devices, terminal devices in 5G networks, or future public land mobile communication networks. Terminal devices in a network (PLMN), etc., are not limited to this in the embodiments of this application.

[0118] By way of example and not limitation, in this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those that focus on a specific type of application function and require the use of other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0119] In this embodiment, the device for implementing the functions of the terminal device can be the terminal device itself, or it can be any device capable of supporting the terminal device in implementing those functions, such as a chip system. This device can be installed in or used in conjunction with the terminal device. In this embodiment, the chip system can be composed of chips or may include chips and other discrete components. This embodiment only uses the terminal device as an example to illustrate the device for implementing the functions of the terminal device, and does not constitute a limitation on the solution of this embodiment.

[0120] The network device in this application embodiment can be a device for communicating with a terminal device. This network device can include an access network device (i.e., an access network node) or a radio access network device, such as a base station. In this application embodiment, the radio access network device can refer to a radio access network (RAN) node (or device) that connects the terminal device to the wireless network. A base station can broadly encompass, or be replaced by, various names including: NodeB, evolved NodeB (eNB), next-generation NodeB (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), master station, auxiliary station, motor slide retainer (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. A base station can be a macro base station, micro base station, relay node, donor node, or similar entities, or combinations thereof. A base station can also refer to a communication module, modem, or chip installed within the aforementioned equipment or apparatus. A base station can also be a mobile switching center, equipment performing base station functions in D2D, V2X, and M2M communications, network-side equipment in future communication networks, or equipment performing base station functions in future communication networks. A base station can support networks using the same or different access technologies. Optionally, a RAN node can also be a server, wearable device, vehicle, or in-vehicle equipment. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU). The embodiments of this application do not limit the specific technologies or equipment forms used in the network equipment.

[0121] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move depending on the location of the mobile base station. In other examples, a helicopter or drone can be configured as a device to communicate with another base station.

[0122] In some deployments, the access network equipment mentioned in the embodiments of this application may be a device including a CU, or a DU, or a device including both CU and DU, or a device with a control plane CU node (central unit-control plane (CU-CP)) and a user plane CU node (central unit-user plane (CU-UP)) and a DU node. For example, the access network equipment may include gNB-CU-CP, gNB-CU-UP, and gNB-DU.

[0123] In some deployments, multiple RAN nodes collaborate to assist terminals in achieving wireless access, with different RAN nodes each implementing some of the base station's functions. For example, RAN nodes can be CUs, DUs, CU-CPs, CU-UPs, or RUs. CUs and DUs can be configured separately or included in the same network element, such as a BBU. RUs can be included in radio frequency equipment or radio frequency units, such as RRUs, AAUs, or RRHs.

[0124] RAN nodes can support one or more types of fronthaul interfaces, each corresponding to a DU and RU with different functions. If the fronthaul interface between the DU and RU is a Common Public Radio Interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and RU is another type of interface, relative to CPRI, it moves some downlink and / or uplink baseband functions—for example, for downlink, precoding, digital beamforming, or one or more of inverse fast Fourier transform (IFFT) / adding a cyclic prefix (CP)—from the DU to the RU; and for uplink, digital beamforming, or one or more of fast Fourier transform (FFT) / removing CP—from the DU to the RU. In one possible implementation, this interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the splitting methods between DU and RU are different, corresponding to different types (category, Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0125] Taking eCPRI Cat A as an example, for downlink transmission, layer mapping is used as the dividing line. The DU is configured to implement one or more functions preceding layer mapping (i.e., coding, rate matching, scrambling, modulation, and layer mapping itself), while other functions following layer mapping (e.g., resource element (RE) mapping, digital beamforming, or IFFT / CP addition) are implemented in the RU. For uplink transmission, de-RE mapping is used as the dividing line. The DU is configured to implement one or more functions preceding de-mapping (i.e., decoding, rate matching de-matching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping itself), while other functions following de-mapping (e.g., digital beamforming or FFT / CP removal) are implemented in the RU. It is understood that descriptions of the functions of the DU and RU corresponding to various types of eCPRI can be found in the eCPRI protocol and will not be elaborated upon here.

[0126] In one possible design, the processing unit in the BBU used to implement baseband functions is called the baseband high (BBH) unit, and the processing unit in the RRU / AAU / RRH used to implement baseband functions is called the baseband low (BBL) unit.

[0127] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an open RAN (ORAN) architecture, CU can also be called open CU (open-CU, O-CU), DU can also be called open DU (open-DU, O-DU), CU-CP can also be called open CU-CP (open-CU-CP) O-CU-CP, CU-UP can also be called open CU-UP (open-CU-UP, O-CU-UP), and RU can also be called open RU (open-RU, O-RU). Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0128] In this embodiment, the apparatus for implementing the functions of a network device can be a network device itself; it can also be an apparatus capable of supporting the network device in implementing those functions, such as a chip system, hardware circuit, software module, or a hardware circuit plus a software module. This apparatus can be installed in the network device or used in conjunction with the network device. In this embodiment, the example of a network device being used to implement the functions of a network device is provided only and does not constitute a limitation on the solutions described in this embodiment.

[0129] Network devices and / or terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on airplanes, balloons, and satellites. This application does not limit the scenario in which the network devices and terminal devices are located. Furthermore, terminal devices and network devices can be hardware devices, or software functions running on dedicated hardware or general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities that include dedicated or general-purpose hardware devices and software functions. This application does not limit the specific form of the terminal devices and network devices.

[0130] Optionally, the AI ​​node can be deployed in one or more of the following locations within the communication system: access network equipment, terminal equipment, or core network elements. Alternatively, the AI ​​node can also be deployed independently, for example, in a location other than any of the aforementioned devices, such as in the host or cloud server of an over-the-top (OTT) system. The AI ​​node can communicate with other devices in the communication system, which can be, for example, one or more of the following: access network equipment, terminal equipment, or core network elements.

[0131] It is understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, these nodes can be divided based on function, such as different AI nodes being responsible for different functions.

[0132] It can also be understood that AI nodes can be AI network elements or AI modules. AI nodes can be independent devices, or they can be integrated into the same device to implement different functions. Alternatively, they can be network elements in hardware devices, software functions running on dedicated hardware, or virtualization functions instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of the AI ​​nodes described above.

[0133] Figure 3 illustrates a possible application framework in a communication system. As shown in Figure 3, network elements in the communication system are connected via interfaces (e.g., next-generation (NG) interfaces, Xn interfaces) or air interfaces. The NG interface is the interface between the radio access network and the 5G core network. The Xn interface is the interface between access network devices, and the air interface is the interface between access network devices and terminal devices. These network element nodes, such as core network devices, RAN nodes, terminal devices, or one or more devices in operation administration and maintenance (OAM), are equipped with one or more AI modules (only one is shown in Figure 3 for clarity). The access network node can be a single RAN node or can include multiple RAN nodes, for example, including CU, DU, and RU. One or more of the RU, CU, or DU can also be equipped with one or more AI modules. Optionally, the CU can be further divided into CU-CP and CU-UP. One or more AI modules are configured in CU-CP and / or CU-UP.

[0134] AI modules are used to implement corresponding AI functions. AI modules deployed in different network elements can be the same or different. Depending on the parameter configuration, the AI ​​module can implement different functions. The AI ​​module model can be configured based on one or more of the following parameters: structural parameters (e.g., at least one of the following: number of neural network layers, neural network width, inter-layer connections, neuron weights, neuron activation function, or bias in the activation function), input parameters (e.g., at least one of the following: type of input parameter, input dimension, number of input ports), or output parameters (e.g., at least one of the following: type of output parameter, output dimension, number of output ports). The bias in the activation function can also be called the bias of the neural network. Input dimension can refer to the size of an input data set; for example, when the input data is a sequence, the input dimension corresponding to that sequence can indicate the length of the sequence. The number of input ports can refer to the quantity of input data. Similarly, output dimension can refer to the size of an output data set; for example, when the output data is a sequence, the output dimension corresponding to that sequence can indicate the length of the sequence. The number of output ports can refer to the quantity of output data.

[0135] An AI module can have one or more models. A model can infer an output, which includes one or more parameters. The learning, training, or inference processes of different models can be deployed on different nodes or devices, or they can be deployed on the same node or device.

[0136] Network devices can be network devices equipped with one or more AI modules. These network devices can include one or more devices in the core network, RAN, or OAM as shown in Figure 3. For example, the AI ​​module can be a RAN intelligent controller (RIC) as shown in Figure 4, such as a near-real-time RIC (near-RT RIC) or a non-real-time RIC (non-RT RIC). For instance, a near-real-time RIC is located in a RAN node (e.g., in a CU or DU), while a non-real-time RIC is located in the OAM, a cloud server, a core network device, or other access network devices. The RIC can obtain subsets from multiple terminal devices from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU), reassemble them into a dataset, and train based on the dataset. Exemplarily, near-real-time RICs and non-real-time RICs can also be configured as separate network elements, and access network devices can be either near-real-time or non-real-time RICs.

[0137] Figure 4 illustrates another possible application framework in a communication system. In addition to access network nodes (CU, DU, and RU are shown in the figure) and terminals, the communication system shown in Figure 4 also includes an RIC (Regulator-Integrated Circuit). For example, the RIC could be the AI ​​module shown in Figure 3, which can be used to implement AI-related functions. The RIC includes near-real-time RICs and non-real-time RICs. Non-real-time RICs primarily process non-real-time information, such as data that is not sensitive to latency, with latency on the order of seconds. Real-time RICs primarily process near-real-time information, such as data that is relatively sensitive to latency, with latency on the order of tens of milliseconds.

[0138] Near real-time (NRT) RICs are used for model training and inference. For example, they are used to train AI models and then use those models for inference. NRT RICs can obtain network-side and / or terminal-side information from RAN nodes (e.g., CUs, CU-CPs, CU-UPs, DUs, and / or RUs) and / or terminals. This information can be used as training data or inference data. Optionally, the NRT RIC can deliver inference results to RAN nodes and / or terminals. Optionally, inference results can be exchanged between CUs and DUs, and / or between DUs and RUs. For example, the NRT RIC delivers inference results to a DU, which then forwards them to an RU.

[0139] Non-real-time RICs are also used for model training and inference. For example, they can be used to train AI models and then use those models for inference. Non-real-time RICs can obtain network-side and / or terminal-side information from RAN nodes (e.g., one or more of CU, CU-CP, CU-UP, DU, or RU) and / or terminals. This information can be used as training data or inference data, and the inference results can be delivered to the RAN nodes and / or terminals. Optionally, inference results can be exchanged between CUs and DUs, and / or between DUs and RUs; for example, a non-real-time RIC delivers inference results to a DU, which then forwards them to an RU.

[0140] Near real-time RICs and non-real-time RICs can also be configured as separate network elements. Alternatively, near real-time RICs and non-real-time RICs can also be part of other devices. For example, near real-time RICs can be set in RAN nodes (e.g., CU, DU), while non-real-time RICs can be set in OAM, cloud servers, core network devices, or other network devices.

[0141] With the development of AI, AI / ML technology has been introduced into the wireless air interface, for example, for the compression and reconstruction of wireless channel information, beam management and positioning enhancement, and can also be used for data processing in transmitters and / or receivers.

[0142] Figure 5 is a schematic diagram of the AI / ML functional framework for wireless air interfaces. Figure 5 shows the data collection unit, model training unit, management unit, inference unit, and model storage unit, etc.

[0143] The data collection unit can collect data from network devices and / or terminal devices. AI-based air interface or RAN systems require different types of data at various stages of the AI ​​model development process. For example, training the model requires a large amount of diverse data, model inference requires real-time data, and model monitoring requires near real-time data. For wireless or air interface tasks, the data collected by the data collection unit via the wireless air interface includes, but is not limited to, one or more of the following: measurement signals transmitted by network devices, measurement results measured and reported by terminal devices, reference signals transmitted by terminal devices, and measurement results measured by network devices. The data collected by the data collection unit, after preprocessing and other operations, can be used for one or more of the following: model training, model inference, model monitoring, or analysis. The data collected by the data collection unit can be applied to different stages, such as training data for model training, performance monitoring data for monitoring model performance, and so on.

[0144] The model training unit can train or retrain (or update) the model based on the training data from the data collection unit, and can send the trained model or the updated model to the model storage unit for storage.

[0145] The management unit can monitor the model's performance based on monitoring data from the data collection unit, and perform actions such as activating / deactivating, switching, selecting, and rolling back the model (or model function) based on performance. For example, it can send performance feedback or retraining requests to the model training unit. The management unit can also send model transfer requests to the model storage unit to request that the model storage unit send the model to the inference unit.

[0146] The inference unit can perform model inference based on the model from the model storage unit and the inference data from the data collection unit, and output the inference results. The inference unit can also receive instructions from the model management unit.

[0147] Intelligent air interface technology can be understood as the application of AI models, or neural network (NN) models, in wireless communication systems to optimize the performance of wireless communication systems.

[0148] Use cases of AI models in wireless communication include, but are not limited to, channel prediction, channel compression and reconstruction, channel estimation, multi-input multi-output (MIMO) equalization, modulation and demodulation, coding and decoding, etc. These use cases are described below with reference to the accompanying figures.

[0149] Channel Prediction: Figure 6 illustrates the AI ​​model used to implement channel prediction. As shown, the model's input is historical channel information, and its output is the predicted future channel information, i.e., the predicted value of the channel information. Model training data includes historical channel information and the ground truth values ​​of future channel information. Model training can involve updating model parameters with the goal of minimizing the error between the predicted and ground truth values ​​(e.g., minimum square error (MSE)) or maximizing the similarity (e.g., square generalized cosine similarity (SGCS)). The model's performance is evaluated by the error (e.g., MSE) or the similarity (e.g., SGCS) between the predicted and actual future channel information; smaller errors or higher similarities result in better performance.

[0150] Channel compression and reconstruction: This can be achieved through channel compression models and channel reconstruction models. Figure 7 is a schematic diagram of an AI model used to implement channel compression and reconstruction (or reconstruction) functions. As shown in the figure, the input of the channel compression model can be the original channel information, such as the channel matrix, precoding matrix, or eigenvector, and the output can be the compressed bits; the input of the channel reconstruction model (or channel reconstruction model) can be the compressed bits, and the output can be the reconstructed (or rebuilt) channel information. The model training data includes the original channel information. The model training can include updating the model parameters with the goal of minimizing the error (e.g., MSE) or maximizing the similarity (e.g., SGCS) between the reconstructed channel information obtained through the channel compression model and the channel reconstruction model and the original channel information. The performance of the model is evaluated by the error (e.g., MSE) or similarity (e.g., SGCS) between the reconstructed channel and the original channel; the smaller the error or the higher the similarity, the better the performance.

[0151] Channel Estimation: Figure 8 illustrates the AI ​​model used to implement channel estimation. As shown, the model input is the received signal, including multiple resource elements (REs). Some of these REs carry reference signals (RS), also known as pilot signals. These REs can be called reference signal REs or pilot REs. The model output is the estimated channel information corresponding to multiple REs. Model training data can include the channel information corresponding to the received signal or pilot REs, and the channel information of all REs (e.g., true channel information, i.e., the ground truth of channel information). Model training can involve updating model parameters with the goal of minimizing the error (e.g., MSE) or maximizing the similarity (e.g., SGCS) between the estimated channel information and the true channel information. The model performance is evaluated by the error (e.g., MSE) or similarity (e.g., SGCS) between the estimated channel information and the true channel information; the smaller the error or the higher the similarity, the better the performance.

[0152] MIMO equalization: This can be simply called equalization. Figure 9 is a schematic diagram of an AI model used to implement MIMO equalization. As shown, the model's input is the signal (or symbol) received through Nr antennas, and the output is the equalized signal (or symbol) from Nt streams. Here, Nr represents the number of receiving antenna ports, and Nt represents the number of data streams, or transmission layers (it should be understood that Nt data streams are obtained by mapping data onto Nt transmission layers; therefore, a data stream can be considered as data mapped to one transmission layer, such as a symbol). Both Nr and Nt are positive integers greater than or equal to 1. The model training data includes the received signal and the corresponding transmitted symbol, which is then used to generate the equalized symbol output by the model. Model training can include minimizing the error between the equalized symbol and the transmitted symbol (such as MSE or cross entropy (CE)) to update the model parameters. The model's performance is evaluated by the error (such as MSE or CE) between the equalized symbol and the transmitted symbol; the smaller the error, the better the performance.

[0153] Demodulation: Figure 10 is a schematic diagram of the AI ​​model used to implement the demodulation function. As shown in the figure, the input of the model is the received symbol or the equalized symbol, and the output is the demodulated bit corresponding to the symbol (as shown in Figure 10(a)) or the log-likelihood ratio (LLR) corresponding to the demodulated bit (as shown in Figure 10(b)). N in the figure represents N demodulated bits, and N is a positive integer greater than 1. The received symbol can be an un-pre-encoded symbol or a pre-encoded symbol. This model can also be used to implement the equalization function, so the received symbol can be equalized first and then demodulated.

[0154] The model training data includes received symbols and their corresponding transmitted bits (i.e., known bits). The model parameters are updated with the objective of minimizing the error between the estimated demodulated bits and the transmitted bits (such as MSE, CE, or bit error rate (BER)). The model's performance is evaluated by the error between the demodulated bits and the transmitted bits (such as MSE, CE, or BER); the smaller the error, the better the performance.

[0155] Decoding: Figure 11 is a schematic diagram of the AI ​​model used to implement the decoding function. As shown in the figure, the input of the model is the demodulated signal (the bits to be decoded shown in the figure), and the output is the estimated decoded bits. In the figure, N represents N bits to be decoded; K represents K decoded bits; N is greater than K, and N and K are positive integers. The model training data includes the received signal or the demodulated signal, and the corresponding transmitted bits (i.e., known bits). The training of the model may include updating the model parameters with the goal of minimizing the error (such as MSE, CE, or BER) between the estimated decoded bits output by the model and the transmitted bits. The performance of the model is evaluated by the error (such as MSE, CE, or BER) between the estimated decoded bits and the transmitted bits; the smaller the error, the better the performance.

[0156] Figure 12 is a schematic diagram of a wireless communication system including a neural network-based transceiver (including a transmitter and a receiver) provided in an embodiment of this application. In the system shown in Figure 12, the transceiver using a neural network model can optimize the performance of transmitting and receiving signals through data-driven training.

[0157] Transceivers employing neural network models can optimize signal transmission and reception performance through data-driven training. This means optimizing one or more modules within the transceiver used for signal processing, such as channel coding, modulation, waveform shaping, and pilot signals, using the neural network model. In the diagram, the neural network (e.g., a deep neutral network (DNN)) can replace one or more modules in the transmitter and receiver to perform one or more operations. More specifically, a neural network-based transmitter can be used to perform one or more of the following in data transmission processing: channel coding, modulation, waveform shaping, precoding, time-domain signal processing, or reference signal (or pilot) optimization. Time-domain signal processing can include various possible operations, such as digital predistortion, pre-equalization, nonlinear compensation, and beamforming, and the neural network can be used to perform one or more of these operations. A neural network-based receiver can be used to perform one or more of the following in data reception processing: time-domain signal processing, equalization, waveform shaping, demodulation, channel decoding, or RS-based channel estimation. More specific examples can be found in Figures 13 and 14.

[0158] Figure 13 is an example of a transceiver including a neural network based embodiment of this application. The transmitter's neural network can be used to optimize the modulation module and / or time-domain signal processing. The information bits to be transmitted are channel-coded to obtain coded bits. After the coded bits are modulated by the neural network, modulation symbols are obtained. The modulation symbols are then RE-mapped to map them onto physical resources. After inverse fast Fourier transform (IFFT), the time-domain signal is obtained. After further time-domain signal processing by the neural network, the signal is transmitted. The receiver's neural network can be used to optimize the demodulation module and time-domain signal processing. The received signal is processed by the neural network for time-domain signal processing and fast Fourier transform (FFT) to obtain modulation symbols. After demodulation by the neural network, the bits to be decoded are obtained. The bits to be decoded are then channel-decoded to obtain estimated information bits.

[0159] Figure 14 is another example of a transceiver including a neural network based on an embodiment of this application. The transmitter's neural network is used to optimize the reference signal. The transmitter generates a reference signal based on the neural network, which is then processed by IFFT and a time-domain signal neural network before being transmitted. The receiver's neural network is used for channel estimation. The received signal is processed by the time-domain signal neural network and then by FFT to obtain the received reference signal. The neural network can perform channel estimation based on the received reference signal.

[0160] Although not shown in the figures, it can be understood that since waveform shaping is primarily performed after the IFFT or before the FFT, and optionally also before or after the IFFT, the waveform shaping operation in the processing shown in Figures 13 and 14 can be included within the IFFT or FFT, and is not shown separately. Furthermore, neural networks are not limited to performing the modulation and time-domain signal processing shown in Figure 13, and the reference signal optimization, time-domain signal processing, and channel estimation shown in Figure 14; as previously mentioned, neural networks can also be used to optimize waveforms.

[0161] It should be understood that the wireless communication systems with neural network-based transceivers shown in Figures 13 and 14 are merely examples. In practical applications, neural network-based transceivers can also be used to implement more functions, such as precoding and waveform shaping. For the sake of simplicity, detailed illustrations are not provided here.

[0162] Figure 15 is a schematic diagram of the training process of the intelligent air interface provided in an embodiment of this application. As shown in the figure, the training of the neural network model used in the wireless communication system includes the training of the transmitting-end neural network model and the receiving-end neural network model. The transmitting-end neural network model performs forward inference on the training data and outputs the forward inference result (as shown in the figure, z). x The forward inference result is sent to the receiving neural network model; the receiving neural network model performs forward inference and backward gradient calculation based on the forward inference result, and obtains the feedback result (as shown in the figure). The receiving neural network model sends the feedback result to the sending neural network model; the sending neural network model performs backpropagation gradient calculation based on the feedback result, and updates its model parameters based on the gradient obtained from the backpropagation gradient calculation. For example, the t-th round update of the sending neural network model's parameters satisfies:

[0163] Among them, w t+1 It refers to the parameters in round t+1, i.e., the parameters updated in round t; w t It is the parameter for round t; z x It is the output of the sending neural network model corresponding to the x-th training sample, l x η is the loss function corresponding to round t, and X is the number of samples trained in this round; t It is the learning rate; It is the gradient of the loss function with respect to the output of the sending neural network model.

[0164] In other words, during any training round, the sending neural network model first performs forward inference to obtain the forward inference result and intermediate results; the intermediate inference results are cached; then, the loss function is calculated from the last layer, and gradients are calculated layer by layer according to the chain rule. The gradient calculation process requires the results of forward inference from each layer. The gradients calculated from multiple training data are summed and used to update the model parameters of the sending neural network model.

[0165] For example, when applying this training process to an AI model for channel prediction, the loss function can be the error (e.g., MSE), similarity (e.g., SGCS), or a function of similarity (e.g., 1-SGCS) between the predicted and true values ​​of future channel information; when applying this training process to an AI model for channel compression and reconstruction, the loss function can be the error (e.g., MSE), similarity (e.g., SGCS), or a function of similarity (e.g., 1-SGCS) between the reconstructed and original channel information; when applying this training process to an AI model for channel estimation, the loss function can be the channel information corresponding to all estimated REs, and... The loss function can be the error (e.g., MSE), similarity (e.g., SGCS), or a function of similarity (e.g., 1-SGCS) between the real channel information corresponding to all REs; when this training process is applied to an AI model for implementing MIMO equalization, the loss function can be the error between the equalized symbol and the transmitted symbol (e.g., MSE or CE); when this training process is applied to an AI model for implementing demodulation, the loss function can be the estimated error between the demodulated bit and the transmitted bit (e.g., MSE, BER, or CE); when this training process is applied to an AI model for implementing decoding, the loss function can be the estimated error between the decoded bit and the transmitted bit (e.g., MSE, BER, or CE).

[0166] It should be understood that the examples of loss functions applied to AI models with different functionalities above are merely illustrative and should not constitute any limitation on this application. For example, error can also be represented by normalized mean square error (NMSE), and similarity can also be represented by generalized cosine similarity (GCS). This application includes, but is not limited to, these methods.

[0167] As mentioned earlier, the current management of management nodes at the individual model (or function) level is inefficient. In reality, multiple models may have constraints on each other. For example, the activation or deactivation of one model might affect the activation or deactivation of the other, or vice versa. Therefore, managing models by incorporating these constraints can improve model execution efficiency and optimize the performance of the wireless communication system.

[0168] The methods provided in the embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0169] Figure 16 is a schematic flowchart of the communication method 1600 provided in an embodiment of this application. Figure 16 illustrates the method from the perspective of interaction between the second device and the third device. The first device can be a node deploying AI functions, and the second device can act as a management node to manage the AI ​​functions deployed in other devices (such as the first device). The first device can communicate with the third device; the third device can receive information from the first device and send information to the first device. The third device can also communicate with the second device; the third device can receive information from the second device and send information to the second device. For example, the third device can send information received from the second device to the first device.

[0170] As an example, the second device can be on the network side, such as a network device, or a chip or circuit in a network device. The first device can be on the terminal side, such as a terminal device, or a chip or circuit in a terminal device, or an AI entity serving the terminal device, such as a server in an OTT system or a cloud server; the third device can also be on the terminal side, such as a terminal device, or a chip or circuit in a terminal device.

[0171] In another example, the second device can be on the terminal side, such as a terminal device, or a chip or circuit in a terminal device. The first device can be on the network side, such as a network device, or a chip or circuit in a network device, or an AI entity serving the network device, such as a smart network element; the third device can also be on the network side, such as a network device, or a chip or circuit in a network device.

[0172] In another example, the second device can be on the network side, and the first and third devices can also be on the network side. The third device can be the first device or another device besides the first device. In this case, the second device can be a network element or entity on the network side, and the first and third devices can be another network element or entity on the network side.

[0173] It should be understood that the first device and the third device can be different modules within the same device, or they can be different devices, or modules within different devices, without limitation. Furthermore, the first device and the third device can communicate, for example, through device-to-device communication such as wired communication, or through internal pathways within the devices, without limitation.

[0174] It should also be understood that the first device, the second device, or the third device may be replaced by the first entity, the second entity, or the third entity, at least two of which may be different physical devices, or at least two of which may be deployed in the same physical device.

[0175] As shown in Figure 16, method 1600 includes steps 1610 to 1620. Optionally, it also includes step 1630. The various steps in method 1600 are described in detail below.

[0176] In step 1610, the third device determines that one or more AI functions satisfy the first constraint relationship.

[0177] In this application's embodiments, AI functionality can also be referred to as AI feature. An AI functionality or AI feature can be implemented by one or more models. A model can also be used to implement multiple AI functionality. Therefore, the AI ​​functionality in this application's embodiments can also be replaced by AI model or AI feature.

[0178] The aforementioned one or more AI functions may be AI functions supported by the first device. For example, the one or more AI functions may be all AI functions supported by the first device, or they may be some of the AI ​​functions supported by the first device, without limitation. The AI ​​functions supported by the first device mean that the first device has deployed an AI model related to that function, and can use that AI model to implement the corresponding function.

[0179] As an example and not a limitation, the one or more AI functions include one or more of the following: channel estimation, channel prediction, channel compression and reconstruction, coding, modulation, precoding, equalization, demodulation, or decoding. To facilitate the distinction between AI functions and non-AI functions, each function implemented through AI will be named using the format "AI + specific function." Therefore, it can also be stated that the above one or more AI functions include one or more of the following: AI channel estimation, AI channel prediction, AI channel compression and reconstruction, AI coding, AI modulation, AI precoding, AI equalization, AI demodulation, or AI decoding.

[0180] The third device can determine a first constraint relationship between the one or more AI functions, which can be one or more of the following constraints: independence, dependency, substitution, or condition switching.

[0181] 1. Independent relationship:

[0182] An AI function that satisfies the independence relationship is a single AI function. In other words, an AI function satisfies the independence relationship if the activation, deactivation, monitoring, and inference configuration of that AI function are not affected by other AI functions.

[0183] The inference configuration includes configurations for one or more of the following: inference parameters, parameters of the model output data, or parameters of the model itself. Inference parameters may include, for example, parameters of the model input signal, such as RS parameters (e.g., bandwidth, port), and data channel parameters (e.g., bandwidth, port, code rate, modulation and coding scheme). Parameters of the model output data may be, for example, parameters reported by CSI, such as the number of bits. Model parameters may be, for example, one or more of the following: the model itself, model identifier, model weights, and model accuracy. Since more detailed parameters for inference configuration are listed below for different AI functions, they will not be elaborated here. Optionally, inference parameters can also be replaced with parameters of the model input data.

[0184] The explanation of the reasoning configuration in the following text can be found in the explanation here, and will not be repeated here.

[0185] 2. Dependency Relationships:

[0186] A set of AI functions that satisfy a dependency relationship consists of multiple AI functions. This dependency relationship means that the activation of one or more AI functions depends on the activation of another one or more AI functions, or that the activation of one or more AI functions is conditional on the activation of another one or more AI functions. These multiple AI functions that satisfy a dependency relationship can be trained together, for example, through joint training on the same dataset.

[0187] To facilitate understanding, let's take two AI functions as an example. These two AI functions satisfy a dependency relationship, where the later AI function depends on the earlier one. For instance, the activation of the later AI function depends on the activation of the earlier one; that is, if the later AI function is activated, the earlier one is also activated. In other words, the activation of the later AI function requires the earlier one to be activated; if the earlier one is activated, the later AI function can be activated or deactivated. Similarly, the deactivation of the later AI function depends on the deactivation of the earlier one; that is, if the earlier one is deactivated, the later one is also deactivated. In other words, the deactivation of the earlier one requires the later one to be activated; if the later one is deactivated, the earlier one can be activated or deactivated. Furthermore, the inference configuration of the later AI function can also be determined based on the state of the earlier AI function (e.g., activated or deactivated), or in other words, based on whether the earlier AI function is activated (or deactivated).

[0188] Taking two AI functions as an example, assuming AI function a depends on AI function b, then: if AI function b is activated, AI function a is also activated, and inference configuration can be performed on AI function a and AI function b; if AI function b is deactivated, AI function a is also deactivated, and inference configuration can be omitted for AI function a and AI function b.

[0189] 3. Substitution relationship:

[0190] A set of AI functions satisfying a substitution relationship can be multiple AI functions. This substitution relationship means that these multiple AI functions are not activated simultaneously; that is, one or more AI functions are not activated at the same time as another one or more AI functions. Therefore, when multiple AI functions satisfy a substitution relationship, the activation of one or more AI functions depends on the activation of the other one or more AI functions. For ease of understanding, let's take two AI functions as an example: AI function 1 and AI function 2 satisfy a substitution relationship. In this case, AI function 1 and AI function 2 are not activated simultaneously. If AI function 1 is activated, then AI function 2 is not activated; if AI function 2 is activated, then AI function 1 is not activated.

[0191] Multiple AI functions that satisfy the substitution relationship can be multiple AI functions with overlapping performance enhancements. Overlapping performance enhancements from multiple AI functions specifically mean that these multiple AI functions bring the same direction of performance optimization to the wireless communication system; therefore, using all AI functions does not change the performance enhancement compared to using only some of them. For example, when using traditional algorithms for decoding and demodulation, if the assumptions about the channel model and noise are inconsistent with the real channel, it may lead to performance degradation. However, neural networks can learn the channel model and noise of the real channel through deep learning, thereby improving performance. Therefore, AI functions can be used to implement decoding or demodulation. Thus, the assumptions about the channel model and noise used for decoding and demodulation differ from those of traditional algorithms. Therefore, AI functions used for decoding offer performance improvements compared to traditional algorithms; AI functions used for demodulation also offer performance improvements compared to traditional algorithms. However, if AI functions are used simultaneously for decoding and demodulation, there is no performance gain due to the overlapping performance enhancements; therefore, it is not necessary to use AI functions for both decoding and demodulation simultaneously. In other words, the AI ​​function used for decoding and the AI ​​function used for demodulation are not activated at the same time.

[0192] Furthermore, the inference configuration of a subsequent AI function can also be determined based on the state of the preceding AI function (e.g., activated or deactivated), or in other words, based on whether the preceding AI function is activated (or deactivated). For a detailed understanding of how the inference configuration of a subsequent AI function is determined based on the state of the preceding AI function, please refer to the relevant description of dependencies, which will not be repeated here.

[0193] 4. Condition switching relationship:

[0194] A set of AI functions that satisfy the conditional switching relationship consists of multiple AI functions. Multiple AI functions satisfying the conditional switching relationship means that the inference configuration of one or more of these AI functions is determined based on the state (e.g., activated or deactivated) of one or more other AI functions.

[0195] As mentioned earlier, inference configuration includes signal-related configuration and model configuration. Taking two AI functions as an example, the inference configuration for the subsequent AI function when the former is activated can differ from the inference configuration for the subsequent AI function when the former is deactivated. In other words, the activation or deactivation of the former AI function affects the inference configuration of the subsequent AI function; for example, it affects which model the subsequent AI function uses, or the accuracy of the data, etc.

[0196] In summary, different constraints between one or more AI functions affect the activation and / or inference configuration of those AI functions. Therefore, if one or more AI functions satisfy the first constraint, it can also be said that the activation of the one or more AI functions satisfies the first constraint, and / or that the inference configuration of the one or more AI functions satisfies the first constraint.

[0197] It should be noted that when describing the constraints of multiple AI functions above, the sequential relationship between AI functions was introduced, such as "previous AI function" and "next AI function." This sequential relationship refers to the order in which the corresponding functions are executed in the transmitter and / or receiver. For example, "previous AI function" and "next AI function" are relative to two AI functions; from a sequential perspective, the function executed first is the "previous AI function," and the function executed later is the "next AI function."

[0198] Figure 17 illustrates various constraint relationships exemplarily.

[0199] As shown in Figure 17(a), AI function a and AI function b satisfy a dependency relationship. The activation of AI function b depends on the activation of AI function a, that is, if AI function b is activated, then AI function a is also activated.

[0200] As shown in Figure 17(b), AI function c and AI function b satisfy a substitution relationship, meaning that AI function c and AI function b are not activated simultaneously. Therefore, when AI function b is activated, AI function c is not activated, and the output of AI function b can be used as the input of traditional function 2; when AI function b is deactivated, AI function c is activated, and the input of AI function c can be the output of traditional function 1. For ease of distinction, the process of the signal being processed by AI function b and traditional function 2 is shown with solid lines, while the process of the signal being processed by traditional function 1 and AI function c is shown with dashed lines.

[0201] As shown in Figure 17(c), AI functions a and c satisfy a conditional switching relationship. That is, the activation or deactivation of AI function a affects the inference configuration of AI function c. When AI function a is activated, the input of AI function c can be the output of AI function a, and inference configuration 1 can be used on AI function c; when AI function a is deactivated, the input of AI function c can be the output of traditional function a, and inference configuration 2 can be used on AI function c. Inference configuration 1 and inference configuration 2 are different inference configurations.

[0202] Several examples are given below. For ease of distinction and explanation, the specific AI functions implemented through the model will be identified by AI, such as AI channel estimation, which means the function of channel estimation is implemented through the model; AI equalization, which means the function of equalization is implemented through the model, and so on, without further listing.

[0203] Example 1: One or more of the above AI functions include AI channel estimation, and the first constraint relationship is an independence relationship. That is, the activation and / or inference configuration of AI channel estimation is not affected by other AI functions. Therefore, AI channel estimation can be activated independently and can also be configured for inference independently.

[0204] Example 2: One or more of the above AI functions include AI channel estimation and AI equalization, and the first constraint relationship is a dependency relationship. That is, AI channel estimation and AI equalization satisfy a dependency relationship. Since equalization is performed after channel estimation, AI equalization depends on AI channel estimation. Therefore, if AI channel estimation is activated, AI equalization can also be activated; if AI channel estimation is deactivated, AI equalization can also be deactivated. Furthermore, the inference configuration for AI equalization is determined based on the state of AI channel estimation. The inference configuration for AI equalization when AI channel estimation is activated can be different from the inference configuration for AI equalization when AI channel estimation is deactivated.

[0205] Example 3: One or more of the above AI functions include AI channel estimation and AI channel prediction, and the first constraint relationship is a dependency relationship. That is, AI channel estimation and AI channel prediction satisfy a dependency relationship. Since AI channel prediction is executed after AI channel estimation, AI channel prediction depends on AI channel estimation. Therefore, if AI channel estimation is activated, AI channel prediction can also be activated; if AI channel estimation is deactivated, AI channel prediction can also be deactivated. Furthermore, the inference configuration for AI channel prediction is determined based on the state of AI channel estimation. The inference configuration for AI channel prediction when AI channel estimation is activated can be different from the inference configuration for AI channel prediction when AI channel estimation is activated and deactivated.

[0206] Example 4: One or more of the above AI functions include AI channel estimation and AI channel compression and reconstruction. The first constraint relationship is a conditional switching relationship, that is, AI channel estimation and AI channel compression and reconstruction satisfy the conditional switching relationship. Since AI channel compression and reconstruction are performed after AI channel estimation, the inference configuration for AI channel compression and reconstruction is determined according to the state of AI channel estimation. The inference configuration for AI channel compression and reconstruction when AI channel estimation is active can be different from the inference configuration for AI channel compression and reconstruction when AI channel estimation is deactivated.

[0207] Example 5: One or more of the above AI functions include AI channel prediction and AI channel compression and reconstruction. The first constraint relationship is a conditional switching relationship, that is, AI channel prediction and AI channel compression and reconstruction satisfy the conditional switching relationship. Since AI channel compression and reconstruction are executed after AI channel prediction, the inference configuration for AI channel compression and reconstruction is determined according to the state of AI channel prediction. The inference configuration for AI channel compression and reconstruction when AI channel prediction is active can be different from the inference configuration for AI channel compression and reconstruction when AI channel prediction is deactivated.

[0208] Example 6: One or more of the above AI functions include AI equalization and AI decoding. The first constraint relationship is a substitution relationship, that is, AI equalization and AI decoding satisfy the substitution relationship. Specifically, AI equalization and AI decoding cannot be activated simultaneously; for example, AI equalization can be activated without AI decoding, or AI decoding can be activated without AI equalization.

[0209] Example 7: One or more of the above AI functions include AI demodulation and AI decoding. The first constraint relationship is a substitution relationship, that is, AI demodulation and AI decoding satisfy the substitution relationship. Specifically, AI demodulation and AI decoding are not activated simultaneously; for example, AI demodulation is activated but AI decoding is not activated; or AI decoding is activated but AI demodulation is not activated.

[0210] The examples above are merely for ease of understanding, using one or two AI functions as examples to illustrate the first constraint relationship satisfied between various AI functions. In actual implementation, the above one or more AI functions are not limited to satisfying only one constraint relationship, but can also satisfy multiple constraint relationships. In this case, the first constraint relationship can include multiple constraint relationships. Several examples will be given below with reference to Figures 18 and 19.

[0211] Example 8: As shown in Figure 18, Figure 18 is an example of multiple AI functions provided in this application's embodiments satisfying various constraint relationships. The aforementioned one or more AI functions include: AI channel estimation, AI channel prediction, and AI channel compression. Optionally, it also includes AI channel reconstruction. Exemplarily, during the training process or performance monitoring process, AI channel estimation, AI channel prediction, AI channel compression, and reconstruction can be deployed on a first device; or, during the training process, inference process, or performance monitoring process, AI channel estimation, AI channel prediction, and AI channel compression can be deployed on a first device, while AI channel reconstruction can be deployed on another device. In practical applications, AI channel estimation, AI channel prediction, and AI channel compression can be deployed on the terminal side, such as the first device on the terminal side, while AI channel reconstruction can be deployed on another device.

[0212] Specifically, AI channel estimation satisfies an independence relationship, AI channel estimation and AI channel prediction satisfy a dependency relationship, AI channel estimation and AI channel compression and reconstruction satisfy a conditional switching relationship, and AI channel prediction and AI channel compression and reconstruction also satisfy a conditional switching relationship. It is easy to see that if AI channel prediction is activated, AI channel estimation is also activated. Whether AI channel estimation and AI channel prediction are activated affects the inference configuration of AI channel compression and reconstruction. For example, when using AI channel estimation and / or AI channel prediction, AI channel compression and reconstruction can correspond to one set of inference configurations; when not using AI channel estimation and AI channel prediction, AI channel compression and reconstruction can correspond to another set of inference configurations. Therefore, the inference configuration of AI channel compression and reconstruction can be switched based on whether AI channel estimation and AI channel prediction are activated.

[0213] Example 9: As shown in Figure 19, Figure 19 is another example of multiple AI functions provided in the embodiments of this application satisfying various constraints. The aforementioned one or more AI functions include: AI channel estimation, AI equalization, AI demodulation, and AI decoding. Exemplarily, the AI ​​channel estimation, AI equalization, AI demodulation, and AI decoding can be deployed in the first device.

[0214] Specifically, AI channel estimation satisfies an independence relationship, AI channel estimation and AI equalization satisfy a dependency relationship, AI equalization and AI decoding satisfy a substitution relationship, and AI demodulation and AI decoding also satisfy a substitution relationship. It is easy to see that if AI equalization is activated, AI channel estimation is also activated. If either AI equalization or AI demodulation is activated, AI decoding is deactivated; or, if AI decoding is activated, AI equalization and AI demodulation are deactivated. Furthermore, AI equalization and AI demodulation can be implemented using one model or multiple models, without limitation.

[0215] It should be understood that the examples and constraints listed above are merely illustrative. The first constraint satisfied by one or more AI functions can be one of the aforementioned constraints or other constraints. Furthermore, this application does not limit the naming of the various constraints; for example, a dependency relationship can also be called a condition support relationship, and a condition switching relationship can also be called a selection relationship. Additionally, the aforementioned relationships can be named first, second, third, and fourth relationships, respectively. It should also be understood that the aforementioned constraints are listed for ease of understanding only, and the constraints between AI functions can include some of these constraints, as well as other constraints.

[0216] In step 1620, the third device sends first information indicating that a first constraint relationship is satisfied between the one or more AI functions. Correspondingly, the second device receives the first information from the third device.

[0217] The third device can use the first information to instruct the second device that one or more of the aforementioned AI functions satisfy the first constraint relationship.

[0218] One possible implementation of the first information indicating the one or more AI functions is to indicate the identifier of the one or more AI functions. Since the various AI functions supported by the first device are obtained through training, the second device, as a management node, can know the identifiers of the various AI functions supported by the first device in advance. Therefore, when the second device indicates the identifier of the one or more AI functions through the first information, the second device can determine the one or more AI functions indicated by the first information based on the first information.

[0219] One possible implementation of using the first information to indicate the first constraint relationship is to indicate the first constraint relationship itself. For example, a dependency relationship can be indicated by "CONDITIONAL ON", a substitution relationship by "ALTERNATIVE WITH", and a condition switching relationship by "SWITCH ACCORDING". An independence relationship is a single AI function and therefore does not require additional indication.

[0220] Another possible implementation of the first information indicating the first constraint relationship is to indicate an identifier of the first constraint relationship. Since the first constraint relationship is one of multiple constraint relationships, the second device can indicate the first constraint relationship based on the correspondence between the multiple constraint relationships and multiple different identifiers. The correspondence between the multiple constraint relationships and multiple different identifiers can be predefined, such as protocol predefined, or it can be indicated in advance by the second device via signaling, or it can be reported in advance by the third device via signaling; this application does not limit this. Multiple constraint relationships can be indicated by different identifiers. The indication of the first constraint relationship by the first information can be the identifier corresponding to the first constraint relationship.

[0221] Optionally, in step 1630, the second device sends second information for managing at least one of the one or more AI functions that satisfy the first constraint relationship, or in other words, the second information is used to instruct the management of at least one of the one or more AI functions that satisfy the first constraint relationship. Accordingly, the third device receives the second information.

[0222] Optionally, step 1630 can be replaced by the second device sending second information for managing at least one of the one or more AI functions that satisfy the first constraint relationship, or in other words, the second information for instructing the management of at least one of the one or more AI functions that satisfy the first constraint relationship. Accordingly, the first device receives the second information.

[0223] Optionally, step 1630 can be replaced by the fourth device sending second information for managing at least one of the one or more AI functions that satisfy the first constraint relationship, or in other words, the second information for instructing the management of at least one of the one or more AI functions that satisfy the first constraint relationship. Accordingly, the third device and / or the first device receives the second information. For example, the fourth device obtains the aforementioned first information or the second information from the second device.

[0224] For example, the first information is used to determine the management of at least one of one or more AI functions that satisfy the first constraint relationship.

[0225] For example, the first piece of information is used to determine the second piece of information.

[0226] It should be understood that, since one or more of the aforementioned AI functions are deployed in the first device, the first device and the third device can be different modules within the same physical device, or they can be different physical devices. Therefore, the third device can forward the received second information to the first device. When the first device and the third device are different modules within the same physical device, the communication between the third device and the first device can be internal communication within the device; when the first device and the third device are different physical devices, the communication between the third device and the first device can be inter-device communication.

[0227] It should also be understood that the fourth device can be replaced by a fourth entity. This fourth device and the second device can be different modules within the same physical device, or they can be different physical devices. No limitation is made here.

[0228] Managing AI capabilities includes performing one or more of the following actions on the AI ​​capabilities: activation, deactivation, monitoring, or inference configuration.

[0229] Activation and deactivation are relative terms. Activating an AI function means using that AI function; deactivating it means stopping its use. Monitoring an AI function means monitoring its performance. Configuring inference for an AI function means configuring its inference parameters.

[0230] For example, inference parameters can be configured according to AI functions. For instance, for AI channel estimation, inference parameters may include the pattern of the reference signal (such as the channel state information reference signal, CSI-RS) and the reference signal resources (such as CSI-RS resources); for AI channel prediction, inference parameters may include the reference signal resources, prediction time, number of prediction times, and time offset between the prediction time and the time of transmitting (or receiving) the reference signal; for AI channel compression and reconstruction, inference parameters may include the reference signal resources and reporting resources; for AI modulation or demodulation, inference parameters may include the modulation order; for AI coding or decoding, inference parameters may include the coding rate; for AI equalization, inference parameters may include the pattern of the reference signal and the reference signal resources; for AI precoding, inference parameters may include the number of ports of the reference signal (such as the demodulation reference signal, DMRS) and the pattern of the reference signal.

[0231] Among the inference parameters configured for AI channel prediction, the reference signal resource can be periodic; the prediction time specifically refers to the future time corresponding to the predicted channel information; and the number of prediction times specifically refers to the number of prediction times when predicting channel information corresponding to one or more prediction times.

[0232] In this embodiment, the second device can activate, deactivate, monitor, or configure inference for at least one of one or more AI functions that satisfy the first constraint relationship. The following will describe these constraints in conjunction with the various constraint relationships listed above.

[0233] Independence Relationship: For a single AI function that satisfies the independence relationship, the second device can independently activate, deactivate, monitor, or infer configuration. It should be understood that for an AI function that satisfies the independence relationship, the management of that AI function may not be affected by other AI functions. This document mainly details the management of multiple AI functions that satisfy other constraints.

[0234] Dependency Relationship: For multiple AI functions that satisfy a dependency relationship, the second device can simultaneously activate, deactivate, or perform performance monitoring on the multiple AI functions simultaneously, or perform inference configuration on at least one of the multiple AI functions according to the dependency relationship. When the second information is used to perform inference configuration on at least one of the multiple AI functions, the inference configuration on at least one of the multiple AI functions has a corresponding relationship.

[0235] Substitution Relationship: For multiple AI functions that satisfy the substitution relationship, the second device can avoid simultaneous activation of these multiple AI functions and performance monitoring of them simultaneously, thereby avoiding unnecessary monitoring overhead. It can also avoid simultaneous inference configuration of these multiple AI functions, thus avoiding unnecessary configuration overhead. Monitoring overhead may include processing resource overhead from performing monitoring, signaling overhead from configuring monitoring resources, reporting resources, etc., and resource overhead. Configuration overhead may include signaling overhead and resource overhead from configuring monitoring resources, reporting resources, etc.

[0236] Conditional Switching Relationship: For multiple AI functions that satisfy the conditional switching relationship, the second device can determine the inference configuration for another set of AI functions based on the state of some of the AI ​​functions, establishing a corresponding relationship between the inference configurations of these multiple AI functions. Monitoring of multiple AI functions can also be determined based on the conditional switching relationship. Taking two AI functions as an example, when monitoring the latter AI function, the configuration can be based on whether the former AI function is activated. For instance, when the former AI function is activated, monitoring can be configured based on its output; when the former AI function is deactivated, monitoring can be configured based on the output of a traditional function with the same function as the former AI function. When the former AI function is activated or deactivated, the inference configuration for the latter AI function can differ in one or more of the following ways: model input, performance metrics, or threshold values ​​corresponding to the performance metrics. For another example, when monitoring the latter AI function, the inference configuration can be used to configure monitoring of the latter AI function's performance in both the activated and deactivated states of the former AI function.

[0237] The correspondence between the inference configurations of these multiple AI functions can be understood as follows: For a certain function (e.g., channel estimation), implementing it using AI (i.e., the AI ​​function) and implementing it using traditional methods may involve different input configurations (e.g., reference signal patterns, densities, or resources), and the resulting outputs may have different data formats and different precisions. Therefore, the inference configurations of other AI functions that depend on this AI function need to be determined based on whether the AI ​​function is activated. For example, one set of inference configurations is used when the AI ​​function is activated, and another set of inference configurations is used when the AI ​​function is not activated.

[0238] In a specific implementation, the second information can manage at least one of the one or more AI functions by indicating the identifiers of one or more AI functions that satisfy the first constraint relationship. When the second information is used to configure inference for at least one of the multiple AI functions that satisfy the first constraint relationship, and there are multiple AI functions being configured for inference (referred to as target AI functions for ease of distinction and explanation), the second device can send a corresponding inference configuration to the third device according to the constraint relationship satisfied between the multiple AI functions. For ease of distinction and explanation, the AI ​​function being configured for inference among the multiple AI functions will be referred to as the target AI function below; that is, the second information is used to configure inference for at least one target AI function among the multiple AI functions that satisfy the first constraint relationship.

[0239] One possible design is to carry the inference configuration for at least one target AI function among multiple AI functions that satisfy the first constraint relationship in a single signaling message. This would link the inference configurations for the one or more AI functions together, meaning that the inference configurations sent for multiple AI functions via the same signaling message would have a corresponding relationship.

[0240] Another possible design is that the inference configuration of at least one target AI function among multiple AI functions satisfying the first constraint is carried in different signaling messages. In other words, the second information may contain information elements carried in one or more signaling messages. When the at least one target AI function is multiple AI functions, the second information can also indicate the correspondence between the inference configurations of the multiple target AI functions. For example, the second information can indicate the identifiers of one or more other target AI functions that have a first constraint relationship with one target AI function in the inference configuration of that target AI function, thereby associating the inference configurations of the multiple target AI functions. Alternatively, the second information can also indicate the same identifier in the inference configurations of the multiple target AI functions satisfying the first constraint relationship, such as the identifier corresponding to the first constraint relationship, or the identifier of an AI function group (see description below), etc., without limitation.

[0241] It should be understood that the second device may perform inference configuration on at least one of the one or more AI functions using the aforementioned second information before inference is required, or it may perform inference configuration on at least one of the one or more AI functions using the second information after receiving the first information and determining the first constraint relationship between the one or more AI functions. That is, the first device may not necessarily perform inference after receiving the second information. For example, it may perform inference using the AI ​​function based on the inference configuration after receiving an activation instruction for a certain AI function.

[0242] Since the second device can determine the first constraint relationship between one or more AI functions supported by the first device based on the first information, it can use this as a reference to manage at least one of the one or more AI functions. Therefore, it can avoid issuing instructions that do not match the constraint relationship between the AI ​​functions in the first device, improve the efficiency of model operation, and thus optimize the performance of the wireless communication system.

[0243] In one possible design, one or more AI functions that satisfy a certain constraint relationship can be defined as an AI function group, or a combination of AI functions. Therefore, one possible implementation of the first information used to indicate that one or more AI functions satisfy the first constraint relationship is that the first information is used to indicate the AI ​​functions included in each of the one or more AI function groups, and the constraint relationship satisfied between the AI ​​functions in each AI function group.

[0244] In other words, each AI function group can correspond to a constraint relationship. For example, when one or more AI functions satisfy a certain constraint relationship, those one or more AI functions can belong to one AI function group; when one or more AI functions satisfy multiple constraint relationships, those one or more AI functions can belong to multiple AI function groups. It can be understood that at least one AI function group among the one or more AI function groups indicated by the first information includes one or more AI functions that satisfy the first constraint relationship.

[0245] Optionally, the first information is also used to indicate the identifier of each AI function group in the one or more AI function groups mentioned above.

[0246] When the third device indicates one or more AI function groups via the first information, it can indicate an identifier for each AI function group. That is, for each AI function group, the first information can indicate the identifier of the AI ​​function group, the one or more AI functions included in the AI ​​function group, and the constraint relationships satisfied by the one or more AI functions.

[0247] Figure 20 exemplarily illustrates multiple AI functions deployed in a first device and their respective AI function groups. As shown in Figure 20, a signal is input from AI function a, processed by AI function a, and output. The output of AI function a becomes the input of AI function b, processed by AI function b, and output. The output of AI function b becomes the input of AI function c, processed by AI function c, and output. AI functions a and b form AI function group 1; AI functions a, b, and c form AI function group 2; and AI functions b and c form AI function group 3. The first information, when used to indicate these three AI function groups, can indicate the identifier of each AI function group, the identifier of the AI ​​functions included in each AI function group, and the constraint relationships satisfied by each AI function group.

[0248] Table 1 exemplarily illustrates the first information used to indicate the one or more AI function groups, the AI ​​functions included in each AI function group, and the constraints they satisfy. For ease of distinction, the identifiers of the AI ​​function groups are represented by numbers, and the identifiers of the AI ​​functions are represented by letters in Table 1.

[0249] Table 1

[0250] As shown in Table 1, AI function group 1 includes AI function a and AI function b, which satisfy constraint relationship 1; AI function group 2 includes AI function a, AI function b and AI function c, which satisfy constraint relationship 2; AI function group 3 includes AI function b and AI function c, which satisfy constraint relationship 3; ...

[0251] It should be understood that Table 1 is only an example. In actual implementation, the identifier of each AI function group, the AI ​​functions included in each AI function group, and the constraints satisfied by each AI function group may not necessarily be presented in the form of a table.

[0252] The second device can manage the AI ​​functions supported by the first device based on the identifiers of each AI function group indicated by the first information, the included AI functions, and the constraints they satisfy. When indicating which AI functions to manage via the second information, the second information can directly indicate the identifier of the managed AI function group. Indicating the identifier of the AI ​​function group saves on overhead compared to individually instructing each AI function. For example, if the second device wants to activate multiple AI functions that satisfy dependencies, it can directly indicate the identifier of the AI ​​function group to which these multiple AI functions belong via the second information.

[0253] In one possible design, the aforementioned one or more AI function groups are determined by one or more AI functions supported by the first device. In other words, the one or more AI function groups can be determined based on all AI functions supported by the first device. For example, the first information can be capability information, or information elements contained in the capability information. The specific process under this design can be seen in the following description in conjunction with Figure 21.

[0254] In another possible design, the aforementioned one or more AI function groups are determined based on attribute information related to each AI function among one or more AI functions supported by the first device. In other words, the one or more AI function groups can be some or all of the AI ​​functions supported by the first device. The specific process under this design can be seen in the following description in conjunction with Figure 22.

[0255] Specifically, the AI ​​function-related attribute information can refer to the attributes of the training samples used to train the model implementing the AI ​​function, the attributes of the inference samples supported by the model implementing the AI ​​function, or the device state when generating samples (including training samples and / or inference samples). Examples include, but are not limited to, the beam characteristics, transmit power, precoding, and transmit antenna spacing of the signals used as training samples and / or inference samples. Beam characteristics can include, but are not limited to, beam shape, directivity, energy concentration, and coherence, and can be affected by transmit power, elevation angle, azimuth, and beam pointing. This AI function-related attribute information can be indicated by an associated ID; the specific usage will be explained below and will not be detailed here.

[0256] The third device can determine the currently available AI functions based on the current inference configuration and the attribute information related to each AI function supported by the first device.

[0257] Furthermore, the third device can receive status information from the second device. The aforementioned one or more AI function groups can be determined based on the attribute information and status information related to each AI function among the one or more AI functions supported by the first device. This status information may include signal configuration and the configuration or attributes of the device itself. The status information may also be called configuration information, or attribute information of the second device related to the AI ​​function.

[0258] For example, this status information may correspond to attribute information related to the AI ​​functions supported by the first device mentioned above. This status information can be indicated by an association identifier, which can be used to indicate the current status of the second device. As an example, the current status of the second device may be related to one or more of the following: baseband processing, radio frequency (RF), or antenna. This status information can be used to indicate one or more of the following: beam characteristics of the reference signal resources currently used by the second device (i.e., an example of the baseband processing status), transmit power (i.e., an example of the RF status), precoding (i.e., an example of the baseband processing status), and the spacing of the transmit antennas (i.e., an example of the antenna status). A set of reference signal resources with the same association identifier has the same or similar beam properties. Different association identifiers can be used to indicate different states. For example, different association identifiers corresponding to beam configuration (such as association identifier #1, association identifier #2, association identifier #3) can correspond to different beam characteristics; different association identifiers corresponding to transmit power configuration (such as association identifier #1, association identifier #2, association identifier #3) can correspond to different transmit powers; different association identifiers corresponding to precoding configuration (such as association identifier #1, association identifier #2, association identifier #3) can correspond to different precoding; and different association identifiers corresponding to antenna spacing configuration (such as association identifier #1, association identifier #2, association identifier #3) can correspond to different transmit antenna spacings.

[0259] For ease of distinction and explanation, the association identifier used above to indicate the attribute information related to the AI ​​function supported by the first device is denoted as association identifier #1, and the association identifier used here to indicate the status information is denoted as association identifier #2. When association identifier #1 is the same as the above association identifier #2, it means that the current status of the second device is the same as or similar to the attribute related to the AI ​​function, and the AI ​​function can be used; when association identifier #1 is different from the above association identifier #2, it means that the current status of the second device is different from the attribute related to the AI ​​function, and the AI ​​function can be not used.

[0260] Figure 21 is another schematic flowchart of the communication method provided in an embodiment of this application.

[0261] In step 2110, the second device sends a capability query message to the third device to trigger the third device to report the AI ​​functions supported by the first device.

[0262] For example, the second device is a network device, the third device is a terminal device, and the capability query message can specifically be a UE Capability Query message. The UE Capability Query message is used to trigger the terminal device to report the AI ​​functions supported by the terminal side. The terminal side can refer to the AI ​​functions deployed in the terminal device or other devices besides the terminal device.

[0263] In step 2120, the third device sends capability information to the second device. This capability information is used to indicate the AI ​​function groups supported by the first device, the identifier of each AI function group, the AI ​​functions included in each AI function group, the inference parameters of each AI function group, and the constraint relationships satisfied by each AI function group.

[0264] For example, the first device is a terminal device, or an AI entity serving the terminal device, such as a server in an OTT system or a cloud server. Therefore, the AI ​​function group supported by the first device is the AI ​​function group supported by the terminal side. The second device is a network device, and the third device is a terminal device. The terminal device can report the AI ​​functions supported by the terminal side to the network device through UE Capability Information. In the embodiments of this application, the UE Capability Information can be used to indicate one or more AI function groups supported by the terminal side, the AI ​​functions included in each AI function group and their inference configuration, and the constraints satisfied by each AI function group.

[0265] The following example, using AI features supported by the terminal side, illustrates the format of UE capability information:

[0266] Here, FunctionalitySet represents a set of functions, and the AI ​​functions it includes can be represented by SEQUENCE{}, which represents an ordered set of fields. AI Function 1 (AIFunctionality-1), AI Function 2 (AIFunctionality-2), AI Function 3 (AIFunctionality-3), and AI Function 4 (AIFunctionality-4) are all AI features supported on the terminal side. “AIFunctionality-x-Para” (x is 1, 2, 3, or 4) represents the parameters of AI function x, which belong to the inference configuration of AI function x. “CONDITIONAL ON” means that the function before “CONDITIONAL ON” depends on the function after “CONDITIONAL ON”; “ALTNATIVE WITH” means that the two functions before and after “ALTNATIVE WITH” can be replaced; “SWITCH ACCORDING” means that the inference configuration of the function before “SWITCH ACCORDING” switches based on the state of the function after “SWITCH ACCORDING”. In the example above, AI function 1 is an independent AI function; AI function 2 depends on AI function 1; AI function 3 and AI function 2 are interchangeable; the inference configuration of AI function 4 is determined based on the state of AI function 1, and can be switched depending on whether AI function 1 is activated or not.

[0267] It should be understood that the format of the UE capability information shown above is only an example. The functionality can also be replaced with feature, and the constraints between functions can also be indicated in other ways, such as by using different identifiers to indicate different constraints.

[0268] In step 2130, the second device sends status information to the third device.

[0269] For example, if the second device is a network device and the third device is a terminal device, the status information can be carried in the radio resource control (RRC) reconfiguration message sent by the network device. For a description of the status information, please refer to the relevant content above, which will not be repeated here.

[0270] In this example, since the third device indicates the AI ​​function group supported by the first device to the second device through the capability information, the second device can send the corresponding status information based on the AI ​​function group supported by the first device.

[0271] In step 2140, the third device reports the available AI function groups to the second device.

[0272] The third device can determine the available AI function groups based on the attribute information related to the AI ​​functions supported by the first device and the status information from the second device. The specific content of the third device's determination of the available AI function groups has been explained above in conjunction with the associated identifiers, and will not be repeated here.

[0273] In step 2150, the second device sends second information to the third device, which is used to manage the AI ​​function group available to the first device.

[0274] For example, the second device is a network device, the third device is a terminal device, and the second information is carried in the RRC reconfiguration message. The network device can manage the AI ​​function groups available on the terminal side through this RRC reconfiguration message, such as activating them or configuring inference.

[0275] The following example illustrates the signaling used by a network device to configure inference with an AI function group:

[0276] Among them, Config represents configuration. The other fields can be understood by referring to the explanation in the example of UE capability information above, and will not be repeated here.

[0277] It should be understood that the inference configuration format shown above is only an example. The functionality can also be replaced with feature, and the constraints between functions can also be indicated in other ways, such as by using different identifiers to indicate different constraints.

[0278] As mentioned earlier, the activation and inference configuration of each AI function can be determined based on the constraints between the AI ​​functions. For example, the inference parameters for AI feature 1 can be configured independently; the activation and inference configuration of AI feature 2 depends on the activation of AI feature 1; the activation of AI feature 3 is interchangeable with that of AI feature 2, and the corresponding inference parameters are used; AI feature 4 is configured by selecting the corresponding inference parameters based on whether AI feature 1 is activated or not.

[0279] Figure 22 is another schematic flowchart of the communication method provided in the embodiments of this application.

[0280] In step 2210, the second device sends a capability query message to the third device to trigger the third device to report the AI ​​functions supported by the first device.

[0281] Step 2210 is similar to step 2110 above, and can be referred to step 2110, so it will not be repeated here.

[0282] In step 2220, the third device sends capability information to the second device, which indicates the AI ​​functions and inference configurations supported by the first device.

[0283] For example, the first device is a terminal device, or an AI entity serving the terminal device, such as a server in an OTT system or a cloud server. Therefore, the AI ​​function group supported by the first device is the AI ​​function group supported by the terminal side. The second device is a network device, and the third device is a terminal device. This capability information is used to indicate the AI ​​functions and inference configuration supported by the terminal side.

[0284] Unlike the process shown in Figure 21, the third device can report all the AI ​​functions supported by the first device to the third device through capability information. However, the constraint relationships of these AI functions have not yet been combined based on the constraint relationships.

[0285] In step 2230, the second device sends status information to the third device.

[0286] Step 2230 is similar to step 2130 above, and can be referred to step 2130, so it will not be repeated here.

[0287] In step 2240, the third device reports the available AI function groups to the second device.

[0288] The third device can determine the available AI functions based on the attribute information related to the AI ​​functions supported by the first device and the status information from the second device. It can also combine these AI functions by considering the constraints between them, and report the available AI function groups to the second device. It is understood that AI function groups may exist under certain status conditions; therefore, the available AI function groups determined by the third device are based on the constraints between the AI ​​functions supported by the first device, the attribute information related to each AI function, and the status information.

[0289] For a detailed explanation of how the third device determines the available AI functions based on the attribute information related to the AI ​​functions supported by the first device and the status information from the second device, please refer to the description in step 2130 above, which will not be repeated here.

[0290] In step 2250, the second device sends second information to the third device, which is used to manage the AI ​​function group available to the first device.

[0291] Step 2250 is similar to step 2150 above, and can be referred to step 2150, so it will not be repeated here.

[0292] It should be understood that the processes illustrated above with reference to Figures 21 and 22 are only two possible implementations. This application does not limit the timing of the transmission of the first information or the signaling used to carry the first information. As long as the third device indicates to the second device one or more constraints satisfied by AI functions, it should fall within the protection scope of this application.

[0293] Based on the above scheme, the third device indicates to the second device, through the first information, the first constraint relationship satisfied between one or more AI functions, which facilitates the second device's management of the one or more AI functions based on the first constraint relationship. For example, when the second device activates and / or configures inference for AI functions supported by the first device, it can refer to the constraint relationship between the AI ​​functions. For multiple AI functions satisfying the dependency relationship, the AI ​​function group to which these multiple AI functions belong can be directly activated; for multiple AI functions satisfying the substitution relationship, they are not activated simultaneously, nor are they configured inference simultaneously; for multiple AI functions satisfying the condition switching relationship, inference configuration is performed for the other one or more AI functions based on the state of one or more of the AI ​​functions. As another example, when the second device monitors the AI ​​functions supported by the first device, it can refer to the constraint relationship between the AI ​​functions. For multiple AI functions satisfying the dependency relationship, joint performance monitoring can be performed; for multiple AI functions satisfying the substitution relationship, performance monitoring is not performed simultaneously; for multiple AI functions satisfying the condition switching relationship, the output of one or more AI functions in the activated or deactivated state can be used to monitor the other one or more AI functions, and so on. Therefore, the management of the AI ​​functions supported by the first device by the third device and the constraint relationship between each AI function can be adapted to avoid the situation where the first device cannot implement it, or the AI ​​function is abnormally operated or the reasoning result is inaccurate due to implementation according to the instructions of the third device. Thus, the efficiency of model management can be improved, which is conducive to improving the running efficiency of the model and optimizing the performance of the wireless communication system.

[0294] The methods provided in the embodiments of this application have been described in detail above with reference to several accompanying drawings. The apparatus provided in the embodiments of this application will now be described with reference to the accompanying drawings.

[0295] Figures 23 to 25 are schematic block diagrams of possible apparatuses provided in embodiments of this application. These apparatuses can be used to implement the functions of the second or third apparatus in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments.

[0296] Figures 23 and 24 are schematic diagrams of possible communication devices provided in the embodiments of this application. These communication devices can be used to implement the functions of the terminal devices and network devices in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments. In the embodiments of this application, the communication device can be the access network device as shown in Figure 1 or Figure 2, the core network device as shown in Figure 3, the access network node as shown in Figure 4, or the terminal device as shown in Figure 1 or Figure 2, or the terminal as shown in Figure 3 or Figure 4.

[0297] As shown in Figure 23, the communication device 2300 includes a processing module 2310 and a communication module 2320. The communication device 2300 is used to implement the function of the third device in the method embodiments shown in Figures 16, 21, or 22.

[0298] In an optional embodiment, the communication device 2300 can be used to implement the function of the third device in the embodiments shown in FIG16, FIG21 or FIG22.

[0299] When the communication device 2300 is used to implement the function of the third device in the embodiment shown in FIG16: the processing module 2310 is used to determine that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by the first device; the communication module 2320 is used to send first information, the first information being used to indicate that one or more AI functions satisfy the first constraint relationship.

[0300] Optionally, the communication module 2320 is also configured to receive second information for managing at least one AI function among one or more AI functions that satisfy the first constraint relationship, the management including one or more of the following: activation, deactivation, monitoring or inference configuration.

[0301] Optionally, the communication module 2320 is also used to receive status information from the second device.

[0302] In another alternative embodiment, the communication device 2300 can be used to implement the function of the second device in the embodiments shown in FIG16, FIG21 or FIG22.

[0303] When the communication device 2300 is used to implement the function of the second device in the embodiment shown in FIG16: the communication module 2320 is used to receive first information, the first information being used to indicate that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by the first device; optionally, the communication module 2320 is also used to send second information, the second information being used to manage at least one of the one or more AI functions that satisfy the first constraint relationship, the management including one or more of the following: activation, deactivation, inference configuration or monitoring.

[0304] Optionally, the communication module 2320 is also used to send status information.

[0305] A more detailed description of the processing module 2310 and the communication module 2320 can be found in the relevant descriptions in the method embodiments shown in Figures 16, 21 and 22.

[0306] It should be noted that the communication module can also be called a transceiver module, transceiver unit, transceiver, transceiver device, or transceiver apparatus, etc. The processing module can also be called a processor, processing board, processing unit, or processing apparatus, etc. Optionally, the communication module is used to execute the sending and receiving operations of the first or second communication device in the above method. The device in the communication module that implements the receiving function can be considered as the receiving module, and the device in the communication module that implements the sending function can be considered as the sending module; that is, the communication module can include both a receiving module and a sending module.

[0307] It should also be noted that, in one possible design, the aforementioned processing module and / or communication module can be implemented through virtual modules. For example, the processing module can be implemented through software functional units or virtual devices, and the communication module can be implemented through software functions or virtual devices. In another possible design, the processing module or communication module can also be implemented through physical devices. For example, if the device is implemented using a chip / chip circuit, the communication module can be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operation) and output operations (corresponding to the aforementioned sending operation); the processing module can be an integrated processor, a microprocessor, or an integrated circuit.

[0308] The module division in this embodiment is illustrative and represents only one logical functional division; in actual implementation, other division methods may be used. Furthermore, the functional modules in the various examples of this embodiment can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0309] Figure 24 is a schematic diagram of a communication device provided in another embodiment of this application. As shown in Figure 24, the communication device 2400 includes a processing circuit 2410 and a communication circuit 2420. The processing circuit 2410 and the communication circuit 2420 are coupled to each other. The processing circuit 2410 may be one or more processors, or all or part of the circuitry in one or more processors used for control or processing functions. It is understood that when the communication device 2400 is a network device or a terminal device, the communication circuit 2420 may be a transceiver circuit, a transceiver, a communication interface, or an input / output interface. When the communication device 2400 is a chip for a network device or a terminal device, the communication circuit 2420 may be an input / output interface, a communication interface, or an input / output circuit. Optionally, the communication device 2400 may further include a memory 2430 for storing instructions executed by the processor, or storing input data required by the processor to execute instructions, or storing data generated after the processor executes instructions.

[0310] When the communication device 2400 is used to implement the methods shown in FIG16, FIG21 and FIG22, the processing circuit 2410 is used to implement the functions of the processing module 2310, and the communication circuit 2420 is used to implement the functions of the communication module 2320.

[0311] When the aforementioned communication device is a chip applied to a terminal device, the chip implements the functions of the terminal device in the above method embodiments. The chip can receive information from other modules (such as a radio frequency module or antenna) in the terminal device, information sent by the network device to the terminal device; or, the chip can send information to other modules (such as a radio frequency module or antenna) in the terminal device, information sent by the terminal device to the network device.

[0312] When the aforementioned communication device is a module applied to a network device (such as a base station), the module implements the functions of the network device in the above method embodiments. This module can receive information from other modules (such as radio frequency modules or antennas) in the network device, and this information may be sent from the terminal device to the network device; alternatively, the module can send information to other modules (such as radio frequency modules or antennas) in the network device, and this information may be sent from the network device to the terminal device. The module here can be the baseband chip of the base station, or a DU (Digital Unit) or other modules. The DU here can be a DU under an Open Radio Access Network (O-RAN) architecture.

[0313] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), graphics processing units (GPUs), neural processing units (NPUs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0314] For example, the processor used for AI can be one or more of the following: graphics processing unit (GPU), neural processing unit (NPU), tensor processing unit (TPU), and data processing unit (DPU).

[0315] For example, one possible implementation of a processor for AI could be the AI ​​processor 2500 shown in Figure 25. Figure 25 is a schematic diagram of the structure of an AI processor provided in an embodiment of this application.

[0316] As shown in Figure 25, the AI ​​processor 2500 may include one or more of the following: an AI core, a digital vision pre-processing (DVPP) module, a task scheduler (TS), an L3 cache, an AI CPU, a control CPU, an L2 cache, a universal serial bus (USB) interface, a network interface card (NIC), a peripheral component interconnect express (PCIe) interface (PCIe is a high-speed serial computer expansion bus standard), a double data rate (DDR) / high bandwidth memory (HBM) interface, a general purpose input / output (GPIO) / inter-integrated circuit (I2C) bus, etc. It is understood that the specific meanings of these terms are well known to those skilled in the art and will not be elaborated upon here.

[0317] The terms “unit”, “module”, etc., used in this specification may be used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution.

[0318] The method steps in the embodiments of this application can be implemented in hardware or in software instructions executable by a processor. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, compact disc read-only memory (CD-ROM), or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. The storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Alternatively, the ASIC can reside in a base station or terminal. The processor and storage medium can also exist as discrete components in the base station or terminal.

[0319] This application also provides a communication system, which includes the second and third devices described in the embodiments of this application, and optionally further includes the first device.

[0320] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., a solid-state drive). The computer-readable storage medium includes instructions or program code that, when executed on a computing device, cause the computing device to perform the methods provided above.

[0321] This application also provides a computer program product, which may be a software or program product containing instructions capable of running on a computing device or stored on any usable medium. When the instructions are executed on the computing device, the computing device performs the methods provided above, or performs the functions of the apparatus provided above.

[0322] This application also provides a chip including at least one processor, which, when program instructions are executed by the at least one processor, causes the at least one processor to perform the method executed by the second device or the method executed by the third device in the above method embodiments.

[0323] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0324] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0325] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0326] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0327] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0328] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0329] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A communication method characterized by comprising: The method comprises: determining that one or more artificial intelligence (AI) functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a first device; sending first information, the first information being used to indicate that the one or more AI functions satisfy the first constraint relationship.

2. The method of claim 1, wherein, The one or more AI functions satisfying the first constraint relationship comprises that activation of the one or more AI functions satisfies the first constraint relationship, and / or that inference configuration of the one or more AI functions satisfies the first constraint relationship.

3. The method of claim 1 or 2, wherein, The method further comprises: receiving second information, the second information being used to manage at least one AI function of the one or more AI functions satisfying the first constraint relationship, the management comprising one or more of the following: activation, deactivation, monitoring, or inference configuration.

4. The method of claim 3, wherein, The second information is used to perform inference configuration on at least one AI function of the one or more AI functions satisfying the first constraint relationship, the inference configuration on the at least one AI function having a corresponding relationship.

5. The method of claim 4, wherein, The second information further indicates the corresponding relationship.

6. The method of any one of claims 1 to 5, wherein, The one or more AI functions belong to one AI function group of one or more AI function groups, each AI function group of the one or more AI function groups comprising one or more AI functions supported by the first device, and one or more AI functions in each AI function group of the one or more AI function groups satisfying one of a plurality of constraint relationships: an independent relationship, a dependent relationship, a substitution relationship, or a conditional switching relationship; An AI function satisfying the independent relationship is one AI function, and AI functions satisfying the dependent relationship, the selection relationship, or the substitution relationship, the conditional switching relationship are multiple AI functions; wherein in the multiple AI functions satisfying the dependent relationship, activation of one AI function depends on activation of another AI function; the multiple AI functions satisfying the substitution relationship are not activated at the same time; and in the multiple AI functions satisfying the conditional switching relationship, inference configuration of one or more AI functions is determined according to whether another or more AI functions are activated.

7. The method of claim 6, wherein, The first information used to indicate that the one or more AI functions satisfy the first constraint relationship comprises: The first information is used to indicate one or more AI functions included in each AI function group of the one or more AI function groups and constraint relationships satisfied by the one or more AI functions.

8. The method of claim 7, wherein, The first information is further used to indicate an identifier of each AI function group of the one or more AI function groups.

9. The method of claim 7 or 8, wherein, The one or more AI function groups are determined based on one or more AI functions supported by the first device.

10. The method of claim 7 or 8, wherein, The one or more AI function groups are determined based on attribute information related to each AI function of the one or more AI functions supported by the first device.

11. The method of claim 10, wherein, The method further comprises: receiving state information from a second device; The one or more AI function groups are determined based on attribute information related to each of the one or more AI functions supported by the first device, including that the one or more AI function groups are determined based on the state information and the attribute information related to each of the one or more AI functions supported by the first device.

12. The method of claim 11, wherein, The state information of the second device is used to indicate a state related to one or more of baseband processing, radio frequency, or antenna.

13. The method of any one of claims 1 to 12, wherein, The one or more AI functions include one or more of channel estimation, channel prediction, channel compression and reconstruction, encoding, modulation, precoding, equalization, demodulation, or decoding.

14. The method of claim 13, wherein, The one or more AI functions include the channel estimation and the equalization, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the equalization is a dependent relationship. Or The one or more AI functions include the channel estimation and the equalization, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the equalization is a dependent relationship. Or The one or more AI functions include the channel estimation and the channel prediction, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the channel prediction is a dependent relationship. Or The one or more AI functions include the channel estimation and the channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the channel compression and reconstruction is a conditional switching relationship. Or The one or more AI functions include the channel prediction and the channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between the channel prediction and the channel compression and reconstruction is a conditional switching relationship. Or The one or more AI functions include the equalization and the decoding, and the first information is used to indicate that the first constraint relationship satisfied between the equalization and the decoding is a substitution relationship. Or The one or more AI functions include the demodulation and the decoding, and the first information is used to indicate that the first constraint relationship satisfied between the demodulation and the decoding is a substitution relationship.

15. A method of communication, comprising: Including: Receiving first information, the first information being used to indicate that one or more artificial intelligence (AI) functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a first device; Sending second information, the second information being used to manage at least one of the one or more AI functions satisfying the first constraint relationship, the management including one or more of activation, deactivation, inference configuration, or monitoring.

16. A method of communication, comprising: Including: Receiving first information, the first information being used to indicate that one or more AI functions satisfy a first constraint relationship, the one or more AI functions being AI functions supported by a device; and determining management of at least one of the one or more AI functions satisfying the first constraint relationship, the management including activation, deactivation, inference configuration, or monitoring.

17. The method of claim 16, wherein, The first information is used for determination of second information, and the second information is used for indicating management of at least one AI function of the one or more AI functions satisfying the first constraint relationship.

18. The method of claim 15 or 17, wherein, The second information is used for inference configuration of at least one AI function of the one or more AI functions satisfying the first constraint relationship, and the inference configuration of the at least one AI function has a corresponding relationship.

19. The method of claim 18, wherein, The second information further indicates the corresponding relationship.

20. The method of any one of claims 15 to 19, wherein, The one or more AI functions satisfy a first constraint relationship, including that activation of the one or more AI functions satisfies the first constraint relationship, and / or inference configuration of the one or more AI functions satisfies the first constraint relationship.

21. The method of any one of claims 15 to 20, wherein, The one or more AI functions belong to one AI function group of one or more AI function groups, each AI function group of the one or more AI function groups includes one or more AI functions supported by the first device, and one or more AI functions in each AI function group of the one or more AI function groups satisfy one of a plurality of constraint relationships, including an independent relationship, a dependent relationship, a substitution relationship, or a conditional switching relationship. An AI function satisfying the independent relationship is one AI function, and AI functions satisfying the dependent relationship, the selection relationship, the substitution relationship, or the conditional switching relationship are multiple AI functions; wherein in the multiple AI functions satisfying the dependent relationship, activation of one AI function depends on activation of another AI function; the multiple AI functions satisfying the substitution relationship are not activated at the same time; and in the multiple AI functions satisfying the conditional switching relationship, inference configuration of one or more AI functions is determined according to whether another or more AI functions are activated.

22. The method of claim 21, wherein, The first information is used for indicating that one or more AI functions satisfy a first constraint relationship, including: The first information is used for indicating AI functions included in each AI function group of the one or more AI function groups and constraint relationships satisfied by the AI functions, and one AI function group of the one or more AI function groups includes the one or more AI functions.

23. The method of claim 22, wherein, The first information is further used for indicating an identifier of each AI function group of the one or more AI function groups.

24. The method of claim 22 or 23, wherein, The one or more AI function groups are determined based on one or more AI functions supported by the first device.

25. The method of claim 22 or 23, wherein, Before receiving the first information, the method further includes: sending state information; and The first information is determined according to the state information and attribute information related to each AI function of one or more AI functions supported by the first device.

26. The method of claim 25, wherein, The state information is used for indicating a state related to one or more of baseband processing, radio frequency, or antenna.

27. The method of any one of claims 15 to 26, wherein, The one or more AI functions include one or more of channel estimation, channel prediction, channel compression and reconstruction, encoding, modulation, precoding, equalization, demodulation, or decoding.

28. The method of claim 27, wherein, The one or more AI functions are channel estimation, and the first information is used for indicating that the channel estimation satisfies the first constraint relationship as an independent relationship. or The one or more AI functions comprise the channel estimation and the equalization, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the equalization is a dependency relationship. Or The one or more AI functions comprise the channel estimation and the channel prediction, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the channel prediction is a dependency relationship. Or The one or more AI functions comprise the channel estimation and the channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between the channel estimation and the channel compression and reconstruction is a conditional switching relationship. Or The one or more AI functions comprise the channel prediction and the channel compression and reconstruction, and the first information is used to indicate that the first constraint relationship satisfied between the channel prediction and the channel compression and reconstruction is a conditional switching relationship. Or The one or more AI functions comprise the equalization and the decoding, and the first information is used to indicate that the first constraint relationship satisfied between the equalization and the decoding is a substitution relationship. Or The one or more AI functions comprise the demodulation and the decoding, and the first information is used to indicate that the first constraint relationship satisfied between the demodulation and the decoding is a substitution relationship.

29. A communications device, characterized by A module or unit for implementing the method of any of claims 1-14, or a module or unit for implementing the method of any of claims 15-28.

30. A communications device, characterized by One or more processors and a communication circuit are included, the communication circuit is used for at least one of input or output of signals by the communication device; the one or more processors are used to implement the method of any of claims 1-14, or to implement the method of any of claims 15-28.

31. The apparatus of claim 30, wherein, The communication device is a network device or a terminal device, or a chip for the network device or the terminal device.

32. A communication system, characterized by Comprising: A second device for performing the method of any of claims 15-28; A third device for performing the method of any of claims 1-14.

33. The system of claim 32, wherein, The system further comprises a first device supporting an artificial intelligence (AI) function.

34. A readable storage medium characterized by, A program or instruction for storing, when the program or instruction is executed, the method of any of claims 1-14 is implemented, or the method of any of claims 15-28 is implemented.

35. A computer program product, characterised in that, A program or instruction for storing, when the program or instruction is executed, the method of any of claims 1-14 is implemented, or the method of any of claims 15-28 is implemented.