A service configuration method, communication system, and communication device

Abstract

This application provides a service configuration method, a communication system, and a communication device. In this method, a first network element can determine, based on information about service units already deployed on a second network element, whether a target service unit can be deployed on that second network element, and the information (e.g., third information and topology information) that needs to be sent to the second network element when deploying the target service unit. This enables the first network element to deploy service units on the second network element as needed, which is beneficial to improving the flexibility of service deployment.

Description

Technical Field

[0001] This application relates to the field of communications, and more particularly to a service configuration method, a communication system, and a communication device. Background Technology

[0002] With the development of communication technology, artificial intelligence (AI) and machine learning (ML) technologies are being applied to assist wireless communication. Current standards have proposed several AI / ML frameworks to enhance data collection or to enhance air interface AI. However, as the application scenarios of AI / ML expand, the current AI / ML frameworks have limited functionality and cannot achieve on-demand deployment of AI / ML functions, thus restricting the expansion of AI / ML capabilities. Summary of the Invention

[0003] This application provides a service configuration method, a communication system, and a communication device to improve the flexibility of service deployment.

[0004] In a first aspect, this application provides a service configuration method, which can be executed by a first network element or by a component of the first network element (e.g., a processor, chip, or chip system). Taking the first network element as an example, the first network element obtains first information from a second network element, the first information including information on at least one service unit already deployed by the second network element, the service unit being used to implement one or more functions of an access network device or terminal device; the first network element determines information on a target service unit to be deployed by the second network element, the information of the target service unit including second information and topology information, the second information including information on multiple service units constituting the target service unit, the topology information being used to indicate the connection topology between the multiple service units constituting the target service unit; the first network element sends third information and topology information to the second network element, the third information being determined based on the first and second information, the third information and topology information being used to instruct the second network element to deploy the target service unit, the target service unit being used to implement the target service.

[0005] In this embodiment, the first network element can determine whether a target service unit can be deployed on the second network element based on the information of the service units already deployed on the second network element, and the information that needs to be sent to the second network element when deploying the target service unit (e.g., third information and topology information). This enables the first network element to deploy service units on the second network element as needed, which is beneficial to improving the flexibility of service deployment.

[0006] In one possible implementation, the first network element obtains first information from the second network element, including: the first network element sending a first message to the second network element, the first message being used to request information on the service units deployed by the second network element; and the first network element receiving a second message from the second network element, the second message including the first information.

[0007] In this embodiment, the first network element triggers the second network element to send the first information to the first network element through the first message. This enables the first network element to obtain the first information when it needs it, which helps to save the signaling overhead of transmitting the first information and also helps to save the storage overhead of the first network element in storing the first information.

[0008] In one possible implementation, the first message includes service unit type information, which indicates the type of service unit queried through the first message; the service unit corresponding to the service unit information contained in the first message is of the same type as the service unit indicated by the service unit type information in the first message.

[0009] For example, the type information of the service unit includes: first type information and / or second type information, wherein the first type information is used to indicate a first type service unit and the second type information is used to indicate a second type service unit; wherein, a second type service unit consists of at least two different first type service units.

[0010] In this embodiment, by carrying the type information of the service unit in the first message, the first network element can query the information of the service unit of the desired type on demand, so that the second network element only reports the service unit that matches the type information, which helps to improve query efficiency and save signaling overhead.

[0011] In one possible implementation, the information of the service unit includes information of the first type of service unit;

[0012] The information for the first type of service unit includes at least one of the following:

[0013] Identification information, used to uniquely identify the first type of service unit; or,

[0014] Functional description information, used to describe the functions of the first type of service unit; or,

[0015] Input / output information, used to indicate the inputs and outputs of the first type of service unit; or,

[0016] Equipment capability requirement information, used to indicate the software and / or hardware capability requirements of the equipment deploying the first type of service unit; or,

[0017] Gradient computation capability information, used to indicate whether the first type of service unit supports backward gradient computation; or...

[0018] Executable file information.

[0019] In one possible implementation, the input / output information includes input identification information and output identification information; the input / output information further includes at least one of the following:

[0020] Input format information, used to indicate the format of the input data for the first type of service unit; or,

[0021] Output format information, used to indicate the format of the output data for the first type of service unit; or...

[0022] Input memory layout information, used to indicate how the input data of the first type of service unit is arranged in memory; or,

[0023] Output memory layout information to indicate how the output data of the first type of service unit is arranged in memory; or...

[0024] Input quantization bit width, used to indicate the quantization bit width of the input data for the first type of service unit; or,

[0025] Output quantization bit width, used to indicate the quantization bit width of the output data of the first type of service unit.

[0026] In one possible implementation, the information of the service unit includes information of the second type of service unit;

[0027] The information for the second type of service unit includes at least one of the following:

[0028] Identification information, used to uniquely identify the second type of service unit; or,

[0029] Functional description information, used to describe the functions of the second type of service unit; or,

[0030] Input / output information, used to indicate the inputs and outputs of the second type of service unit; or,

[0031] Topology information, used to indicate the connection topology between multiple service units constituting a second type of service unit; or,

[0032] Functional parameter information, used to indicate the functional parameters used by the second type of service unit when implementing service functions.

[0033] In one possible implementation, the second information includes information about a first service unit constituting the target service unit and information about a second service unit; if the gradient computation capability information of the first service unit indicates that the first service unit supports backward gradient computation, and if the gradient computation capability information of the second service unit indicates that the second service unit supports backward gradient computation, then the target service unit supports backward gradient computation.

[0034] In this embodiment, by configuring each service unit constituting the target service unit to support back gradient calculation, the target service unit can also support back gradient calculation, which is beneficial for the target service unit to achieve online model training and improves the flexibility of applying AI technology to the communication system.

[0035] In one possible implementation, the second message further includes device capability information of the second network element, which indicates the software and / or hardware capabilities of the second network element. The first network element determines information about the target service unit to be deployed by the second network element, including: the first network element determining second information based on the first information and the device capability information of the second network element; the second information includes information about the first service unit constituting the target service unit and information about the second service unit; the capabilities indicated by the capability information of the second network element include the capabilities indicated by the device capability requirement information of the first service unit and the capabilities indicated by the device capability requirement information of the second service unit.

[0036] In this embodiment, the first network element will consider the device capability information of the second network element when determining the information of the target service unit. The capabilities of the second network element must support the capability requirements of the service unit, which helps to ensure the stable operation of the target service unit after deployment in the second network element and improves the reliability of service deployment.

[0037] In one possible implementation, the first type of service unit includes a radio frequency service unit, a physical layer service unit, or a media access control (MAC) layer service unit.

[0038] In one possible implementation, the second type of service unit includes a model management service unit, a data management service unit, a performance monitoring service unit, a perception service unit, a computing service unit, a model storage service unit, or a data storage service unit.

[0039] In one possible implementation, the first network element is a management node and the second network element is an access network device; or, the first network element is a management node and the second network element is a terminal device; or, the first network element is an access network device and the second network element is a terminal device.

[0040] Secondly, this application provides a service configuration method, which can be executed by a second network element or by a component of the second network element (e.g., a processor, chip, or chip system). Taking the second network element as an example, the second network element sends first information to a first network element. The first information includes information about at least one service unit deployed by the second network element, which is used to implement one or more functions of an access network device or terminal device. Then, the second network element receives third information and topology information from the first network element. The third information is determined based on the first and second information. The second information includes information about multiple service units constituting the target service unit, and the topology information is used to indicate the connection topology between the multiple service units constituting the target service unit. Then, the second network element deploys the target service unit based on the third information, the first information, and the topology information.

[0041] In one possible implementation, the second network element sends first information to the first network element, including: the second network element receiving a first message from the first network element, the first message being used to request information about the service units deployed by the second network element; and the second network element sending a second message to the first network element, the second message including the first information.

[0042] In one possible implementation, the first message includes service unit type information, which indicates the type of service unit queried through the first message; the service unit corresponding to the service unit information contained in the first message is of the same type as the service unit indicated by the service unit type information in the first message.

[0043] In one possible implementation, the type information of the service unit includes: first type information and / or second type information, wherein the first type information is used to indicate a first type service unit and the second type information is used to indicate a second type service unit; wherein a second type service unit consists of at least two different first type service units.

[0044] In one possible implementation, the information of the service unit includes information of the first type of service unit;

[0045] The information for the first type of service unit includes at least one of the following:

[0046] Identification information, used to uniquely identify the first type of service unit; or,

[0047] Functional description information, used to describe the functions of the first type of service unit; or,

[0048] Input / output information, used to indicate the inputs and outputs of the first type of service unit; or,

[0049] Equipment capability requirement information, used to indicate the software and / or hardware capability requirements of the equipment deploying the first type of service unit; or,

[0050] Gradient computation capability information, used to indicate whether the first type of service unit supports backward gradient computation; or...

[0051] Executable file information.

[0052] In one possible implementation, the input / output information includes input identification information and output identification information; the input / output information further includes at least one of the following:

[0053] Input format information, used to indicate the format of the input data for the first type of service unit; or,

[0054] Output format information, used to indicate the format of the output data of the first type of service unit; or...

[0055] Input memory layout information, used to indicate how the input data of the first type of service unit is arranged in memory; or,

[0056] Output memory layout information to indicate how the output data of the first type of service unit is arranged in memory; or...

[0057] Input quantization bit width, used to indicate the quantization bit width of the input data for the first type of service unit; or,

[0058] Output quantization bit width, used to indicate the quantization bit width of the output data of the first type of service unit.

[0059] In one possible implementation, the information of the service unit includes information of the second type of service unit;

[0060] The information for the second type of service unit includes at least one of the following:

[0061] Identification information, used to uniquely identify the second type of service unit; or,

[0062] Functional description information, used to describe the functions of the second type of service unit; or,

[0063] Input / output information, used to indicate the inputs and outputs of the second type of service unit; or,

[0064] Topology information, used to indicate the connection topology between multiple service units constituting a second type of service unit; or,

[0065] Functional parameter information is used to indicate the functional parameters used by the second type of service unit when implementing service functions.

[0066] In one possible implementation, the second information includes information about a first service unit constituting the target service unit and information about a second service unit; the gradient computation capability information of the first service unit indicates that the first service unit supports inverse gradient computation, and the target service unit supports inverse gradient computation when the gradient computation capability information of the second service unit indicates that the second service unit supports inverse gradient computation.

[0067] In one possible implementation, the second message further includes device capability information of the second network element, which is used to indicate the software and / or hardware capabilities of the second network element; the target service unit includes a first service unit and a second service unit, and the capabilities indicated by the capability information of the second network element include the capabilities indicated by the device capability requirement information of the first service unit and the capabilities indicated by the device capability requirement information of the second service unit.

[0068] In one possible implementation, the first type of service unit includes a radio frequency service unit, a physical layer service unit, or a media access control (MAC) layer service unit.

[0069] In one possible implementation, the second type of service unit includes a model management service unit, a data management service unit, a performance monitoring service unit, a perception service unit, a computing service unit, a model storage service unit, or a data storage service unit.

[0070] In one possible implementation, the first network element is a management node and the second network element is an access network device; or, the first network element is a management node and the second network element is a terminal device; or, the first network element is an access network device and the second network element is a terminal device.

[0071] It should be noted that there are many other specific implementation methods in this application, and you can refer to the specific implementation methods and their beneficial effects in the first aspect, which will not be repeated here.

[0072] Thirdly, embodiments of this application provide a communication device, which can be a first network element in the foregoing embodiments or a chip within the first network element. The communication device may include a processing module and a transceiver module. When the communication device is a first network element, the processing module may be a processor, and the transceiver module may be a transceiver. The first network element may also include a storage module, which may be a memory. The storage module stores instructions, and the processing module executes the instructions stored in the storage module to cause the first network element to perform the method in the first aspect or any embodiment of the first aspect. When the communication device is a chip within the first network element, the processing module may be a processor, and the transceiver module may be an input / output interface, pin, or circuit, etc. The processing module executes the instructions stored in the storage module to cause the first network element to perform the method in the first aspect or any embodiment of the first aspect. The storage module may be a storage module within the chip (e.g., a register, cache, etc.), or a storage module located outside the chip within the first network element (e.g., a read-only memory, random access memory, etc.).

[0073] Fourthly, embodiments of this application provide a communication device, which can be a second network element in the foregoing embodiments or a chip within the second network element. The communication device may include a processing module and a transceiver module. When the communication device is a second network element, the processing module may be a processor, and the transceiver module may be a transceiver. The second network element may also include a storage module, which may be a memory. The storage module stores instructions, and the processing module executes the instructions stored in the storage module to cause the second network element to perform the method in the second aspect or any embodiment of the second aspect. When the communication device is a chip within the second network element, the processing module may be a processor, and the transceiver module may be an input / output interface, pin, or circuit, etc. The processing module executes the instructions stored in the storage module to cause the second network element to perform the method in the second aspect or any embodiment of the second aspect. The storage module may be a storage module within the chip (e.g., a register, cache, etc.), or a storage module located outside the chip within the second network element (e.g., a read-only memory, random access memory, etc.).

[0074] Fifthly, this application provides a communication device including modules, units, or means for implementing the methods described in the foregoing aspects. The device may be an integrated circuit chip. The integrated circuit chip includes a processor. The processor is coupled to a memory for storing programs or instructions that, when executed by the processor, cause the communication device to perform the methods described in any of the embodiments of the foregoing aspects.

[0075] Sixthly, embodiments of this application provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods described in any of the foregoing embodiments.

[0076] In a seventh aspect, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the methods described in any of the preceding embodiments.

[0077] Eighthly, embodiments of this application provide a communication system, which includes a first network element performing the first aspect and any embodiment thereof, and a second network element performing the second aspect and any embodiment thereof.

[0078] Ninthly, this application provides a communication system, the communication system comprising:

[0079] The system includes a management node and at least one network element, the at least one network element comprising a first network element, which deploys a model training service unit that supports backpropagation gradient calculation. Specifically, the first network element is configured to receive a model training request from the management node, the model training request including functional parameters used by the model training service unit when training the model; the first network element is also configured to perform model training on the training data based on the functional parameters through the model training service unit to obtain a target model; and the first network element is further configured to send a model training response to the management node, the model training response including the target model output by the model training service unit.

[0080] In this approach, after the management node deploys the model training service unit in the first network element, it can trigger the model training service unit in the first network element to train the model based on the training data and output the target model through a model training request. Then, the first network element can send the target model to the management node. This enables the management node to trigger the first network element to perform model training on demand, which is beneficial to improving the flexibility of model training compared to the traditional approach of training models only on specific devices.

[0081] In one possible implementation, the functional parameters include: training hyperparameters, training loss function, and training stopping condition.

[0082] In one possible implementation, the first network element includes a second network element and a third network element, and the model training service unit includes an encoder service unit and a decoder service unit. The second network element deploys the encoder service unit, and the third network element deploys the decoder service unit. The encoder service unit supports backpropagation gradient calculation, and the decoder service unit also supports backpropagation gradient calculation. Specifically, the second network element sends a call request to the third network element to request that the encoder and decoder service units execute model training. The second network element also sends forward response information to the third network element, which is used by the encoder service unit to instruct the decoder service unit on the features generated during the forward propagation of model training. The third network element sends backward response information to the second network element, which corresponds to the forward response information and is used by the decoder service unit to instruct the encoder service unit on the gradients generated during the backward propagation of model training.

[0083] In this embodiment, since both the encoder service unit deployed in the second network element and the decoder service unit deployed in the third network element support inverse gradient calculation, after establishing the topological connection between the encoder service unit and the decoder service unit, the various service units cooperate to achieve online model training. Compared with the traditional approach of performing offline model training only on a specific device, this approach improves the flexibility of model training in the communication system and facilitates the implementation of flexible, scalable, and evolvable air interface model training services.

[0084] In one possible implementation, at least one network element further includes a fourth network element, which is equipped with a data retrieval service unit for outputting training data, including first training data and second training data. The fourth network element is also used to send the first training data to a second network element, which is used by the encoder service unit to generate forward response information. The fourth network element is also used to send the second training data to a third network element, which is used by the decoder service unit to generate reverse response information.

[0085] In one possible implementation, at least one network element further includes a fifth network element, which is equipped with a performance monitoring service unit for monitoring the performance of the target model; the first network element is further configured to send a performance monitoring request to the fifth network element, the performance monitoring request including a monitoring period and monitoring parameters, the monitoring parameters being used by the performance monitoring service unit to determine whether the target model meets the training stopping condition; the fifth network element is configured to send a first instruction message to the first network element if the training stopping condition is met, the first instruction message being used to instruct the model training service unit to stop model training.

[0086] In one possible implementation, at least one network element further includes a sixth network element, which is equipped with a model storage service unit; the sixth network element is used to receive the target model output by the model training service unit, and the model storage service unit is used to store the target model.

[0087] In one possible implementation, the fourth network element is also equipped with a data storage service unit, which is used to store training data. Attached Figure Description

[0088] Figure 1A An example diagram of the system architecture for the service configuration method provided in this application;

[0089] Figure 1B Another example diagram of the system architecture for the service configuration method provided in this application;

[0090] Figure 1C Another example diagram of the system architecture for the service configuration method provided in this application;

[0091] Figure 1D Another example diagram of the system architecture for the service configuration method provided in this application;

[0092] Figure 2A An example diagram of the service-based AI / ML framework provided for this application;

[0093] Figure 2B An example diagram of the services provided in the application;

[0094] Figure 2C Another example diagram of the service-based AI / ML framework provided for this application;

[0095] Figure 3 A flowchart of the service configuration method provided in this application;

[0096] Figure 4A Another flowchart of the service configuration method provided in this application;

[0097] Figure 4B An example diagram illustrating the service configuration method provided in this application;

[0098] Figure 5 An example diagram of the service invocation method provided in this application;

[0099] Figure 6 A schematic diagram of the communication device provided in this application;

[0100] Figure 7 Another schematic diagram of the communication device provided in this application. Detailed Implementation

[0101] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0102] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such terms are interchangeable where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0103] It should be understood that the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be single or multiple. Additionally, the character " / " in this document generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, "at least one of the following" or similar expressions in this document are used to represent any combination of the listed items; for example, at least one of A, B, and / or C can represent the following six situations: A alone, B alone, C alone, A and B simultaneously, B and C simultaneously, A and C simultaneously, and A, B, and C simultaneously, where A, B, and C can be single or multiple.

[0104] To facilitate understanding, the communication systems and application scenarios to which the service configuration method proposed in this application is applicable will be introduced below:

[0105] The technical solutions provided in this application can be applied to various communication systems, such as: 5th generation (5G) or new radio (NR) systems, long term evolution (LTE) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, wireless local area network (WLAN) systems, satellite communication systems, and future communication systems; or, the technical solutions provided in this application can also be applied to integrated systems of the aforementioned multiple systems. Furthermore, 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), Internet of Things (IoT) communication systems, or other communication systems.

[0106] In the aforementioned communication system, one network element can send signals to or receive signals from another network element. These signals may include information, signaling, or data. The aforementioned network element can also be replaced by entities, network entities, devices, communication equipment, communication modules, nodes, communication nodes, etc. For example, the communication system may 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. For example, Figure 1A This is a schematic diagram of a communication system applicable to the service configuration method in embodiments of this application. For example... Figure 1A As shown, the communication system 100 may include at least one network device, such as Figure 1A The network device 110 shown; in addition, the communication system 100 may also include at least one terminal device, such as Figure 1A The terminal devices 120 and 130 are shown. The network device 110 can communicate with the terminal devices (such as terminal devices 120 and 130) via a wireless link.

[0107] In the embodiments of this application, the terminal device may also be referred to as user equipment (UE), terminal, wireless terminal device, mobile terminal (MT) device, subscriber unit, subscriber station, mobile station (MS), mobile station, remote station, remote terminal device, access terminal device, user terminal device, user agent, user equipment, wireless communication device, user agent, or user apparatus, etc. The terminal device can be a device that provides voice and / or data, such as a handheld device with wireless connectivity, vehicle-mounted device, 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 wireless modems or 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.

[0108] 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.

[0109] The network device in this application embodiment can be a device for communicating with a terminal device. This network device can refer to a radio access network (RAN) node (or device) or base station that connects the terminal device to the wireless network. Currently, some common examples of access network nodes (or devices) include: Node B (NB), evolved Node B (eNB or eNodeB), generation node B (gNB) in 5G new radio (NR) systems, nodes in future communication systems (e.g., xNodeB), transmission reception point (TRP), transmitting point (TP), transmission measurement function (TMF), radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), and access point (AP), etc. Furthermore, in network architectures such as cloud radio access network (CloudRAN) or open radio access network (ORAN), access network equipment can be devices that include centralized units (CUs) (also known as control units) and / or distributed units (DUs). In this type of RAN equipment, which includes both CUs and DUs, the protocol layers of the gNB in ​​the NR system are separated. Some protocol layer functions are centrally controlled by the CU, while the remaining partial or complete protocol layer functions are distributed in the DU, which is centrally controlled by the CU. The separation of CUs and DUs can be based on the protocol stack. For example, one possible separation method is to deploy the radio resource control (RRC), service data adaptation protocol (SDAP), and packet data convergence protocol (PDCP) layers in the CU, and the remaining radio link control (RLC), media access control (MAC), and physical (PHY) layers in the DU.The CU and DU are connected via the F1 interface. The CU, representing its gNB, connects to the core network via the NG interface, and the CU, representing its gNB, connects to other gNBs (or other CUs) via the Xn interface. In the actual deployment of RAN equipment, in addition to the logical gNB composed of CU and DU, the RAN equipment also includes RU (not shown in the figure). RU is a hardware unit that includes some PHY layer functions and / or antenna equipment. Optionally, RU can be configured independently of antenna equipment (e.g., antenna line device (ALD) (also known as antenna linear device)) or integrated with antenna equipment. For example, in a 5G NR system, the aforementioned RU can be an active antenna unit (AAU), that is, a processing unit integrated with a remote radio unit (RRU) (or remote radio head (RRH)) and antenna equipment.

[0110] It should be noted that in practical applications, there may be multiple ways to deploy access network equipment, and this application is not limited to any particular method. For example, 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 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, network equipment may include gNB-CU-CP, gNB-CU-UP, and gNB-DU. As another example, in some deployments, multiple RAN nodes cooperate to assist terminals in achieving wireless access, with different RAN nodes implementing some functions of the base station. For example, RAN nodes may be CU, DU, CU-CP, CU-UP, or RU, etc. CU and DU may be set up separately, or they may be included in the same network element; for example, CU and DU may be included in a BBU. RU may be included in radio frequency equipment or radio frequency units; for example, RU may be included in an RRU, AAU, or RRH. For example, the processing unit used to implement baseband functions in the BBU is called the baseband high (BBH) unit, and the processing unit used to implement baseband functions in the RRU / AAU / RRH is called the baseband low (BBL) unit.

[0111] It should also be understood that a RAN node can support one or more types of fronthaul interfaces, with different fronthaul interfaces corresponding to DUs and RUs 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, it will implement some of the downlink and / or uplink baseband functions relative to the CPRI. For example, for downlink, one or more of precoding, beamforming (BF), or inverse fast Fourier transform (IFFT) / cyclic prefix addition (CP) are moved from the DU to the RU; for uplink, one or more of beamforming (BF), or fast Fourier transform (FFT) / cyclic prefix removal (CP) are moved from the DU to the RU. In one possible implementation, the interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the partitioning between the DU and RU differs, corresponding to different categories (Cats) of eCPRI, such as eCPRI Cat A, B, C, D, E, and F. Taking eCPRI Cat A as an example, for downlink transmission, the partitioning is based on layer mapping. The DU is configured to implement one or more functions preceding and following layer mapping (i.e., coding, rate matching, scrambling, modulation, and one or more of layer mapping), while other functions following layer mapping (e.g., resource element (RE) mapping, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix addition (CP)) are moved to the RU for implementation.For uplink transmission, the 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 one or more functions in de-RE mapping). Other functions following de-mapping (e.g., one or more functions in digital BF or fast Fourier transform (FFT) / CP removal) are moved to 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.

[0112] In this embodiment, the apparatus for implementing the functions of a network device can be a network device itself, or an apparatus capable of supporting the network device in implementing those functions (e.g., a chip system, hardware circuit, software module, or hardware circuit plus software module). This apparatus can be installed in the network device or used in conjunction with the network device. In this embodiment, only the apparatus for implementing the functions of an access network device is described as a network device, and this does not constitute a limitation on the solutions of this embodiment.

[0113] It should be noted that 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; they can also be software functions running on dedicated or general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform); or they can be entities that include dedicated or general-purpose hardware devices and software functions. This application does not limit the specific form of terminal devices and network devices.

[0114] Furthermore, to support artificial intelligence (AI) technology in wireless networks, the communication system also includes a management node. This management node is used to deploy AI-enabled service units within the network to implement AI functions in wireless communication. In some scenarios, this management node is also referred to as an agent. AI-enabled service units can be deployed in one or more of the following locations within the communication system: access network devices, terminal devices, or core network devices, or the AI ​​node can be deployed independently (e.g., in a host or cloud server of an over-the-top (OTT) system). For example, Figure 1B This is a schematic diagram of another communication system applicable to the service configuration method in the embodiments of this application. Compared to Figure 1ARegarding the communication system 100 shown, Figure 1B The communication system 200 shown also includes a management node 140. The management node 140 is used to deploy service units with AI capabilities in the network device 110 or terminal device 120 / 130, so that the network device 110 or terminal device 120 / 130 can perform AI-related operations, such as building training datasets or training AI models.

[0115] Optionally, a service unit with AI capabilities can be understood as an AI module. For example, Figure 1C This is a schematic diagram of a possible application framework in a communication system. For example... Figure 1C As shown, network elements in a communication system are connected via interfaces (e.g., NG interfaces or Xn interfaces) or air interfaces. These network element nodes (e.g., access network nodes (RAN nodes), terminal equipment, core network equipment, or one or more devices in OAM) are equipped with one or more AI modules (for clarity, ...). Figure 1C (Only one is shown in the image). The access network node (or device) can be a single RAN node or include multiple RAN nodes, such as CU and DU. The CU and / 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 models are configured in the CU-CP and / or CU-UP. It should be understood that the AI ​​module is used to implement the corresponding AI function. AI modules deployed in different network elements can be the same or different. Depending on the different parameter configurations, the AI ​​module can implement different functions. An AI module can have one or more AI models. The learning, training, or inference processes of different AI models can be deployed in different nodes or devices, or they can be deployed in the same node or device. The terminal device or access network device in this application can be equipped with one or more of the aforementioned AI modules.

[0116] For example, Figure 1D This is a schematic diagram of a possible application framework in a communication system. For example... Figure 1D As shown, the communication system includes a RAN intelligent controller (RIC), which includes a near-real-time RIC (near-RT RIC) and a non-real-time RIC (non-RT RIC). Figure 1B The functions of the management node shown can be provided by Figure 1DThe RIC implementations shown include near real-time RICs and non-real-time RICs. Non-real-time RICs primarily handle non-real-time information, such as data that is not sensitive to latency (latency in the order of seconds). Near real-time RICs primarily handle near real-time information, such as data that is relatively sensitive to latency (latency in the order of tens of milliseconds). Furthermore, near real-time and non-real-time RICs can also be configured as separate network elements. Optionally, near real-time and non-real-time RICs can also be part of other devices. For example, near real-time RICs can be located in RAN nodes (e.g., CUs or DUs), while non-real-time RICs can be located in OAMs, cloud servers, core network devices, or other network devices; this application does not impose any restrictions.

[0117] To facilitate understanding of the embodiments of this application, the basic concepts of artificial intelligence (AI) involved in this application will first be explained, which will not limit the scope of protection of the embodiments of this application.

[0118] 1. Machine learning (ML):

[0119] Machine learning is a crucial technological approach to achieving AI. AI endows machines with human-like intelligence, using computer hardware and software to simulate certain intelligent human behaviors, including machine learning and other methods. Machine learning refers to learning models or rules from raw data, such as neural networks, decision trees, and support vector machines. Machine learning can be categorized into supervised learning, unsupervised learning, and reinforcement learning.

[0120] Supervised learning, based on collected sample values ​​and labels, uses machine learning algorithms to learn the mapping relationship between sample values ​​and labels, and expresses this learned mapping relationship using a machine learning model. The process of training the machine learning model is the process of learning this mapping relationship. For example, in signal detection, the noisy received signal is the sample, and the corresponding real constellation point is the label. Machine learning aims to learn the mapping relationship between samples and labels through training, that is, to enable the machine learning model to learn a signal detector. During training, the model parameters are optimized by calculating the error between the model's predicted values ​​and the real labels. Once the mapping relationship is learned, it can be used to predict the sample label of each new sample. The mapping relationship learned in supervised learning can include linear mappings and nonlinear mappings. Based on the type of label, the learning task can be divided into classification tasks and regression tasks.

[0121] Unsupervised learning relies solely on collected sample values, using algorithms to discover inherent patterns within the samples. One type of unsupervised learning algorithm uses the samples themselves as supervisory signals; that is, the model learns the mapping relationship from sample to sample, which is called self-supervised learning. During training, model parameters are optimized by calculating the error between the model's predictions and the samples themselves. Self-supervised learning can be used for signal compression and decompression recovery applications; common algorithms include autoencoders and generative adversarial networks.

[0122] Reinforcement learning, unlike supervised learning, is a type of algorithm that learns problem-solving strategies through interaction with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have explicit "correct" action labels. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust its decision actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user based on the total system throughput feedback from the wireless network, aiming to achieve a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the environment state and the optimal decision action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action." Reinforcement learning training is achieved through iterative interaction with the environment.

[0123] Deep neural networks (DNNs) are a specific implementation of machine learning. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems rely on extensive expert knowledge to design communication modules, while DNN-based deep learning communication systems can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0124] Based on their construction method, DNNs can be divided into feedforward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs). FNNs can be neural networks where neurons in adjacent layers are completely connected pairwise, which makes FNNs typically require a large amount of storage space and have high computational complexity.

[0125] CNNs are neural networks specifically designed to process data with a grid-like structure. For example, time-series data (discrete sampling along the time axis) and image data (two-dimensional discrete sampling) can both be considered grid-like data. CNNs do not use all the input information at once for computation; instead, they use a fixed-size window to extract a portion of the information for convolution operations, which significantly reduces the computational cost of model parameters. Furthermore, depending on the type of information extracted by the window (such as people and objects in an image representing different types of information), each window can use different convolution kernels, allowing CNNs to better extract features from the input data.

[0126] Recurrent Neural Networks (RNNs) are a type of distributed neural network (DNN) that utilizes feedback time-series information. Their input includes the current input value and their own output value from the previous time step. RNNs are well-suited for acquiring temporally correlated sequence features, and are particularly applicable to applications such as speech recognition and channel coding / decoding.

[0127] AI models refer to function models that map inputs of a certain dimension to outputs of a certain dimension, and their parameters can be obtained through machine learning training. For example, f(X) = aX 2 +b is a quadratic function model, which can be viewed as an AI model. a and b correspond to the parameters of this model and can be obtained through machine learning training. In machine learning, the data used for model training, validation, and / or testing can form a dataset or training dataset. The quantity and / or quality of data in the dataset or training dataset will affect the effectiveness of machine learning. Model training involves selecting an appropriate loss function (which measures the difference between the model's predictions and the true values) and using optimization algorithms to train the model parameters to minimize the loss function value. Model testing involves evaluating the model's performance using test data after training. Model application involves using the trained model to solve real-world problems.

[0128] A neural network, or artificial neural network, is a mathematical model that mimics the behavioral characteristics of animal neural networks to perform distributed parallel information processing. It is a special form of AI model.

[0129] 2. Model Training:

[0130] Model training involves selecting an appropriate function (such as a loss function) and using optimization algorithms to train the model parameters so that the difference between the model's predicted values ​​and the ground truth (or target values, labels) tends to be minimized.

[0131] For example, model training methods include, but are not limited to, supervised learning, self-supervised learning, and knowledge distillation.

[0132] 3. Model files and model parameters:

[0133] Model files and / or model parameters can be used to determine the model. Optionally, the model in this application may refer to the model itself, or it may refer to the model files and / or model parameters used to determine the model.

[0134] The model file can be used to indicate the model structure, which may include, but is not limited to, FNN, CNN, or RNN. The model file can have a fixed format, such as a standard predefined format, or a format pre-negotiated by both ends of the interface. Model parameters can refer to parameters in the neural network model, such as, but not limited to, the number of layers in the neural network, the type and weights of neurons in each layer, etc. This application does not limit the method of distributing model parameters.

[0135] Take DNN as an example. The idea behind DNN comes from the neuronal structure of the brain. Each neuron can perform a weighted summation operation on its inputs and then use the result of the weighted summation operation to generate the output through a non-linear function. For example, the input of a neuron is x = [x0, x1, ..., x...]. N-1 The weights corresponding to the inputs are w = [w0, w1, ..., w] N-1 The bias of the weighted summation is b. The nonlinear function f() can take many forms; for example, the nonlinear function f() can be the maximum value function max{0, x}. Then the effect of a neuron's execution is... Where N is a positive integer, and n is a positive integer greater than or equal to 0 and less than or equal to (N-1). The weights of the weighted summation operation of neurons in a neural network and the nonlinear function are called the parameters of the neural network. The parameters of all neurons in a neural network constitute the parameters of the neural network.

[0136] A DNN typically has multiple neural network layers, including an input layer, one or more hidden layers, and an output layer. Generally, the first layer is the input layer, the last layer is the output layer, and the layers in between are hidden layers. Each layer contains multiple neurons. Layers are fully connected; that is, any neuron in the i-th layer is connected to any neuron in the (i+1)-th layer. The input layer processes the received values ​​(i.e., the DNN's input) through neurons and then passes them to the hidden layers. Similarly, the hidden layers pass the computation results to the final output layer, producing the DNN's output. This application does not limit the structure and parameters used in the AI ​​model.

[0137] One of the model structure or model parameters can be predefined, while the other can be sent by the sender (e.g., the network side). Alternatively, both the model structure and model parameters can be sent by the sender (e.g., the network side). This application does not impose any restrictions on this.

[0138] Sending a model can refer to sending a model file and / or model parameters, while receiving a model can refer to receiving a model file and / or model parameters.

[0139] To apply AI technology to wireless communication, current standards propose several AI / ML frameworks to enhance data collection or improve air interface AI. However, as AI / ML application scenarios expand, the current AI / ML frameworks have limited functionality and cannot enable on-demand deployment of AI / ML features, thus restricting the expansion of AI / ML capabilities.

[0140] In response, this application provides a service-based AI / ML framework that service-izes the functions to be implemented, so that the services corresponding to the functions can be flexibly deployed in one or more network elements in the communication system, and then the specific functions can be realized through the interaction of one or more network elements.

[0141] It should be noted that deploying services corresponding to functions within a network element refers to deploying service units corresponding to the service in the network element (e.g., access network equipment or terminal equipment) that needs to use the function through a management node. This enables the network element to implement the corresponding function based on the deployed service units, i.e., to obtain or provide the service corresponding to the service unit. The service units involved in this application embodiment refer to software units or hardware / software modules that can be deployed within a network element (e.g., access network equipment or terminal equipment) to implement one or more functions. For example, the service unit is used to implement air interface functions (i.e., functions related to wireless communication between the access network equipment and the terminal equipment), data management functions, and model management functions, etc., which are not limited in this embodiment.

[0142] Figure 2A An example diagram of a service-based AI / ML framework provided for embodiments of this application. For example... Figure 2A As shown, the AI / ML framework provided in this application primarily service-orientedizes air interface functions, model management functions, and data management functions, thereby achieving flexible AI / ML functionality. The service-oriented interfaces can be scheduled or configured by management nodes / agents.

[0143] Figure 2B An example diagram illustrating the services provided in an embodiment of this application. For example... Figure 2B As shown, a service can be composed of one or more services connected in series, parallel, or branched. For example, Figure 2BThe service 0 shown is composed of services 0.0, 0.1 and 0.2 connected in series or in parallel.

[0144] Figure 2C Another example diagram of a service-based AI / ML framework provided for embodiments of this application. (See diagram below.) Figure 2C As shown, the AI / ML framework is mainly divided into management nodes (also known as agents or related modules) and services. Among them, the management node is the core of the AI / ML framework, and the services include basic services, advanced services, and storage services.

[0145] The basic service is implemented by basic service units. Taking the air interface basic service unit as an example, the basic service is related to the hardware structure of the network element to which the service unit is deployed. Specifically, the basic service includes at least radio frequency (RF) service, physical layer service (PHY layer service), and medium access control layer service (MAC layer service). It can be understood that the basic service includes at least the aforementioned three types of services, which can construct the functions of a traditional air interface. Furthermore, the aforementioned three types of services can be further subdivided. For example, the RF service requires the network element to include RF-related hardware modules. For instance, the RF service unit may include an RF amplification service unit, an automatic gain control service unit, a digital-to-analog converter service unit, an analog-to-digital converter service unit, a clock synchronization service unit, an antenna service unit, etc. For example, PHY layer services require network elements to include physical layer signal processing modules. These PHY layer service units may include channel coding service units, channel decoding service units, rate matching service units, modulation service units, demodulation service units, layer mapping service units, precoding service units, resource mapping service units, waveform generation service units, reference signal generation service units, channel estimation service units, and channel equalization service units. Similarly, MAC layer services require network elements to deploy relevant MAC layer functional modules. These MAC layer service units may include resource scheduling service units, channel mapping service units, and hybrid automatic repeat request (HARQ) service units. It should be noted that a basic service can be implemented by a single service unit. For example, deploying a basic service unit within a network element enables that network element to implement or obtain the basic service corresponding to that basic service unit. A basic service can be implemented by a service unit consisting of multiple basic service units connected in series or in parallel. For example, multiple basic service units can be deployed in a network element, and these multiple basic service units can be connected in series to form a larger service unit, so that the network element can implement or obtain the basic service corresponding to the connected basic service unit.

[0146] Advanced services encompass the core functionalities of the AI / ML framework and are implemented by advanced service units. These units primarily include function / model management service units and data management service units. The function / model management service unit includes function / model inference service units, function / model training service units, and performance monitoring service units. The data management service unit includes data collection service units and data retrieval service units. Optionally, advanced service units may also include performance monitoring service units, perception service units, and computation service units. It should be noted that advanced service units can be constructed from one or more basic service units connected in series or parallel, or from at least one basic service unit and at least one advanced service unit connected in series or parallel. Optionally, traditional air interface services built using basic service units can be advanced service units.

[0147] Storage services are implemented by service units related to storage media, requiring network elements to have storage-related media or modules. For example, storage service units include model storage service units and data storage service units. In some scenarios, storage services can also be considered advanced services.

[0148] Specifically, in Figure 2C In the AI / ML framework shown, after receiving task information, the management node can decompose the task information into information about service units to be deployed through the planning module. This information is then sent to the network elements that need the services, enabling the network elements to deploy the service units based on this information, and subsequently solve the task based on the deployed service units. It should be noted that the planning module can be located within the management node, in a computing device or server connected to the management node, or in other devices with computing or planning functions; this embodiment is not limited to these options.

[0149] The following is combined Figure 3 The main process of the service configuration method provided in this application is described below:

[0150] like Figure 3The diagram shows a flowchart of an embodiment of a service configuration method provided in this application. This service configuration method is illustrated using the interaction between a first network element and a second network element as an example. Of course, the entity executing the actions of the first network element in this method can also be a device, module, or chip within the first network element; similarly, the entity executing the actions of the second network element in this method can also be a device, module, or chip within the second network element. This embodiment does not specifically limit this. For example, the first network element is a management node, and the second network element is an access network device; that is, the management node sends configuration information to the access network device to deploy a service unit on the access network device; or, the first network element is a management node, and the second network element is a terminal device; that is, the management node sends configuration information to the terminal device to deploy a service unit on the terminal device; or, the first network element is an access network device, and the second network element is a terminal device; that is, the access network device sends configuration information to the terminal device to deploy a service unit on the terminal device.

[0151] For example, such as Figure 3 As shown, the service configuration method includes the following steps:

[0152] Step 301: The first network element obtains the first information of the second network element.

[0153] The first information includes information about at least one service unit deployed in the second network element. The service unit is used to implement one or more functions of the access network device or terminal device. It can be understood that a service unit refers to a software unit or a combination of hardware and software modules that can be deployed within a network element (e.g., access network device or terminal device) to implement one or more functions. For example, the service unit is used to implement air interface functions (i.e., functions related to wireless communication between the access network device and the terminal device), etc., but this embodiment is not limited to this.

[0154] Optionally, the information for a service unit may include at least one of the following:

[0155] Identification information is used to uniquely identify service units. For example, if a second network element has deployed multiple service units, each service unit has a unique identification information to distinguish between different service units.

[0156] Functional description information describes the function of the service unit. For example, the functional description information for the NR modulation service unit is to complete the mapping from coded bits to symbols. Another example is the functional description information for the RF amplification service unit, which describes adjusting the amplification factor of the RF amplifier by calling the RF amplifier's hardware interface.

[0157] Input / output information is used to indicate the input and output of the service unit. This input / output information includes input identification information and output identification information. Optionally, the input / output information may also include at least one of the following: input format information, indicating the format of the service unit's input data; output format information, indicating the format of the service unit's output data; input memory layout information, indicating the arrangement of the service unit's input data in memory; output memory layout information, indicating the arrangement of the service unit's output data in memory; input quantization bit width, indicating the quantization bit width of the service unit's input data; and output quantization bit width, indicating the quantization bit width of the service unit's output data. In other words, the input / output information indicates the format, memory layout, and quantization bit width of the input and output data to ensure data format consistency when different service units are connected in series. For example, if service unit 1 and service unit 2 are connected in series, the format, memory layout, and quantization bit width of the output data of service unit 1 and the input data of service unit 2 must be consistent. For example, taking the NR modulation service unit as an example, the input data format is {modulation order m, int}, {coded bit stream, N x m, bool}; the output data format is {symbol stream, N x 1, int}; the quantization bit width of both the input and output data is 8 bits. Since the service unit's information includes input and output information indicating the format of the input and output data, memory layout, and quantization bit width, it implements the definition of the hardware abstraction layer. This allows the framework to support alignment of heterogeneous hardware / services from different vendors, which is beneficial for improving the flexibility of the combined deployment of basic units.

[0158] Equipment capability requirements information indicates the software and / or hardware capability requirements of the equipment deploying the service unit. For example, this includes software and hardware platform information, indicating the software and hardware platforms the equipment deploying the service unit possesses, such as operating systems, programming languages, frameworks, libraries, and containers; accelerator information, indicating whether the equipment supports acceleration devices such as graphics processing units (GPUs), neural processing units (NPUs), field-programmable gate arrays (FPGAs), or application-specific integrated circuits (ASICs); and computing power information, indicating the type and magnitude of computing power required for the normal operation of the service unit.

[0159] Gradient computation capability information is used to indicate whether a service unit supports backpropagation gradient computation, that is, whether the service unit supports forward propagation computation and backpropagation computation. If a service unit supports backpropagation gradient computation, then the service unit has the ability to train a model; if the service unit does not support backpropagation gradient computation, then the service unit does not have the ability to train a model. In some scenarios, "supporting backpropagation gradient computation" is also referred to as "differentiable," and "not supporting backpropagation gradient computation" is also referred to as "non-differentiable." This embodiment and subsequent embodiments will only describe it using the description of "whether backpropagation gradient computation is supported."

[0160] An executable file is a file that can be loaded and executed by an operating system. The presentation of executable programs varies across different operating system environments. In this embodiment, the executable file is used to trigger the network element that receives service unit information to deploy the service unit.

[0161] Complexity information indicates the number of floating point operations per second (FLOPS) or multiply-accumulate operations (MACs) performed by the service unit. FLOPS indicates single-precision (32-bit) or double-precision (64-bit) addition or multiplication, while MACs indicate the number of multiplications and additions performed by the service unit. For example, a complexity of {150 MACs} means that the sum of the number of multiplications and additions is 150.

[0162] Open / closed source information indicates whether the service unit's information is open source or closed source. When the service unit's information is closed source, it includes an executable file, which triggers the network element that receives the service unit's information to deploy the service unit. When the service unit's information is open source, it may not include an executable file, but it includes source code, which can be compiled into an executable file.

[0163] Optionally, the service unit information may also include the service unit name, the service unit supplier identifier, etc.

[0164] Optionally, if the service unit is a high-level service unit, the information of the service unit, in addition to identification information, functional description information, and input / output information, also includes at least one of the following:

[0165] Topology information is used to indicate the connection topology between multiple service units constituting a higher-level service unit. For example, if service unit 0 is composed of service unit 0.0, service unit 0.1, and service unit 0.2 connected in series, then the topology information of service unit 0 indicates that service unit 0.0 is connected to service unit 0.1, service unit 0.1 is connected to service unit 0.2, and the data transmission direction is from service unit 0.0 to service unit 0.1 and then to service unit 0.2. Optionally, the topology information can be represented by the identification information of the service units. For example, if the topology information includes {ID of service unit 0.0, ID of service unit 0.1, ID of service unit 0.2}, it indicates that service unit 0.0 is connected to service unit 0.1, service unit 0.1 is connected to service unit 0.2, and the data transmission direction is from service unit 0.0 to service unit 0.1 and then to service unit 0.2.

[0166] Functional parameter information indicates the functional parameters used by the high-level service unit when implementing its service functions. In one example, the high-level service unit is the model training service unit. The functional parameter information for the model training service unit includes: training hyperparameters, training loss function, and training stopping conditions. Training hyperparameters (or simply hyperparameters) are parameters set before training a machine learning or deep learning model. Hyperparameters remain unchanged during training and are used to control the model's learning process and structure. Common hyperparameters include the number of layers in the neural network, the number of neurons, and activation functions. The training loss function (or simply loss function, also known as the cost function) measures the difference between the model's predicted values ​​and the true values. The training stopping condition is the condition under which model training stops. For example, the training stopping condition is when the model reaches a set accuracy requirement; that is, training can stop when the neural network training reaches the preset accuracy requirement. Or, training can stop when the model's loss function value reaches an acceptable range. Another example is when the training stopping condition is when the model reaches the maximum number of iterations; that is, training will also stop when the training reaches the set maximum number of iterations, which helps prevent the model from overfitting or getting trapped in local optima. In another example, the advanced service unit is the performance monitoring service unit. The functional parameter information of the performance monitoring service unit includes the monitoring period and monitoring parameters. The monitoring period indicates the duration between two consecutive monitoring actions; the monitoring parameters indicate the object being monitored. For example, monitoring parameters could be the accuracy of the dataset, system throughput, or system throughput rate.

[0167] It should be noted that since different service units are used to implement different service functions, in practical applications, the information contained in a service unit can be added or reduced based on the specific service unit. Examples will not be listed here.

[0168] For example, taking the NR modulation service unit as the basic service unit, the information of the NR modulation service unit includes the following:

[0169] {Basic Service Name: {NR Modulation}

[0170] Service Description: {Complete the mapping from encoded bits to symbols}

[0171] Basic Service ID: {1}

[0172] Supplier: {HW}

[0173] Input format: {{modulation order m, int}, {coded bit stream, N x m, bool}}

[0174] Output format: {symbolic stream, Nx1, int}

[0175] Input memory layout: {Storage space occupied by various variable types (e.g., double type occupies 8 bytes), array storage order (e.g., m*n array stored row-first, then column-first), storage method contiguous} (optional)

[0176] Output memory layout: {Storage space occupied by various variable types (e.g., double type occupies 8 bytes), array storage order (e.g., m*n array stored row-first, then column-first), storage method contiguous} (optional)

[0177] Quantization bit width: {8 bits}

[0178] Input ID: {1}

[0179] Output ID: {2}

[0180] The basic service unit is applicable to the following (hardware and software) environments: {all}

[0181] Complexity: {150 MACs}

[0182] Open / Closed Source: {Open Source}

[0183] Gradient computation capability: {Supports inverse gradient computation}

[0184] Executable file: {mod_cpu.bin, ...}}

[0185] For example, taking the NR transmitter service unit as an example of an advanced service unit, the information of the NR transmitter service unit includes the following:

[0186] {Advanced Service Name: {NR Transmitter}

[0187] Service Description: {Complete the mapping from bits to time-domain transmitted symbols}

[0188] Premium Service ID: {0}

[0189] Management node ID: {agnet0}

[0190] Basic service topology: {service 0->service 1->service N}

[0191] Input description (referencing the input description of service 0)

[0192] Output description (referencing the output description of service N)

[0193] Functional parameters: Maximum output power, Power Class, Maximum Power Reduction (MPR), Additional Maximum Power Reduction (A-MPR), Adjacent Channel Leakage Ratio (ACLR), and Spectrum Emission Mask (SEM).

[0194] Furthermore, in this embodiment, the first network element can obtain the first information from the second network element through any of the following implementation methods.

[0195] In one possible implementation, the first information is obtained through a request-response mechanism. For example, the first network element sends a first message to the second network element, the first message being used to request information about the service units deployed by the second network element; after receiving the first message, the second network element sends a second message to the first network element, the second message including the first information.

[0196] In another possible implementation, the first information is obtained through a subscription mechanism. For example, the first network element sends a first message to the second network element. The first message is a subscription request message, which requests the second network element to periodically or conditionally send the first information to the first network element. Based on the first network element's subscription request message, the second network element periodically or conditionally sends a second message containing the first information to the first network element.

[0197] In another possible implementation, the second network element proactively reports the first information to the first network element. For example, after the second network element deploys a service unit, it sends information about the deployed service unit to the first network element.

[0198] In practical applications, the first network element can obtain the first information from the second network element using any of the aforementioned implementation methods; this embodiment is not limited to any of these methods.

[0199] It should be noted that if a request-response mechanism or a subscription mechanism is adopted, the first network element can carry information in the first message to indicate that the information of the service unit reported by the second network element is what the first network element expects to obtain, and is not necessarily the information of all service units deployed by the second network element.

[0200] In one possible implementation, the first message includes service unit type information, which indicates the type of service unit queried through the first message. This can be understood as the service unit type information indicating which types of service unit information the first network element expects the second network element to report, so that the first message sent by the second network element to the first network element indicates the type of service unit information expected by the first network element. In other words, the service unit corresponding to the service unit information included in the first message is of the same type as the service unit indicated by the service unit type information in the first message.

[0201] In one example, if the service unit type is divided into two types, the service unit type information may include first type information and second type information, wherein the first type information indicates a first type service unit, and the second type information indicates a second type service unit. Optionally, if a second type service unit consists of at least two different first type service units, the first type service unit can be understood as described above. Figure 2C Corresponding to the basic service units described in the examples, for example, the first type of service unit includes radio frequency service units, physical layer service units, or media access control (MAC) layer service units; the second type of service unit can be understood as the aforementioned Figure 2C The advanced service units described in the corresponding examples, such as the second type of service units, include model management service units, data management service units, performance monitoring service units, perception service units, or computing service units.

[0202] In another example, if the service unit types are divided into three types, the service unit type information can include first type information, second type information, and third type information. The first type information indicates a first-type service unit, the second type information indicates a second-type service unit, and the third type information indicates a third-type service unit. Optionally, if a second-type service unit consists of at least two different first-type service units, and the third-type service unit is used to implement storage functionality, then the first-type service units can be understood as described above. Figure 2C Corresponding to the basic service units described in the examples, for example, the first type of service unit includes radio frequency service units, physical layer service units, or media access control (MAC) layer service units; the second type of service unit can be understood as the aforementioned Figure 2CCorresponding to the advanced service units described in the examples, for instance, the second type of service unit includes model management service unit, data management service unit, performance monitoring service unit, perception service unit, or computing service unit; the third type of service unit can be understood as those mentioned above. Figure 2C The corresponding storage service units described in the examples, for example, include model storage service units or data storage service units.

[0203] It should be noted that, based on the different application scenarios of the AI / ML framework provided in this application embodiment, the service units that the AI / ML framework may involve can be more finely divided according to the needs of the scenario. Examples will not be listed here. It should also be noted that the type information of the service unit carried in the first message can be partial or complete type information. For example, if the service unit type is divided into two types, the first message can carry either the first type information or the second type information, or it can carry both. As another example, if the service unit type is divided into three types, the first message can carry any one of the first type information, the second type information, and the third type information, or it can carry any two of the first type information, the second type information, and the third type information simultaneously. This embodiment does not limit the specific type of type information carried in the first message.

[0204] In this embodiment, by carrying the type information of the service unit in the first message, the first network element can query the information of the service unit of the desired type on demand, so that the second network element only reports the service unit that matches the type information, which helps to improve query efficiency and save signaling overhead.

[0205] Furthermore, in one possible implementation, the second message also includes device capability information of the second network element, which indicates the software and / or hardware capabilities possessed by the second network element. For example, the device capability information includes software and hardware platform information, indicating the software and hardware platforms supported by the second network element, such as operating systems, programming language types, frameworks, libraries, and containers; accelerator information, indicating whether the second network element supports acceleration devices such as graphics processing units (GPUs), neural processing units (NPUs), field-programmable gate arrays (FPGAs), or application-specific integrated circuits (ASICs); and computing power information, indicating the type and magnitude of computing power that the second network element can provide.

[0206] Step 302: The first network element determines the information of the target service unit to be deployed by the second network element.

[0207] Here, the target service unit refers to a service unit capable of solving the task. For example, the first network element determines the service unit that needs to be deployed in the second network element based on the received task information, so that the second network element can complete the task indicated by the aforementioned task information. For example, the task information indicates improving the throughput of the target area; the first network element analyzes the task information through the planning module and obtains a solution to configure the transceivers in that area as AI transceivers, that is, it is necessary to deploy an AI sending service unit or an AI receiving service unit in the second network element.

[0208] The target service unit information includes second information and topology information. The second information includes information about the multiple service units constituting the target service unit, and the topology information indicates the connection topology between the multiple service units constituting the target service unit. For example, if service unit 1, service unit 2, and service unit 3 are connected in series to form the target service unit, then the second information includes information about service unit 1, service unit 2, and service unit 3. The topology information indicates that service unit 1 is connected to service unit 2, service unit 2 is connected to service unit 3, and the data transmission direction is from service unit 1 to service unit 2 and then to service unit 3. Optionally, the topology information can be represented by the identification information of the service units or by a connection diagram; this implementation is not limited. For an explanation of the specific content included in the service unit information, please refer to the relevant description in step 301 above; it will not be repeated here.

[0209] In one possible implementation, the first network element considers the device capability information of the second network element when determining the information of the target service unit. For example, the first network element determines the second information based on the first information and the device capability information of the second network element. If the second information includes information of the first service unit and information of the second service unit constituting the target service unit, then the capability indicated by the capability information of the second network element includes the capability indicated by the device capability requirement information of the first service unit and the capability indicated by the device capability requirement information of the second service unit. That is, the capability of the second network element needs to support the capability requirements of the service unit, which helps to ensure the stable operation of the target service unit after deployment in the second network element and improves the reliability of service deployment.

[0210] Furthermore, in one possible implementation, if each service unit constituting the target service unit supports backward gradient calculation, then the target service unit also supports backward gradient calculation, meaning the target service unit can support model training. For example, the second information includes information about the first service unit constituting the target service unit and information about the second service unit; if the gradient calculation capability information of the first service unit indicates that the first service unit supports backward gradient calculation, and if the gradient calculation capability information of the second service unit indicates that the second service unit supports backward gradient calculation, then the target service unit supports backward gradient calculation. In this implementation, by configuring each service unit constituting the target service unit to support backward gradient calculation, so that the target service unit also supports backward gradient calculation, it is beneficial for the target service unit to achieve online model training and improves the flexibility of applying AI technology to the communication system.

[0211] Optionally, the first network element will also determine third information, which is the information in the second information that needs to be sent to the second network element. The third information may contain the complete second information or only a portion of the second information.

[0212] In one possible implementation, the third information is determined based on the first and second information. For example, the third information is obtained by excluding the first information from the second information. That is, after determining the second information, the first network element excludes the first information from the second information to obtain the third information, in order to reduce signaling overhead.

[0213] Step 303: The first network element sends third information and topology information to the second network element; correspondingly, the second network element receives the third information and topology information from the first network element.

[0214] Among them, the third information and topology information are used to instruct the second network element to deploy the target service unit, and the target service unit is used to implement the target service.

[0215] Step 304: The second network element deploys the target service unit based on the third information and topology information.

[0216] For example, after receiving the third information and topology information, the second network element deploys the target service unit based on the third information and topology information.

[0217] In one example, if the third information contains complete second information, then the second network element deploys the target service unit based on the third information and topology information.

[0218] In another example, if the third information is obtained by excluding the first information from the second information, the second network element deploys the target service unit based on the third information, the first information, and the topology information.

[0219] In this embodiment, the first network element can determine whether a target service unit can be deployed on the second network element based on the information of the service units already deployed on the second network element, and the information that needs to be sent to the second network element when deploying the target service unit (e.g., third information and topology information). This enables the first network element to deploy service units on the second network element as needed, which is beneficial to improving the flexibility of service deployment.

[0220] The following is combined Figure 4A An embodiment of the service configuration method provided in this application is described. In this embodiment, a first network element is taken as the management node, and a second network element is taken as the access network device and a terminal device. Furthermore, an example is taken where AI transmitter service and AI receiver service are deployed on the access network device and the terminal device respectively. Figure 4A As shown, the service configuration method includes the following steps:

[0221] Step 401: The access network device sends its first information to the management node; correspondingly, the management node receives the first information from the access network device.

[0222] The first information of the access network device includes information about at least one service unit that has been deployed on the access network device. For an explanation of the service unit information, please refer to the relevant description in step 301 above; it will not be repeated here.

[0223] In this embodiment, the management node can obtain the first information of the access network device from the access network device through a request-response mechanism, a subscription mechanism, or the access network device can actively send its first information to the management node. This embodiment does not limit the method. The way the management node obtains the first information of the access network device is similar to the way the first network element obtains the first information of the second network element in step 301 above. Please refer to the relevant description in step 301 above for details, which will not be repeated here.

[0224] For example, such as Figure 4BAs shown, the first information of the access network device includes information about multiple basic service units deployed by the access network device. These include information about the channel coding (CEC) service unit, the bit-to-symbol mapping service unit, the layer mapping service unit, the precoding service unit, the resource element mapping (RE) service unit, and the inverse fast fourier transform (IFFT) service unit. The information of the channel coding (CEC) service unit includes information about the AI ​​coding (AI Enc) service unit, the rate matching service unit, and the scrambling service unit.

[0225] Step 402: The terminal device sends its first information; correspondingly, the access network device receives the first information of the terminal device, or the management node receives the first information of the terminal device.

[0226] The first information of the terminal device includes information about at least one service unit that has been deployed on the terminal device. For an explanation of the service unit information, please refer to the relevant description in step 301 above; it will not be repeated here.

[0227] The terminal device sends its first information to the access network device. After receiving the first information from the terminal device, the access network device can cache the first information to help it determine the information of the service unit to be deployed. Alternatively, the access network device can send the first information from the terminal device to the management node so that the management node can determine the information of the service unit to be deployed.

[0228] In this embodiment, the access network device can obtain the first information of the terminal device from the terminal device through a request-response mechanism, a subscription mechanism, or the terminal device can actively send its first information to the access network device. This embodiment does not limit this. The method by which the access network device obtains the first information of the terminal device is similar to the method by which the first network element obtains the first information of the second network element in step 301 above. Please refer to the relevant description in step 301 above for details, which will not be repeated here. Furthermore, the management node can obtain the first information of the terminal device from the access network device through a request-response mechanism, a subscription mechanism, or the access network device can actively send the first information of the terminal device to the management node after receiving it. This embodiment does not limit this. The specific implementation method is similar to the method by which the first network element obtains the first information of the second network element in step 301 above. Please refer to the relevant description in step 301 above for details, which will not be repeated here.

[0229] For example, such as Figure 4B As shown, the first information of the terminal device includes information about multiple basic service units deployed on the terminal device. These include information about the channel decoding (Dnc) service unit, the symbol-to-log-likelihood rate (LLR) mapping service unit, the resource element demapping (RE) service unit, and the fast fourier transform (FFT) service unit. The information from the channel decoding (Dnc) service unit includes information from the de-rate matching service unit and the de-scramble service unit.

[0230] It should be noted that in this embodiment, there is no explicit time order limitation between steps 401 and 402. Step 401 can be executed first and then step 402, or step 402 can be executed first and then step 401, or steps 401 and 402 can be executed simultaneously. This embodiment does not limit this.

[0231] Step 403: The access network device sends the first association information to the management node; correspondingly, the management node receives the first association information from the access network device.

[0232] The first associated information is used to indicate the necessary information for the access network device to obtain AI transmitter services. For example, the first associated information includes real-time status information such as the access network device's computing power usage and resource usage.

[0233] In this embodiment, the management node can obtain the first association information from the access network device through a request-response mechanism, a subscription mechanism, or the access network device can actively send the first association information to the management node; this embodiment is not limited in this respect. The management node can determine which service units to deploy as target service units based on the first association information, and can also use the first association information to verify the feasibility of deploying a certain service unit on the access network device.

[0234] It should be noted that in this embodiment, step 403 is an optional step. Whether the access network device sends the first association information is related to the type of service unit to be deployed by the access network device.

[0235] Step 404: The terminal device sends the second association information; correspondingly, the access network device receives the second association information, or the management node receives the second association information.

[0236] The second associated information is used to indicate the necessary information for the terminal device to obtain AI receiver services. For example, the second associated information includes real-time status information such as the terminal device's computing power usage and resource usage.

[0237] In this embodiment, the terminal device sends the second association information to the access network device. After receiving the second association information, the access network device can cache the second association information so that it can use the second association information to verify the feasibility of deploying a certain service unit on the terminal device, or the access network device can use the second association information to determine which service units need to be deployed in the terminal device as target service units. The access network device can also send the second association information to the management node so that the management node can use the second association information to verify the feasibility of deploying a certain service unit on the terminal device, or the management node can use the second association information to determine which service units need to be deployed in the terminal device as target service units.

[0238] It should be noted that in this embodiment, step 404 is an optional step. Whether the terminal device sends the second association information is related to the type of service unit to be deployed on the terminal device.

[0239] It should also be noted that in this embodiment, there is no explicit time order limitation between steps 403 and 404. Step 403 can be executed first and then step 404 can be executed first and then step 403 can be executed, or steps 403 and 404 can be executed simultaneously. This embodiment does not limit this.

[0240] In this embodiment, if the access network device sends the first information of the terminal device obtained from the terminal device to the management node, then the management node, the access network device, and the terminal device will execute steps 405, 406, and 407, that is... Figure 4A Option 1 is shown.

[0241] Step 405: The management node determines the information of the transmitter service unit to be deployed on the access network equipment and the information of the receiver service unit to be deployed on the terminal equipment.

[0242] For example, if the management node receives a task message indicating a need to increase throughput in the target area, the management node, through its planning module, determines that the solution requires configuring AI transceiver services to replace traditional transceiver services. This means configuring AI transmitter services on access network devices and AI receiver services on terminal devices. Then, the management node determines the information of the AI ​​transmitter service units to be deployed on the access network devices and the AI ​​receiver service units to be deployed on the terminal devices.

[0243] The information of the AI ​​transmitter service unit includes information about each service unit constituting the AI ​​transmitter and the topology information of the AI ​​transmitter service unit. For example, such as... Figure 4B As shown, the information of each service unit constituting the AI ​​transmitter includes information on the channel coding (channel Enc) service unit, the bit-to-symbol mapping service unit, the layer mapping service unit, the precoding service unit, the resource element mapping (REmapping) service unit, and the IFFT service unit. The topology information of the AI ​​transmitter is used to indicate the connection relationships of the aforementioned service units constituting the AI ​​transmitter service units. For example, the topology information of the AI ​​transmitter indicates that the AI ​​coding service unit, the rate matching service unit, and the scrambling service unit are sequentially connected to constitute the channel coding service unit, and that the channel coding service unit, the bit-to-symbol mapping service unit, the layer mapping service unit, the precoding service unit, the REmapping service unit, and the IFFT service unit are sequentially connected.

[0244] The information of the AI ​​receiver service unit includes information about each service unit constituting the AI ​​receiver and the topology information of the AI ​​receiver service unit. For example, such as... Figure 4BAs shown, the information of each service unit constituting the AI ​​receiver includes, in addition to, information about the de-rate matching service unit, the de-scramble service unit, the symbol-to-LLR mapping service unit, the resource element demapping service unit, and the FFT service unit, information about the AI ​​decoding service unit. The topology information of the AI ​​receiver is used to indicate the connection relationships of the aforementioned service units constituting the AI ​​receiver service units. For example, the topology information of the AI ​​receiver indicates that the de-scrambling service unit, the de-rate matching service unit, and the AI ​​decoding service unit are sequentially connected to constitute the channel decoding service unit, and that the FFT service unit, the RE demapping service unit, the symbol-to-LLR mapping service unit, and the channel decoding service unit are sequentially connected.

[0245] Step 406: The management node sends the topology information of the AI ​​transmitter service unit to the access network device; correspondingly, the access network device receives the topology information of the AI ​​transmitter service unit from the management node.

[0246] The topology information of the AI ​​transmitter service unit is used to indicate the connection relationships among the various service units that constitute the AI ​​transmitter service unit. For a specific example, please refer to the relevant description in step 405 above; it will not be repeated here.

[0247] It should be noted that since the access network equipment has already deployed the basic service units that constitute the AI ​​transmitter service unit, the management node only needs to send the topology information connecting the aforementioned basic service units to the access network equipment.

[0248] Optionally, the management node also sends third information of the AI ​​transmitter service unit to the access network device; correspondingly, the access network device receives the third information of the AI ​​transmitter service unit from the management node.

[0249] In one implementation, the third information of the AI ​​transmitter service unit can be information obtained by excluding the first information of the access network device from the information of the AI ​​transmitter service unit. That is, the third information of the AI ​​transmitter service unit is information about basic service units not deployed by the access network device. For example, if the access network device has some undeployed basic service units, the management node sends information about these undeployed basic service units to the access network device. For example, if the access network device does not deploy a layer mapping service unit, the management node needs to send the layer mapping service unit information in addition to the topology information of the AI ​​transmitter service unit to the access network device.

[0250] Step 407: The management node sends the topology information of the AI ​​receiver service unit to the terminal device; correspondingly, the terminal device receives the topology information of the AI ​​receiver service unit from the management node.

[0251] The topology information of the AI ​​receiver service unit is used to indicate the connection relationships among the various service units that constitute the AI ​​receiver service unit. For a specific example, please refer to the relevant description in step 405 above; it will not be repeated here.

[0252] Optionally, the management node also sends third information of the AI ​​receiver service unit to the terminal device; correspondingly, the terminal device receives the third information of the AI ​​receiver service unit from the management node.

[0253] In one implementation, the third information of the AI ​​receiver service unit can be information obtained by excluding the terminal's first information from the information of the AI ​​receiver service unit; that is, the third information of the AI ​​receiver service unit is information from basic service units not deployed on the terminal device. For example, such as... Figure 4B As shown, if the terminal device does not deploy an AI decoding service unit, the management node needs to send the AI ​​receiver service unit's topology information to the terminal device, as well as the AI ​​decoding service unit's information.

[0254] It should be noted that in this embodiment, there is no explicit time order requirement between steps 406 and 407. Step 406 can be executed before step 407, or step 407 can be executed before step 406, or steps 406 and 407 can be executed simultaneously. This embodiment does not impose any restrictions. After the access network device executes step 406, the access network device will execute step 412. After the terminal device executes step 407, the terminal device will execute step 413.

[0255] In this embodiment, if the access network device has the capability to deploy service units for the terminal device, the management node, the access network device, and the terminal device will execute steps 408, 409, 410, and 411, that is... Figure 4A Option 2 is shown.

[0256] Step 408: The management node determines the information of the AI ​​transmitter service unit to be deployed on the access network device.

[0257] Step 409: The management node sends the topology information of the AI ​​transmitter service unit to the access network device; correspondingly, the access network device receives the topology information from the management node's AI transmitter service unit.

[0258] Optionally, the management node sends the third information of the AI ​​transmitter service unit to the access network device; correspondingly, the access network device receives the third information from the management node's AI transmitter service unit.

[0259] Step 410: The access network device determines the information of the AI ​​receiver service unit to be deployed on the terminal device.

[0260] Step 411: The access network device sends the topology information of the AI ​​receiver service unit to the terminal device; correspondingly, the terminal device receives the topology information of the AI ​​receiver service unit from the access network device.

[0261] Optionally, the access network device sends the third information of the AI ​​receiver service unit to the terminal device; correspondingly, the terminal device receives the third information of the AI ​​receiver service unit from the access network device.

[0262] In this embodiment, the difference between Option 2 (i.e., steps 408 to 411) and Option 1 (i.e., steps 405 to 407 above) is that in Option 2, the access network device can collect the first information of the terminal device and has the ability to deploy service units for the terminal device. Therefore, the access network device can determine the information of the AI ​​receiver service unit to be deployed on the terminal device based on the first information of the terminal device.

[0263] After the access network device executes step 409, it will execute step 412. After the terminal device executes step 411, it will execute step 413.

[0264] Step 412: Deploy the AI ​​transmitter service unit on the access network equipment.

[0265] For example, such as Figure 4B As shown, the access network device connects to the deployed service units of the access network device based on the topology information of the received AI transmitter service unit, thereby realizing the deployment of the AI ​​transmitter service unit.

[0266] Step 413: Deploy the AI ​​receiver service unit on the terminal device.

[0267] For example, such as Figure 4B As shown, the terminal device deploys the AI ​​decoding service unit based on the information received from the AI ​​decoding service unit, and connects the service units already deployed on the terminal device based on the topology information received from the AI ​​receiver service unit, thereby realizing the deployment of the AI ​​receiver service unit.

[0268] In this embodiment, the management node can deploy AI transmitter service units on the access network devices and AI receiver service units on the terminal devices, enabling intelligent transmission and reception of data between the access network devices and the terminal devices. This improves the flexibility of data transmission and reception between the access network devices and the terminal devices.

[0269] Furthermore, this application embodiment also provides a communication system, which includes a management node and at least one network element. The management node can deploy service units with AI functions in the aforementioned at least one network element, enabling the service units deployed in the aforementioned at least one network element to achieve online model training, thereby improving the flexibility of model training. The aforementioned network element can be an access network device, a terminal device, an over-the-top (OTT) server, a simulator with digital twin functionality, or a large computer. Specifically, to support native AI training, this embodiment proposes defining a logical gradient layer in the basic service unit, that is, configuring the function of supporting back gradient calculation in the information of the basic service unit. When each basic service unit included in the target service unit contains a logical gradient layer, that is, when each basic service unit supports back gradient calculation, the target service unit can support online real-time data training. For ease of understanding, the following is combined with... Figure 5 The process of implementing online model training in the service unit is described below:

[0270] like Figure 5 As shown, the communication system includes a management node and at least one network element. The at least one network element includes a first network element, which deploys a model training service unit 01. Specifically, the first network element is used to receive a model training request from the management node, the model training request including functional parameters used by the model training service unit 01 when training the model; furthermore, the first network element is also used to perform model training on the training data based on the functional parameters by the model training service unit 01 to obtain a target model; furthermore, the first network element is also used to send a model training response to the management node, the model training response including the target model output by the model training service unit 01.

[0271] Optional functional parameters include: training hyperparameters, training loss function, and training stopping condition. For an explanation of functional parameters, please refer to the introduction of functional parameters in step 301 above; it will not be repeated here.

[0272] In one possible implementation, the model training service unit 01 includes an encoder service unit 011 and a decoder service unit 012. The encoder service unit 011 and decoder service unit 012 can be deployed in the same network element, for example, in the aforementioned first network element; or, the encoder service unit 011 and decoder service unit 012 can be deployed in different network elements, for example, the encoder service unit 011 is deployed in a second network element, and the decoder service unit 012 is deployed in a third network element. This embodiment is not limited to this.

[0273] The following example illustrates the deployment of encoder service unit 011 and decoder service unit 012 in the second and third network elements, respectively. Specifically, the second network element sends a call request to the third network element, which requests the encoder service unit 011 and decoder service unit 012 to perform model training. For example, encoder service unit 011 in the second network element sends a call request to decoder service unit 012 in the third network element, which triggers model training. Furthermore, the second network element also sends forward response information to the third network element. This forward response information is used by the encoder service unit to instruct the decoder service unit on the features generated during the forward propagation of model training. The third network element sends backward response information to the second network element. This backward response information corresponds to the forward response information and is used by the decoder service unit to instruct the encoder service unit on the gradients generated during the backward propagation of model training. For example, the encoder service unit 011 in the second network element generates forward response information based on the training data. The second network element sends the forward response information to the third network element through the interface between network elements. The decoder service unit 012 of the third network element generates backward response information based on the training data. The third network element sends the backward response information to the encoder service unit 011 through the interface between network elements, so that the encoder service unit 011 and the decoder service unit 012 can output the target model through model training.

[0274] In one possible implementation, a network element in the communication system is further equipped with a data retrieval service unit 013. For example, at least one network element in the communication system further includes a fourth network element, which is equipped with the data retrieval service unit 013. It should be noted that the fourth network element may be included in the first network element, or it may be included in the second or third network element, or it may be independent of the first, second, or third network element. Specifically, the data retrieval service unit 013 is used to output training data, which includes first training data and second training data. The fourth network element is used to send the first training data to the second network element, and the first training data is used by the encoder service unit 011 to generate forward response information; the fourth network element is also used to send the second training data to the third network element, and the second training data is used by the decoder service unit 012 to generate reverse response information.

[0275] It should be noted that the first training data can be periodically sent by the data retrieval service unit 013 to the encoder service unit 011, or it can be sent by the data retrieval service unit 013 after receiving a request from the encoder service unit 011; this embodiment is not limited. The second training data can be periodically sent by the data retrieval service unit 013 to the decoder service unit 012, or it can be sent by the data retrieval service unit 013 after receiving a request from the encoder service unit 011; this embodiment is not limited.

[0276] In one possible implementation, a network element in the communication system further deploys a performance monitoring service unit 02. For example, at least one network element in the communication system also includes a fifth network element, which deploys the performance monitoring service unit 02 for monitoring the performance of the target model. It should be noted that the fifth network element may be included in the first network element or may be independent of the first network element; this embodiment is not limited to this.

[0277] Specifically, the first network element is also used to send a performance monitoring request to the fifth network element. The performance monitoring request includes a monitoring period and monitoring parameters. The monitoring parameters are used by the performance monitoring service unit to determine whether the target model meets the training stop condition. The fifth network element is used to send a first instruction message to the first network element when the training stop condition is met. The first instruction message is used to instruct the model training service unit to stop model training.

[0278] In one possible implementation, at least one network element further includes a sixth network element, which is equipped with a model storage service unit. The sixth network element is used to receive the target model output by the model training service unit, and the model storage service unit is used to store the target model.

[0279] In one possible implementation, the fourth network element is also equipped with a data storage service unit, which is used to store training data.

[0280] In this embodiment, since the service units deployed by the management node in at least one network element all support inverse gradient calculation, after the topological connection between the various service units is established, the various service units cooperate with each other to achieve online model training. This is beneficial to improving the flexibility of model training in the communication system and to achieving flexible, scalable, and evolvable air interface services.

[0281] like Figure 6 The diagram shown is a structural schematic of a communication device 60 provided in this embodiment. It should be understood that the aforementioned... Figure 3 The first network element or the second network element in the corresponding method embodiment can be based on this embodiment. Figure 6 The structure of the communication device 60 shown. (The aforementioned...) Figure 4A The management node, access network device, or terminal device in the corresponding method embodiment can be based on this embodiment. Figure 6 The structure of the communication device 60 shown.

[0282] like Figure 6 As shown, the communication device 60 may include a processor 601, a memory 603, and a communication interface 602. The processor 601 is coupled to the memory 603, and the processor 601 is coupled to the communication interface 602.

[0283] The aforementioned communication interface 602 is connected to other communication devices via a communication link. For example, the communication interface 602 can be a network interface between the access network device and the network interface, such as the S1 interface.

[0284] The aforementioned processor 601 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. Processor 601 can refer to a single processor or include multiple processors; the specific meaning is not limited here.

[0285] Furthermore, the aforementioned memory 603 is primarily used to store software programs and data. Memory 603 can exist independently and be connected to processor 601. Optionally, memory 603 can be integrated with processor 601, for example, integrated within one or more chips. Memory 603 can store program code that executes the technical solutions of the embodiments of this application, and its execution is controlled by processor 601. The various types of computer program code being executed can also be considered as drivers for processor 601. Memory 603 can include volatile memory, such as random-access memory (RAM); memory can also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory 603 can also include combinations of the above types of memory. Memory 603 can refer to a single memory or can include multiple memories. For example, memory 603 is used to store various types of data, such as service unit type information. For details, please refer to the relevant descriptions in the previous embodiments; they will not be repeated here.

[0286] In one design, the communication device 60 is used to perform the aforementioned... Figure 3 The method corresponding to the first network element in the embodiment. The first network element can be a management node or an access network device. Specifically, the communication interface 602 is used to acquire first information of the second network element, the first information including information on at least one service unit already deployed by the second network element, the service unit being used to implement one or more functions of the access network device or terminal device; the processor 601 is used to determine information on the target service unit to be deployed by the second network element, the information on the target service unit including second information and topology information, the second information including information on multiple service units constituting the target service unit, the topology information being used to indicate the connection topology between the multiple service units constituting the target service unit; the communication interface 602 is also used to send third information and topology information to the second network element, the third information being determined based on the first information and the second information, the third information and the topology information being used to instruct the second network element to deploy the target service unit.

[0287] In one possible implementation, the communication interface 602 is used to send a first message to the second network element, the first message being used to request information on the service units deployed by the second network element; and to receive a second message from the second network element, the second message including the first information.

[0288] In another design, the communication device 60 is used to perform the aforementioned... Figure 3 The method corresponding to the second network element in the embodiment. The second network element can be an access network device or a terminal device. The communication interface 602 is used to send first information to the first network element, the first information including information on at least one service unit deployed by the second network element, the service unit being used to implement one or more functions of the access network device or the terminal device; and to receive third information and topology information from the first network element, the third information being determined based on the first information and the second information, the second information including information on multiple service units constituting the target service unit, and the topology information being used to indicate the connection topology between the multiple service units constituting the target service unit; the processor 601 is used to deploy the target service unit based on the third information, the first information, and the topology information.

[0289] In one possible implementation, the communication interface 602 is specifically used to receive a first message from a first network element, the first message being used to request information on the service units deployed by the second network element; and to send a second message to the first network element, the second message including the first information.

[0290] In one possible implementation of the two aforementioned designs, the first message includes service unit type information, which indicates the type of service unit queried through the first message; the service unit corresponding to the service unit information contained in the first message is of the same type as the service unit indicated by the service unit type information in the first message. Optionally, the service unit type information includes: first type information and / or second type information, where the first type information indicates a first type service unit and the second type information indicates a second type service unit; wherein, a second type service unit consists of at least two different first type service units.

[0291] In one possible implementation of the two designs described above, the information of the service unit includes the information of the first type of service unit;

[0292] The information for the first type of service unit includes at least one of the following:

[0293] Identification information, used to uniquely identify the first type of service unit; or,

[0294] Functional description information, used to describe the functions of the first type of service unit; or,

[0295] Input / output information, used to indicate the inputs and outputs of the first type of service unit; or,

[0296] Equipment capability requirement information, used to indicate the software and / or hardware capability requirements of the equipment deploying the first type of service unit; or,

[0297] Gradient computation capability information, used to indicate whether the first type of service unit supports backward gradient computation; or...

[0298] Executable file information.

[0299] In one possible implementation of the two aforementioned designs, the input / output information includes input identification information and output identification information; the input / output information also includes at least one of the following:

[0300] Input format information, used to indicate the format of the input data for the first type of service unit; or,

[0301] Output format information, used to indicate the format of the output data of the first type of service unit; or...

[0302] Input memory layout information, used to indicate how the input data of the first type of service unit is arranged in memory; or,

[0303] Output memory layout information to indicate how the output data of the first type of service unit is arranged in memory; or...

[0304] Input quantization bit width, used to indicate the quantization bit width of the input data for the first type of service unit; or,

[0305] Output quantization bit width, used to indicate the quantization bit width of the output data of the first type of service unit.

[0306] In one possible implementation of the two designs described above, the information of the service unit includes information of the second type of service unit;

[0307] The information for the second type of service unit includes at least one of the following:

[0308] Identification information, used to uniquely identify the second type of service unit; or,

[0309] Functional description information, used to describe the functions of the second type of service unit; or,

[0310] Input / output information, used to indicate the inputs and outputs of the second type of service unit; or,

[0311] Topology information, used to indicate the connection topology between multiple service units constituting a second type of service unit; or,

[0312] Functional parameter information is used to indicate the functional parameters used by the second type of service unit when implementing service functions.

[0313] In one possible implementation of the two aforementioned designs, the second information includes information about the first service unit constituting the target service unit and information about the second service unit; the gradient computation capability information of the first service unit indicates that the first service unit supports inverse gradient computation, and the target service unit supports inverse gradient computation when the gradient computation capability information of the second service unit indicates that the second service unit supports inverse gradient computation.

[0314] In one possible implementation of the two aforementioned designs, the second message further includes device capability information of the second network element, which is used to indicate the software and / or hardware capabilities of the second network element; the processor 601 is further configured to determine the second information based on the first information and the device capability information of the second network element, the second information including information of the first service unit constituting the target service unit and information of the second service unit, the capabilities indicated by the capability information of the second network element including the capabilities indicated by the device capability requirement information of the first service unit and the capabilities indicated by the device capability requirement information of the second service unit.

[0315] It should be noted that the specific implementation method and beneficial effects of this embodiment can be referred to the method of the first network element or the second network element in the above embodiments, and will not be repeated here.

[0316] like Figure 7As shown, this application also provides a communication device 70. The communication device 70 can be a terminal device, an access network device, or a management node, or it can be a component of the terminal device, access network device, or management node (e.g., an integrated circuit, a chip, etc.). The communication device 70 can also be other communication modules used to implement the methods in the method embodiments of this application.

[0317] The communication device 70 may include a processing module 701 (or processing unit). Optionally, it may also include an interface module 702 (or transceiver unit or transceiver module) and a storage module 703 (or storage unit). The interface module 702 is used to enable communication with other devices. The interface module 702 may be, for example, a transceiver module or an input / output module.

[0318] In one possible design, such as Figure 7 One or more modules may be implemented by one or more processors, or by one or more processors and memory; or by one or more processors and transceivers; or by one or more processors, memory, and transceivers. This application does not limit the implementation in this way. The processors, memory, and transceivers can be configured individually or integrated into one unit.

[0319] The communication device 70 has the functions of the terminal device described in the embodiments of this application. For example, the communication device 70 includes modules, units, or means corresponding to the steps involved in the terminal device described in the embodiments of this application. These functions, units, or means can be implemented by software, hardware, or hardware executing corresponding software, or a combination of software and hardware. Further details can be found in the corresponding descriptions in the foregoing method embodiments. Please refer to the preceding text for specific details. Figure 6 The implementation method of the communication device 60 in the corresponding embodiment to realize the function of the second network element.

[0320] Alternatively, the communication device 70 may have the functions of the access network device described in the embodiments of this application. For example, the communication device 70 includes modules, units, or means corresponding to the steps involved in the access network device described in the embodiments of this application. These functions, units, or means can be implemented in software, hardware, or hardware executing corresponding software, or a combination of software and hardware. Further details can be found in the corresponding descriptions in the foregoing method embodiments. Please refer to the preceding text for specific details. Figure 6 The corresponding embodiment describes how the communication device 60 implements the function of the first network element or the second network element.

[0321] Alternatively, the communication device 70 may have the functionality of a management node as described in the embodiments of this application. For example, the communication device 70 may include modules, units, or means corresponding to the steps involved in the management node described in the embodiments of this application. These functions, units, or means may be implemented in software, hardware, or hardware executing corresponding software, or a combination of software and hardware. Further details can be found in the corresponding descriptions in the foregoing method embodiments. Please refer to the preceding text for specific details. Figure 6 The corresponding embodiment describes an implementation method for the communication device 60 to perform the function of the first network element.

[0322] Some or all of the steps of the communication method in the embodiments of this application can be implemented by a graphics processing unit (GPU) or a neural network processing unit (NPU), or by a GPU or NPU in conjunction with other processors.

[0323] Furthermore, this application provides a computer program product comprising one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. For example, implementing the aforementioned... Figure 3 The methods related to the first network element. For example, implementing the methods described above. Figure 3 Methods related to the second network element. For example, implementing the methods described above. Figure 4A Methods related to terminal devices. For example, implementing the methods described above. Figure 4A Methods related to access network devices. For example, implementing the methods described above. Figure 4AThe method relates to the management node in the system. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can store, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, magnetic tape), optical media (e.g., digital versatile disc (DVD)), or semiconductor media (e.g., solid-state disk (SSD)).

[0324] Furthermore, this application also provides a computer-readable storage medium storing a computer program that is executed by a processor to perform the aforementioned functions. Figure 3 Methods related to the first network element in the network.

[0325] Furthermore, this application also provides a computer-readable storage medium storing a computer program that is executed by a processor to perform the aforementioned functions. Figure 4A Methods related to the second network element in the network.

[0326] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0327] Those skilled in the art will clearly 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.

Claims

1. A service configuration method, executed by a first network element or a chip within the first network element, characterized in that, include: The first network element obtains first information from the second network element, the first information including information on at least one service unit deployed by the second network element, the service unit being used to implement one or more functions of the access network device or terminal device; The first network element determines the information of the target service unit to be deployed by the second network element. The information of the target service unit includes second information and topology information. The second information includes information of multiple service units constituting the target service unit. The topology information is used to indicate the connection topology between the multiple service units constituting the target service unit. The first network element sends third information and the topology information to the second network element. The third information is determined based on the first information and the second information. The third information and the topology information are used to instruct the second network element to deploy the target service unit.

2. The method according to claim 1, characterized in that, The first network element obtains the first information of the second network element, including: The first network element sends a first message to the second network element, the first message being used to request information on the service units deployed by the second network element; The first network element receives a second message from the second network element, the second message including the first information.

3. The method according to claim 2, characterized in that, The first message includes service unit type information, which indicates the type of service unit queried through the first message; the service unit corresponding to the service unit information in the first message is of the same type as the service unit indicated by the service unit type information in the first message.

4. The method according to claim 3, characterized in that, The type information of the service unit includes: first type information and / or second type information, wherein the first type information is used to indicate a first type service unit and the second type information is used to indicate a second type service unit; wherein, a second type service unit consists of at least two different first type service units.

5. The method according to any one of claims 1 to 4, characterized in that, The information of the service unit includes information of the first type of service unit; The information for the first type of service unit includes at least one of the following: Identification information, used to uniquely identify the first type of service unit; or, Functional description information, used to describe the function of the first type of service unit; or, Input / output information, used to indicate the inputs and outputs of the first type of service unit; or, Equipment capability requirement information, used to indicate the software and / or hardware capability requirements of the equipment deploying the first type of service unit; or, Gradient computation capability information, used to indicate whether the first type of service unit supports inverse gradient computation; or... Executable file information.

6. The method according to claim 5, characterized in that, The input / output information includes input identification information and output identification information; the input / output information also includes at least one of the following: Input format information, used to indicate the format of the input data for the first type of service unit; or, Output format information, used to indicate the format of the output data of the first type of service unit; or... Input memory layout information, used to indicate how the input data of the first type of service unit is arranged in memory; or, Output memory layout information to indicate how the output data of the first type of service unit is arranged in memory; or... Input quantization bit width, used to indicate the quantization bit width of the input data of the first type of service unit; or, Output quantization bit width, used to indicate the quantization bit width of the output data of the first type of service unit.

7. The method according to any one of claims 1 to 6, characterized in that, The information of the service unit includes information of the second type of service unit; The information for the second type of service unit includes at least one of the following: Identification information, used to uniquely identify the second type of service unit; or, Functional description information, used to describe the function of the second type of service unit; or, Input / output information, used to indicate the inputs and outputs of the second type of service unit; or, Topology information, used to indicate the connection topology between multiple service units constituting the second type of service unit; or, Functional parameter information is used to indicate the functional parameters used by the second type of service unit when implementing service functions.

8. The method according to claim 5, characterized in that, The second information includes information about the first service unit and the second service unit that constitute the target service unit; If the gradient computation capability information of the first service unit indicates that the first service unit supports inverse gradient computation, and if the gradient computation capability information of the second service unit indicates that the second service unit supports inverse gradient computation, then the target service unit supports inverse gradient computation.

9. The method according to claim 5, characterized in that, The second message also includes device capability information of the second network element, which is used to indicate the software and / or hardware capabilities of the second network element; The first network element determines the information of the target service unit to be deployed by the second network element, including: The first network element determines the second information based on the first information and the equipment capability information of the second network element. The second information includes information of the first service unit constituting the target service unit and information of the second service unit. The capability indicated by the capability information of the second network element includes the capability indicated by the equipment capability requirement information of the first service unit and the capability indicated by the equipment capability requirement information of the second service unit.

10. The method according to any one of claims 4 to 9, characterized in that, The first type of service unit includes a radio frequency service unit, a physical layer service unit, or a media access control (MAC) layer service unit.

11. The method according to any one of claims 4 to 9, characterized in that, The second type of service unit includes a model management service unit, a data management service unit, a performance monitoring service unit, a perception service unit, a computing service unit, a model storage service unit, or a data storage service unit.

12. The method according to any one of claims 1 to 11, characterized in that, The first network element is a management node, and the second network element is an access network device; or... The first network element is a management node, and the second network element is a terminal device; or, The first network element is an access network device, and the second network element is a terminal device.

13. A service configuration method, executed by a second network element or a chip within the second network element, characterized in that, include: The second network element sends first information to the first network element. The first information includes information about at least one service unit deployed by the second network element. The service unit is used to implement one or more functions of the access network device or terminal device. The second network element receives third information and topology information from the first network element. The third information is determined based on the first information and the second information. The second information includes information about multiple service units constituting the target service unit. The topology information is used to indicate the connection topology between the multiple service units constituting the target service unit. The second network element deploys the target service unit based on the third information, the first information, and the topology information.

14. The method according to claim 13, characterized in that, The second network element sends first information to the first network element, including: The second network element receives a first message from the first network element, the first message being used to request information on the service units deployed by the second network element; The second network element sends a second message to the first network element, the second message including the first information.

15. The method according to claim 14, characterized in that, The first message includes service unit type information, which indicates the type of service unit queried through the first message; the service unit corresponding to the service unit information in the first message is of the same type as the service unit indicated by the service unit type information in the first message.

16. The method according to claim 15, characterized in that, The type information of the service unit includes: first type information and / or second type information, wherein the first type information is used to indicate a first type service unit and the second type information is used to indicate a second type service unit; wherein, a second type service unit consists of at least two different first type service units.

17. The method according to any one of claims 13 to 16, characterized in that, The information of the service unit includes information of the first type of service unit; The information for the first type of service unit includes at least one of the following: Identification information, used to uniquely identify the first type of service unit; or, Functional description information, used to describe the function of the first type of service unit; or, Input / output information, used to indicate the inputs and outputs of the first type of service unit; or, Equipment capability requirement information, used to indicate the software and / or hardware capability requirements of the equipment deploying the first type of service unit; or, Gradient computation capability information, used to indicate whether the first type of service unit supports inverse gradient computation; or... Executable file information.

18. The method according to claim 17, characterized in that, The input / output information includes input identification information and output identification information; the input / output information also includes at least one of the following: Input format information, used to indicate the format of the input data for the first type of service unit; or, Output format information, used to indicate the format of the output data of the first type of service unit; or... Input memory layout information, used to indicate how the input data of the first type of service unit is arranged in memory; or, Output memory layout information to indicate how the output data of the first type of service unit is arranged in memory; or... Input quantization bit width, used to indicate the quantization bit width of the input data of the first type of service unit; or, Output quantization bit width, used to indicate the quantization bit width of the output data of the first type of service unit.

19. The method according to any one of claims 13 to 18, characterized in that, The information of the service unit includes information of the second type of service unit; The information for the second type of service unit includes at least one of the following: Identification information, used to uniquely identify the second type of service unit; or, Functional description information, used to describe the function of the second type of service unit; or, Input / output information, used to indicate the inputs and outputs of the second type of service unit; or, Topology information, used to indicate the connection topology between multiple service units constituting the second type of service unit; or, Functional parameter information is used to indicate the functional parameters used by the second type of service unit when implementing service functions.

20. The method according to claim 19, characterized in that, The second information includes information about the first service unit and the second service unit that constitute the target service unit; If the gradient computation capability information of the first service unit indicates that the first service unit supports inverse gradient computation, and if the gradient computation capability information of the second service unit indicates that the second service unit supports inverse gradient computation, then the target service unit supports inverse gradient computation.

21. The method according to claim 19, characterized in that, The second message also includes device capability information of the second network element, which is used to indicate the software and / or hardware capabilities of the second network element; The target service unit includes a first service unit and a second service unit. The capability indicated by the capability information of the second network element includes the capability indicated by the equipment capability requirement information of the first service unit and the capability indicated by the equipment capability requirement information of the second service unit.

22. The method according to any one of claims 16 to 21, characterized in that, The first type of service unit includes a radio frequency service unit, a physical layer service unit, or a media access control (MAC) layer service unit.

23. The method according to any one of claims 16 to 21, characterized in that, The second type of service unit includes a model management service unit, a data management service unit, a performance monitoring service unit, a perception service unit, a computing service unit, a model storage service unit, or a data storage service unit.

24. The method according to any one of claims 13 to 23, characterized in that, The first network element is a management node, and the second network element is an access network device; or... The first network element is a management node, and the second network element is a terminal device; or, The first network element is an access network device, and the second network element is a terminal device.

25. A communication system, characterized in that, include: The system includes a management node and at least one network element, wherein the at least one network element includes a first network element, and the first network element is deployed with a model training service unit. The first network element is used to receive a model training request from the management node, the model training request including the functional parameters used by the model training service unit when training the model; The first network element is also used to perform model training on the training data based on the functional parameters through the model training service unit to obtain the target model; The first network element is further configured to send a model training response to the management node, the model training response including the target model output by the model training service unit.

26. The communication system according to claim 25, characterized in that, The functional parameters include: training hyperparameters, training loss function, and training stopping condition.

27. The communication system according to claim 25 or 26, characterized in that, The first network element includes a second network element and a third network element, and the model training service unit includes an encoder service unit and a decoder service unit. The second network element is equipped with the encoder service unit, and the third network element is equipped with the decoder service unit. The second network element is used to send a call request to the third network element, which is used to request the encoder service unit and the decoder service unit to perform model training; The second network element is also used to send forward response information to the third network element, wherein the forward response information is used by the encoder service unit to instruct the decoder service unit on the features generated during the forward propagation of model training; The third network element is used to send reverse response information to the second network element. The reverse response information corresponds to the forward response information. The reverse response information is used by the decoder service unit to instruct the encoder service unit on the gradient generated during the backpropagation process of model training.

28. The communication system according to claim 27, characterized in that, The at least one network element further includes a fourth network element, the fourth network element being deployed with a data retrieval service unit, the data retrieval service unit being used to output training data, the training data including first training data and second training data; The fourth network element is further configured to send first training data to the second network element, wherein the first training data is used by the encoder service unit to generate the forward response information; The fourth network element is also used to send second training data to the third network element, and the second training data is used by the decoder service unit to generate the reverse response information.

29. The communication system according to claim 25 or 26, characterized in that, The at least one network element further includes a fifth network element, wherein the fifth network element is equipped with a performance monitoring service unit, and the performance monitoring service unit is used to monitor the performance of the target model. The first network element is also used to send a performance monitoring request to the fifth network element. The performance monitoring request includes a monitoring period and monitoring parameters. The monitoring parameters are used by the performance monitoring service unit to determine whether the target model meets the training stopping condition. The fifth network element is used to send a first instruction message to the first network element when the training stop condition is met. The first instruction message is used to instruct the model training service unit to stop model training.

30. The communication system according to any one of claims 25 to 29, characterized in that, The at least one network element further includes a sixth network element, wherein the sixth network element is deployed with a model storage service unit; The sixth network element is used to receive the target model output by the model training service unit, and the model storage service unit is used to store the target model.

31. The communication system according to claim 28, characterized in that, The fourth network element is also equipped with a data storage service unit, which is used to store the training data.

32. A communication device, characterized in that, Including processor and memory; The memory stores computer programs; The processor invokes the computer program to cause the communication device to perform the method as described in any one of claims 1 to 12.

33. A communication device, characterized in that, Including processor and memory; The memory stores computer programs; The processor invokes the computer program to cause the communication device to perform the method as described in any one of claims 13 to 24.

34. A computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method as claimed in any one of claims 1 to 12; or, to perform the method as claimed in any one of claims 13 to 24.

Images