Data processing method and related equipment

Abstract

The invention provides a data processing method and related equipment, and is applied to the field of communication. The method is applied to a first network element, the first network element provides a first AI service, in the method, the first network element receives a first request message, and the first request message indicates to call the first AI service; the first network element determines, according to the first request message, a second network element on which the first AI service depends; wherein the input data of the target model for realizing the first AI service has a first constraint; the first constraint comprises a data type constraint; the second network element provides a second AI service, and the output result of the second AI service meets the first constraint; the first network element calls a second AI service of a second network element to obtain first data; and the first network element processes the first data through the target model to obtain a first output result. According to the embodiment of the invention, the efficiency of the AI service can be improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a data processing method and related equipment. Background Art

[0002] With the development of artificial intelligence technology, artificial intelligence (AI) based on machine learning is increasingly widely used in communication networks. In a communication network, the device that provides AI services is called an AI service provider, and the device that calls AI services is called an AI service consumer. The AI ​​service consumer can obtain the prediction output of unknown attributes through the AI ​​service of the AI ​​service provider. After obtaining the prediction output of the above unknown attributes, the AI ​​service consumer determines whether certain measures need to be taken according to its own business logic to achieve specific business purposes. For example, in order to realize the intelligence of 5G mobile networks, a network data analysis function (NWDAF) network element is defined. As an AI service provider, NWDAF predicts the output results including predictions of future data. As an AI service consumer, the processing service network element adjusts the network according to the results of the NWDAF prediction output to optimize the operation of the network. For example, according to the prediction output results output by NWDAF, the service network element determines that the packet loss rate of the message transmission of the terminal's ongoing service exceeds the threshold, which will affect the service experience of the user using the terminal, so the service network element will take corresponding measures, such as increasing the proportion of retransmission packets.

[0003] In the communication network, the core network and access network, as well as the user plane, control plane and management plane constitute different domains respectively. These domains can have their own AI service providers. In order to better meet business needs, AI service consumers in one domain sometimes need to call the AI ​​services of AI service providers in another domain. An existing method of calling AI services located in other domains is that some domains are configured with exposure functions (EF), and some domains are configured with AI agents. When an AI service consumer needs to call an AI service located in a different domain, the AI ​​service consumer sends a request to the AI ​​agent in its domain, and the AI ​​agent calls the AI ​​service of the AI ​​service provider in the domain where the EF is located through the EF of other domains.

[0004] However, the AI ​​agent in the prior art only calls the AI ​​service of the AI ​​service provider, but cannot realize the collaboration between multiple AI service providers, which may cause low efficiency of AI services. Therefore, when there are multiple AI service providers in the communication system, how to realize the collaboration between AI service providers to improve the efficiency of AI services is an urgent problem to be solved. Summary of the invention

[0005] The embodiments of the present application provide a data processing method and related equipment, which are applied in the field of communications, and are used to achieve collaboration between AI service providers, thereby improving the efficiency of AI services.

[0006] In a first aspect of the present application, a data processing method is provided, which is executed by a first network element, or the method is executed by some components (such as a processor, a chip or a chip system) in the first network element, or the method can also be implemented by a logic module or software that can implement all or part of the functions of the first network element. In the first aspect and its possible implementation, the data processing method is described as being executed by the first network element, and the first network element can be a terminal device or a network device. In the method, the first network element, as an AI service provider, can provide a first AI service. The first network element can implement the first AI service through a target model, and the input data of the target model needs to meet a first constraint, and the first constraint includes: a data type constraint; the first network element receives a first request message indicating that it calls the first AI service. The first network element determines the second network element required to implement the first AI service according to the first request message. The second network element, as an AI service provider, can provide a second AI service, and the output result of the second AI service meets the above-mentioned first constraint. The first network element indicates to call the second AI service of the second network element to obtain the first data, and then the first network element processes the first data through the target model to obtain the first output result.

[0007] Based on the above technical solution, the first network element and the second network element are both AI service providers. The first network element needs to implement the first AI service through the target model, and the input data of the target model needs to meet the first constraint. The first network element needs to rely on the second network element to provide the second AI service to implement the first AI service, and the output result of the second AI service meets the first constraint. Therefore, the first data obtained by the second AI service can be directly processed through the target model to obtain the first output result of the target model, thereby enabling the AI ​​service provider to make full use of the AI ​​services of other AI service providers to assist in the implementation of its own AI services, which improves the efficiency of AI services.

[0008] In a possible implementation manner of the first aspect, the second network element may determine the second network element on which the first AI service depends in the following manner: the first network element sends a first constraint to an index library, the index library determines indication information of the second network element based on the first constraint, and sends it to the first network element, and correspondingly, the first network element receives the indication information of the second network element sent by the index library.

[0009] Based on the above technical solution, the first network element obtains the indication information of the second network element that meets the first constraint through the index library, so that it can communicate with the second network element according to the indication information.

[0010] Optionally, the indication information of the second network element may be access address information of the second network element, or a network element identifier of the second network element, or access address information of an AI service of the second network element.

[0011] Optionally, the second network element may also determine the second network element on which the first AI service depends in the following manner: a storage repository is configured in the first network element, and the first network element determines the indication information of the second network element in the storage repository according to the first constraint.

[0012] Optionally, the index library may also send to the first network element: the granularity supported by the output result of the second AI service registered by the second network element in the index library.

[0013] Based on the above technical solution, since the granularity supported by the output result of the second AI service is not necessarily the same as the first constraint, for example, the first constraint is that the time granularity of the data is s, while the supported granularity is that the time granularity of the data is s, and the spatial granularity is the cell. The granularity information included in the supported granularity may be more comprehensive than the granularity information included in the first constraint. Therefore, by sending the granularity supported by the second AI service to the first network element, the first network element can have a more comprehensive understanding of the granularity characteristics of the output result of the second AI service.

[0014] Optionally, before the first network element receives the first request message, the AI ​​service consumer or AI service provider determines the first AI service to be called according to the business to be executed, and the constraint conditions that the first AI service needs to meet, and the AI ​​service consumer or AI service provider sends a request message to the index library, the request message carries the query condition, the query condition includes: the identifier of the first AI service and the constraint condition; and then the index library determines the first network element that meets the query condition according to the query condition. At this time, the index library can directly send the first request message to the first network element to request to call the first AI service of the first network element, or the index library can first send a response message to the AI ​​service consumer or AI service provider, the response message carries the indication information of the first network element, and after receiving the response message, the AI ​​service consumer or AI service provider sends the first request message to the first network element to request to call the first AI service of the first network element. Optionally, the indication information of the first network element can be the access address information of the first network element, or the network element identifier of the first network element, or the access address information of the AI ​​service of the first network element.

[0015] In a possible implementation manner of the first aspect, the first constraint includes: a data granularity constraint, and the data granularity constraint is carried in the first request message.

[0016] Based on the above technical solution, since the inference prediction of the model is closely related to the autocorrelation of the data, and the autocorrelation of the data is closely related to the granularity of the data, the granularity of the data will affect the inference prediction of the model. In this implementation, the first constraint is defined to include the data granularity constraint, which also defines that the output result of the second AI service needs to meet the granularity requirements of the target model, and the target model performs data inference prediction by processing the output result of the second AI service. Therefore, this limitation is conducive to improving the accuracy of the model data inference prediction.

[0017] Based on the above technical solution, the first request message is used to indicate the call of the first AI service of the first network element, and the first request message carries the data granularity constraint. Therefore, the first network element can adaptively call the AI ​​service that meets the data granularity constraint according to the first request message, so that the AI ​​service of the called first network element is more in line with the needs of the sender of the first request message.

[0018] Optionally, when the data is used to describe some features required to implement the first AI service and the data can help the model make better predictions, the data granularity constraint is that the data is greater than or equal to the first granularity; when the data is current data that needs to be collected and is used to input the model to predict some attributes, the data granularity constraint is that the data is less than or equal to the first granularity.

[0019] Based on the above technical solution, selecting data with a larger granularity as data for describing features can achieve the determination of some features from a larger granularity while eliminating the noise caused by data with too small a granularity; selecting data with a smaller granularity as the current data for the input model can make the data more refined, which is conducive to more accurate prediction of some attributes.

[0020] Optionally, the data granularity constraint includes: time granularity constraint, space granularity constraint, and object granularity constraint. Among them, the time granularity constraint indicates the time span covered by the data, or the update cycle; the space granularity constraint indicates the scope of the network space to which the data is applicable; and the object granularity constraint indicates the scope of objects covered by the data.

[0021] In a possible implementation manner of the first aspect, the first network element sends a second request message to the second network element, where the second request message indicates calling a second AI service of the second network element, and the second request message carries a data granularity constraint.

[0022] Based on the above technical solution, since the second request message carries the data granularity constraint, the second network element can determine the second AI service that meets the data granularity constraint according to the second request message, so that the first data finally obtained can meet the granularity requirements of the first network element.

[0023] Optionally, the second request message may also carry an identifier of the target object, which is used to instruct the second network element to process data related to the target object. The identifier of the target object may be derived from the first request message, that is, the first request message carries the identifier of the target object to instruct the first network element to predict unknown attributes of the target object; the identifier of the target object may also be determined according to the needs of implementing the first AI service, for example: collecting data of the target object, or requiring features related to the target object.

[0024] Optionally, the second request message may further carry an identifier of a second AI service to indicate calling the second AI service of the second network element.

[0025] In a possible implementation manner of the first aspect, the first constraint includes: a confidence level of the data is not less than a first confidence level.

[0026] Based on the above technical solution, since the accuracy of the model's reasoning and prediction is closely related to the confidence of the input data, based on the above technical solution, the target model performs data reasoning and prediction by processing the output data of the second AI service, that is, the output data of the second AI service is used as the input data of the target model. The first constraint is limited to include the confidence constraint of the data, which means that the output result of the called second AI service must meet the confidence constraint of the target model, and the input data of the model must meet the confidence constraint, which ensures the accuracy of the model data reasoning and prediction to a certain extent.

[0027] In a possible implementation manner of the first aspect, the first network element sends first registration information to the index library, where the first registration information includes: a granularity supported by an output result of the first AI service.

[0028] Based on the above technical solution, the first network element sends the registration information of the first AI service to the index library, where the registration information includes the granularity supported by the output data of the first AI service, so that the index library can determine whether the first AI service meets the granularity constraints given by the AI ​​service consumer or the AI ​​service provider based on the registration information.

[0029] Optionally, when the first network element sends the first registration information to the index library, it can also send indication information of the first network element at the same time, so that when it is determined that the first network element meets the constraints sent by the AI ​​service provider or the AI ​​service consumer, the index library can then determine the indication information of the first network element and send the indication information to the AI ​​service provider or the AI ​​service consumer.

[0030] In a possible implementation manner of the first aspect, the first registration information further includes: a confidence level supported by an output result of the first AI service.

[0031] Based on the above technical solution, the registration information includes the confidence level supported by the output data of the first AI service, so that the index library can determine whether the first AI service meets the confidence level constraint given by the AI ​​service consumer based on the registration information.

[0032] Optionally, the first registration information further includes: the granularity of the input data of the first AI service. Specifically, the input data may be data used to describe some features required to implement the first AI service, and the input data may also be current data that needs to be collected.

[0033] Optionally, the first registration information further includes: an identifier of the first AI service, or a data type of output data of the first AI service.

[0034] Optionally, the first registration information also includes: the precision, recall, and F1 value supported by the target model on which the first AI service depends.

[0035] In a possible implementation of the first aspect, the first request message carries a second constraint, where the second constraint is a granularity constraint on an output result of the first AI service. After receiving the first request message, the first network element determines, based on the first request message, from multiple models a target model whose output satisfies the second constraint.

[0036] Based on the above technical solution, the first request message carries the second constraint, which actually indicates that the first network element outputs a result that satisfies the second constraint, and the first network element selects a target model that outputs a result that satisfies the second constraint. The first network element implements the first AI service through the target model, so the output result of the first AI service meets the requirements of the sender of the first request message.

[0037] Optionally, the AI ​​service consumer or AI service provider determines a target object for which unknown attributes need to be predicted based on the business to be executed, and carries an identifier of the target object in the first request message.

[0038] Optionally, the target model needs to satisfy a third constraint, which includes: the precision of the model is not lower than the first precision; the recall rate of the model is not lower than the first recall rate, and the F1 value of the model is not lower than the first score.

[0039] Optionally, when the first AI service is a switching threshold prediction service, the first data includes: a first prediction result of wireless resources of an adjacent cell of the target object, a second prediction result of the movement trajectory of the target object, and a third prediction result of the type of service run by the target object. The first network element determines a target cell whose air interface resource occupancy rate is lower than a threshold from the adjacent cells based on the first prediction result; the first network element adjusts the switching threshold of the target cell based on the second prediction result to obtain a first switching threshold; the first network element adjusts the first switching threshold based on the third prediction result to obtain a second switching threshold; the first network element uses the indication information of the target cell and the second switching threshold as the first output result.

[0040] Based on the above technical solution, the first network element takes the cell whose air interface resource occupancy rate is lower than the threshold as the target cell for switching, so that after switching to the target cell, the air interface resources of the target object are relatively sufficient; the first network element adjusts the switching threshold according to the prediction result of the motion trajectory of the target object, so that the switching threshold adapts to the trajectory of the target object; the first network element adjusts the switching threshold according to the prediction result of the service type of the target object, so that the switching threshold adapts to the service type of the target object.

[0041] Optionally, in order to obtain the above-mentioned first switching threshold, the first network element lowers the switching threshold of the target cell that meets the first condition, and / or increases the switching threshold of the target cell that does not meet the first condition according to the second prediction result, to obtain the first switching threshold, and the first condition is: being located on the motion trajectory indicated by the second prediction result.

[0042] Based on the above technical solution, the first network element can increase the switching threshold of the target cell that is not located on the predicted trajectory of the target object, or lower the switching threshold of the target cell located on the predicted trajectory of the target object according to the prediction result of the motion trajectory of the target object, so that when the signal strength of multiple adjacent cells meets the requirements, they are more inclined to switch to the cell located on the predicted trajectory of the target object.

[0043] Optionally, in order to obtain the above-mentioned second switching threshold, the first network element increases the first switching threshold of the target cell that meets the second condition, and / or lowers the first switching threshold of the first cell that does not meet the second condition according to the third prediction result, and the second condition is: switching to the target cell can make the performance and / or experience of the target service type better than the preset value, and the target service type is the service type indicated by the third prediction result.

[0044] Based on the above technical solution, the first network element can give priority to the target cell that can achieve better performance and / or experience for the service type running the target object according to the prediction result of the service type of the target object, thereby lowering the switching threshold of the target cell, or increasing the switching threshold of other target cells, so that when the signal strength of multiple adjacent cells meets the requirements, they are more inclined to switch to the cell that can achieve better performance and / or experience of the predicted service type.

[0045] The second aspect of the present application provides a data processing method, which is executed by a third network element, or the method is executed by some components in the third network element (such as a processor, a chip or a chip system, etc.), or the method can also be implemented by a logic module or software that can realize all or part of the functions of the third network element. In the first aspect and its possible implementation, the data processing method is described as being executed by a third network element, and the third network element can be a terminal device or a network device. In this method, the network elements are based on the following architecture: the third network element can control the index library, the first network element provides a first AI service, and the input data of the target model used to implement the first AI service has a first constraint, and the first constraint includes: data type constraint; the second network element provides a second AI service, and assists the first network element in implementing the first AI service through the second AI service. Based on the above architecture, the third network element receives the first constraint sent by the first network element; the third network element determines the indication information of the second network element in the index library according to the first constraint; the third network element sends the indication information of the second network element to the first network element, so that the first network element obtains the first data by calling the second AI service of the second network element; the first network element processes the first data through the target model to obtain the first output result.

[0046] Based on the above technical solution, the third network element determines the indication information of the second network element required by the first network element by calling the index library according to the received first constraint, and sends the indication information of the second network element to the first network element, so that the first network element can call the second AI service of the second network element.

[0047] Optionally, the index library may be configured in a third network element, and the index library may also be a server, and the third network element may call this server.

[0048] In a possible implementation manner of the second aspect, the third network element receives first registration information from the first network element, where the first registration information includes: a granularity supported by an output result of the first AI service.

[0049] Optionally, when the third network element receives the first registration information of the first network element, it may also simultaneously receive the indication information of the first network element sent by the first network element, wherein the indication information of the first network element may be the access address information of the first network element, or the network element identifier of the first network element, or the access address information of the AI ​​service of the first network element.

[0050] Based on the above technical solution, the registration information of the first network element includes: the granularity supported by the output data of the first AI service, so that the index library can judge whether the first AI service meets the granularity constraint given by the AI ​​service consumer based on the registration information.

[0051] In a possible implementation manner of the second aspect, the first registration information further includes: a confidence level supported by an output result of the first AI service.

[0052] Based on the above technical solution, the registration information of the first network element includes the confidence supported by the output data of the first AI service, so that the index library can judge whether the first AI service meets the confidence constraint given by the AI ​​service consumer based on the registration information.

[0053] In a possible implementation of the second aspect, before the third network element receives the first constraint sent by the first network element, the third network element receives registration information of an AI network element, where the AI ​​network element includes the second network element, and the registration information of the AI ​​network element includes: data types supported by output results of the AI ​​service provided by the AI ​​network element; the third network element determines, in the index library, indication information of the second network element that satisfies the first constraint based on the registration information of the AI ​​network element.

[0054] It can be understood that the third network element can determine the second network element that meets the first constraint in the AI ​​network element according to the registration information of the AI ​​network element in the index library, and then determine the indication information of the second network element according to the registration information of the second network element.

[0055] Based on the above technical solution, the registration information of the AI ​​network element includes: the data types supported by the output results of the AI ​​service provided by the AI ​​network element. When the first constraint includes a granularity constraint, the third network element can select a second network element that meets the first constraint from these AI network elements based on the registration information, so that the second network element finally selected by the third network element meets the requirements of the first network element.

[0056] Optionally, the registration information of the AI ​​network element also includes: the confidence level supported by the output results of the AI ​​service provided by the AI ​​network element.

[0057] Optionally, the registration information of the AI ​​network element also includes: the granularity supported by the input data of the AI ​​service provided by the AI ​​network element. Specifically, the input data may be data used to describe some features required to implement the AI ​​service, and the input data may also be current data that needs to be collected.

[0058] Optionally, the first registration information also includes: an identifier of the AI ​​service provided by the AI ​​network element.

[0059] Optionally, the registration information of the AI ​​network element also includes: the precision, recall rate, and F1 value supported by the target model on which the AI ​​service of the AI ​​network element depends.

[0060] The third aspect of the present application provides a communication device, which can implement the method in the first aspect or any possible implementation of the first aspect. The device includes corresponding units or modules for executing the above method. The units or modules included in the device can be implemented by software and / or hardware. For example, the device can be a network device, or the device can be a component in the network device (such as a processor, a chip or a chip system, etc.), or the device can also be a logical module or software that can implement all or part of the functions of the network device.

[0061] The device includes a transceiver unit and a processing unit; the transceiver unit is used to receive a first request message, the first request message indicating the call of a first AI service; the processing unit is used to determine, according to the first request message, a second network element on which the first AI service depends; wherein the input data of the target model used to implement the first AI service has a first constraint; the first constraint includes: a data type constraint; the second network element provides a second AI service, and the output result of the second AI service satisfies the first constraint; the first data is obtained by calling the second AI service of the second network element; the first data is processed through the target model to obtain a first output result.

[0062] In a possible implementation of the third aspect, the processing unit is specifically used to: call the transceiver unit to send the first constraint to the index library; obtain the indication information of the second network element obtained according to the first constraint and sent by the index library and received by the transceiver unit.

[0063] In a possible implementation manner of the third aspect, the first constraint includes: a data granularity constraint, and the data granularity constraint is carried in the first request message.

[0064] In a possible implementation manner of the third aspect, the processing unit is specifically used to call the transceiver unit to send a second request message to the second network element, where the second request message carries the data granularity constraint, and the second request message indicates calling the second AI service of the second network element.

[0065] In a possible implementation manner of the third aspect, the first constraint includes: a confidence level of the data is not less than a first confidence level.

[0066] In a possible implementation manner of the third aspect, the transceiver unit is further used to send first registration information to the index library, where the first registration information includes: a granularity supported by an output result of the first AI service.

[0067] In a possible implementation manner of the third aspect, the first registration information also includes: a confidence level supported by an output result of the first AI service.

[0068] In a possible implementation of the third aspect, the first request message carries a second constraint, which is a granularity constraint of an output result obtained by the first AI service, and the processing unit is further used to determine, from multiple models, the target model whose output satisfies the second constraint based on the first request message.

[0069] The third aspect of the present application provides a communication device, which can implement the method in the first aspect or any possible implementation of the first aspect. The device includes corresponding units or modules for executing the above method. The units or modules included in the device can be implemented by software and / or hardware. For example, the device can be a network device, or the device can be a component in the network device (such as a processor, a chip or a chip system, etc.), or the device can also be a logical module or software that can implement all or part of the functions of the network device.

[0070] Among them, the device includes a transceiver unit and a processing unit; the transceiver unit is used to receive a first constraint sent by a first network element, wherein the first network element provides a first AI service; the input data of the target model used to implement the first AI service has a first constraint; the first constraint includes: data type constraint; the processing unit is used to determine the indication information of the second network element in the index library according to the first constraint, the second network element provides a second AI service, and the output result of the second AI service satisfies the first constraint; the transceiver unit is also used to send the indication information of the second network element to the first network element, so that the first network element calls the second AI service of the second network element to obtain the first data; the first network element processes the first data through the target model to obtain the first output result. These modules can also perform the corresponding functions in the above-mentioned second aspect method example, please refer to the detailed description in the method example for details, which will not be repeated here.

[0071] In a possible implementation manner of the fourth aspect, the transceiver unit is further used to receive first registration information of the first network element, where the first registration information includes: a granularity supported by an output result of the first AI service.

[0072] In a possible implementation manner of the fourth aspect, the first registration information also includes: a confidence level supported by an output result of the first AI service.

[0073] In a possible implementation manner of the fourth aspect, the transceiver unit is further used to: receive registration information of an AI network element, the AI ​​network element includes the second network element, and the registration information of the AI ​​network element includes: a data type supported by an output result of an AI service provided by the AI ​​network element; and call the processing unit to determine, in an index library according to the registration information of the AI ​​network element, indication information of the second network element that satisfies the first constraint.

[0074] In a fifth aspect, the present application provides a communication device, which may be the first network element in the above method embodiment or the third network element in the above method embodiment, or a chip set in the first network element or the third network element. The communication device includes a communication interface and a processor, and optionally, also includes a memory. The memory is used to store computer programs or instructions, and the processor is coupled to the memory and the communication interface. When the processor executes the computer program or instruction, the communication device executes the method executed by the first network element or the third network element in the above method embodiment.

[0075] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, which is used to store one or more computer-executable instructions. When the computer-executable instructions are executed by a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect, or the processor executes the method described in the second aspect or any possible implementation of the second aspect.

[0076] The seventh aspect of the embodiments of the present application provides a computer program product (or computer program). When the computer program product is executed by the processor, the processor executes the method of the first aspect or any possible implementation of the first aspect, or the processor executes the method of the second aspect or any possible implementation of the second aspect.

[0077] An eighth aspect of an embodiment of the present application provides a communication system, which includes the communication device of the third aspect and the communication device of the fourth aspect, and / or the communication system includes the communication device of the fifth aspect and the communication device of the sixth aspect.

[0078] Among them, the technical effects brought about by any design method in the third to eighth aspects can refer to the technical effects brought about by different implementation methods in the first to second aspects mentioned above, and will not be repeated here.

[0079] It should be understood that, for components in a device, the “sending” mentioned above may be referred to as “output” and the “receiving” may be referred to as “input”. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 A schematic diagram of a communication system in this application;

[0081] Figure 2 It is a schematic diagram of the working mode of AI technology in communication system;

[0082] Figure 3 A schematic diagram of the data processing method provided in this application;

[0083] Figure 4a is a schematic diagram of a process for determining a signal measurement threshold and a measurement period in Example 1;

[0084] Figure 4b is a schematic diagram of the process of determining the switching decision in Example 1;

[0085] Figure 5 is a schematic diagram of Example 2;

[0086] Figure 6 A schematic diagram of a communication device provided for this application;

[0087] Figure 7 This is a schematic diagram of a first network element structure provided in an embodiment of the present application;

[0088] Figure 8 This is a schematic diagram of a third network element structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0089] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0090] The following are some terms related to the embodiments of the present application for explanation.

[0091] 1. Terminal device: It can be a wireless terminal device that can receive network device scheduling and instruction information. The wireless terminal device can be a device that provides voice and / or data connectivity to users, or a handheld device with wireless connection function, or other processing devices connected to a wireless modem.

[0092] The terminal device can communicate with one or more core networks or the Internet via a radio access network (RAN). The terminal device can be a mobile terminal device, such as a mobile phone (or "cellular" phone, mobile phone), a computer and a data card. For example, it can be a portable, pocket-sized, handheld, computer-built-in or vehicle-mounted mobile device that exchanges language and / or data with the radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), tablet computers (Pads), computers with wireless transceiver functions, and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station (MS), a remote station, an access point (AP), a remote terminal device, an access terminal device, a user terminal device (user terminal), a user agent, a subscriber station (SS), a customer premises equipment (CPE), a terminal, a user equipment (UE), a mobile terminal (MT), etc. The terminal device may also be a wearable device and a next generation communication system, for example, a terminal device in a 5G communication system or a terminal device in a future evolved public land mobile network (PLMN), etc.

[0093] 2. Network equipment: It can be equipment in a wireless network, for example, a network equipment can be a radio access network (RAN) node (or equipment) that connects a terminal device to a wireless network, which can also be called a base station. At present, some examples of RAN equipment are: a new generation Node B (gNodeB) in a 5G communication system, a transmission reception point (TRP), an evolved Node B (eNB), a radio network controller (RNC), a Node B (NB), a base station controller (BSC), a base transceiver station (BTS), a home base station (e.g., home evolved Node B, or home Node B, HNB), a base band unit (BBU), or a wireless fidelity (Wi-Fi) access point (AP), etc. In addition, in a network structure, a network equipment can include a centralized unit (CU) node, a distributed unit (DU) node, or a RAN device including a CU node and a DU node. The network device includes the first network element, the second network element and the third network element of the present application.

[0094] In addition, in other possible cases, the network device may be other devices that provide wireless communication functions for the terminal device. The embodiments of the present application do not limit the specific technology and specific device form used by the network device. For the convenience of description, the embodiments of the present application do not limit.

[0095] The network device may also include a core network device, which may include, for example, an access and mobility management function (AMF), a user plane function (UPF) or a session management function (SMF).

[0096] In the embodiment of the present application, the device for realizing the function of the network device may be a network device, or may be a device capable of supporting the network device to realize the function, such as a chip system, which may be installed in the network device. In the technical solution provided in the embodiment of the present application, the technical solution provided in the embodiment of the present application is described by taking the device for realizing the function of the network device as an example that the network device is used as the device.

[0097] 3. The terms "system" and "network" in the embodiments of the present application can be used interchangeably. "At least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the objects associated with each other are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, "at least one of A, B and C" includes A, B, C, AB, AC, BC or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in the embodiments of the present application are used to distinguish multiple objects, and are not used to limit the order, timing, priority or importance of multiple objects.

[0098] The following is an example of the system architecture of the embodiment of the present application.

[0099] It can be understood that the system architecture described in the embodiment of the present application is for more clearly illustrating the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application.

[0100] It should be understood that the present application can be applied to a 5G system, or other communication systems in future communication networks, such as a 6G system.

[0101] Figure 1 A schematic diagram of a communication system in this application. Figure 1In the example of 6G system, the system includes AI service consumers, index library, wireless AI service provider, management plane AI service provider, control plane AI service provider and user plane AI service provider. The management plane (MP) AI service provider is an AI functional entity located in the management plane, called MP AI entity; the wireless AI service provider is an AI functional entity located in the access network (AN), called RAN AI entity; the control plane AI service provider and the user plane AI service provider are AI functional entities located in the core network, wherein the control plane AI service provider is referred to as CN CP AI entity and the user plane AI service provider is referred to as CN UP AI entity. Each of the above AI service providers collects new data that has recently occurred from the network and predicts and outputs the inference results through the machine learning (ML) model. The AI ​​services provided by the above AI service providers can be called by AI service consumers, or they can be called by other AI service providers as assistants of AI service providers. The AI ​​service consumer calls the AI ​​services supported by the above AI service providers to obtain the inference results of the AI ​​service.

[0102] Optional, in Figure 1 In the system, the system may also include a terminal AI service provider and an application AI service provider. The terminal AI service provider is called a UE AI entity, and the application AI service provider is called an APP AI entity.

[0103] In addition, this application is not limited to Figure 1 The communication scenario shown is further described below with respect to other communication scenarios involved in the present application.

[0104] In one possible implementation, the present application can be applied to a 5G system. In a 5G system, the management plane AI service provider can be a management data analytics service (MDAS), the core network AI service provider can be an NWDAF, and the wireless AI service provider can be a RAN ML function. AI service consumers can be access and mobility management functions (AMF), session management functions (SMF), policy control functions (PCF), user plane functions (UPF), etc., or wireless network equipment, such as a new radio (NR) NodeB, or a UE or an application server.

[0105] The AI ​​technology involved in the embodiments of the present application is described in more detail below.

[0106] In the above-mentioned 5G systems, AI technology based on machine learning has been increasingly widely used. And it is foreseeable that in the future 6G systems, AI technology will be further developed and applied.

[0107] Figure 2 It is a schematic diagram of the working mode of AI technology in the communication system. Figure 2 In the AI ​​industry, the working mode of AI technology is mainly divided into two stages:

[0108] The first stage is the training stage: a large amount of historical data is collected as training sample data. Each set of training sample data contains the values ​​of several dimensional attributes. Through machine learning methods, an ML model can be trained and fitted. When the ML model inputs the values ​​of some dimensional attributes, it can fit the values ​​of another part of the dimensional attributes in the training sample data.

[0109] The second stage is the inference stage: collecting new data currently occurring from the network, that is, the values ​​of some known dimensional attributes as input. This process is also called network data collection, and predicting and outputting the unknown values ​​of another part of the dimensional attributes through the ML model. This process is also called output prediction or inference result.

[0110] The above two stages can be implemented by the AI ​​service provider. After the AI ​​service provider outputs the unknown value of another dimension attribute, it will send the unknown value to the AI ​​service consumer. After receiving the unknown value, the AI ​​service consumer will determine whether certain measures need to be taken according to the business logic to achieve specific business goals.

[0111] A network is often divided into two parts: AN and CN. The main function of AN is to connect UE to the network. For example, RAN is a typical AN, which can connect UE to the network through wireless. The main function of CN is to provide user authentication and authorization management, assist in service routing, interconnection between different users, billing and other services.

[0112] In communication systems, there are generally UP and CP. UP transmits real user service data, such as voice data or packet service data. CP transmits control signaling to control the establishment, maintenance and release of a call or connection. In addition, the equipment or network elements that make up the communication system also need to be configured, troubleshoot, alarm prompt, performance measurement, etc. These are all important parts of network management and are all the responsibility of MP.

[0113] Whether AN or CN, whether MP, CP or MP, they can all have AI service providers in their respective domains. These AI service providers have their own service scopes, and their cooperation with each other can often achieve more functions, making the network more intelligent. However, the existing technology lacks collaboration solutions between AI service providers, especially data collaboration and service collaboration between AI service providers. To this end, when there are multiple AI service providers in the communication system, how to achieve collaboration between AI service providers to improve the efficiency of AI services is an urgent problem to be solved.

[0114] In order to solve the above problems, in an embodiment of the present application, when the first network element needs to perform the first AI service, the first network element can call the second AI service of the second network element, wherein the input data of the target model for implementing the first AI service must satisfy the first constraint, and the output result of the second AI service also needs to satisfy the first constraint. By calling the second AI service through the second network element, the first network element can obtain the first data, and then process the first data through the target model of the first AI service to obtain the first output result. Therefore, based on the embodiment of the present application, the first network element can assist the implementation of its own first AI service through the second AI service of the second network element, thereby realizing collaboration between the first network element and the second network element.

[0115] The following is an example of the method flow of the embodiment of the present application.

[0116] See also Figure 3 , Figure 3 A schematic diagram of a data processing method provided in this application, the method comprising the following steps:

[0117] It should be noted that Figure 3The method is illustrated by taking the first network element as an example of the execution subject, but the present application does not limit the execution subject of the method. Specifically, Figure 3 In the corresponding embodiments, the execution subject in S301-S304 is the first network element. The execution subject may also be a chip, a chip system, or a processor that supports the first network element to implement the method, or a logic module or software that can implement all or part of the functions of the first network element.

[0118] S301. The first network element receives a first request message, wherein the first request message indicates to call a first AI service of the first network element;

[0119] It is understandable that the first network element is an AI service provider that provides a first AI service. For example, the first network element may be an NWDAF network element, or the first network element may be an MDAS, etc.

[0120] Before step S301, the first network element will register the AI ​​services that it can provide in the index library. After that, the index library can determine whether the first network element meets a certain query condition based on the information registered by the first network element. The following is a specific description:

[0121] In a possible implementation, the first network element sends first registration information to the index library, where the first registration information includes: the granularity supported by the output result of the first AI service.

[0122] It should be noted that the first network element may be able to implement the first AI service through multiple models, but the granularity of the results output by each model may be different. The above-mentioned supported granularity refers to the granularity of the output results of all models of the first network element that can implement the first AI service. For example, the first network element can implement the first AI service through model A or B. Model A outputs prediction data every second, and model B outputs prediction data every minute. The granularities supported by the first network element registered in the index library are: s and min, which indicates that the first AI service of the first network element can output both second-level prediction results and minute-level prediction results.

[0123] It should be noted that the above index library may be configured in a network element, and the above index library may also be a server, and this server may be called by a certain network element, so the actual executor of the above steps is the network element that controls the index library.

[0124] Optionally, the granularity includes: time granularity, space granularity, and object granularity. Among them, the time granularity indicates the time span covered by the data, or the update cycle. The space granularity indicates the scope of the network space to which the data is applicable. The object granularity indicates the scope of objects covered by the data.

[0125] For example, the time granularity may be milliseconds, seconds, minutes, hours, etc., and the spatial granularity may be beams, grids, sectors, cells, tracking areas (TAs), service areas, etc. The object granularity may be a single object, a group of objects, etc.

[0126] It should be noted that the granularity does not mean that the output result can only be of that granularity. Granularity is a constraint on the unit of measurement of the output result. For example, when the granularity is s, the output result of the model can be data every 1-N seconds, where N is a positive integer, such as data every 1 second or data every 2 seconds.

[0127] In a possible implementation, the first registration information further includes: a confidence level supported by an output result of the first AI service.

[0128] Optionally, the first registration information also includes: the granularity of input data of the first AI service.

[0129] Optionally, the first registration information also includes: an identifier of the first AI service.

[0130] Optionally, the first registration information also includes: a data type of an output result of the first AI service.

[0131] Optionally, the first registration information also includes: the precision, recall, and F1 value supported by the target model on which the first AI service depends.

[0132] It is understandable that there is a correlation between the above confidence and the above granularity. For example, the prediction of the data for the next minute may have a larger error than the prediction of the data for the next second, so the confidence of the data is smaller.

[0133] Optionally, the first registration information also includes: indication information of the first network element.

[0134] Optionally, when the first network element sends the first registration information to the index library, it may also send indication information of the first network element at the same time. The indication information of the first network element may be access address information of the first network element, or a network element identifier of the first network element, or service access address information of a first AI service of the first network element.

[0135] If it is not limited to the scope of the first network element, in fact, all AI network elements will register the AI ​​services they can provide in the index library, and will also send their own indication information to the index library. The AI ​​network element is a network element that can provide AI services, and the AI ​​network element includes the first network element and the second network element in this application. As for the registration process of the AI ​​network element, reference can be made to the registration process of the first network element, and the process of the AI ​​network element sending the indication information of the AI ​​network element to the index library can refer to the process of the first network element sending the indication information of the AI ​​network element to the index library, which will not be repeated here.

[0136] In one possible implementation, an AI service consumer or an AI service provider determines the first AI service that needs to be called and the constraints that the first AI service needs to meet based on the business to be executed. The AI ​​service consumer or the AI ​​service provider sends a request message to the index library, and the request message carries the constraints.

[0137] In a possible implementation, after receiving the above constraint conditions, the index library determines the network element that meets the constraint conditions according to the registration information of the AI ​​network element. In the embodiment of the present application, the first network element is used to refer to the network element that meets the condition. The index library sends a response message to the above AI service consumer or AI service provider, and the response message carries the indication information of the first network element. After receiving the response message, the AI ​​service consumer or AI service provider sends a first request message to the first network element to request to call the first AI service of the first network element.

[0138] Optionally, the index library may directly send a first request message to the first network element to call the first AI service.

[0139] There are many possible situations for the information carried in the first request message, which are described in detail below:

[0140] In a possible implementation, the first request message carries a data granularity constraint, where the data granularity constraint is used to limit the granularity of input data of a model on which the first AI service depends.

[0141] It should be noted that the so-called data granularity constraint means that the input data must be at a specific unit level. For example, network element A determines that AI threshold prediction requires data with a granularity of cell level based on business needs. Therefore, it sends a request message to the corresponding network element, and this request message carries the cell to instruct the network element to obtain cell-level data.

[0142] Optionally, when the data is used to describe some features required to implement the first AI service, the data granularity constraint is that the data is greater than or equal to the first granularity, and when the data is current data that needs to be collected and is used to input the model to predict some attributes, the data granularity constraint is that the data is less than or equal to the first granularity. Further, the first granularity can be carried in the first request message. For example, the first request message carries the parameter s. As mentioned above, the granularity is a constraint on the unit of the data, so s indicates that the unit of the first granularity is s.

[0143] In a possible implementation, the first request message carries a confidence constraint, where the confidence constraint is: the confidence of the data is not less than the first confidence.

[0144] It should be noted that the above confidence constraint limits the confidence of the output data of the first AI service, which limits the possibility that the true value falls around the data, and it gives the credibility range of the data. For example, the service to be performed by network element A is to make a switching decision, and it is necessary to determine the alternative cells and switching thresholds for switching, so it is necessary to call the switching threshold prediction AI service. And according to business needs, network element A determines that data A with a confidence level of 85% is also required for AI threshold prediction. Data A is used to input the model to predict the data. Network element A sends a request message to the corresponding network element, and this request message carries 85% to instruct the network element to obtain data A with a confidence level of 85%.

[0145] In a possible implementation, the first request message carries a second constraint. The second constraint is used to constrain an output result of the first AI service.

[0146] Optionally, the second constraint includes: a granularity constraint on output data of the first AI service.

[0147] For example, network element A needs a channel traffic prediction service, and it needs prediction data of the channel traffic every second in the future, and network element B can provide services that meet these requirements, so it sends a first request message to network element B. The first request message carries S to instruct network element B to output the prediction data of the channel traffic every second in the future.

[0148] Optionally, the second constraint includes: a confidence constraint on output data of the first AI service.

[0149] For example, network element A needs a channel traffic prediction service, and the confidence of the predicted data must not be less than 95%, and network element B can provide services that meet these requirements, so it sends a request message to network element B, which also carries 95% to instruct network element B to output data with a confidence level of not less than 95%.

[0150] Optionally, the first request message may also carry a precision constraint, a recall constraint or an F1 value constraint of the target model.

[0151] For example, network element A needs a channel traffic prediction service, and the accuracy of the channel traffic prediction model must not be less than 90%, the recall rate must not be less than 85%, or the F1 value must not be less than 82%; and network element B can provide services that meet these requirements, so it sends a first request message to network element B, and the first request message carries 90%, which is used to instruct network element B to select a target model with an accuracy rate of not less than 90%; the first request message carries 85%, which is used to instruct network element B to select a target model with a recall rate of not less than 85%; or the first request message carries 82%, which is used to instruct network element B to select a target model with an F1 value of not less than 82%.

[0152] Optionally, the AI ​​service consumer or AI service provider determines a target object for which unknown attributes need to be predicted based on the business to be executed, and carries an identifier of the target object in the first request message.

[0153] It is understandable that the above-mentioned target object can be a network device or a terminal device.

[0154] Optionally, the indication information of the first network element includes: an identifier of the first network element; an Internet protocol (IP) address of the first network element, and a website address of the AI ​​service of the first network element.

[0155] From the perspective of the first network element, in one possible implementation, after receiving the first request message, the first network element determines, based on the first request message, from a plurality of models an output target model that satisfies a constraint condition, where the constraint condition is the constraint condition indicated by the first request message.

[0156] After receiving the first request message, the first network element needs to call the first AI service. As an AI service provider, it is difficult for the first network element to implement the first AI service by relying solely on its own model. Therefore, it needs to use the AI ​​services of other AI service providers to assist in the implementation of its own AI service. The assisting AI service provider is referred to as an AI service collaborator hereinafter.

[0157] S302. The first network element determines, based on the first request message, a second network element on which the first AI service depends; wherein the implementation of the first AI service requires the use of a target model, the input data of the target model has a first constraint, and the second network element provides a second AI service, and the output result of the second AI service satisfies the first constraint; the first constraint includes: a data type constraint.

[0158] It is understandable that the above-mentioned second network element refers to the AI ​​service collaborator of the first network element.

[0159] As mentioned above, the first network element needs to rely on AI service collaborators to assist in the implementation of its own AI service, and the AI ​​service collaborators required by the first network element have the following possibilities: one possibility is that when the first network element needs to collect certain current data, these current data can be input into the target model to infer and predict unknown attributes. At this time, if there are some AI service providers that output this type of data, compared with the first network element collecting data by itself, directly using the output data of other AI service providers to replace the data that originally needs to be collected can save a certain amount of time. Another possibility is that the first network element needs to improve the accuracy of model data prediction by inputting some new features. These new features are actually data used to describe features, hereinafter referred to as feature data. At this time, if there are some AI service providers that output this type of data, the first network element can obtain the output data of other AI service providers as the feature data. As for other possibilities, they are not listed one by one here, but no matter which possibility, the first network element needs certain types of data, and the first network element can assist in the implementation of its own AI service through AI service providers that can output this type of data. Therefore, the first constraint needs to include a data type constraint. The data type constraint is used to limit the data type of the output data of the AI ​​service collaborator to certain specific types. For example, a data type constraint may limit the output data of an AI service collaborator to data that predicts the resource occupancy rate of a neighboring area.

[0160] Among many network elements that can provide AI services, the first network element can search for an AI service collaborator whose output data meets the data type constraint by one or more of the following methods:

[0161] In one possible implementation, the first network element sends a first constraint to the index library. As mentioned above, the first constraint includes a data type constraint. The index library determines the indication information of the second network element based on the first constraint and sends it to the first network element. Correspondingly, the first network element receives the indication information of the second network element.

[0162] It is understandable that, as mentioned above, the index library records the registration information of the AI ​​service of each AI network element, so it can determine the second network element that meets the first constraint based on the registration information. And the index library records the indication information of each AI network element, or the registration information of each AI network element is associated with the indication information of the AI ​​network element, so after determining the second network element, the index library can find the indication information of the corresponding second network element.

[0163] Optionally, a storage repository is configured in the first network element, and the first network element determines the indication information of the second network element in the storage repository according to a first constraint. As mentioned above, the first constraint includes a data type constraint.

[0164] It is understandable that, similar to the index library, the storage library also records the registration information of the AI ​​services of each AI network element and the indication information of the AI ​​network element, so the indication information of the second network element can be determined in the storage library according to the first constraint. The registration process of each AI network element in the storage library can refer to the registration process of each AI network element in the index library, and the process of each AI network element sending the indication information of the AI ​​network element to the storage library can refer to the process of each AI network element sending the indication information of the AI ​​network element to the storage library, which will not be repeated here.

[0165] More specifically, the first network element may notify the index library or storage library in the following manner: the output data of the AI ​​service provided by the AI ​​service provider required by the first network element must meet the data type constraint:

[0166] In one possible implementation, the first network element sends a data type identifier to the index library or storage library to instruct the index library or storage library to find an AI service provider that outputs data of this type. For example, the first network element sends an identifier of signal strength prediction data to the index library or storage library. Accordingly, before this, each AI service provider has registered the data type supported by its output data in the index library or storage library.

[0167] In a possible implementation, the first network element knows that a certain AI service can output data of the type required by the first network element, and the second AI service is used to refer to the AI ​​service below. Therefore, the first network element sends the identifier of the second AI service to the index library or storage library to instruct the index library or storage library to query the AI ​​service provider that can provide the second AI service. Correspondingly, before this, each AI service provider has registered the identifier of the AI ​​service it supports in the index library or storage library.

[0168] It can be understood that both the data type identifier and the identifier of the second AI service are actually used as data type constraints, and they are both included in the first constraint and sent to the second network element together with other constraints through the second request message.

[0169] In order to meet the data granularity requirements of different users, in a communication system, there may be multiple AI service providers that can output a certain type of data, but the granularity of the data they output is different. The granularity required by the first network element may be a specific granularity, so the first network element needs to notify the index library or storage library to select an AI service provider that can output this granularity from these AI service providers. Specifically, the following methods can be used:

[0170] In one possible implementation, the first constraint includes: a data granularity constraint, thereby notifying the index library or the storage library to select a second network element whose output data of the supported AI service satisfies the data granularity constraint.

[0171] Optionally, when the data is used to describe some features required to implement the first AI service, the data granularity constraint is that the data is greater than or equal to the first granularity; when the data is current data that needs to be collected and is used to input a model to predict some attributes, the data granularity constraint is that the data is less than or equal to the first granularity.

[0172] Although different AI service providers may output the same type of data, they may use different models, and thus the confidence of the output data may be different. For example, the granularity of the output data of different models is different, and thus the confidence of the output data is different. For example, the prediction of the data for the next second may have a smaller error than the data for the next minute, and thus the data confidence is higher. However, the first network element may need high-confidence data due to the requirements of the first AI service, so the first network element needs to inform the index library or repository of its own requirements for data confidence. Specifically, this can be done in the following ways:

[0173] In a possible implementation, the first constraint includes: the confidence of the data is not less than a first confidence, so that the index library or storage library selects the second network element that meets the confidence constraint.

[0174] For example, network element A can provide a handover threshold prediction AI service. In order to implement this service, network element A also needs to collect data A with a confidence level of 85%, which is used to input the model to predict the data. Therefore, it sends a first constraint to the index library or storage library, which carries 85% to instruct the network element to obtain data A with a confidence level of 85%.

[0175] It should be noted that both the data granularity constraint and the confidence constraint may come from the first request message, or may be determined by the first network element according to the need to implement the first AI service. Taking the data granularity constraint as an example, the first request message sends a granularity s. As mentioned above, the granularity s is used to limit the unit of data to s, that is, the input data of the first AI service is data every 1-N seconds, such as data every 1 second, or data every 2 seconds, etc. For another example, the first network element needs to collect data every N seconds to implement the first AI service, where N is a positive integer. The data is used to input the model to predict some attributes. The first network element can determine the first granularity to be s accordingly, and the data granularity constraint of the input data to be collected is that the unit of the data is less than or equal to s, for example, data in units of ms can be collected.

[0176] From the perspective of the index library or storage library, after receiving the first constraint, the index library or storage library determines the AI ​​service provider that meets the first constraint based on the registration information of the AI ​​network element. The embodiment of the present application uses the second network element to refer to the AI ​​service provider that meets the conditions, that is, the aforementioned AI service collaborator.

[0177] After that, since the granularity information registered by each AI network element is not necessarily the same as the first constraint, for example, the first constraint is that the time granularity of the data is s, and the registered granularity information is that the time granularity of the data is s, and the spatial granularity is the cell. In other words, the registered granularity information may be more comprehensive than the granularity information included in the first constraint. In order to allow the first network element to have a more comprehensive understanding of the granularity characteristics of the output results of the second AI service, the index library or storage library sends the granularity supported by the second AI service to the first network element. Specifically, it can be done in the following ways:

[0178] Optionally, the index library or storage library sends to the first network element: the granularity supported by the output result of the second AI service registered by the second network element in the index library or storage library.

[0179] S303. The first network element calls the second AI service through the second network element to obtain the first data;

[0180] It can be understood that the above-mentioned first data is the output result of the second AI service.

[0181] In order to call the second AI service, the first network element interacts with the second network element in the following manner: the first network element sends a second request message to the second network element, instructing the second network element to call the second AI service, and the second network element calls the second AI service according to the second request message.

[0182] There are many possible situations for the information carried in the second request message, which are described in detail below:

[0183] In order to meet the data granularity requirements of different users, AI service providers usually have multiple models that can output a certain type of data, and the granularity of the output data of these models may be different. Therefore, the first network element needs to inform the second network element of the granularity it needs, so that the second network element can select a model that can output data with this granularity. Specifically, the first network element can achieve this in the following ways:

[0184] In a possible implementation, the second request message carries a data granularity constraint, thereby instructing the second network element to select a model whose output data satisfies the data granularity constraint.

[0185] Since the first network element usually does not need data from all devices, for example, the first network element is a base station, and it needs signal strength prediction data of the user equipment connected to it. If the second network element obtains data from all devices, it will not only take a lot of time, but also be unnecessary. Therefore, the first network element needs to inform itself of which devices' data it needs. Specifically, the first network element can inform the second network element in the following ways:

[0186] Optionally, the second request message may also carry an identifier of the target object.

[0187] It can be understood that the identifier of the target object may be derived from the first request message, that is, the first request message carries the identifier of the target object to instruct the first network element to predict the unknown attributes of the target object. It may also be that the implementation of the first AI service requires the collection of data of certain target objects, or requires certain characteristics of the target objects, so that the first network element determines the identifier of the corresponding target object based on this.

[0188] As mentioned above, in order to meet the data confidence requirements of different users, AI service providers usually have multiple models that can output a certain type of data, and the confidence of the output data of these models may be different. Therefore, the first network element needs to inform the second network element of the confidence it needs, so that the second network element can select a model that can output data with this confidence. Specifically, the first network element can achieve this in the following ways:

[0189] In a possible implementation, the second request message carries a confidence constraint, thereby instructing the second network element to select a model whose output data satisfies the confidence constraint.

[0190] Referring to the above reasons, the first network element needs to inform the second network element of its requirements for the precision, recall or F1 value of the model, which can be informed in the following ways:

[0191] Optionally, the second request message may also carry a precision constraint, a recall constraint or an F1 value constraint, thereby instructing the second network element to select a model that satisfies the precision constraint, the recall constraint or the F1 value constraint.

[0192] From the perspective of the second network element, after receiving the second request message sent by the first network element, the second network element calls the second AI service, obtains the first data, and sends the first data to the first network element.

[0193] S304. The first network element processes the first data through the target model to obtain a first output result.

[0194] It can be understood that, since the first data is the output result of the target model, it satisfies the input constraint of the target model, that is, the first constraint, and therefore the first data can be input into the target model and processed by the target model.

[0195] It is understandable that the above first output result is the output result of the target model, but this solution does not limit the first output result to be the output result of the first AI service. For example, after obtaining the first output result, the first network element may obtain the output data of other AI service providers, and then continue to process the output data through another model. It is also possible that the first network element processes the first output result through other models and finally obtains the second output result, and the second output result is the output result of the first AI service.

[0196] based on Figure 3 In the technical solution shown, the first network element and the second network element are both AI service providers. The first network element needs to implement the first AI service through the target model, and the input data of the target model needs to meet the first constraint. The first network element needs to rely on the second network element to provide the second AI service to implement the first AI service, and the output result of the second AI service meets the first constraint. Therefore, the first data obtained by the second AI service can be directly processed through the target model to obtain the first output result of the target model, thereby enabling the AI ​​service provider to make full use of the AI ​​services of other AI service providers to assist in the implementation of its own AI services, which improves the efficiency of AI services.

[0197] It can be seen from the above technical solutions that the technical solutions provided by this application have the following beneficial effects: this application can realize deep collaboration of data of AI service providers, thereby realizing efficient network data collection, and full utilization of the output results of each AI service, thereby improving the accuracy and operating efficiency of AI services in the communication system.

[0198] The following two examples are combined to Figure 3 The method shown is illustrated by way of example.

[0199] The following two examples are Figure 3 Possible implementations of the method shown. Example 1 is an example of a situation where the first constraint in step S302 includes: data granularity constraint and data type constraint, and Example 2 is an example of a situation where the first constraint in step S302 includes: data granularity constraint, data confidence constraint and data type constraint.

[0200] Example 1

[0201] Embodiment 1 is an example of the application of the present application to intelligent switching. The wireless base station or the switching decision control unit of the base station acts as an AI service consumer, and the RAN AI entity acts as an AI service provider. The wireless base station or the switching decision control unit of the base station discovers and calls the AI ​​service of the RAN AI entity, and determines the optimal UE measurement threshold and UE measurement period based on the prediction results output by the AI ​​service. The conditions that the UE measurement threshold and UE measurement period must meet are: while ensuring that the necessary cell switching can be triggered in a timely manner, the number of measurement reports reported is minimized. The switching decision control unit achieves the purpose of saving signaling overhead and reducing energy consumption of network equipment and UE by determining the UE measurement threshold and UE measurement period that meet this condition. In order to provide the AI ​​service required by the switching decision control unit, the RAN AI entity will further call the AI ​​service provided by the CN AI entity to obtain the prediction results of the UE service type and movement trajectory, and at the same time, combined with the prediction results of the wireless resource occupancy rate of the adjacent cells output by its own AI service, dynamically determine the switching threshold of each target cell. When multiple cells meet the signal strength requirements and available wireless resources, the RAN AI entity selects the most reasonable target cell from these multiple cells based on the prediction results of the UE service type and movement trajectory, and dynamically determines the switching threshold to achieve on-demand switching to the most reasonable cell to connect the UE.

[0202] Figure 4a is a schematic diagram of the process of determining the signal measurement threshold and measurement period in Example 1, Figure 4a The specific process includes:

[0203] 401. The RAN AI entity, the CN CP AI entity, the CN UP AI entity, the MP AI entity, and the UE AI entity register supported AI services in the index library and send their respective indication information;

[0204] Specifically, after the AI ​​service is turned on, the RAN AI entity sends an AI service registration message to the index library, which carries the following registration information: the identifier of the switching threshold prediction service, the identifier of the UE signal strength prediction service, and the identifier of the neighboring cell resource occupancy prediction service, as well as the granularity supported by the output results of each AI service. Among them, the granularity supported by the output result of the AI ​​service for switching threshold prediction is s / UE, indicating that the output result of the AI ​​service is for a single UE, and the output result is updated in seconds. The output result of the AI ​​service for UE signal strength prediction supports two granularity prediction results, namely ms / beam / UE and s / UE. Among them, ms / beam / UE indicates that the output result of the AI ​​service is for a single UE, the output result is in beam units, and the output result is updated in milliseconds. The granularity supported by the output result of the AI ​​service for neighboring cell resource occupancy prediction is s / cell, and s / cell indicates that the output result of the AI ​​service is for a single UE, and the output result is in cells.

[0205] CN CP AI sends an AI service registration message to the index library, which carries the following registration information: the identifier of the UE trajectory prediction service and the granularity supported by the output results of the AI ​​service. The granularity is min / cell / UE, indicating that the UE trajectory prediction service performs minute-level cell location prediction for a single UE.

[0206] The CN UP AI sends an AI service registration message to the index library, which carries the following registration information: the identifier of the UE service type prediction service and the granularity supported by the output result of the AI ​​service. The granularity is min / cell / UE.

[0207] The MP AI entity registers the base station resource occupancy prediction service it supports with the index library, with a granularity of min / base station.

[0208] The UE AI entity sends an AI service registration message to the index library, which carries the following registration information: the identifier of the UE downlink signal strength prediction service and the granularity supported by the output result of the AI ​​service. The granularity is ms / beam / UE, indicating that the prediction analysis result is the beam signal strength at the ms level for a single UE.

[0209] When each of the above AI entities sends an AI service registration message to the index library, it may also send indication information of each AI entity at the same time, wherein the indication information of the AI ​​entity may be the access address information of the AI ​​entity, or the identifier of the AI ​​entity, or the access address information of the AI ​​service of the AI ​​entity.

[0210] 402. The handover decision control unit determines the AI ​​service to be called and the corresponding constraint conditions according to the signal strength measurement task to be performed;

[0211] The task to be performed by the handover decision control unit is to send the latest measurement command to each terminal in seconds and set different measurement report thresholds and measurement cycles. Therefore, the handover decision control unit determines that it needs to call an AI service that can predict signal strength, and the corresponding constraint is the second-level prediction of a single UE.

[0212] 403. The switching decision control unit sends an AI service query request message to the index library;

[0213] The switching decision control unit sends an AI service query request to the index library, which carries the identifier of the signal strength prediction service and the granularity constraint s / UE as the query condition, indicating that the index library is queried for AI service entities that support outputting second-level signal strength prediction for a single UE.

[0214] 404. The index library sends an AI service query response message to the switching decision control unit;

[0215] The index library determines that the RAN AI entity meets the query condition of step 403 based on the registration information obtained in step 401, and sends an AI service query response message to the switching decision control unit. The message carries the service access address information of the RAN AI entity, and may also carry the identifier of the signal strength prediction service and the granularity s / UE supported by the service.

[0216] 405. The handover decision control unit sends an AI service request message to the RAN AI entity;

[0217] The request message carries the identifier of the signal strength prediction service, the identifier of the target UE, the granularity constraint s / UE that the output data of the service needs to meet, and the identifier of the target UE instructing the RAN AI entity to process the target UE.

[0218] 406. The RAN AI entity determines a UE signal strength prediction model according to the constraint condition;

[0219] The RAN AI entity selects a model that supports second-level signal strength prediction for a single UE according to the identification and granularity constraint of the signal strength prediction service in step 406 .

[0220] 407. The RAN AI entity sends an AI service response message or notification to the handover decision control unit;

[0221] The AI ​​service response message or notification carries the output result of the signal strength prediction service. Specifically, the RANAI entity collects the current signal strength measurement data of the target UE and outputs the predicted value of the UE signal strength for the next second through the model. The RAN AI entity continuously sends AI service messages or notifications to the switching decision control unit at a period of 1 second, carrying the predicted value of the UE signal strength.

[0222] 408. The handover decision control unit determines a UE signal measurement threshold and a measurement period according to the prediction result.

[0223] When the predicted value of the UE signal strength decreases to a preset value, the handover decision control unit starts to make a handover decision.

[0224] Figure 4b is a schematic diagram of the process of determining the switching decision in Example 1, Figure 4b The specific process includes:

[0225] 409. The switching decision control unit determines the required AI service and corresponding constraints;

[0226] Since the handover decision control unit needs to make a handover decision, the handover decision control unit needs to call a handover threshold prediction service, the corresponding granularity of which is s / UE.

[0227] 410. The switching decision control unit sends a query request message to the index library AI service;

[0228] The AI ​​service query request message carries a query condition, which is the identifier of the switching threshold service and the granularity constraint s / UE, indicating that an AI service entity that supports outputting the second-level switching threshold prediction result of a single UE is queried in the index library.

[0229] 411. The index library sends an AI service query response message to the switching decision control unit;

[0230] Specifically, the index library determines that the RAN AI entity meets the query condition of step 410 based on the registration information obtained in step 401, and then sends an AI service query response message to the switching decision control unit. The message carries the service access address information of the RAN AI entity, and may also carry the switching threshold service identifier, and the granularity s / UE supported by the output result of the service.

[0231] 412. The handover decision control unit sends an AI service request message to the RAN AI entity;

[0232] The AI ​​service request message carries the handover threshold AI service identifier and the granularity constraint s / UE. Optionally, it can also carry the model feature granularity constraints min / UE and min / cell, indicating that the handover threshold prediction AI model used needs to use the feature data of a single UE at the minute level and the feature data of a single cell at the minute level.

[0233] 413. The RAN AI entity determines a handover threshold prediction model and a handover cell prediction model according to the constraint conditions;

[0234] Specifically, the RAN AI entity selects a model that outputs a second-level UE switching threshold according to the AI ​​service request message of step 412. At the same time, the model uses the motion trajectory feature data and service type feature data of a single UE at the minute level, as well as the resource occupancy feature data of a single adjacent cell at the minute level.

[0235] 414. The RAN AI entity determines the AI ​​service to be invoked based on the input data required by the model;

[0236] Specifically, the RAN AI entity uses the characteristic data of a single cell at the minute level according to the needs of the model to determine whether to call its own adjacent cell resource occupancy prediction service, that is, it can obtain the prediction results of the adjacent cell resource occupancy rate of the target UE from itself every minute; the RAN AI entity uses the characteristic data of a single UE at the minute level according to the needs of the model to determine whether it needs to query the AI ​​service entity that can output the movement trajectory and service type prediction results of a single UE at the minute level.

[0237] 415. The RAN AI entity sends an AI service query request to the index library;

[0238] The AI ​​service query request carries the UE trajectory AI service identifier and the granularity constraint min / UE as query conditions, and carries the UE service type AI service identifier and the granularity constraint min / UE as query conditions.

[0239] 416. The index library sends an AI service query response message to the RAN AI entity;

[0240] Specifically, the index library determines, based on the registration information obtained in step 0, that the CN CP AI entity satisfies the UE trajectory AI service identifier and the granularity constraint min / UE; and that the CN UP AI entity satisfies the UE service type AI service identifier and the granularity constraint min / UE.

[0241] 417. The RAN AI entity sends an AI service request message to the CN CP AI entity and the CN UP AI entity.

[0242] The RAN AI entity regards the CN CP AI entity and the CN UP AI as AI service collaborators, and uses the output data of their AI services as part of the feature data required by its own model.

[0243] 418. The CN CP AI entity and the CN UP AI entity respectively send an AI service response message or notification to the RAN AI entity;

[0244] The AI ​​service response message or notification carries a first prediction result of the wireless resource occupancy rate of the neighboring cells of the target object, a second prediction result of the movement trajectory of the target object, and a third prediction result of the type of service running by the target object.

[0245] 419. The RAN AI entity processes the data carried in the received response message or notification through the model to obtain a handover threshold;

[0246] First, the RAN AI entity determines a cell whose air interface resource occupancy rate is lower than a threshold from the neighboring cells according to the first prediction result, and uses the cell as a target cell.

[0247] Then, the RAN AI entity adjusts the switching threshold of the target cell according to the second prediction result to obtain a first switching threshold. Specifically, the RAN AI entity lowers the switching threshold of the target cell located on the motion trajectory indicated by the second prediction result, and / or increases the switching threshold of the target cell not located on the motion trajectory indicated by the second prediction result, to obtain the first switching threshold.

[0248] Afterwards, the RAN AI entity adjusts the first switching threshold according to the third prediction result to obtain the second switching threshold. Specifically, the RAN AI entity increases the first switching threshold of the target cell that meets the second condition, and / or lowers the first switching threshold of the first cell that does not meet the second condition according to the third prediction result. The second condition is: switching to the target cell can make the performance and / or experience of the target service type better than the preset value, and the target service type is the service type indicated by the third prediction result.

[0249] Finally, the RAN AI entity uses the indication information of the target cell and the second switching threshold as the output result of the AI ​​service.

[0250] 420. The RAN AI entity sends a response message or notification to the handover decision control unit AI service;

[0251] Specifically, the RAN AI entity continuously sends an AI service response to the handover decision control unit, carrying the one or more handover target cells and the corresponding second handover threshold.

[0252] 421. The handover decision control unit sends a measurement command to the UE;

[0253] The handover decision control unit determines the appropriate signal measurement threshold and measurement period according to the prediction result, and re-issues the measurement command to the UE. If the latest value determined in step 408 changes, the handover decision control unit re-issues the measurement command to the UE, updating the measurement threshold and measurement period to the latest value.

[0254] 422. The UE sends a measurement report to the handover decision control unit.

[0255] The UE reports a measurement report based on the latest measurement threshold and measurement period. The measurement report includes the signal strength of the current cell and the adjacent cells.

[0256] It is understandable that step 421 and step 422 are not necessarily performed after step 409 , and step 421 and step 422 will be performed as long as the latest value determined in step 408 changes.

[0257] 423. The handover decision control unit decides whether to perform handover according to the handover threshold and the measurement report.

[0258] The handover decision control unit determines whether to perform handover according to the handover target cell and the handover threshold in step 420 and the signal strength measurement report of the UE in step 422 .

[0259] The beneficial effect of this implementation is that because the handover decision control unit selects the most reasonable target cell based on the UE service type and movement trajectory, as well as the neighboring wireless resource prediction results, and dynamically determines the handover threshold, a better handover can be achieved. In this process, the AI ​​service provider uses the output data of the AI ​​service collaborator as part of the feature data for its own analysis and reasoning, thereby improving the accuracy of the AI ​​service.

[0260] Example 2

[0261] Embodiment 2 is an example of the application of the technical solution of the present application to ensure intelligent service experience. The Policy Control Function (PCF) acts as an AI service consumer, and the CN CP AI entity acts as an AI service provider to provide experience analysis and prediction services to the PCF. The PCF discovers and calls the service experience analysis service of the CN CP AI entity, and determines the optimal service quality assurance strategy and parameters based on the prediction results output by the experience analysis. In order to provide experience analysis and prediction services to the PCF, the CN CP AI entity determines the AI ​​service output results of the CN UP AI entity, RAN AI entity, and APP AI entity using the constraints based on the data collected as needed, the corresponding granularity constraints, and the confidence constraints, instead of the actual network data measurement and collection, thereby improving the data collaboration efficiency between AI service providers.

[0262] Figure 5 is a schematic diagram of Example 2, Figure 5 The specific process includes:

[0263] 501. The RAN AI entity, the CN CP AI entity, the CN UP AI entity, and the APP AI entity register supported AI services in the index library and send their respective indication information;

[0264] Each AI service provider registers the AI ​​services it can provide with the index library, where the registration information includes the various granularities and confidence levels supported by the output results of the AI ​​services it can provide. Specifically, after the RAN AI entity turns on the channel state prediction service, it sends an AI service registration message to the index library. The message carries the granularity ms / UE and the confidence level 95%; the granularity s / UE and the confidence level 85%. This means that the result of the channel state prediction output is for a single UE, the spatial range is the beam level, the prediction result is the channel state in the next few milliseconds, and the confidence level of the prediction result relative to the subsequent facts can reach 95%. If the prediction result is the channel state 1 second in the future, the confidence level of the prediction result relative to the subsequent facts can reach 85%. The RAN AI entity also registers the cell air interface resource prediction service it supports with the index library, with a granularity of min / cell and a confidence level of 95%. Similarly, the CN CP AI entity registers the service experience analysis prediction service with a granularity of min / UE and a confidence level of 90%. The CN UP AI entity registers the service flow rate and message delay prediction services it supports with the index library, with the granularity of s / UE and the confidence level of 90%. The APP AI entity registers the service flow rate prediction services it supports with the index library, with the granularity of s / UE and the confidence level of 95%.

[0265] When each of the above AI entities sends an AI service registration message to the index library, it may also send indication information of each AI entity at the same time, wherein the indication information of the AI ​​entity may be the access address information of the AI ​​entity, or the identifier of the AI ​​entity, or the access address information of the AI ​​service of the AI ​​entity.

[0266] 502. PCF sends an AI service query request message to the index library;

[0267] The AI ​​service query request message carries the identifier of the business experience analysis and prediction service, as well as the constraints of the service: min / UE, and the confidence level of the output result is not less than 90%.

[0268] 503. The index library sends an AI service query response message to PCF;

[0269] Based on the registration information obtained in step 501, the index library sends an AI service query response message to the PCF, which carries the indication information of the CN UP AI entity and, optionally, the identifier of the business experience analysis and prediction service and the granularity supported by the output results of the service registered by the CN UPAI entity.

[0270] 504. PCF sends an AI service request message to the CN CP AI entity;

[0271] The AI ​​service request message carries the service experience analysis service identifier and the service constraint min / UE, and the confidence level of the output result is not less than 90%.

[0272] The AI ​​service request message also carries constraints on the model's input data, which are: the business flow rate prediction data and message delay prediction data input by the model need to meet the s / UE granularity and a confidence level of not less than 90% respectively, and the channel state prediction data input by the model needs to meet the s / UE granularity and a confidence level of not less than 80%.

[0273] 505. Determine the AI ​​service to be called based on the input data required to implement the AI ​​service model according to the constraint condition;

[0274] The CN CP AI entity selects a suitable prediction model according to the service identification and constraint conditions of the service experience analysis in step 504, and determines the data to be collected and the corresponding granularity constraints according to the data of each dimension that needs to be input into the model and the granularity constraints. Specifically, the CN CP AI entity determines that the service experience analysis needs to collect the channel state data, application flow rate data and message delay data of the UE every second.

[0275] The CN CP AI entity uses other AI service providers as AI service collaborators, and uses the output results of the AI ​​service collaborators that meet the confidence constraint and are less than or equal to the granularity of the step 504 constraint to replace the collected data, that is, it decides to use the output results of the AI ​​service of the AI ​​service collaborator as the current data to be collected, and the output results need to meet the constraints in step 504. For example, the CN CP AI entity determines that the service flow rate, message delay and channel status of the target UE need to be collected every second. If the predicted output of other AI service collaborators is used to replace the above-mentioned collected data, the service flow rate prediction and message delay prediction output by the AI ​​service collaborator need to meet the s / UE granularity and a confidence level of not less than 90% respectively, and the channel status meets s / UE and a confidence level of not less than 80%.

[0276] 506. The CN CP AI entity sends an AI service query request message to the index library;

[0277] The CN CP AI entity sends an AI service query request to the index library, which carries the service flow rate service identifier, s / UE, and a confidence level of not less than 90% as query conditions. The AI ​​service query request also carries the message delay service identifier, s / UE, and a confidence level of not less than 90% as query conditions. In addition, the AI ​​service query request also carries the channel state prediction service identifier, s / UE, and a confidence level of not less than 80% as query conditions.

[0278] 507. The index library sends an AI service query response message to the CN CP AI entity;

[0279] The index library sends an AI service query response message to the CN CP AI based on the registration information obtained in step 501, which carries the identifiers of the AI ​​services of the CN UP AI entity, RAN AI entity and APP AI entity that meet the above query conditions, as well as the granularity and confidence supported by the output results of these AI services.

[0280] 508. The CN CP AI entity sends an AI service request message to the APP AI entity, the CN UP AI entity and the RAN AI entity respectively;

[0281] The CN CP AI entity determines that the confidence of the application flow rate prediction output by the APP AI entity at the s / UE granularity is greater than that of the result output by the CN UP AI entity, so it decides to select the application flow rate prediction output by the APP AI entity instead of data collection. The CN CP AI determines to use the channel state prediction result output by the RAN AI entity and the message delay prediction result output by the CN UP AI entity. The CN CP AI entity then sends an AI service request message to the APP AI entity, the CN UP AI entity, and the RAN AI entity respectively.

[0282] 509. The APP AI entity, the CN UP AI entity, and the RAN AI entity send an AI service response message or notification to the CN CP AI entity;

[0283] The AI ​​service response message or notification carries the output result of the corresponding AI service.

[0284] 510. The CN CP AI entity processes the data carried in the received response message or notification through the model to obtain an experience analysis result;

[0285] The CN CP AI entity uses the output results sent by the APP AI entity, the CN UP AI entity and the RAN AI in 509 to replace the data to be collected, uses the model to perform service experience analysis reasoning, and outputs the experience analysis results.

[0286] 511. The CN CP AI entity sends an AI service response message or notification to the PCF;

[0287] The AI ​​service response message or notification carries the experience analysis result obtained in step 510.

[0288] 512. PCF adjusts policy parameters.

[0289] The PCF adjusts policy parameters according to the experience analysis result received in step 511, for example, when the experience is below a certain threshold, allocates a larger allowed data rate to the UE, etc., so as to achieve intelligent service experience guarantee.

[0290] The beneficial effect of this implementation is that the AI ​​service provider obtains the output results of the AI ​​service collaborator that meet the constraint conditions, and uses the output results as the data that originally needs to be collected, thereby improving the operational efficiency of data collection.

[0291] The above describes the embodiment of the present application from the perspective of the method. The following describes the communication device in the embodiment of the present application from the perspective of specific device implementation.

[0292] See also Figure 6, the embodiment of the present application provides a communication device 600, which can implement the functions of the first network element (or the third network element) in the above method embodiment, and thus can also achieve the beneficial effects of the above method embodiment. In the embodiment of the present application, the communication device 600 can be the first network element (or the third network element), or it can be an integrated circuit or component inside the first network element (or the third network element), such as a chip. The following embodiments are described by taking the communication device 600 as the first network element (or the third network element) as an example.

[0293] In a possible implementation, when the device 600 is used to execute the method executed by the first network element in any of the aforementioned embodiments, the device 600 includes a processing unit 602 and a transceiver unit 601; the transceiver unit 601 is used to receive a first request message, and the first request message indicates calling a first AI service; the processing unit 602 is used to determine, according to the first request message, a second network element on which the first AI service depends; wherein the input data of the target model used to implement the first AI service has a first constraint; the second network element provides a second AI service, and the output result of the second AI service satisfies the first constraint; the first constraint includes: a data type constraint; calling the second AI service of the second network element to obtain first data; processing the first data through the target model to obtain a first output result.

[0294] In a possible implementation, the processing unit 602 is specifically used to: call the transceiver unit 601 to send the first constraint to the index library; obtain the indication information of the second network element obtained according to the first constraint and sent by the index library and received by the transceiver unit 601.

[0295] In a possible implementation, the first constraint includes: a data granularity constraint, and the data granularity constraint is carried in the first request message.

[0296] In a possible implementation, the processing unit 602 is specifically used to call the transceiver unit 601 to send a second request message to the second network element, where the second request message carries the data granularity constraint, and the second request message instructs the second network element to call the second AI service.

[0297] In a possible implementation, the first constraint includes: the confidence level of the data is not less than a first confidence level.

[0298] In a possible implementation, the transceiver unit 601 is further used to send first registration information to the index library, where the first registration information includes: the granularity supported by the output result of the first AI service.

[0299] In a possible implementation, the first registration information also includes: a confidence level supported by an output result of the first AI service.

[0300] In one possible implementation, the first request message carries a second constraint, which is a granularity constraint of the output result obtained by the first AI service. The processing unit 602 is also used to determine the target model whose output satisfies the second constraint from multiple models based on the first request message.

[0301] In a possible implementation, when the device 600 is used to execute the method executed by the third network element in any of the aforementioned embodiments, the device 600 includes a processing unit 602 and a transceiver unit 601; the transceiver unit 601 is used to receive a first constraint sent by a first network element, wherein the first network element provides a first AI service; the input data of the target model for implementing the first AI service has a first constraint; the first constraint includes: a data type constraint, the processing unit 602 is used to determine the indication information of the second network element in the index library according to the first constraint, the second network element provides a second AI service, and the output result of the second AI service satisfies the first constraint; the transceiver unit 601 is also used to send the indication information of the second network element to the first network element, so that the first network element calls the second AI service of the second network element to obtain the first data; the first network element processes the first data through the target model to obtain the first output result. These modules can also perform the corresponding functions in the above-mentioned second aspect method example, please refer to the detailed description in the method example for details, which will not be repeated here.

[0302] In a possible implementation, the transceiver unit 601 is further used to receive first registration information of the first network element, where the first registration information includes: the granularity supported by the output result of the first AI service.

[0303] In a possible implementation, the first registration information also includes: a confidence level supported by an output result of the first AI service.

[0304] In one possible implementation, the transceiver unit 601 is also used to: receive registration information of an AI network element, where the AI ​​network element includes the second network element, and the registration information of the AI ​​network element includes: the data type supported by the output result of the AI ​​service provided by the AI ​​network element; and call the processing unit 602 to determine the indication information of the second network element that satisfies the first constraint in the index library according to the registration information of the AI ​​network element.

[0305] It should be noted that the information execution process and other contents of the units of the above-mentioned communication device 600 can be specifically referred to the description in the method embodiment shown in the above-mentioned application, and will not be repeated here.

[0306] See also Figure 7 , an embodiment of the first network element in the embodiment of the present application includes:

[0307] Figure 7 It is a schematic diagram of a first network element structure provided in an embodiment of the present application. The first network element 700 may include one or more central processing units (CPU) 701 and a memory 705. The memory 705 stores one or more application programs or data.

[0308] The memory 705 may be a volatile storage or a persistent storage. The program stored in the memory 705 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processor 701 may be configured to communicate with the memory 705 and execute a series of instruction operations in the memory 705 on the first network element 700.

[0309] The first network element 700 may also include one or more power supplies 702, one or more wired or wireless network interfaces 703, one or more input and output interfaces 704, and / or one or more operating systems, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0310] The central processor 701 can execute the operations performed by the first network element in the aforementioned embodiment, and the details are not repeated here.

[0311] See also Figure 8 , an embodiment of the third network element in the embodiment of the present application includes:

[0312] Figure 8 It is a schematic diagram of a third network element structure provided in an embodiment of the present application. The third network element 800 may include one or more central processing units (CPU) 801 and a memory 805. The memory 805 stores one or more application programs or data.

[0313] The memory 805 may be a volatile storage or a persistent storage. The program stored in the memory 805 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processor 801 may be configured to communicate with the memory 805 and execute a series of instruction operations in the memory 805 on the third network element 800.

[0314] The third network element 800 may also include one or more power supplies 802, one or more wired or wireless network interfaces 803, one or more input and output interfaces 804, and / or one or more operating systems, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0315] The central processor 801 can execute the operations performed by the third network element in the aforementioned embodiment, and the details are not repeated here.

[0316] An embodiment of the present application also provides a computer-readable storage medium, including instructions, which, when executed on a computer, enable the computer to execute the method in the aforementioned embodiment.

[0317] The embodiments of the present application also provide a computer program product including instructions, which, when executed on a computer, enables the computer to execute the method in the aforementioned embodiments.

[0318] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0319] An embodiment of the present application also provides a communication system, and the network system architecture includes the communication device in any of the above embodiments (including the first network element and the third network element).

[0320] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

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

[0322] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0323] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. A data processing method, characterized in that: Applied to a first network element, the first network element providing a first AI service, the method comprising: The first network element receives a first request message, where the first request message indicates calling the first AI service; The first network element determines, according to the first request message, a second network element on which the first AI service depends; wherein the input data of the target model used to implement the first AI service has a first constraint; the second network element provides a second AI service, and an output result of the second AI service satisfies the first constraint; the first constraint includes: a data type constraint; The first network element calls the second AI service of the second network element to obtain first data; The first network element processes the first data through the target model to obtain a first output result.

2. The data processing method according to claim 1, characterized in that: The determining the second network element on which the first AI service depends includes: The first network element sends the first constraint to an index library; The first network element receives indication information of the second network element obtained according to the first constraint and sent by the index library.

3. The data processing method according to claim 1 or 2, characterized in that: The first constraint includes: a data granularity constraint, and the data granularity constraint is carried in the first request message.

4. The data processing method according to claim 3, characterized in that: The first network element calling the second AI service of the second network element includes: The first network element sends a second request message to the second network element, where the second request message carries the data granularity constraint, and the second request message indicates calling the second AI service of the second network element.

5. The data processing method according to any one of claims 1 to 4, characterized in that: The first constraint includes: the confidence level of the data is not less than a first confidence level.

6. The data processing method according to any one of claims 2 to 5, characterized in that: Before the first network element receives the first request message, the method further includes: The first network element sends first registration information to the index library, where the first registration information includes: a granularity supported by an output result of the first AI service.

7. The data processing method according to claim 6, characterized in that: The first registration information also includes: the confidence level supported by the output result of the first AI service.

8. The data processing method according to any one of claims 1 to 7, characterized in that: The first request message carries a second constraint, where the second constraint is a granularity constraint of an output result obtained by the first AI service. After the first network element receives the first request message, before the first network element determines, according to the first request message, a second network element on which the first AI service depends, the method further includes: The first network element determines, based on the first request message, from a plurality of models and outputs the target model that satisfies the second constraint.

9. A data processing method, characterized in that: include: The third network element receives the first constraint sent by the first network element, wherein the first network element provides a first AI service; input data of a target model for implementing the first AI service has a first constraint; the first constraint includes: a data type constraint; The third network element determines, in the index library according to the first constraint, indication information of the second network element, the second network element provides a second AI service, and an output result of the second AI service satisfies the first constraint; The third network element sends instruction information of the second network element to the first network element, so that the first network element calls the second AI service of the second network element to obtain first data; the first network element processes the first data through the target model to obtain a first output result.

10. The data processing method according to claim 9, characterized in that: Before the third network element receives the first constraint sent by the first network element, the method further includes: The third network element receives first registration information of the first network element, where the first registration information includes: a granularity supported by an output result of the first AI service.

11. The data processing method according to claim 10, characterized in that: The first registration information also includes: the confidence level supported by the output result of the first AI service.

12. The data processing method according to any one of claims 9 to 11, characterized in that: Before the third network element receives the first constraint sent by the first network element, the method further includes: The third network element receives registration information of an AI network element, the AI ​​network element includes the second network element, and the registration information of the AI ​​network element includes: a data type supported by an output result of an AI service provided by the AI ​​network element; The third network element determines, according to the first constraint, indication information of the second network element in the index library, including: The third network element determines, in the index library according to the registration information of the AI ​​network element, the indication information of the second network element that satisfies the first constraint.

13. A communication device, characterized in that: including a transceiver unit and a processing unit; The transceiver unit is configured to receive a first request message, where the first request message indicates calling the first AI service; The processing unit is used to determine, according to the first request message, a second network element on which the first AI service depends; wherein the input data of the target model used to implement the first AI service has a first constraint; the second network element provides a second AI service, and an output result of the second AI service satisfies the first constraint; the first constraint includes: a data type constraint; calling the second AI service of the second network element to obtain first data; processing the first data through the target model to obtain a first output result.

14. The communication device according to claim 13, characterized in that: The processing unit is specifically used for: Calling the transceiver unit to send the first constraint to the index library; Acquire indication information of the second network element obtained according to the first constraint and sent by the index library and received by the transceiver unit.

15. The communication device according to claim 13 or 14, characterized in that: The first constraint includes: a data granularity constraint, and the data granularity constraint is carried in the first request message.

16. The communication device according to claim 15, characterized in that: The processing unit is specifically configured to call the transceiver unit to send a second request message to the second network element, where the second request message carries the data granularity constraint, and the second request message indicates calling the second AI service of the second network element.

17. The communication device according to any one of claims 13 to 16, characterized in that: The first constraint includes: the confidence level of the data is not less than a first confidence level.

18. The communication device according to any one of claims 14 to 17, characterized in that: The transceiver unit is further used to send first registration information to the index library, where the first registration information includes: the granularity supported by the output result of the first AI service.

19. The communication device according to claim 18, characterized in that: The first registration information also includes: the confidence level supported by the output result of the first AI service.

20. The communication device according to any one of claims 13 to 19, characterized in that: The first request message carries a second constraint, where the second constraint is a granularity constraint of an output result obtained by the first AI service. The processing unit is further used to determine, from multiple models, the target model whose output satisfies the second constraint based on the first request message.

21. A communication device, characterized in that: including a transceiver unit and a processing unit; The transceiver unit is used to receive a first constraint sent by a first network element, wherein the first network element provides a first AI service; input data of a target model for implementing the first AI service has a first constraint; the first constraint includes: a data type constraint; The processing unit is configured to determine, in the index library according to the first constraint, indication information of a second network element, where the second network element provides a second AI service, and an output result of the second AI service satisfies the first constraint; The transceiver unit is also used to send indication information of the second network element to the first network element, so that the first network element calls the second AI service of the second network element to obtain first data; the first network element processes the first data through the target model to obtain a first output result.

22. The communication device according to claim 21, characterized in that The transceiver unit is further used to receive first registration information of the first network element, where the first registration information includes: a granularity supported by an output result of the first AI service.

23. The communication device according to claim 22, characterized in that: The first registration information also includes: the confidence level supported by the output result of the first AI service.

24. The communication device according to any one of claims 21 to 23, characterized in that: The transceiver unit is also used for: receiving registration information of an AI network element, the AI ​​network element including the second network element, the registration information of the AI ​​network element including: a data type supported by an output result of an AI service provided by the AI ​​network element; The processing unit is called to determine, in an index library according to the registration information of the AI ​​network element, indication information of the second network element that satisfies the first constraint.

25. A communication device, characterized in that: include: A processor and a memory, wherein the processor is coupled to the memory; The memory is used to store programs; The processor is configured to execute the program in the memory so that the communication device executes the method according to any one of claims 1 to 8, or the communication device executes the method according to any one of claims 9 to 12.

26. A communication system, characterized in that: The communication system comprises the communication device according to claim 13 and the communication device according to claim 21.

27. A computer-readable storage medium comprising instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 8, or enable the computer to execute the method according to any one of claims 9 to 12.

28. A computer program product comprising instructions, which, when executed on a computer, causes the computer to perform the method as claimed in any one of claims 1 to 8, or causes the computer to perform the method as claimed in any one of claims 9 to 12.

Images