Network node, server, and program for mobile communication network
By dynamically controlling the delivery format of analysis services in mobile communication networks, the system addresses resource constraints in NFs, enabling efficient use of learning models and improving analysis accuracy.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-03-01
- Publication Date
- 2026-03-13
AI Technical Summary
In mobile communication networks, the second mode of analysis using a fine-tuned learning model in network functions (NFs) requires significant computing resources, which may not be available in all NFs, limiting its applicability.
A mechanism is implemented to dynamically control the delivery format of analysis services by determining whether to use the first or second mode based on the NF's computing resources and availability, allowing the system to select the appropriate learning model for each NF.
This approach ensures that NFs with insufficient resources can still utilize analysis services effectively, while those with sufficient resources can leverage the more accurate second mode, optimizing resource usage and improving analysis accuracy.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to a technique for using a learning model in a mobile communication system.
Background Art
[0002] Non-Patent Document 1 discloses a configuration for data analytics services in a mobile communication network. In the following description, a network function (NF) that provides a service is denoted as a producer NF, and an NF that uses (subscribes to) the service provided by the producer NF is denoted as a consumer NF. Note that the producer NF can also operate as a consumer NF at the same time.
[0003] According to Non-Patent Document 1, a network data analytics function (NWDAF), which is a producer NF that provides data analytics services, is provided in the core network of a mobile communication network. The consumer NF obtains analysis results by subscribing to the data analytics services provided by the NWDAF.
[0004] The consumer NF is implemented, for example, in an external data network (DN) connected to the mobile communication network, such as a server arranged on the Internet. The server is operated, for example, by a service provider that provides a specific service, such as a video distribution service, to a wireless device (WD) of the mobile communication network. The service provider can obtain analysis results such as the perceived quality of experience (QoE) of users using the WD that uses the service of the service provider by subscribing the data analytics services to the server (NF) operated by the service provider.
[0005] The data analysis service can be provided in either the first mode shown in Figure 1(A) or the second mode shown in Figure 1(B). In the first mode, only the NWDAF uses the learning model. The NWDAF obtains analysis results by taking information held by the mobile communication network (hereinafter referred to as NW information) as input to the learning model and transmits the obtained analysis results to the NF. In the second mode, the NWDAF uses the first learning model and the NF uses the second learning model. The first learning model held by the NWDAF and the second learning model held by the NF can be generated from the learning model held by the NWDAF in the first mode. In the second mode, the NWDAF obtains intermediate information by taking NW information as input to the first learning model and transmits the obtained intermediate information to the NF. The NF then obtains analysis results by taking the intermediate information as input to the second learning model. [Prior art documents] [Non-patent literature]
[0006] [Non-Patent Document 1] 3GPP(registered trademark) TS23.288, V18.3.0, September 2023 [Overview of the project] [Problems that the invention aims to solve]
[0007] In the second mode, the analysis accuracy can be improved because the second learning model can be fine-tuned in the NF. On the other hand, in the second mode, processing using the second learning model must be performed in the NF, which increases the computing resources required for the NF. For example, if the NF does not have sufficient computing resources, the second mode may not be available at all.
[0008] This disclosure provides a mechanism for appropriately controlling the delivery format of analytical services. [Means for solving the problem]
[0009] According to one aspect of this disclosure, the server includes a transmission means for transmitting to the network node whether it is possible to use a mode in which an analysis service provided by a network node of a mobile communication network is obtained by acquiring intermediate information from the network node and using the intermediate information as input to a learning model to obtain analysis results. [Effects of the Invention]
[0010] According to this disclosure, a mechanism is provided to appropriately control the delivery method of the analysis service. [Brief explanation of the drawing]
[0011] [Figure 1] A diagram illustrating the delivery model of the analysis service. [Figure 2] system configuration diagram. [Figure 3] A diagram showing an example sequence. [Figure 4] A flowchart for determining the delivery method. [Figure 5] A diagram showing an example of a network node configuration. [Figure 6] A diagram showing an example server configuration. [Modes for carrying out the invention]
[0012] The embodiments will be described in detail below with reference to the attached drawings. Note that the following embodiments do not limit the invention as defined in the claims, and not all combinations of features described in the embodiments are essential to the invention. Two or more features from the multiple features described in the embodiments may be combined arbitrarily. Furthermore, identical or similar configurations will be given the same reference numeral, and redundant descriptions will be omitted.
[0013] Figure 2 is a diagram of the communication system according to this embodiment. According to Figure 2, the core network 100 of the mobile communication network includes a network node 1 that implements NWDAF. In the following description, network node 1 will also be referred to as NWDAF. The core network 100 also includes various network functions of a different type than NWDAF.
[0014] NWDAF provides analysis services. To provide analysis services, NWDAF stores, or is configured to access, a storage device that stores, one or more first learning models and one or more second learning models corresponding to each of the first learning models. Each learning model is identified by an identifier. The identifier may be configured to identify a pair of first learning models and the second learning models used with those first learning models. In the first mode, NWDAF generates analysis results to send to NF by using the first learning models and the second learning models corresponding to those first learning models. In the second mode, NWDAF generates intermediate information to send to NF by using the first learning models. NF then generates analysis results by using the second learning models that are paired with the first learning models used by NWDAF.
[0015] In Figure 2, Server 3 is operated by a service provider that provides services such as video distribution to wireless devices (WDs) in a mobile communication network, and is located on an external data network (DN) connected to the core network 100, such as the internet. Server 3 implements an NF (Consumer NF) that utilizes the analysis services provided by NWDAF. In the following description, Server 3 will also be referred to as NF.
[0016] Figure 3 is a sequence diagram when starting to use the analysis service provided by NWDAF. In S10, the NF sends an analysis request message to the NWDAF. The analysis request message includes information necessary to identify the learning model to be used for the analysis, for example, specific information indicating the analysis result (analysis content) to be obtained. Further, the analysis request message may include availability information indicating whether the second mode can be used for the NF. The availability information can be pre-set for the NF by, for example, the service provider operating the NF. When the availability information indicates that the second mode is applicable, the analysis request message further includes resource information indicating the amount of computer resources of the NF that can be used for the inference process using the second learning model.
[0017] In S11, the NWDAF determines whether to apply the first mode or the second mode to the NF based on the information received together with the analysis request message from the NF. If it is determined to apply the second mode, in S12, the NWDAF sends the second learning model that outputs the analysis result required by the NF to the NF. Thereafter, the NWDAF generates intermediate information based on the first learning model that is a pair of the second learning model distributed to the NF and sends it to the NF, and the NF obtains the analysis result based on the intermediate information from the NWDAF.
[0018] Note that if it is determined to apply the first mode, the process of S12 is not performed. If it is determined to apply the first mode, the NWDAF sends the analysis result required by the NF to the NF by using a pair of the first learning model and the second learning model that output the analysis result required by the NF.
[0019] Figure 4 is a flowchart of the determination process executed by the NWDAF in S11 of FIG. 3. In S20, the NWDAF determines whether it is indicated in the approval information received from the NF that the application of the second mode is possible. If the application of the second mode is not possible, the NWDAF determines to apply the first mode in S23. If the application of the second mode is possible, the NWDAF selects a second learning model that outputs the analysis result required by the NF from the second learning model stored in the NWDAF or the storage device. The analysis result required by the NF is indicated by specific information. Then, the NWDAF S21 determines whether there is a second learning model executable by the NF in the selected second learning model. The second learning model executable by the NF is a second learning model whose required amount of computer resources for execution is less than or equal to the amount of computer resources available to the NF. The amount of computer resources available to the NF is indicated in the resource information.
[0020] If there is no second learning model executable by the NF, the NWDAF determines to apply the first mode in S23. On the other hand, if there is a second learning model executable by the NF and the option S24 is not executed, the NWDAF determines to apply the second mode in S22. When it is determined to apply the second mode, the NWDAF transmits the second learning model executable by the NF to the NF in S12 of FIG. 3. If there are multiple second learning models executable by the NF, the NWDAF selects one second learning model from the multiple second learning models executable by the NF by an arbitrary method and transmits it to the NF.
[0021] With the above configuration, for an NF that does not want to use the second mode or an NF for which the second mode cannot be applied due to the limitation of the computer resources of the NF, the first mode is applied, the use of the second mode is permitted, and for an NF for which the second mode can be applied, the second mode can be applied. Therefore, the provision form of the analysis service can be appropriately controlled.
[0022] Subsequently, the option S24This explains how to perform this action. In this case, NWDAF will S24 In this step, NWDAF determines whether the amount of information in the intermediate information is less than or equal to the amount of information in the analysis result for each of the one or more second learning models that can be executed by NF. If there is a second learning model in which the amount of information in the intermediate information is less than or equal to the amount of information in the analysis result, NWDAF decides to apply the second mode. In this case, NWDAF sends the second learning model in which the amount of information in the intermediate information is less than or equal to the amount of information in the analysis result to NF. On the other hand, if the amount of information in the intermediate information is greater than the amount of information in the analysis result for all of the one or more second learning models that can be executed by NF, NWDAF decides in S23 to apply the first mode.
[0023] Optional S24 By performing this action, it is possible to suppress the increase in the amount of information sent from NWDAF to NF.
[0024] Furthermore, if NF already possesses a second learning model, NF can indicate its applicability in the feasibility information of the analysis request message and include the identifier of the second learning model possessed by NF in the analysis request message. If NF possesses a second learning model, NWDAF can determine to apply the second mode. In this case, NWDAF does not need to transmit the second learning model in S12 of Figure 3. NWDAF also generates intermediate information using the first learning model corresponding to the second learning model notified by NF.
[0025] Furthermore, if there are multiple other NWDAFs within the mobile communication network, or if the NF has acquired a second learning model from another mobile communication network, the NWDAF may not be able to acquire the first learning model corresponding to the second learning model notified by the NF. Therefore, if the NWDAF cannot acquire the first learning model corresponding to the second learning model held by the NF, the NWDAF will determine to apply the first mode.
[0026] <Device configuration> Figure 5 is a configuration diagram of network node 1 according to several embodiments. 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 Figure 5 can be realized by one or more processors executing computer programs stored in the one or more memory devices. Network node 1 can also be realized as a single device. Alternatively, network node 1 can be realized as multiple devices that can communicate with each other. Note that Figure 5 shows only the parts necessary for understanding the embodiments, and network node 1 may include other functional blocks not shown.
[0027] The processing unit 11 performs processing necessary to collect network information and provide analysis services such as inference and generation of intermediate information based on the network information. The transmitting unit 14 and receiving unit 13 perform transmission and reception processing with other networks. For example, the receiving unit 13 is configured to receive availability information from the server 3 requesting the use of the analysis service, indicating whether it is possible to use the second mode.
[0028] The determination unit 12 determines whether to apply the first mode or the second mode to the server 3. For example, if the determination unit 12 receives information from the server 3 indicating that it is not possible to use the second mode, it determines to apply the first mode. If it receives information indicating that it is possible to use the second mode, it may determine whether to apply the first mode or the second mode based on the amount of computing resources available on the server 3 for processing using the learning model. Specifically, if the amount of computing resources available on the server 3 is equal to or greater than the amount of computing resources required for processing using the second learning model, the determination unit 12 may determine to apply the second mode. Otherwise, it may determine to apply the first mode.
[0029] The determination unit 12 can also be used to determine which mode to apply to the server 3 based on the amount of information in the intermediate information and the amount of information in the analysis results. For example, the determination unit 12 can determine that if the amount of computing resources available on the server 3 is greater than or equal to the amount of computing resources required for processing using the second learning model, and the amount of information in the intermediate information is less than or equal to the amount of information in the analysis results, then the second mode should be applied; otherwise, the first mode should be applied.
[0030] Furthermore, if the server 3 receives an identifier for the second learning model along with information indicating whether the second mode can be used, the determination unit 12 determines whether the processing unit 11 can generate intermediate information to be input to the second learning model identified by the identifier and send it to the server 3. In other words, the determination unit 12 determines whether the processing unit 11 can use the first learning model corresponding to the second learning model identified by the identifier. If the determination unit 12 determines that it cannot generate intermediate information to be input to the second learning model identified by the identifier and send it to the server 3, it determines that the first mode should be applied. fixed Otherwise, it may be determined that the second mode should be applied.
[0031] Figure 6 is a configuration diagram of server 3 according to several embodiments. 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. Each functional block shown in Figure 6 can be realized by one or more processors executing computer programs stored in the one or more memory devices. Server 3 can also be realized as a single device. Alternatively, server 3 can be realized as multiple devices that can communicate with each other. Note that Figure 6 shows only the parts necessary to understand the embodiments, and server 3 may include other functional blocks that are not shown.
[0032] The processing unit 31 performs the processing necessary to use the analysis service provided by the network node 1. The transmitting unit 33 and the receiving unit 32 perform transmission and reception processing with the network node 1. For example, the transmitting unit 33 sends availability information to the network node 1 indicating whether it is possible to use the second mode when using the analysis service. Furthermore, when the transmitting unit 33 sends availability information indicating that it is possible to use the second mode, it also sends resource information to the network node 1 indicating the amount of computing resources available on the server 3 for processing using the second learning model. Furthermore, if the processing unit 31 has a second learning model, the transmitting unit 33 sends the identifier of the second learning model held by the processing unit 32 to the network node 1 along with the availability information indicating that it is possible to use the second mode.
[0033] Furthermore, according to this disclosure, a computer program is provided that, when executed on one or more processors of a device having one or more processors, causes the device to function as a network node 1 or server 3, and a computer-readable storage medium storing the computer program is provided. In addition, according to this disclosure, a method is provided for the process shown in Figures 3 and 4, for the process to be executed by network node 1, for the process to be executed by server 3, for the process to be executed by network node 1 / server 3, and a computer-readable storage medium storing the computer program is provided.
[0034] In summary, a mechanism is provided to appropriately control the delivery format of the analysis service. Therefore, it becomes possible to contribute to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs): "Build resilient infrastructure, promote sustainable industrialization and foster innovation."
[0035] The invention is not limited to the embodiments described above, and various modifications and changes are possible within the scope of the gist of the invention. [Explanation of symbols]
[0036] 31: Processing unit, 32: Receiving unit, 33: Transmitting unit
Claims
1. a server, A server comprising a transmitting means for transmitting to the network node whether or not it is possible to use a mode in which, when using an analysis service provided by a network node of a mobile communication network, intermediate information is acquired from the network node and the mode in which analysis results are acquired by using the intermediate information as input to a learning model.
2. The server described in claim 1, wherein when transmitting the availability information indicating that the mode can be used, the transmitting means transmits resource information to the network node indicating the amount of computer resources available to the server for processing using the learning model.
3. The server of claim 1, wherein, when the server has the learning model, the transmitting means transmits to the network node the availability information indicating that the mode can be used and an identifier of the learning model held by the server.
4. The server of claim 1 , wherein the network node is a node implementing a Network Data Analysis Function (NWDAF).
5. A program that, when executed by one or more processors of a device having one or more processors, causes the device to function as the server according to any one of claims 1 to 4.
6. A network node of a mobile communication network that provides an analysis service, comprising: a receiving means for transmitting intermediate information from a server requesting the use of the analysis service to the server and receiving availability information indicating whether a second mode for obtaining an analysis result in the server can be used by using the intermediate information as an input for a learning model; a determination means for determining to apply to the server a first mode in which the network node transmits an analysis result to the server when the availability information indicating that the second mode is not available is received, and for determining whether to apply to the server the first mode or the second mode based on the amount of computer resources available to the server for processing using the learning model when the availability information indicating that the second mode is available is received; A network node comprising:
7. The network node described in claim 6, wherein the determination means determines that the second mode is to be applied to the server if the amount of computer resources available to the server is equal to or greater than the amount of computer resources required for processing using the learning model, and otherwise determines that the first mode is to be applied to the server.
8. The network node described in claim 6, wherein the determination means determines that the second mode is to be applied to the server if the amount of computer resources available to the server is greater than or equal to the amount of computer resources required for processing using the learning model and the amount of information of the intermediate information is less than or equal to the amount of information of the analysis result, and otherwise determines that the first mode is to be applied to the server.
9. The network node described in claim 6, wherein when the determination means receives from the server an identifier of the learning model together with the availability information indicating that the second mode can be used, the determination means determines to apply the second mode to the server if the intermediate information to be used as input to the learning model identified by the identifier can be sent to the server, and determines to apply the first mode to the server if the intermediate information to be used as input to the learning model identified by the identifier cannot be sent to the server.
10. The network node of claim 7 , wherein the network node is a node implementing a Network Data Analysis Function (NWDAF).
11. A program that, when executed by one or more processors of a device having one or more processors, causes the device to function as a network node according to any one of claims 6 to 10.