Network node of mobile communication network, server, and computer-readable storage medium

Abstract

In the present invention, a network node of a mobile communication network comprises: a processing means that, using first information collected in the mobile communication network as an input for a first learning model, acquires intermediate information, which is an output of the first learning model, , and performs processing for adding a first identifier to the intermediate information; and a transmission means that transmits the intermediate information and the first identifier added to the intermediate information to a server allocated to a network outside the mobile communication network.

Description

Network node, server and computer-readable storage medium for mobile communication network

[0001] The present disclosure relates to a technology for utilizing learning models in a mobile communication system.

[0002] Non-Patent Document 1 discloses a configuration for a data analysis service in a mobile communication network. According to Non-Patent Document 1, an application function (AF), which is a service consumer, establishes a subscription by subscribing to a data analysis service provided by a mobile communication network. The AF may be implemented in an external data network (DN) connected to the mobile communication network, such as a server located on the Internet. The core network of the mobile communication network includes a network data analysis function (NWDAF). The NWDAF is one of various network functions (NFs) located in the core network. The NWDAF performs analysis using a learning model and transmits the analysis results to the AF with which the subscription is established.

[0003] A server implementing the AF is operated by, for example, a service provider that provides a specific service, such as a video distribution service, to wireless devices (WDs) in a mobile communication network. By subscribing the AF (server) to a data analysis service, the service provider can obtain, for example, analysis results on the quality of experience (QoE) of users of WDs who use the service of the service provider from the mobile communication network.

[0004] In addition, to estimate the quality of experience related to a service, information held by the NFs of the mobile communication network (hereinafter referred to as NW information) and information held by the service provider, such as information about the application for using the service (hereinafter referred to as APP information), are usually used. Therefore, the NWDAF performs inference using the NW information collected from each NF in the mobile communication network and the APP information notified by the AF as inputs to a learning model, and notifies the AF of the result (analysis result).

[0005] Here, there are various indicators for quality of experience, and the QoE indicator that the NWDAF can provide to the AF as an analysis result may differ from the QoE indicator that the service provider wants to provide. To solve this problem, a configuration in which the AF performs analysis using a learning model may be considered. However, in this case, the NWDAF must transmit network information to the AF in order for the AF to perform inference. However, the network information is also data that directly indicates the structure, status, etc. of the mobile communication network, and there are cases in which it is not permitted to disclose the network information itself to the outside.

[0006] 3GPP TS23.288, V18.3.0, September 2023 3GPP S2-2306424

[0007] In order to obtain analysis results in the AF without transmitting the network information itself to the AF, for example, the vertical federated learning (VFL) configuration disclosed in Non-Patent Document 2 can be applied. Specifically, as shown in FIG. 1 , a first learning model is arranged in the NWDAF, and a second learning model is arranged in the AF. The NWDAF generates intermediate information using the network information as input to the first learning model and transmits the intermediate information to the AF. The AF performs inference using the intermediate information acquired from the NWDAF and the APP information collected by the AF as input to the second learning model. By applying the VFL configuration, the information transmitted from the NWDAF to the AF can be treated as intermediate information different from the network information, while allowing the AF to perform inference. Note that the first learning model and the second learning model can be generated based on a learning model that performs inference using the network information and APP information as input, for example.

[0008] In the configuration of Fig. 1, in order to update (learn) the first learning model and the second learning model, the AF needs to determine the error in the output of the second learning model to update the second learning model, and also needs to calculate the error in the intermediate information and feed this error in the intermediate information back to the NWDAF. Note that the NWDAF updates the first learning model based on the error in the intermediate information.

[0009] However, non-patent document 1 does not define a procedure for placing two mutually associated learning models (the above-mentioned first learning model and second learning model) in an NWDAF and an AF that has established a subscription to the services provided by the NWDAF, nor a procedure for updating the two learning models based on VFL.

[0010] According to one aspect of the present disclosure, a network node of a mobile communication network includes a processing means that performs processing to obtain intermediate information that is the output of a first learning model by using first information collected in the mobile communication network as input to the first learning model, and to assign a first identifier to the intermediate information, and a transmission means that transmits the intermediate information and the first identifier assigned to the intermediate information to a server located in an external network of the mobile communication network.

[0011] According to the present disclosure, a mechanism is provided that enables VFL by a network node in a mobile communication network and a server outside the mobile communication network.

[0012] Other features and advantages of the present invention will become apparent from the following description taken in conjunction with the accompanying drawings, in which the same or similar elements are designated by the same reference numerals.

[0013] FIG. 1 is an explanatory diagram of VFL, a system configuration diagram, a diagram showing an example of a sequence, a diagram showing an example of a sequence, a diagram showing an example of a configuration of a network node, and a diagram showing an example of a configuration of a server.

[0014] Hereinafter, the embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments do not limit the scope of the invention as claimed, and not all combinations of features described in the embodiments are necessarily essential to the invention. Two or more of the features described in the embodiments may be arbitrarily combined. Furthermore, the same reference numerals are used for the same or similar components, and redundant explanations will be omitted.

[0015] <First Embodiment> Fig. 2 is a configuration diagram of a communication system according to this embodiment. According to Fig. 2, a core network 100 of a mobile communication network includes a network node 1. The network node 1 may be a device that implements the above-described NWDAF function. Although not shown, the core network 100 also includes various network functions of types different from the NWDAF. Also, in Fig. 2, a server 3 is located in an external data network (DN) connected to the core network 100, such as the Internet. The server 3 is, for example, a device that implements the above-described AF, and may be operated by a service provider that provides services such as video distribution services to wireless devices (WDs) in the mobile communication network.

[0016] In this embodiment, two learning models that operate in cooperation are used so that the server 3 can perform inference based on NW information that the network node 1 can collect and APP information that the server 3 has. One of the two learning models, a first learning model, is arranged in the network node 1, and the other, a second learning model, is arranged in the server 3. The network node 1 inputs the NW information to the first learning model and acquires intermediate information that is the output of the first learning model. The network node 1 transmits the intermediate information to the server 3. The server 3 inputs the APP information and the intermediate information from the network node 1 to the second learning model, thereby acquiring analysis results such as the QoE of users of WD that are output by the second learning model. Note that the APP information may include information that the server 3 has as well as information that the server 3 collects from sources other than the mobile communication network.

[0017] 3 is a sequence diagram according to this embodiment. In FIG. 3, the server 3 and the network node 1 are assumed to transmit and receive messages directly. However, the server 3 and the network node 1 may also be configured to transmit and receive messages via another device, for example, a device that implements a network publishing function (NEF). In this case, the other device may either directly forward a message received from either the server 3 or the network node 1 to the other, or may transmit a separate message to the other that conveys the contents of the message received from either the server 3 or the network node 1.

[0018] S10 and S11 correspond to a process (subscription establishment process) in which the server 3 establishes a subscription for the analysis service with the mobile communication network. In S10, the server 3 sends a subscription request message to the network node 1. The subscription request message may be, for example, Nnwdaf_AnalyticsSubscription_Subscribe.

[0019] In this embodiment, the subscription request message of S10 includes a parameter "VFL" to indicate to the network node 1 that, instead of receiving the analysis result by the network node 1, the network node 1 uses the first learning model, the server 3 uses the second learning model, and the server 3 requests to receive intermediate information from the network node 1. In response to the subscription request message, the network node 1 sends a subscription acceptance message to the server 3 in S11. The subscription acceptance message includes a subscription ID for identifying the subscription to be established.

[0020] In this embodiment, the network node 1 determines the subscription ID and notifies the server 3. However, the server 3 may determine the subscription ID. In this case, the server 3 can include the subscription ID in the subscription request message of S10. In this case, the processing of S11 can be omitted.

[0021] In this embodiment, the server 3 has a learning model that is the basis for the first learning model and the second learning model. The learning model outputs, for example, an inference result of QoE based on NW information and APP information. In S12, the server 3 generates a first learning model to be used by the network node 1 and a second learning model to be used by the server 3 based on the learning model. The first learning model outputs intermediate information based on the NW information. The second learning model outputs an inference result based on the intermediate information and APP information. Note that the processing of S12 can also be performed before the start of the processing of FIG. 3.

[0022] S13 and S14 correspond to a distribution process of the learned model. In S13, the network node 1 sends a model request message to the server 3 requesting the first learned model. The model request message includes a subscription ID that allows the server 3 to recognize for which subscription the model is being requested. In response to the model request message, the server 3 sends the first learned model to the network node 1 in S14. The message for sending the first learned model to the network node 1 also includes a subscription ID that allows the network node 1 to recognize for which subscription the model is being requested.

[0023] In this embodiment, the network node 1 transmits a model request message to the server 3, and in response, the server 3 transmits the first learning model to the network node 1. However, the server 3 may be configured to transmit the first learning model to the network node 1 without a request from the network node 1. In that case, the processing of S13 is omitted.

[0024] Steps S15 and onward correspond to inference and learning model update processing. In step S15, the network node 1 transmits to the server 3 a message including intermediate information output by the first learning model by inputting the current NW information into the first learning model. At this time, the network node 1 includes in the message an information ID, which is an identifier for identifying the intermediate information, and a subscription ID, which indicates which subscription the intermediate information relates to. According to FIG. 3 , the network node 1 sets the information ID of the intermediate information transmitted in step S15 to "1." In step S16, the server 3 uses the second learning model based on the intermediate information with the information ID "1" and the current APP information to infer, for example, the QoE of a user of a specific WD.

[0025] In S17, the network node 1 transmits the intermediate information output by the first learning model by inputting the NW information at that time to the server 3. According to Fig. 3, the network node 1 sets the information ID of the intermediate information transmitted in S17 to "2". In S18, the server 3 uses the second learning model based on the intermediate information with the information ID "2" and the APP information at that time to infer, for example, the QoE of a user of a specific WD.

[0026] In S19, the server 3 evaluates the error of the QoE inference result in S16, for example, by backpropagation, and updates the second learning model while determining the error of the intermediate information whose information ID is "1". Then, in S20, the server 3 transmits a message including error information indicating the error of the intermediate information to the network node 1. At this time, the server 3 includes the information ID and subscription ID in the message so that the network node 1 can identify which intermediate information the error indicated by the error information belongs to. Since the error information transmitted in S20 indicates the error of the intermediate information whose information ID is "1" received in S15, an information ID of value "1" is included in S20. Although not shown, the network node 1 updates the first learning model based on the error of the intermediate information received in S20. Furthermore, although not shown, the processing corresponding to S19 and S20 is also performed on the intermediate information whose information ID is "2". In addition, if the information ID is determined not to be unique within a certain subscription, but to be unique within all subscriptions processed by network node 1, there is no need to include the subscription ID in the messages at S15, S17 and S20.

[0027] As described above, in this embodiment, a subscription ID is determined when a subscription is established, and the learning model is distributed using this subscription ID. With this configuration, the first learning model used by the network node 1 and the second learning model used by the server 3 are each associated with the same subscription ID. In other words, two learning models that operate in coordination are associated via the subscription ID. Therefore, one of the two learning models that operate in coordination can be placed in the network node 1, and the other can be placed in the server 3. In addition, an information ID is included in each piece of intermediate information transmitted by the network node 1. This allows the network node 1 to identify which intermediate information the error information corresponds to, thereby enabling VFL.

[0028] Second Embodiment Next, the second embodiment will be described, focusing on the differences from the first embodiment. In the first embodiment, the server 3 has a learning model that is the basis for the first learning model and the second learning model, and the server 3 transmits the first learning model to the network node 1. In this embodiment, the network node 1 has a learning model that is the basis for the first learning model and the second learning model, and the network node 1 transmits the second learning model to the server 3.

[0029] 4 is a sequence diagram according to this embodiment. S50 and S11 correspond to a process (subscription establishment process) in which the server 3 establishes a subscription for an analysis service between the server 3 and the mobile communication network. In S50, the server 3 transmits a subscription request message to the network node 1. The subscription request message includes an analysis ID and an application ID (App ID) in addition to the parameter "VFL." The analysis ID indicates the content to be analyzed (inferred), and the App ID indicates the application used in the service to be analyzed (inferred). The network node 1 determines the learning model to use based on the combination of the analysis ID and the App ID. Therefore, in S51, the network node 1 generates a first learning model and a second learning model based on the learning model determined based on the combination of the analysis ID and the App ID.

[0030] S52 and S53 correspond to the distribution process of the learning model. In S52, the server 3 sends a model request message to the network node 1 requesting a second learning model. The model request message includes a subscription ID that allows the network node 1 to recognize for which subscription the model is being requested. In response to the model request message, the network node 1 sends the second learning model to the server 3 in S53. The message for sending the second learning model to the server 3 also includes a subscription ID that allows the server 3 to recognize for which subscription the model is being requested.

[0031] In this embodiment, the network node 1 determines the subscription ID and notifies the server 3, but the server 3 may determine the subscription ID. In this embodiment, the server 3 transmits a model request message to the network node 1, and in response, the network node 1 transmits the second learning model to the server 3. However, the network node 1 may transmit the second learning model to the server 3 without a request from the server 3. In this case, the processing of S52 is omitted.

[0032] The processing from S15 onwards is the same as in the first embodiment.

[0033] As described above, in this embodiment, a subscription ID is determined when a subscription is established, and the training model is distributed using this subscription ID. With this configuration, the first training model used by the network node 1 and the second training model used by the server 3 are each associated with the same subscription ID. In other words, two training models that operate in coordination are associated via the subscription ID. Therefore, one of the two training models that operate in coordination can be placed in the network node 1, and the other can be placed in the server 3. Furthermore, by including the subscription ID in the model request message, information that identifies the training model, i.e., the analysis ID and App ID, can be omitted from the model request message. In addition, an information ID is included in each piece of intermediate information transmitted by the network node 1. This makes it possible to identify which intermediate information the error information corresponds to, thereby enabling VFL.

[0034] <Device Configuration> Fig. 5 is a configuration diagram of a network node 1 according to some embodiments. The network node 1 includes, for example, one or more processors and one or more memory devices. The one or more memory devices may include volatile memory devices and non-volatile memory devices. Each functional block shown in Fig. 5 may be realized by one or more processors executing a computer program stored in the one or more memory devices. Furthermore, the network node 1 may be realized by a single device. Alternatively, the network node 1 may be realized by multiple devices capable of communicating with each other.

[0035] The collection unit 11 collects network information from each device in the mobile communication network. The processing unit 12 manages subscriptions for the services provided. When a subscription is established, a subscription ID (second identifier) ​​is assigned as described above. The processing unit 12 has a first learning model. In the first embodiment, the first learning model is received from the server 3. In the second embodiment, the first learning model is generated by the processing unit 11 based on the learning model possessed by the processing unit 11. The processing unit 11 inputs network information into the first learning model, thereby acquiring intermediate information that is the output of the first learning model. The processing unit 11 then performs a process of assigning an information ID (first identifier) ​​to each piece of intermediate information output by the first learning model.

[0036] The transmitting unit 14 transmits the intermediate information and the information ID assigned to the intermediate information to the server 3. The receiving unit 13 receives the error information and the information ID assigned to the error information from the server 3. When the receiving unit 13 receives the error information, the processing unit 12 determines, based on the information ID assigned to the error information, which of the intermediate information transmitted to the server 3 the error indicates, and updates the first learning model.

[0037] FIG. 6 is a configuration diagram of a server 3 according to some embodiments. The server 3 includes, for example, one or more processors and one or more memory devices. The one or more memory devices may include volatile memory devices and non-volatile memory devices. The functional blocks shown in FIG. 6 may be realized by one or more processors executing computer programs stored in the one or more memory devices. The server 3 may also be realized by a single device. Alternatively, the server 3 may be realized by multiple devices capable of communicating with each other.

[0038] The processing unit 31 manages subscriptions to services. When a subscription is established, a subscription ID (second identifier) ​​is assigned as described above. The processing unit 31 has a second learning model. In the first embodiment, the second learning model is generated by the processing unit 31 based on the learning model held by the processing unit 31. In the second embodiment, the second learning model is received from the network node 1.

[0039] The receiving unit 32 receives the intermediate information and the information ID (first identifier) ​​assigned to the intermediate information from the network node 1. When the receiving unit 32 receives the intermediate information, the processing unit 31 performs inference by inputting the received intermediate information and APP information into a second learning model. The processing unit 31 then updates the second learning model based on the inference error and performs processing to determine the error of the intermediate information that was input to the second learning model. The processing unit 31 also determines the information ID assigned to the intermediate information in which an error was determined, and notifies the transmitting unit 33 of the error information indicating the determined error and the determined information ID. The transmitting unit 33 transmits the error information and information ID notified by the processing unit 31 to the network node 1.

[0040] The present disclosure provides a computer program, which, when executed by one or more processors of an apparatus having one or more processors, causes the apparatus to function as the network node 1 or the server 3, and a computer-readable storage medium storing the computer program. Furthermore, the present disclosure provides a method executed by the network node 1 or the server 3 with respect to the processes shown in Figures 3 and 4, a computer program causing the network node 1 / server 3 to execute the method shown in Figures 3 and 4, and a computer-readable storage medium storing the computer program.

[0041] The invention is not limited to the above-described embodiment, and various modifications and variations are possible within the scope of the gist of the invention.

[0042] This application claims priority based on Japanese Patent Application No. 2023-185751, filed on October 30, 2023, the entire contents of which are incorporated herein by reference.

Claims

1. A network node of a mobile communications network, comprising: a processing means for performing a process of acquiring intermediate information which is the output of a first learning model by using first information collected in the mobile communications network as an input of the first learning model, and assigning a first identifier to the intermediate information; and a transmission means for transmitting the intermediate information and the first identifier assigned to the intermediate information to a server located in an external network of the mobile communications network.

2. A network node as described in claim 1, further comprising a receiving means for receiving error information and the first identifier assigned to the error information from the server, wherein the processing means determines, based on the first identifier assigned to the error information, which of a plurality of pieces of intermediate information transmitted to the server the error indicated by the error information corresponds to.

3. The network node of claim 2, wherein the server subscribes to a service provided by the network node, and performs inference by using the intermediate information as input to a second learning model, and the error in the intermediate information is based on an error in the inference.

4. The network node of claim 3, wherein the first learning model is received from the server in response to a request message including a second identifier that identifies the subscription sent to the server.

5. The network node of claim 3, wherein the first learning model is received from the server together with a second identifier that identifies the subscription.

6. The network node of claim 3, wherein the second learning model is sent to the server in response to receiving a request message from the server including a second identifier that identifies the subscription.

7. The network node of claim 3, wherein the second learning model is transmitted to the server together with a second identifier that identifies the subscription.

8. A network node according to any one of claims 1 to 7, wherein the network node is a node implementing a Network Data Analysis Function (NWDAF).

9. A computer-readable storage medium having stored thereon a program which, when executed by one or more processors of an apparatus having one or more processors, causes the apparatus to function as a network node according to any one of claims 1 to 8.

10. A server that subscribes to services provided by a network node of a mobile communication network, comprising: a receiving means for receiving from the network node intermediate information output by a first learning model by using first information collected in the mobile communication network as input to the first learning model, and a first identifier assigned to the intermediate information; a processing means for performing inference by using the intermediate information and second information as input to a second learning model, updating the second learning model based on an error in the inference, and performing a process of determining an error in the intermediate information; and a transmitting means for transmitting error information indicating the error in the intermediate information and the first identifier assigned to the intermediate information to the network node.

11. The server of claim 10, wherein the second learning model is received from the network node in response to a request message including a second identifier that identifies the subscription sent to the network node.

12. The server of claim 10, wherein the second learning model is received from the network node together with a second identifier that identifies the subscription.

13. The server of claim 10, wherein the first learning model is transmitted to the network node in response to receiving a request message from the network node including a second identifier that identifies the subscription.

14. The server of claim 10, wherein the first learning model is transmitted to the network node together with a second identifier that identifies the subscription.

15. A computer-readable storage medium having stored thereon a program that, when executed by one or more processors of an apparatus having one or more processors, causes the apparatus to function as a server according to any one of claims 10 to 14.

Images