Information processing apparatus

Abstract

An information processing apparatus of the present invention executes an application. The information processing apparatus includes a control unit. The control unit issues, directly and / or via a network exposure function, a request for processing a learning model used in the application to a device having a network data analysis function for a mobile network.

Description

Technical Field

[0001] The present disclosure relates to an information processing device. Background Art

[0002] The first standard of the fifth generation mobile communication system (so-called 5G) was formulated as Rel-15 in 2018. In addition, services corresponding to 5G started in Japan in March 2020. The fifth generation mobile communication system has features of enhanced mobile broadband (eMBB), ultra-reliable low latency communication (URLLC), and massive machine type communication (mMTC).

[0003] Rel-18, which will be worked on from June 2022, is the first release in Beyond 5G (B5G), and one of the new advances is the application of artificial intelligence (AI) technology to 5G.

[0004] Reference List

[0005] Patent Literature

[0006] Patent Document 1: JP 2022-520279 A Summary of the invention

[0007] Technical issues

[0008] Due to the ultra-high speed, ultra-reliable low-latency communication and massive machine-type communication characteristics of 5G, real-time transmission of 4K or 8K moving images is expected. In addition, use cases such as voice interaction applications using AI / machine learning (ML) are increasing. As a method for processing AI / ML, for example, a method in which processing is implemented in the cloud in addition to the client, or a method in which distributed processing is implemented on the client and the cloud can be considered. In addition to applications, it is also desirable to have a mechanism for guaranteeing latency in 5G communications through end-to-end (E2E) including AI / ML processing.

[0009] Patent document 1 described above discloses a technology for maintaining data collected by a network data analysis function (NWDAF) of a 5G system in a network function repository entity. However, there is no mention of AI / ML processing in Patent document 1.

[0010] Therefore, the present disclosure provides a mechanism that can further reduce latency in 5G communications including AI / ML processing in end-to-end (E2E).

[0011] It should be noted that the foregoing problem or purpose is only one of multiple problems or purposes that can be solved or achieved by multiple embodiments disclosed in this specification.

[0012] Solutions to the Problem

[0013] The information processing device of the present disclosure executes an application. The information processing device includes a control unit. The control unit issues a request for processing of a learning model used in the application through a network exposure function and / or directly to a device with a network data analysis function of a mobile network. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a diagram showing a configuration example of a neural network (NN) model.

[0015] Figure 2 is a diagram showing a configuration example of a deep neural network (DNN) model.

[0016] Figure 3 is a diagram showing an example of AI / ML processing through a 5G system.

[0017] Figure 4 is a diagram showing another example of AI / ML processing through a 5G system.

[0018] Figure 5 is a diagram showing the configuration of a network architecture of a 5G system.

[0019] Figure 6 is a block diagram showing a configuration example of an information processing device according to an embodiment of the present disclosure.

[0020] Figure 7 is a diagram showing a configuration example of a base station device according to an embodiment of the present disclosure.

[0021] Figure 8 is a block diagram showing a configuration example of a wireless communication device according to an embodiment of the present disclosure.

[0022] Fig. 9 is a diagram showing an example of a connectivity model for edge computing in a 5G system.

[0023] Fig.10 is a diagram showing an architecture example of an edge application server for non-roaming periods.

[0024] Fig.11 is a diagram showing another architecture example of an edge application server for non-roaming periods.

[0025] Fig.12 is a diagram showing an example of a QoS architecture for 5GS.

[0026] Fig.13 is a diagram showing a configuration example of an architecture for data collection of a network data analysis function.

[0027] Fig.14is a diagram showing a configuration example of an architecture for acquiring network analysis information.

[0028] Fig.15 is a diagram showing a configuration example of an architecture for storing collected data and network analysis information.

[0029] Fig.16 is a diagram showing a configuration example of an architecture for acquiring an AI / ML model.

[0030] Fig.17 is a diagram showing a processing example in which the AF acquires analysis information from the NWDAF.

[0031] Fig.18 is a diagram showing a processing example in which the AF acquires analysis information from the NWDAF.

[0032] Fig.19 is a diagram showing a configuration example of an application utilizing AI / ML according to an embodiment of the present disclosure.

[0033] Fig. 20 is a diagram showing a configuration example of an application utilizing AI / ML according to an embodiment of the present disclosure.

[0034] Fig.21 It is a diagram of an example of a method for providing information of a 5G system through an NEF and issuing a request to the 5G system according to an embodiment of the present disclosure.

[0035] Fig. 22 is a diagram showing an example of a method of providing information of a 5G system through an NEF and issuing a request to the 5G system according to an embodiment of the present disclosure.

[0036] Fig.23 is a sequence diagram illustrating an example of a PDU session establishment process according to an embodiment of the present disclosure.

[0037] Fig.24 is a sequence diagram illustrating another example of a PDU session establishment process according to an embodiment of the present disclosure.

[0038] Fig.25 is a sequence diagram illustrating another example of a PDU session establishment process according to an embodiment of the present disclosure.

[0039] Fig.26 : is a flowchart showing a processing example when the operation of the learning model of the application according to the embodiment of the present disclosure is distributed through the client and the cloud. DETAILED DESCRIPTION

[0040] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. It should be noted that in the present specification and the accompanying drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and thus repeated descriptions thereof are omitted.

[0041] In addition, in the present specification and the drawings, similar components of the embodiments may be distinguished from each other by appending different letters or numbers after the same reference numerals. It should be noted that, in the case where it is not particularly necessary to distinguish similar components from each other, only the same reference numerals are appended.

[0042] One or more embodiments (including examples, modified examples and application examples) described below can be implemented independently. On the other hand, at least a portion of each of the multiple embodiments described below can be appropriately combined with at least a portion of another embodiment for implementation. The multiple embodiments can include novel features that are different from each other. Therefore, the multiple embodiments can contribute to solving different purposes or problems, and can show different effects.

[0043] <<0. Introduction>>

[0044] As described above, due to the ultra-high speed, ultra-reliable low-latency communication and massive machine-type communication characteristics of 5G, real-time transmission of 4K or 8K moving images is expected. In addition, use cases using AI / ML such as voice interaction applications are increasing. As a method for processing AI / ML, for example, a method of implementing processing in the cloud in addition to the client, or a method of implementing distributed processing on the client and the cloud can be considered. In addition to applications, it is also desirable to have a mechanism for guaranteeing latency in 5G communications through E2E including AI / ML processing.

[0045] Conventionally, techniques for maintaining data collected by the NWDAF of a 5G system in a network function repository entity are known, but procedures for causing the NWDAF of a 5G system to process AI / ML models processed by applications have not been considered.

[0046] In addition, depending on the application, multiple PDU sessions are established for one application. Assume a case where a data service through E2E is provided via multiple PDU sessions. At this time, it is desirable to have a mechanism for ensuring synchronization between multiple PDU sessions.

[0047] Therefore, the information processing device according to the technology proposed in the present invention executes the application. The information processing device issues a request to a device with a network data analysis function (NWDAF) of a mobile network for executing a process of a learning model used in the application. The information processing device issues a request to the device through a network exposure function and / or directly.

[0048] As a result, when the information processing device causes the cloud server to process all or part of the application including the AI / ML model through wireless communication, the information processing device can cause the NWDAF of the 5G system to process the AI / ML model. As a result, the 5G system can ensure the processing of the AI / ML model including wireless communication through E2E.

[0049] In addition, the information processing device can cause the edge application server to handle the processing of the AI / ML model to be processed by the cloud server. As a result, the information processing device can reduce the delay in data communication that occurs in association with the processing of the AI / ML model.

[0050] In addition, the information processing device can control the distributed processing of the AI / ML model based on the processing power of the wireless communication device. This makes it possible to protect the privacy of the data handled by the wireless communication device and reduce the load on the cloud server.

[0051] In addition, the information processing device can control the distributed processing of AI / ML models based on the quality of wireless communication. As a result, even in a dynamically changing wireless communication environment, the 5G system can still guarantee the processing of AI / ML models including wireless communication through E2E.

[0052] <<1. AI / ML models and their operational processing>>

[0053] Figure 1 1 is a diagram showing a configuration example of a neural network (NN) model. A neural network is used in machine learning (ML) and the like.

[0054] like Figure 1 As shown in , a neural network includes layers called input layers, hidden layers (or intermediate layers), and output layers. Each layer includes at least one node.

[0055] Each node is connected by edges. Each layer has a function called an activation function, and each edge is weighted.

[0056] Machine learning is one of the artificial intelligence methods that enables computers to learn and perform recognition, determination, or estimation similarly to humans. The processing performed by machine learning includes two types of processing, including learning processing and determination processing for recognition, determination, or estimation.

[0057] In the learning process, a device (not shown) implementing the learning causes the computer to learn the neural network model using the learning data and optimize the weighting factor of each edge. The model extracted as a result of the learning process is a learned model.

[0058] In the determination process, a device (not shown) implementing the determination inputs unknown data to the learned model. As a result of the operation process, the learned model outputs a result of identification, determination or estimation for the unknown data.

[0059] Figure 2 is a diagram showing a configuration example of a deep neural network (DNN) model. A deep neural network is used in deep learning.

[0060] A deep neural network model includes multiple hidden layers. In machine learning, the number of hidden layers of a neural network model is limited by the computational power of a computer. On the other hand, in deep learning, a DNN model with a larger number of hidden layers is used due to the improvement of the computational power of a computer. By using a learned DNN model learned using a large amount of data, the accuracy of recognition, determination, or estimation is improved.

[0061] Examples of general algorithms used in deep learning include

[0062] - Convolutional Neural Networks (CNN);

[0063] - Recurrent Neural Networks (RNN);

[0064] -Fully connected neural network;

[0065] -Long Short-Term Memory (LSTM); and

[0066] -Autoencoders.

[0067] In CNN, the hidden layer includes layers called convolutional layers and pooling layers. In the convolutional layer, filtering by convolution operation is performed, and data called feature maps are extracted. In the pooling layer, information of the feature map output from the convolutional layer is compressed and down-sampled.

[0068] RNN has a network structure in which the value of a hidden layer is cyclically input to the hidden layer. In RNN, for example, short-cycle time series data is processed.

[0069] In a fully connected neural network, all intermediate layers are fully connected layers (all nodes between layers are connected to each other). Fully connected neural networks are mainly used in the field of speech recognition.

[0070] By introducing parameters called memory cells that maintain the state of the intermediate layer into the output of the intermediate layer of RNN, LSTM can maintain the influence of the output in the distant past. In other words, LSTM processes time series data with longer periods than RNN.

[0071] In an autoencoder, low-dimensional features that can reproduce input data through unsupervised learning are extracted. Autoencoders are effective for noise removal, dimensionality reduction, etc.

[0072] Representative technical fields in which deep learning is used include four fields: image recognition, speech recognition, natural language processing, and anomaly detection by robots. In image recognition, deep learning is used for applications such as tagging people in social network services (SNS) and autonomous driving. In speech recognition, deep learning is applied to smart speakers, etc. In natural language processing, deep learning is applied to search and automatic translation through browsers. In anomaly detection by robots, deep learning is used in airports, railways, manufacturing sites, etc.

[0073] Figure 3 is a diagram showing an example of AI / ML processing through a 5G system. Figure 3 The operation of AI / ML models is shown being processed by the cloud.

[0074] like Figure 3 As shown in , the wireless communication device 10 as a client sends the input data of the AI / ML model to the cloud server 40 (an example of an application server to be described later) through the base station device 20 and the core network 30 of the 5G system.

[0075] The cloud server 40 inputs each input data received from the wireless communication device 10 to each node of the input layer of the AI / ML model and implements the operation of the AI / ML model. The cloud server 40 returns the operation result of the output layer of the AI / ML model to the wireless communication device 10 through the core network 30 and the base station device 20 of the 5G system.

[0076] Figure 4 is a diagram showing another example of AI / ML processing through a 5G system. Figure 4 The operation of the AI / ML model is shown in a distributed manner by the client and the cloud. Here, the AI / ML model is divided into a first AI / ML model and a second AI / ML model.

[0077] like Figure 4 As shown in , the wireless communication device 10 as a client inputs each input data to each node of the input layer of the divided first AI / ML model, and implements the operation of the first AI / ML model. The wireless communication device 10 sends the data of each node of the output layer of the first AI / ML model as intermediate processing data to the cloud server 40 through the base station device 20 and the core network 30 of the 5G system.

[0078] The cloud server 40 inputs each intermediate processing data received from the wireless communication device 10 to each node of the input layer of the divided second AI / ML model, and implements the operation of the second AI / ML model. The cloud server 40 returns the operation result of the output layer of the second AI / ML model to the wireless communication device 10 through the core network 30 and the base station device 20 of the 5G system.

[0079] <<2. System Configuration Example>>

[0080] <2.1、5G system architecture>

[0081] Figure 5 5G system is a diagram showing the configuration of a network architecture of a 5G system. Hereinafter, the 5G system is referred to as 5GS (5G system).

[0082] like Figure 5 As shown in FIG. 5 , 5GS includes user equipment (UE) 10, (R) AN 20 and 5G core (5GC) 30. It should be noted that 5GC 30 is also referred to as NG core (NGC) or core network. In addition, the mark of (R) AN indicates a base station device including a radio access network (RAN) and an access network (AN). In addition, UE 10 is the wireless communication device described above.

[0083] By connecting to the 5GS via the Internet through an application server (AS) 40 that processes the application, the UE 10 can use the application through the 5G service.

[0084] When an entity providing an application (e.g., a service provider) has a contract such as a service layer agreement (SLA) with a public land mobile network (PLMN) operator providing 5G services, the application server 40 is arranged in the 5GC 30 as a DN 340. Alternatively, the application server 40 and the DN 340 may be connected via a dedicated line or a virtual private network (VPN). It should be noted that the application server 40 is also referred to as a cloud server, or may be provided in the form of an edge server.

[0085] The control plane function group of 5GS includes access and mobility management function (AMF) 301, network exposure function (NEF) 302, and network repository function (NRF) 303. In addition, the control plane function group includes network slice selection function (NSSF) 304, policy control function (PCF) 305, session management function (SMF) 306, and unified data management (UDM) 307. The control plane function group includes application function (AF) 308, authentication server function (AUSF) 309, and UE radio capability management function (UCMF) 310. The control plane function group includes edge application server discovery function (EASDF) 311, location management function (LMF) 312, and NWDAF 313. As described above, the control plane function group includes multiple network functions (NFs).

[0086] Here, the AF 308 may operate as an NF that processes a control plane of the application server 40. The AF 308 may be installed in a physically identical device as an entity that is logically different from the application server 40, for example, in the application server 40. In addition, the AF 308 operates as an NF that processes a control plane for 5GS applications and may be arranged in the 5GC 30.

[0087] The UDM 307 includes a unified data repository (UDR) that maintains and manages subscription information, and a front end (FE) unit that processes subscription information. The AMF 301 implements mobility management. The SMF 306 implements session management.

[0088] UCMF 310 maintains UE radio capability information corresponding to all UE radio capability IDs in the PLMN. UCMF 310 is responsible for assigning UE radio capability IDs assigned by each PLMN.

[0089] It should be noted that the NSSF 304, the NEF 302, the PCF 305, the EASDF 311, the LMF 312, and the NWDAF 313 will be described later.

[0090] Namf is a service-based interface provided by AMF 301. Nsmf is a service-based interface provided by SMF 306. Nnef is a service-based interface provided by NEF 302.

[0091] Npcf is a service-based interface provided by PCF 305. Nudm is a service-based interface provided by UDM 307. Naf is a service-based interface provided by AF 308. Nnrf is a service-based interface provided by NRF 303.

[0092] Nnssf is a service-based interface provided by NSSF 304. Nausf is a service-based interface provided by AUSF 309. Nucmf is a service-based interface provided by UCMF 310.

[0093] Neasdf is a service-based interface provided by EASDF 311. Nlfm is a service-based interface provided by LMF 312. Nnwdaf is a service-based interface provided by NWDAF 313.

[0094] By issuing a request or subscription to a service provided by another network function, each NF can receive a response or notification from the service. That is, each NF exchanges information with another NF through a request / response or subscription / notification via each service-based interface.

[0095] The user plane function (UPF) 330 has a function of user plane processing. The data network (DN) 340 has a function of realizing connection to services unique to the mobile network operator (MNO), the Internet, and third-party services. The UPF 330 acts as a transmission processing unit for user plane data processed by the application server 40. The UPF 330 also acts as a gateway connected to the (R)AN 20.

[0096] Here, the 5GS may configure each NF of the 5GC 30 through virtualization or containerization, or may install each NF in a cloud server. In addition, the 5GS may dynamically set each NF in a reconfigurable manner using software defined networking (SDN).

[0097] (R)AN 20 has a function of achieving connection to RAN and connection to AN other than RAN. (R)AN 20 includes a base station called gNB or ng-eNB. RAN may also be called next generation (NG) RAN.

[0098] In addition, the functions of the (R)AN 20 are divided into a central unit (CU) that processes a packet data convergence protocol (PDCP) sublayer and higher L2 / L3 functions, and a distributed unit (DU) that processes a radio link control (RLC) sublayer and lower L2 / L1 functions. The functions of the (R)AN 20 can be arranged in a distributed manner through an F1 interface.

[0099] In addition, the functions of the DU are divided into a radio unit (RU) that processes the LOW PHY sublayer and the radio unit (radio) and a DU that processes the RLC, medium access control (MAC) and HIGH PHY sublayer. For example, the functions of the DU can be arranged in a distributed manner through a fronthaul that complies with the evolved common public radio interface (eCPRI).

[0100] Here, 5GS can configure the functions of CU and / or DU through virtualization or container, and the functions can be installed in the cloud server. In addition, 5GS can dynamically set the functions of CU and / or DU in a reconfigurable manner using SDN.

[0101] UE 10 and AMF 301 exchange information with each other via reference point N1. (R)AN 20 and AMF 301 exchange information with each other via reference point N2. SMF 306 and UPF 330 exchange information with each other via reference point N4.

[0102] <2.2. Configuration Example of Information Processing Equipment>

[0103] Next, we will refer to Figure 6 A configuration example of the information processing device 300 according to an embodiment of the present disclosure is described. Figure 6 is a block diagram showing a configuration example of the information processing device 300 according to an embodiment of the present disclosure.

[0104] The information processing device 300 is a device that implements the functions of the NF and AF 308 of the core network 30 and the application server 40. The information processing device 300 is, for example, a server device. The information processing device 300 may be a device generally referred to as a cloud server or an edge server.

[0105] like Figure 6 As shown in , the information processing device 300 includes a communication unit 31, a storage unit 32, and a control unit 33. It should be noted that Figure 6 The configuration shown in the figure is a functional configuration, and the hardware configuration may be different from Figure 6 In addition, the functions of the information processing device 300 may be implemented in a plurality of physically separated configurations in a distributed manner. For example, the information processing device 300 may include a plurality of server devices.

[0106] The communication unit 31 is a communication interface for communicating with other devices. The communication unit 31 may be a network interface or a device connection interface. For example, the communication unit 31 may be a local area network (LAN) interface such as a network interface card (NIC), or may be a USB interface including a universal serial bus (USB) host controller, a USB port, etc. In addition, the communication unit 31 may be a wired interface or a wireless interface. The communication unit 31 acts as a communication device of the information processing device 300. The communication unit 31 communicates with the base station device 20, another NF node or an AN node under the control of the control unit 33.

[0107] The storage unit 32 is a data readable / writable storage device such as a dynamic random access memory (DRAM), a static random access memory (SRAM), a flash memory, or a hard disk. The storage unit 32 serves as a storage device of the information processing device 300.

[0108] The control unit 33 is a controller that controls each unit of the information processing device 300. The control unit 33 is implemented by a processor such as a central processing unit (CPU), a microprocessing unit (MPU), or a graphics processing unit (GPU). For example, the control unit 33 is implemented by executing various programs stored in a storage device inside the information processing device 300 by the processor using a random access memory (RAM) or the like as a work area. It should be noted that the control unit 33 can be implemented by an integrated circuit such as an application-specific integrated circuit (ASIC) or a field programmable gate array (FPGA). Any of the CPU, MPU, GPU, ASIC, and FPGA can be regarded as a controller.

[0109] <2.3. Configuration example of base station equipment>

[0110] Next, the base station device 20 will be described. The base station device 20 is a communication device that operates a cell and provides a wireless communication service to one or more wireless communication devices 10 located within the coverage area of ​​the cell. The cell is operated according to any wireless communication system such as LTE or NR. The base station device 20 is connected to a core network. The core network is connected to a packet data network through a gateway device. In addition, the base station device 20 operates a beam that can be identified by a synchronization signal / PBCH block (SSB), and transmits and receives data to / from one or more wireless communication devices 10 through one or more beams.

[0111] It should be noted that the base station device 20 may include an assembly of multiple physical or logical devices. For example, the base station device 20 in the embodiment of the present disclosure may be divided into multiple devices of a baseband unit (BBU) and a RU, and may be interpreted as an assembly of multiple devices. In addition or alternatively, the base station device 20 in the embodiment of the present disclosure may be any one or both of the BBU and the RU. The BBU and the RU may be connected via a predetermined interface (e.g., eCPRI). In addition or alternatively, the RU may be referred to as a remote radio unit (RRU) or a radio point (RD). In addition or alternatively, the RU may correspond to the gNB-DU described later. In addition or alternatively, the BBU may correspond to the gNB-CU described later. Alternatively, the RU may be connected to the gNB-DU described later. In addition, the BBU may correspond to a combination of the gNB-CU and the gNB-DU described later. In addition or alternatively, the RU may be a device formed integrally with an antenna. The antenna included in the base station device 20 (e.g., an antenna formed integrally with the RU) may adopt an advanced antenna system and support MIMO (e.g., FD-MIMO) or beamforming. In the advanced antenna system, the antenna included in the base station device 20 (eg, an antenna integrated with the RU) may include, for example, 64 transmit antenna ports and 64 receive antenna ports.

[0112] In addition, a plurality of base station devices 20 may be connected to each other. One or more base station devices 20 may be included in a radio access network (RAN). That is, the base station device 20 may be simply referred to as a RAN, a RAN node, an AN, or an AN node. The RAN in LTE is referred to as an enhanced universal terrestrial RAN (EUTRAN). The RAN in NR is referred to as an NGRAN. The RAN in W-CDMA (UMTS) is referred to as an UTRAN. The base station device 20 of LTE is referred to as an evolved node B (eNodeB) or an eNB. That is, the EUTRAN includes one or more eNodeBs (eNBs). In addition, the base station device 20 of NR is referred to as a gNodeB or a gNB. That is, the NGRAN includes one or more gNBs. In addition, the EUTRAN may include a gNB (en-gNB) connected to a core network (EPC) in a communication system (EPS) of LTE. Similarly, the NGRAN may include an ng-eNB connected to a core network (5GC) in a 5G communication system (5GS). In addition or alternatively, when the base station device 20 is an eNB, a gNB, etc., the base station device 20 may be referred to as a 3GPP access. In addition or alternatively, when the base station device 20 is a wireless access point (e.g., an access point of WiFi (registered trademark)), the base station device 20 may be referred to as non-3GPP access. In addition or alternatively, the base station device 20 may be an optical extension device referred to as a remote radio head (RRH). In addition or alternatively, when the base station device 20 is a gNB, the base station device 20 may be referred to as a combination of the gNB central unit (CU) and the gNB distributed unit (DU) described above or any one of them. The gNB CU hosts multiple upper layers (e.g., RRC, SDAP, and PDCP) of the access layer for communication with the UE. On the other hand, the gNB-DU hosts multiple lower layers (e.g., RLC, MAC, and PHY) of the access layer. That is, among the messages and information described later, RRC signaling (e.g., various SIBs including MIB and SIB1, RRCSetup messages, and RRCReconfiguration messages) may be generated by the gNB CU, and the DCI and various physical channels (e.g., PDCCH and PBCH) described later may be generated by the gNB-DU. Alternatively, for example, in RRC signaling, some configurations (setting information) such as IE: cellGroupConfig may be generated by the gNB-DU, and the rest of the configurations may be generated by the gNB-CU. These configurations (setting information) may be sent and received to / from the F1 interface described later. The base station device 20 may be configured to be able to communicate with another base station device 20. For example, when the multiple base station devices 20 are a combination of eNBs or eNBs and en-gNBs, the respective base station devices 20 may be connected to each other via an X2 interface.In addition or alternatively, when the multiple base station devices 20 are a combination of gNB or gn-eNB and gNB, the devices can be connected to each other through the Xn interface. In addition or alternatively, when the multiple base station devices 20 are a combination of gNB CU and gNB DU, the devices can be connected to each other through the F1 interface described above. The messages / information described later (RRC signaling or DCI information and physical channels) can be transmitted between multiple base station devices 20 (for example, through the X2, Xn or F1 interface).

[0113] In addition, as described above, the base station device 20 can be configured to manage multiple cells. The cell provided by the base station device 20 is called a service cell. The service cell includes a primary cell (PCell) and a secondary cell (SCell). When providing dual connectivity (e.g., EUTRA-EUTRA dual connectivity, EUTRA-NR dual connectivity (ENDC), EUTRA-NR dual connectivity with 5GC, NR-EUTRA dual connectivity (NEDC) or NR-NR dual connectivity) for a UE (e.g., a wireless communication device 10), the PCell provided by the leading node (MN) and zero or one or more SCells are referred to as a leading cell group. In addition, the service cell may include a primary secondary cell or a primary SCG cell (PSCell). That is, when providing dual connectivity for a UE, the PSCell provided by the secondary node (SN) and zero or one or more SCells are referred to as a secondary cell group (SCG). Unless a special configuration (e.g., PUCCH on SCell) is made, the physical uplink control channel (PUCCH) is sent in the PCell and PSCell, but not in the SCell. In addition, radio link failure is also detected in PCell and PSCell, but not in SCell (does not need to be detected). As described above, PCell and PSCell have special roles in (multiple) serving cells and are therefore also referred to as special cells (SpCell). One downlink component carrier and one uplink component carrier can be associated with one cell. In addition, the system bandwidth corresponding to one cell can be divided into multiple bandwidth parts. In this case, one or more bandwidth parts (BWP) can be set for the UE, and one bandwidth part can be used for the UE as an active BWP. In addition, the radio resources (e.g., frequency band, parameter set (subcarrier spacing) and time slot configuration) that can be used by the wireless communication device 10 can be different according to the cell, component carrier or BWP.

[0114] Figure 7 is a diagram showing a configuration example of the base station device 20 according to an embodiment of the present disclosure. The base station device 20 is a communication device (radio system) that performs wireless communication with the wireless communication device 10. The base station device 20 is a type of information processing device.

[0115] The base station device 20 includes a communication unit 210, a storage unit 220, a network communication unit 230, and a control unit 240. It should be noted that Figure 7 The configuration shown in the figure is a functional configuration, and the hardware configuration may be different from Figure 7 The configuration shown in . In addition, the functions of the base station device 20 can be implemented in a plurality of physically separated devices in a distributed manner. For example, as described above, the functions of the base station device 20 are distributed to the CU and DU or the CU, DU and RU.

[0116] The communication unit 210 is a wireless communication interface for wireless communication with other communication devices (e.g., the wireless communication device 10 and another base station device 20). The communication unit 210 is a wireless transceiver that operates under the control of the control unit 240. The communication unit 210 can be compatible with a variety of wireless access systems. For example, the communication unit 210 can be compatible with both NR and LTE. The communication unit 210 can be compatible with another cellular communication system such as W-CDMA or cdma2000. In addition, the communication unit 210 can also be compatible with a wireless LAN communication system in addition to the cellular communication system. Of course, the communication unit 210 can be compatible with only one wireless access system.

[0117] The communication unit 210 includes a reception processing unit 211, a transmission processing unit 212, and an antenna 213. The communication unit 210 may include a plurality of reception processing units 211, a plurality of transmission processing units 212, and a plurality of antennas 213. It should be noted that when the communication unit 210 is compatible with a plurality of wireless access systems, each unit of the communication unit 210 may be configured separately for each wireless access system. For example, when the base station device 20 is compatible with NR and LTE, the reception processing unit 211 and the transmission processing unit 212 may be configured separately for NR and LTE.

[0118] The reception processing unit 211 processes an uplink signal received through the antenna 213. The reception processing unit 211 includes a wireless reception unit 211a, a demultiplexing unit 211b, a demodulation unit 211c, and a decoding unit 211d.

[0119] The wireless receiving unit 211a performs down-conversion, removal of unnecessary frequency components, control of amplification level, orthogonal demodulation, conversion to digital signals, removal of guard intervals, extraction of frequency domain signals by fast Fourier transform, etc. on the uplink signal. For example, it is assumed that the wireless access system of the base station device 20 is a cellular communication system such as LTE. At this time, the demultiplexing unit 211b separates the uplink channel (such as a physical uplink shared channel (PUSCH) or a physical uplink control channel (PUCCH)) and the uplink reference signal from the signal output from the wireless receiving unit 211a. The demodulation unit 211c demodulates the received signal using a modulation system such as binary phase shift keying (BPSK) or quadrature phase shift keying (QPSK) with respect to the modulation symbol of the uplink channel. The modulation system used by the demodulation unit 211c can be multi-level QAM, such as 16 quadrature amplitude modulation (16QAM), 64QAM or 256QAM. The decoding unit 211 d performs decoding processing on the coded bits of the demodulated uplink channel. The decoded uplink data and uplink control information are output to the control unit 240 .

[0120] The transmission processing unit 212 performs transmission processing on the downlink control information and the downlink data. The transmission processing unit 212 includes a coding unit 212a, a modulation unit 212b, a multiplexing unit 212c, and a wireless transmission unit 212d.

[0121] The coding unit 212a encodes the downlink control information and downlink data input from the control unit 240 using a coding system such as block coding, convolution coding or turbo coding. Here, the coding can be implemented using a polarization code or a low-density parity check code (LDPC code). The modulation unit 212b modulates the coded bits output from the coding unit 212a by a predetermined modulation system such as BPSK, QPSK, 16QAM, 64QAM or 256QAM. The multiplexing unit 212c multiplexes the modulation symbols and downlink reference signals of each channel and arranges the multiplexed results in a predetermined resource unit. The wireless transmission unit 212d performs various types of signal processing on the signal from the multiplexing unit 212c. For example, the wireless transmission unit 212d implements processing such as conversion to the time domain by fast Fourier transform, adding a guard interval, generating a baseband digital signal, converting to an analog signal, orthogonal modulation, up-conversion, removing redundant frequency components and power amplification. The signal generated by the transmission processing unit 212 is transmitted from the antenna 213 .

[0122] The storage unit 220 is a storage device capable of reading and writing data, such as a DRAM, an SRAM, a flash memory, or a hard disk. The storage unit 220 serves as a storage device of the base station device 20.

[0123] The network communication unit 230 is a communication interface for communicating with other devices (e.g., another base station device 20). For example, the network communication unit 230 is a LAN interface such as a NIC. The network communication unit 230 may be a USB interface including a USB host controller, a USB port, etc. In addition, the network communication unit 230 may be a wired interface or a wireless interface. The network communication unit 230 acts as a network communication device of the base station device 20. The network communication unit 230 communicates with other devices under the control of the control unit 240.

[0124] The control unit 240 is a controller that controls each unit of the base station device 20. The control unit 240 is implemented by a processor such as a CPU, an MPU, or a GPU, for example. For example, the control unit 240 is implemented by executing various programs stored in a storage device inside the base station device 20 by the processor using a RAM or the like as a work area. It should be noted that the control unit 240 can be implemented by an integrated circuit such as an ASIC or an FPGA. Any of the CPU, MPU, GPU, ASIC, and FPGA can be regarded as a controller.

[0125] <2.4. Configuration Example of Wireless Communication Device>

[0126] Will refer to Figure 8 A configuration example of the wireless communication device 10 according to an embodiment of the present disclosure is described. Figure 8 1 is a block diagram showing a configuration example of the wireless communication device 10 according to an embodiment of the present disclosure.

[0127] The wireless communication device 10 is a wireless communication device that implements wireless communication with the base station device 20. The wireless communication device 10 is, for example, a mobile phone, a smart device (smart phone or tablet device), a personal digital assistant (PDA), or a personal computer. The wireless communication device 10 may be a head-mounted display, VR glasses, etc. having the function of wirelessly transmitting and receiving data.

[0128] In addition, the wireless communication device 10 may be able to perform sidelink communication with another wireless communication device 10. The wireless communication device 10 may be able to use automatic retransmission techniques such as hybrid automatic repeat request (HARQ) when implementing sidelink communication. The wireless communication device 10 may be able to perform non-orthogonal multiple access (NOMA) communication with the base station device 20. It should be noted that the wireless communication device 10 may also be able to perform NOMA communication in communication (sidelink) with another wireless communication device 10. In addition, the wireless communication device 10 may be able to perform low power wide area (LPWA) communication with other communication devices (e.g., a base station device 20 and another wireless communication device 10). In addition, the wireless communication used by the wireless communication device 10 may be wireless communication using millimeter waves. It should be noted that the wireless communication used by the wireless communication device 10 (including sidelink communication) may be wireless communication using radio waves or may be wireless communication using infrared or visible light (optical wireless).

[0129] The wireless communication device 10 can implement communication by connecting to multiple base station devices or multiple cells at the same time. For example, when one base station device 20 can provide multiple cells, the wireless communication device 10 can perform carrier aggregation by using one cell as a pCell and another cell as an sCell. In addition, when each of the multiple base station devices 20 can provide one or more cells, the wireless communication device 10 can use one or more cells managed by one base station device (MN (e.g., MeNB or MgNB)) as a pCell or a pCell and (multiple) sCells and use one or more cells managed by another base station device 20 (SN (e.g., SeNB or SgNB)) as a pCell (PSCell) or a pCell (PSCell) and (multiple) sCells to achieve dual connectivity (DC). DC can be referred to as multi-connectivity (MC).

[0130] It should be noted that when a communication area is supported by cells of different base station devices 20 (multiple cells with different cell identifiers or the same cell identifier), multiple cells can be bundled, and the base station device 20 and the wireless communication device 10 can communicate with each other through carrier aggregation (CA) technology, dual connection (DC) technology or multi-connection (MC) technology. Alternatively, through the cells of different base station devices 20, the wireless communication device 10 and multiple base station devices 20 can communicate with each other through coordinated multi-point transmission and reception (CoMP) technology.

[0131] The wireless communication device 10 includes a communication unit 110, a storage unit 120, a network communication unit 130, an input / output unit 140, and a control unit 150. It should be noted that Figure 7The configuration shown in the figure is a functional configuration, and the hardware configuration may be different from Figure 7 Furthermore, the functionality of the wireless communication device 10 may be implemented in a distributed manner in a plurality of physically separate configurations.

[0132] The communication unit 110 is a signal processing unit for wireless communication with other wireless communication devices (e.g., a base station device 20 and another wireless communication device 10). The communication unit 110 operates under the control of the control unit 150. The communication unit 110 may be a wireless transceiver corresponding to one or more wireless access systems. For example, the communication unit 110 may be compatible with both NR and LTE. In addition to NR and LTE, the communication unit 110 may be compatible with W-CDMA and cdma2000. In addition, the communication unit 110 may be compatible with communication using NOMA.

[0133] The communication unit 110 includes a reception processing unit 111, a transmission processing unit 112, and an antenna 113. The communication unit 110 may include a plurality of reception processing units 111, a plurality of transmission processing units 112, and a plurality of antennas 113. The configurations of the communication unit 110, the reception processing unit 111, the transmission processing unit 112, and the antenna 113 are similar to the configurations of the communication unit 210, the reception processing unit 211, the transmission processing unit 212, and the antenna 214 of the base station device 20.

[0134] The storage unit 120 is a storage device capable of reading and writing data, such as a DRAM, an SRAM, a flash memory, or a hard disk. The storage unit 120 serves as a storage device of the wireless communication device 10.

[0135] The network communication unit 130 is a communication interface for communicating with other devices connected via a network. For example, the network communication unit 130 is a LAN interface such as a NIC. The network communication unit 130 may be a wired interface or a wireless interface. The network communication unit 130 serves as a network communication device of the wireless communication device 10. The network communication unit 130 communicates with other devices under the control of the control unit 150.

[0136] The input / output unit 140 is a user interface for exchanging information with a user. For example, the input / output unit 140 is an operating device for the user to perform various operations, such as a keyboard, a mouse, an operating key, and a touchpad. Alternatively, the input / output unit 140 is a display device such as a liquid crystal display or an organic electroluminescent display (organic EL display). The input / output unit 140 can be an acoustic device such as a speaker or a buzzer. In addition, the input / output unit 140 can be a lighting device such as a light emitting diode (LED) lamp. The input / output unit 140 acts as an input / output device (input device, output device, operating device, or notification device) of the wireless communication device 10.

[0137] The control unit 150 is a controller that controls each unit of the wireless communication device 10. The control unit 150 is implemented by a processor such as a CPU, an MPU, or a GPU, for example. For example, the control unit 150 is implemented by executing various programs stored in a storage device inside the wireless communication device 10 by the processor using a RAM or the like as a work area. It should be noted that the control unit 150 may be implemented by an integrated circuit such as an ASIC or an FPGA. Any of the CPU, MPU, GPU, ASIC, and FPGA may be regarded as a controller.

[0138] <<3. Network Slice Selection Auxiliary Information>>

[0139] Network slicing is a service unit obtained by dividing the communication services provided by 5G according to the communication attributes required for each service (such as data rate, latency, etc.).

[0140] Each network slice is assigned a single network slice selection assistance information (S-NSSAI) as information for assisting in selecting a network slice (network slice selection assistance information). The S-NSSAI includes a set of mandatory slice / service types (SSTs) and optional slice distinguishers (SDs), the former including 8 bits for identifying the slice type and the latter including 24 bits for distinguishing different slices in the same SST.

[0141] The network slice setting information includes one or more configured (NSSAI). The serving public land mobile network (PLMN) may set the configured NSSAI applied for each PLMN in the UE 10. Alternatively, the home PLMN (HPLMN) may set the default configured NSSAI in the UE 10. The UE 10 under the serving PLMN may use the default configured NSSAI only when the configured NSSAI for the serving PLMN is not set in the UE 10.

[0142] It should be noted that the default configured NSSAI may be pre-set in the UE 10. In addition, the UDM 307 of the HPLMN may provide or update the default configured NSSAI using UE parameter update through UDM control plane processing. The configured NSSAI includes one or more S-NSSAIs.

[0143] The requested NSSAI is the NSSAI provided from the UE 10 to the serving PLMN during the registration process. The requested NSSAI should be one of the following:

[0144] -NSSAI is configured by default;

[0145] - NSSAI configured;

[0146] - is permitted to use the NSSAI or a part thereof; and

[0147] - NSSAI obtained by adding one or more S-NSSAI included in the configured NSSAI to the allowed NSSAI or a part thereof.

[0148] The allowed NSSAI is, for example, the NSSAI provided by the serving PLMN to the UE 10 during the registration process. The allowed NSSAI indicates the value(s) of one or more S-NSSAIs available in the current registration area of ​​the current serving PLMN.

[0149] The rejected S-NSSAI indicates a value of an S-NSSAI whose use is not allowed in at least one tracking area in the current registration area of ​​the current serving PLMN, among one or more S-NSSAIs included in the requested NSSAI.

[0150] The subscribed S-NSSAI is an S-NSSAI that can be used by the UE 10 in the PLMN according to the subscription information.

[0151] The NSSF 304 determines the allowed NSSAI and the configured NSSAI, and determines the AMF set that is a candidate list for the AMF 301.

[0152] <<4. Exposure to AF and Requests from AF>>

[0153] The service provider who manages and operates the application server 40 can obtain information from each NF of the 5G system through the NEF 302 within the scope of the SLA with the PLMN operator. The NEF 302 can securely protect and disclose the capabilities and events of each NF to the service provider.

[0154] In addition, the service provider can use AF 308, which is a control plane function for the application server 40, to send an application function request (AF request) to each NF of the 5G system through NEF 302.

[0155] The AF request shall include the mandatory traffic description, target UE identifier(s) and AF transaction identifier.

[0156] Traffic description is information for identifying traffic. Traffic description includes a set of data network name (DNN) and S-NSSAI and application identifier (application ID) or traffic filtering information.

[0157] The DNN corresponds to the Access Point Name (APN) used in 4G or earlier systems.

[0158] The application ID is information for identifying an application that handles traffic on the user plane. The application ID is used by the UPF 330 to identify traffic of the application.

[0159] The traffic filtering information is information for classifying traffic, and is, for example, a 5-tuple including a source IP, a source port number, a destination IP, a destination port number, and a protocol number.

[0160] In addition, the AF request may include information about the location of potential applications according to the conditions. Here, the information about the location of potential applications is provided as a list including a data network access identifier (DNAI) for identifying user plane access to one or more DNs 340 as candidates for implementing the application.

[0161] In addition, the AF request may optionally include the following information, among others:

[0162] - space validity conditions;

[0163] -N6 traffic routing requirements;

[0164] - Possibility of application relocation;

[0165] -UE IP address reservation indication;

[0166] - temporary validity conditions;

[0167] - Information about AF subscriptions to SMF events;

[0168] - Information on IP replacement for the edge application server (EAS) in 5GC 30;

[0169] -User plane delay requirements;

[0170] - information about changes in AF; and

[0171] -Instructions for Edge Application Server Relocation (EAS Relocation).

[0172] The spatial validity condition is provided in the form of a valid area. When the AF request is a request for determining the routing of traffic in the SMF 306, the spatial validity condition indicates that the routing is only applied to the traffic of the UE 10 located at a specific location. When the AF request is a request for notification of an event for registering path management of the user plane, the spatial validity condition indicates that the notification is only applied to the traffic of the UE 10 located at a specific location.

[0173] The information about N6 traffic routing requirements is information provided for each DNAI. The information about N6 traffic routing requirements may include routing profile ID and N6 traffic routing information.

[0174] Here, N6 is a reference point between UPF 330 and DN 340. Routing profile ID is identification information for referring to a policy on routing agreed in advance between AF 308 and 5GC 30. N6 traffic routing information includes information necessary for transmitting traffic to DNAI.

[0175] The application relocation possibility is information indicating whether the application can be relocated after the location of the application is selected by 5GC 30.

[0176] The UE IP address reservation indication indicates whether the IP address of the UE 10 associated with the traffic identified by the traffic description should be retained. When receiving the instruction from the AF 308, the 5GC 30 retains the IP address of the UE 10 by avoiding reselection of the UPF 330 after the UPF 330 is selected.

[0177] The temporary validity condition is provided in the form of a time interval or period indicating the time when the AF request is applied. When the AF request is a request for determination of the routing of traffic in the SMF 306, the temporary validity condition indicates when the routing is applied. When the AF request is a request for notification of an event for registering path management of the user plane, the temporary validity condition indicates when the notification occurs.

[0178] The AF request including information about the AF subscription for SMF events is a request for registering notifications of changes in the user plane path associated with the traffic identified by the traffic description. The request includes a subscription type, a notification target address for receiving event notifications, etc.

[0179] When the subscription type is advance notification, SMF 306 sends a notification of path change before setting a new user plane path. When the subscription type is post notification, SMF 306 sends a notification of path change after setting a new user plane path.

[0180] The information for IP replacement of edge application servers gives instructions on the identifier of the source edge application server and the identifier of the target edge application server for the service through edge computing. The identifiers here are, for example, the IP addresses and port numbers of the source and target edge application servers.

[0181] The user plane delay requirement is the delay in the user plane considered when relocating the target edge application server. At the network location, AF 308 can request the user plane delay requirement from 5GC 30 through an AF request, so that SMF 306 can determine the relocation of PSA-UPF based on the AF request. The network location here is the estimated value of the delay in the user plane between UE 10 and the candidate PDU session anchor (PSA)-UPF for the network location known to SMF 306.

[0182] The information on the AF change is information on the relocation of the AF 308. The information on the AF change includes an AF ID as information for identifying the target AF 308 of the change destination. The instruction to relocate the edge application server is an instruction to relocate the application.

[0183] <<5. Location information management function>>

[0184] In UE-based mode, the UE 10 may obtain assistance data from the LMF 312 and perform measurements and calculations on the position of the Global Navigation Satellite System (GNSS). The UE 10 may, for example, use a well-known method known as Assisted-GNSS.

[0185] It should be noted that, for example, when a signal from a GNSS cannot be received, the UE 10 can detect information about the location by means other than the GNSS. For example, the UE 10 can detect information about its own location using a method called sidelink positioning, WLAN positioning, Bluetooth (registered trademark) positioning, or Terrestrial Beacon System (TBS) positioning.

[0186] In sidelink positioning, the UE 10 measures the sidelink received signal strength indicator (SL RSSI) through sidelink communication with surrounding UEs 10 or roadside units (RSUs). The UE 10 calculates the position based on the information about the position of each UE 10 or roadside unit and the measured value of the SL RSSI. In addition, the UE 10 can measure the round trip time (RTT) with the surrounding UEs 10 or roadside units and calculate its position.

[0187] In WLAN positioning, the UE 10 measures the received signal strength indicator (RSSI) for each WLAN access point. The UE 10 calculates the position based on the information about the known coordinates of each WLAN access point and its measured value. In addition, the UE 10 can measure the RTT with the WLAN access point and calculate the position.

[0188] In Bluetooth positioning, the UE 10 may measure RSSI about each Bluetooth beacon and calculate a position based on information about known coordinates of each Bluetooth beacon and its measured value.

[0189] When the UE 10 includes a GNSS receiver, the UE 10 performs measurements on GNSS such as code phase, Doppler or carrier phase in UE-assisted mode. The UE 10 sends these measurements to the LMF 312, and the LMF 312 calculates the position of the UE 10.

[0190] In addition, for a UE 10 that cannot receive a signal from a GNSS or does not include a GNSS even when the UE 10 includes a GNSS receiver, the LMF 312 may acquire information about the location of the UE 10 through the following positioning techniques:

[0191] - Observed Time Difference of Arrival (OTDOA)

[0192] -Multiple RTT

[0193] - Downlink Angle of Departure (DL AoD)

[0194] - Downlink Time Difference of Arrival (DL TDOA)

[0195] - Uplink Time Difference of Arrival (UL TDOA), or

[0196] -UL Angle of Arrival (AoA). Alternatively, the LMF 312 may acquire information about the location of the UE 10 through a positioning technique using a Cellular ID (CID).

[0197] For example, in OTDOA, the UE 10 receives downlink positioning reference signals (PRS) from multiple transmission points (TPs) and reports measurement values ​​about the physical cell ID, global cell ID, TP ID, and PRS timing through the LTE Positioning Protocol (LPP) to the LMF 312. As a result, the LMF 312 calculates the position of the UE 10 based on information about the known coordinates of the measured TPs and the relative timing of the reported PRS.

[0198] In addition, for example, in positioning using CID, LMF 312 calculates the position of UE 10 based on information about known coordinates of ng-eNB or gNB and the following measurement results reported from UE 10.

[0199] The UE 10 reports, for example, an evolved cellular global identifier (ECGI) or a physical cell ID and measurement results on reference signal received power (RSRP), reference signal received quality (RSRQ), and UE Rx-Tx time difference to the LMF 312. Here, the UE Rx-Tx time difference is defined as the time difference between the reception timing of the UE 10 and the transmission timing of the UE 10.

[0200] <<6. Edge computing technology>>

[0201] Fig. 9 is a diagram showing an example of a connectivity model for edge computing in a 5G system. It should be noted that this figure and the Fig.10 and 11 It is based on the drawings described in the document "3GPP TS23.548".

[0202] In 5GC 30, at least the following three are defined as connectivity models for connections to edge servers that handle edge computing:

[0203] -Distributed anchor points;

[0204] -session breakout; and

[0205] -Multi-PDU sessions.

[0206] Fig. 9 (1) shows an example of a “distributed anchor point.” A “distributed anchor point” is a connectivity model in which a PSA-UPF is installed at a local location near the UE 10 for one PDU session.

[0207] Fig. 9(2) shows an example of “session grooming.” “Session grooming” is a connectivity model in which a PSA UPF (C-PSA UPF) is installed at a central site and one or more PSA UPFs (L-PSA UPFs) are installed at local sites for one PDU session.

[0208] Fig. 9 (3) shows an example of a "multi-PDU session". In a "multi-PDU session", an edge computing application uses a PDU session connected to a PSA UPF installed at a local location. In addition, another application uses a PDU session connected to a PSAUPF installed at a central location.

[0209] Here, some or all of the functions of the application processed by the application server 40 are implemented in the DN 340. For example, in "distributed anchor point", "session grooming" or "multi-PDU session", some or all of the applications are implemented in the local site DN 340 operating as an edge server. That is, the DN 340 installed in the local site operates as an edge application server. Here, some or all of the functions of the application are functions for processing data sent and received through a QoS flow requesting low latency.

[0210] Similarly, according to the function of implementing the application in the DN 340 at the central site or the DN 340 at the local site in a distributed manner, the AF 308 of the control plane processing application server 40 is set in a distributed manner.

[0211] Fig.10 is a diagram showing an example of the architecture of the edge application server 41 for non-roaming periods.

[0212] This architecture is an architecture for connecting DN (local part of DN) 340-2 in the form of "session breakout" in one PDU session, and edge application server (EAS) 41 is connected to DN 340-2 in one PLMN.

[0213] The SMF 306 of the PLMN controls the UPF 330 to connect the PSA UPF (C-PSA UPF) installed at the central site to the edge application server 41. The UPF supports the PSA UPF (L-PSA UPF 330-3) installed at the local site and the uplink classifier (UL CL) / branching point (BP) 330-2. The SMF 306 controls the UPF 330 through N4 as a reference point.

[0214] Here, "session grooming" may generally be referred to as local grooming.

[0215] Fig.11is a diagram showing another architecture example of the edge application server 41 for non-roaming periods.

[0216] This architecture is an architecture for connecting DN (local part of DN) 340-2 in the form of a "distributed anchor point" in a PDU session, and an edge application server (EAS) 41 is connected to the DN 340-2 in a PLMN.

[0217] The SMF 306 of the PLMN controls the installed PSA UPF to connect to the edge application server 41 through the reference point N4.

[0218] SMF 306 may discover candidate edge application servers 41 through EASDF 311 .

[0219] <<7. QoS of 5G system>>

[0220] Fig.12 3GPP TS 38.300 is a diagram showing an example of a QoS architecture of 5GS. The diagram is based on the drawings described in the document "3GPP TS 38.300".

[0221] At the non-access stratum (NAS) layer level, a QoS flow has the finest granularity to distinguish different QoS from each other in a protocol data unit (PDU) session. Within a PDU session, a QoS flow is identified by a QoS flow ID (QFI).

[0222] In addition, in XR or media services, a group of packets transmits the payload of a PDU set. A PDU set includes one or more PDUs that transmit the payload of an information unit generated at the application layer. Examples of the information unit include image frames, video slices for XR or media services, and I-frames, P-frames, and B-frames for moving image data in a group of pictures (GOP) format.

[0223] That is, the packets in the PDU set are treated as data units to be received and decoded within a specific time period. In the QoS flow with the finest granularity in terms of QoS control, data can be identified with finer granularity through the PDU set. In addition to QoS control at the QoS flow level, 5GS can apply QoS control at the PDU set level.

[0224] SMF 306 associates PCC rules with QoS flows based on QoS and service requests. SMF 306 associates QFI to new QoS flows and obtains PCC rules and other information associated with QoS flows from PCF 305. SMF 306 obtains the QoS profile of the QoS flow, instructions about the corresponding UPF (e.g., N4 rules), and QoS rules from the PCC rules.

[0225] The base station (gNB) of the (R)AN 20 may establish at least one data radio bearer (DRB) with each UE 10 together with a PDU session. A DRB is a logical path for transmitting data. In the QoS model of 5G, a guaranteed flow bit rate (GBR) in which a frequency band is guaranteed and a non-guaranteed flow bit rate (non-GBR) in which a frequency band is not guaranteed are supported.

[0226] (R)AN 20 and 5GC 30 guarantee the quality of service by mapping each packet to an appropriate QoS flow and DRB. That is, a two-stage mapping is implemented, that is, mapping between IP flows and QoS flows in the non-access stratum (NAS) and mapping between QoS flows and DRBs in the access stratum (AS).

[0227] At the NAS level, QoS flows are characterized by QoS profiles provided from 5GC 30 to (R)AN 20 and QoS rules provided from 5GC 30 to UE 10.

[0228] The QoS profile is used by the (R)AN 20 to determine the processing method on the radio interface. The QoS rules are used to instruct the UE 10 to implement the mapping between the traffic of the user plane in the uplink and the QoS flow. Therefore, for a multicast multicast / broadcast service (MBS) session, the UE 10 is not provided with the QoS rules of the MBS QoS flow and the QoS parameters at the QoS flow level.

[0229] The QoS profile is provided from the SMF 306 to the (R)AN 20 via the AMF 301 and the reference point N2, or is set in the (R)AN 20 in advance.

[0230] In addition, the SMF 306 may provide one or more QoS rules and (if necessary) QoS flow-level QoS parameters associated with the QoS rules to the UE 10 via the AMF 301 and reference point N1.

[0231] In addition or in lieu thereof, reflective QoS control may be applied to the UE 10. Reflective QoS control is QoS control that monitors the QFI of downlink packets and applies the same mapping to uplink packets.

[0232] Depending on the QoS profile, the QoS flow becomes a GBR QoS flow or a non-GBR QoS flow. The QoS profile of a QoS flow includes, for example, QoS parameters such as a 5G QoS Identifier (5QI) and Allocation and Retention Priority (ARP).

[0233] ARP includes information about priority, preemption capability, and preemption vulnerability.

[0234] Priority defines the relative importance of QoS flows, and indicates that the smallest value of the priority is the most preferred. Preemption capability is an indication that defines whether a specific QoS flow can seize resources that have been assigned to another QoS flow with a lower priority.

[0235] Preemption vulnerability is an index that defines whether a specific QoS flow can vacate resources assigned to the specific QoS flow for another QoS flow with a higher priority. "Enable" or "Disable" is set for preemption capability and preemption vulnerability.

[0236] In a GBR QoS flow, the QoS profile includes:

[0237] - Uplink and downlink guaranteed flow bit rate (GFBR);

[0238] - Uplink and downlink Maximum Flow Bit Rate (MFBR);

[0239] - Uplink and downlink maximum packet loss rate;

[0240] - Delay critical resource types;

[0241] - Notification control;

[0242] etc.

[0243] In non-GBR QoS flows, the QoS profile includes Reflective QoS Attributes (RQA), additional QoS flow information, and the like.

[0244] The Notification Control of QoS parameters indicates whether notification is requested from the (R)AN 20 when a particular QoS flow cannot meet the GFBR. For a particular GBR QoS flow, when the Notification Control is "enabled" and it is determined that the GFBR cannot be met, the (R)AN 20 sends a notification to the SMF 306 indicating this.

[0245] At this time, unless the (R)AN 20 is in a special state requesting the release of RAN resources of the GBR QoS flow, the (R)AN 20 should maintain the QoS flow. Examples of the special state include radio link failure and RAN internal congestion.

[0246] When it is determined that the QoS flow meets the GFBR again, the (R)AN 20 sends a new notification to the SMF 306 indicating this.

[0247] The aggregate maximum bit rate (AMBR) is related to the session AMBR of each PDU session and the UE-AMBR of each UE 10. The session AMBR limits the aggregate bit rate that is expected to be provided for all non-GBR QoS flows of a specific PDU session and is managed by the UPF 330. The UE-AMBR limits the aggregate bit rate that is expected to be provided for all non-GBR QoS flows of a specific UE 10 and is managed by the (R)AN 20.

[0248] 5QI is related to QoS features and provides a policy for setting node-specific parameters to each QoS flow. Standardized or preset QoS features for 5G can be found from 5QI, and no explicit signaling is implemented. QoS features indicated by signal can be included as part of the QoS profile.

[0249] The QoS characteristics include information about resource type, priority, packet delay budget, packet error rate, average window, maximum data burst, etc. The resource type is GBR QoS flow or non-GBR QoS flow. The packet delay budget may include the packet delay budget in 5GC 30.

[0250] At the AS level, DRB defines the packet processing method at the wireless interface (Uu interface). DRB provides the same packet transmission processing for any packet.

[0251] (R)AN 20 maps the QoS flow to the DRB based on the QFI and the QoS profile set in the QFI. (R)AN 20 may establish different DRBs for packets requesting different types of packet transmission processing (see Fig.12 ).

[0252] In addition, the (R)AN 20 can also multiplex multiple QoS flows belonging to the same PDU session into the same DRB (see Fig.12 ).

[0253] In the uplink, the mapping of QoS flows to DRBs is controlled by mapping rules that are signaled in two different ways.

[0254] One method is a method called reflection mapping. In reflection mapping, the UE 10 monitors the QFI of downlink packets for each DRB and applies the same mapping to uplink packets.

[0255] Another method is a method called explicit configuration. In explicit configuration, the mapping rules of QoS flow to DRB mapping are explicitly signaled through RRC.

[0256] In the downlink, QFI is signaled on the Uu interface for Reflective Quality of Service (RQoS) by the (R)AN 20. However, unless the (R)AN 20 and NAS use reflective mapping for the QoS flow carried in a particular DRB, neither the (R)AN 20 nor the NAS will signal the QFI on the Uu interface for that DRB.

[0257] In the uplink, the (R)AN 20 may be configured to signal the QFI to the UE 10 over the Uu interface. In addition, the (R)AN 20 may set a default DRB for each PDU session. When an uplink packet is not suitable for either explicit configuration or reflective mapping, the UE 10 maps the packet to the default DRB for the PDU session.

[0258] For non-GBR QoS flows, the 5GC 30 may send additional QoS flow information parameters related to any QoS flow to the (R)AN 20. This is to give instructions to increase the frequency of a specific traffic relative to another non-GBR QoS flow in the same PDU session.

[0259] How to map multiple QoS flows in a PDU session to one DRB depends on the (R)AN 20. For example, the (R)AN 20 can map the GBR QoS flow and the non-GBR QoS flow to the same DRB or different DRBs. In addition, the (R)AN 20 can map multiple GBR QoS flows to the same DRB or different DRBs.

[0260] In 5G NR, a new Service Data Adaptation Protocol (SDAP) sublayer is introduced for QoS control through QoS flows. Through the SDAP sublayer, the traffic of the QoS flow is mapped to the appropriate DRB. The SDAP sublayer can have multiple SDAP entities. The SDAP sublayer has an SDAP entity for each PDU session on the Uu interface. The SDAP entity is established or released through RRC.

[0261] The QoS flow is identified by the QFI in the PDU session container included in the GPRS Tunneling Protocol (GTP)-U header. The PDU session is identified by the GTP-U Tunnel Endpoint ID (TEID). The SDAP sublayer maps each QoS flow to a specific DRB.

[0262] When receiving the QoS monitoring request from the AF 308, the PCF 305 may generate an authorized QoS monitoring policy and provide the QoS monitoring policy to the SMF 306 by adding the QoS monitoring policy to the PCC rules.

[0263] The SMF 306 may activate end-to-end uplink (UL) / downlink (DL) packet delay measurement between the UE 10 and the PSA-UPF 330 for a QoS flow during a PDU session establishment procedure or a PDU session modification procedure.

[0264] The SMF 306 sends a QoS monitoring request to the UPF 330 through a reference point N4, and sends N2 signaling to request QoS monitoring between the UPF 330 and the (R)AN 20.

[0265] The QoS monitoring request includes monitoring variables determined by the SMF 306 based on the QoS monitoring policy received from the PCF 305 or an authorized QoS monitoring policy pre-set locally.

[0266] The (R)AN 20 measures the delay of the UL / DL packet in the (R)AN 20 part and provides the measured value to the UPF 330 through the reference point N3.

[0267] The UPF 330 calculates the delay of the UL / DL packet at the reference point N3 or N9. The UPF 330 sends the QoS monitoring result to the SMF 306 based on a predetermined condition. Here, the predetermined condition is, for example, sending only once, periodically sending, or event triggering.

[0268] In addition, the UPF 330 may support sending the QoS monitoring result to the AF 308 through the locally located NEF 302. Here, the QoS monitoring result is, for example, a measurement result of the bit rate (e.g., average bit rate or maximum bit rate) of each GBR QoS flow of the target PDU session, a measurement result of the aggregate bit rate of all non-GBR QoS flows of the target PDU session, a measurement result of the packet error rate of the target PDU session, a measurement result of the aggregate bit rate of all non-GBR QoS flows of the target UE 10, a measurement result of the packet error rate of the target UE 10, a measurement result of the delay of UL / DL packets, etc. The delay of UL / DL packets includes the delay of UL / DL packets in the (R)AN 20 part obtained from the (R)AN 20 and the delay of UL / DL packets at the reference point N3 or N9.

[0269] Based on the QoS monitoring policy, SMF 306 may be configured to send the DN performance analysis of NWDAF 313 described later to AF 308 in addition to or in lieu of QoS monitoring.

[0270] The locally located NEF 302 can send QoS monitoring results and / or DN performance analysis with low latency to the locally deployed AF 308. The locally located AF 308 can register a subscription for the sending of low latency QoS monitoring results and / or DN performance analysis in the PCF 305 through the non-local NEF 302 or the locally located NEF 302.

[0271] When the non-local NEF 302 determines that the non-local NEF 302 itself is not suitable to respond to the request from the locally located AF 308 , the non-local NEF 302 may redirect the request from the locally located AF 308 to the locally located NEF 302 .

[0272] The locally located AF 308 can directly register the subscription of the Npcf_PolicyAuthorization_Subscribe service in the PCF 305, directly receive the QoS monitoring results from the UPF 330, and / or directly receive the DN performance analysis from the NWDAF313.

[0273] In addition, the locally located PSA-UPF may provide a Nupf_EventExposure_Notify service. The locally located PSA-UPF may provide Nupf_EventExposure_Notify to a notification target address specified by a session reporting rule received from the SMF 306.

[0274] For example, the local PSA-UPF registers the sending of the QoS monitoring result as the Nupf_EventExposure_Notify service, and specifies the non-local AF 308 or the local NEF 302 as the notification target address. When the local NEF 302 is specified as the notification target address, the local NEF 302 sends the QoS monitoring result to the non-local AF 308.

[0275] In addition, when the PDU session is a session through a time-sensitive networking (TSN) bridge, the packet delay budget for the time-sensitive communication (TSC) QoS flow is the sum of the 5G access network (AN) packet delay budget (PDB) and the core network (CN) PDB. The QoS monitoring mechanism described above can also be applied to TSC QoS flows for bit rate (e.g., average bit rate or maximum bit rate) measurement and end-to-end UL / DL packet delay measurement between the UE 10 and the PSA-UPF 330.

[0276] <<8. Network data analysis function>>

[0277] In the 5G system, the NWDAF 313 (see Figure 5 ) for network automation.

[0278] Fig.13 This figure and the following description are diagrams showing a configuration example of an architecture for data collection for a network data analysis function. Figures 14 to 18 It is based on the drawings described in the document "3GPP TS23.288".

[0279] Fig.13 The NWDAF 313-1 shown in FIG. 313-1 can use each NF 313-2 of the 5G system as a data source to collect data through the Nnf interface. Here, each NF 313-2 is Figure 5 Each NF of the 5GC 30 shown in FIG. In addition, the NWDAF 313-1 can collect data through a data collection coordination function (DCCF) 313-3. It should be noted that the data that the NWDAF 313-1 can collect from another NF 313-2 may be described as a parameter of the 5G system hereinafter.

[0280] The Ndccf interface is an interface prepared for the NWDAF 313-1 to request the DCCF 313-3 to deliver data. The DCCF 313-3 collects data from each NF 313-2 through the Nnf interface and delivers the data to the NWDAF 313-1 through the Ndccf interface. Alternatively, the DCCF 313-3 can use a mechanism for collecting data from NFs supported by the message transfer framework 313-4 and deliver the collected data to the NWDAF 313-1 through the Nmfas interface data.

[0281] Each NF 313-2 in the 5G system can obtain network analysis information (such as statistical information) from the NWDAF 313-1 through the Nnwdaf interface. In addition, each NF 313-2 in the 5G system can obtain network analysis information through the DCCF 313-3.

[0282] Fig.14 is a diagram showing a configuration example of an architecture for acquiring network analysis information.

[0283] The DCCF 313-3 collects network analysis information from the NWDAF 313-1 through the Ndccf interface, and delivers the network analysis information to each NF 313-2 of the 5G system through the Nnwdaf interface. Alternatively, the DCCF 313-3 may deliver the acquired network analysis information to each NF 313-2 of the 5G system through the Nmfas interface. In this case, a mechanism for acquiring network analysis information from the NWDAF 313-1 supported by the message transfer framework 313-4 is used.

[0284] The NF (eg, NWDAF 313 - 1 ) may store network analysis information (eg, statistical information) in an analysis data repository function (ADRF) 313 - 6 via a Nadrf interface.

[0285] Fig.15 is a diagram showing a configuration example of an architecture for storing collected data and network analysis information.

[0286] The DCCF 313-3 obtains the analysis information of the network from the NWDAF 313-1 through the Ndccf interface, and stores it in the ADRF 313-6 through the Nadrf interface. In addition, the DCCF 313-3 stores the data collected from each NF 313-5 in the ADRF 313-6 through the Nadrf interface. Alternatively, the DCCF 313-3 may request the messaging framework 313-4 to store the acquired data and the network analysis information in the ADRF 313-6 through the Nadrf interface.

[0287] Fig.16 is a diagram showing a configuration example of an architecture for acquiring an AI / ML model.

[0288] The analysis logic function (AnLF) is a logic function of the NWDAF 313 that performs inference using an AI / ML model, acquires analysis information, and discloses the analysis information. The analysis information is, for example, statistical information or prediction information of past events.

[0289] The Model Training Logic Function (MTLF) is a logic function of the NWDAF 313 that learns an AI / ML model and exposes its learning service.

[0290] The NWDAF 313 may include either or both of the AnLF and the MTLF.

[0291] The NWDAF 313 including the AnLF may acquire the learned AI / ML model from another NWDAF 313 including the MTLF through the Nnwdaf interface and use the learned AI / ML model. Here, the Nnwdaf interface is an interface used by the AnLF to request or subscribe to the learned model providing the service of the MTLF.

[0292] Fig.17 is a diagram showing an example of a process in which the AF 308 acquires analysis information from the NWDAF 313 . Fig.17 A procedure for subscribing to or unsubscribing from analysis information of the NWDAF 313 by the NEF 302 is shown.

[0293] NEF 302 controls the analysis exposure mapping in AF 308 by the allowed analysis IDs (step 0). For example, NEF 302 limits the subscription of the analysis ID to a specific AF 308, or limits the notification of the analysis ID to a specific AF 308.

[0294] The AF 308 activates Nnef_AnalyticsExposure_Subscribe or Nnef_AnalyticsExposure_Unsubscribe through the NEF 302, and registers or cancels the delivery of analytical information (step 1). If the delivery of analytical information is permitted for the NEF 302, the NEF 302 performs the following processing.

[0295] The NEF 302 subscribes to or cancels the delivery of analytical information by activating the Nnef_AnalyticsExposure_Subscribe or Nnef_AnalyticsExposure_Unsubscribe service based on a request from the AF 308 (step 2).

[0296] If the NEF 302 registers for delivery of analytics information, the NWDAF 313 activates analytics information or Nnwdaf_AnalyticsSubscription_Notify service and notifies the NEF 302 of a termination request (step 3).

[0297] When receiving the notification from the NWDAF 313, the NEF 302 activates the analytics information or Nnef_AnalyticsExposure_Notify service and notifies the termination request to the AF 308. The NEF 302 can limit the notification to the AF 308 by analyzing the exposure map.

[0298] The AF 308 confirms whether a termination request is included in the Nnef_AnalyticsExposure_Notify message. If a termination request is included, the AF 308 activates the Nnwdaf_AnalyticsSubscription_Unsubscribe service and cancels the delivery of the analytics information (step 4).

[0299] Fig.18 is a diagram showing an example of a process in which the AF 308 acquires analysis information from the NWDAF 313 . Fig.18 A procedure for requesting analysis information of the NWDAF 313 by the NEF 302 / sending analysis information of the NWDAF 313 in response is shown.

[0300] The NEF 302 controls the analysis exposure mapping in the AF 308 by the allowed analysis IDs (step 0). For example, the NEF 302 limits the analysis IDs requested by a specific AF 308, or limits the response of the analysis ID to a specific AF 308.

[0301] The AF 308 activates the Nnef_AnalyticsExposure_Fetch service through the NEF 302 and requests reception of analytical information (step 1). If the request for analytical information is permitted for the NEF 302, the NEF 302 performs the following processing.

[0302] The NEF 302 requests analytics information from the NWDAF 313 by activating the Nnwdaf_AnalyticsInfo_Request service based on the request from the AF 308 (step 2).

[0303] The NWDAF 313 sends the analysis information to the NEF 302 in response (step 3).

[0304] NEF 302 sends the analysis information to AF 308 in response. NEF 302 may limit the response to AF 308 by setting the PLMN operator. AF 308 confirms whether a termination request is included in the Nnef_AnalyticsExposure_Fetch response message, and stops the request for analysis information if the termination request is included (step 4).

[0305] As described previously, the NWDAF 313 outputs the following analysis information, for example, in response to a request from another NF.

[0306] -Slice load level related network data analysis

[0307] - Observation service experience related network data analysis

[0308] -NF load analysis

[0309] -Network performance analysis

[0310] -UE related analysis

[0311] -User data congestion analysis

[0312] -QoS sustainability analysis

[0313] - Diffusion analysis

[0314] -WLAN performance analysis

[0315] -Session management congestion control experience analysis

[0316] - Redundant transmission experience analysis

[0317] -DN Performance Analysis

[0318] As slice load level related network data analysis, NWDAF 313 provides statistical information and forecast information about the load information of network slices, as well as statistical information and forecast information about the load information of network slice instances.

[0319] NWDAF 313 provides the following service experiences as observation service experience related network data analysis:

[0320] -Service experience of network slicing;

[0321] -Application service experience;

[0322] -Service experience of edge applications on the user plane; and

[0323] -Service experience for applications per radio access technology (RAT) or per frequency

[0324] NWDAF 313 provides statistical information and forecast information about the load of each NF of the 5G system as NF load analysis.

[0325] As a network performance analysis, NWDAF 313 provides statistical information and prediction information about network performance, such as the status of gNB, resource usage of gNB, the number of UE 10, communication performance, or mobility performance.

[0326] As UE-related analysis, the NWDAF 313 provides movement analysis of the UE 10, communication analysis information of the UE 10, network data analysis information on expected operation parameters of the UE 10, and network data analysis information on abnormal operation.

[0327] As user data congestion analysis, the NWDAF 313 provides statistical information and prediction information on congestion of user data sent on the user plane, and statistical information and prediction information on congestion of user data sent on the control plane.

[0328] As a QoS sustainability analysis, NWDAF 313 provides statistical information and prediction information on QoS sustainability.

[0329] As a diffusion analysis, the NWDAF 313 provides statistical information and prediction information of the data volume for each segment, each slice, each slice for a specific UE 10 or group, or transaction distribution.

[0330] As a WLAN performance analysis, the NWDAF 313 provides statistical information and prediction information on the performance of the WLAN.

[0331] As a session management congestion control experience analysis, NWDAF 313 provides statistical information of session management congestion control experience (SMCCE).

[0332] As redundant transmission experience related analysis, the NWDAF 313 provides statistical information and prediction information about the experience of redundant transmission.

[0333] As a DN performance analysis, the NWDAF 313 provides statistical information on the performance of the DN service, such as an average traffic rate, a maximum traffic rate, an average packet delay amount, a maximum packet delay amount, or an average packet loss rate.

[0334] <<9. Processing of network functions>>

[0335] Fig.19 and 20 is a diagram showing a configuration example of an application using AI / ML according to an embodiment of the present disclosure. Fig.19 In , the application performs an AI / ML process (in this case, speech recognition). Fig. 20 In the example, the application performs multiple (here, for example, three) AI / ML processes.

[0336] exist Fig.19 In the example, the AI / ML processing of speech recognition is itself an application (e.g., a voice interaction application). Fig. 20 As shown in , it is expected that applications including multiple AI / ML processing will increase in the future, such as applications that implement control by using voice recognition, image recognition, and AI / ML processing of the results of voice recognition and image recognition.

[0337] In addition, if Figure 3 and 4As shown in , when the AI / ML of the application is processed by a cloud server or by a client and a cloud server in a distributed manner, the amount of delay caused by wireless communication occurs. When a cloud server installed outside the 5G system is used, it is difficult to control the delay caused by wireless communication, and there are problems in satisfying the quality of experience (QoE) of the application through E2E. It should be noted that multiple AI / ML processes are implemented in a dependent manner or in parallel.

[0338] As described previously, the NWDAF 313 implements the learning and inference / estimation processing of the AI / ML model for acquiring statistical information and prediction information about the NFs of the 5G system.

[0339] Therefore, in this embodiment, by implementing the AI / ML processing of the application, the NWDAF 313 of the 5G system can control the delay caused by wireless communication.

[0340] Here, the network functions necessary for the NWDAF 313 of the 5G system to implement AI / ML processing of applications originally processed by a cloud server installed outside the 5G system, processing regarding these network functions, etc. will be described.

[0341] Fig.21 3 is a diagram showing an example of a method for providing information of a 5G system through the NEF 302 and making a request to the 5G system according to an embodiment of the present disclosure. Fig.21 In the example, AF 308 obtains 5G system information from NWDAF 313 through NEF 302.

[0342] The NWDAF 313 exposes information about the AI / ML models supported by the NWDAF 313 itself to the AF 308, including the type of AI / ML model, the type of algorithm, setting parameters, etc.

[0343] Here, the type of AI / ML model is, for example, a neural network model or a deep neural network model. The type of algorithm is, for example, a convolutional neural network, a recurrent neural network, or a LSTM.

[0344] In addition, the type of AI / ML model can be defined by the type of AI / ML model and the type of algorithm. The setting parameters are, for example, the range of the number of layers that can be set in a neural network model or a deep neural network model, the range of the number of nodes in each layer, or the parameters of the 5G system that can be set as input data for the input layer.

[0345] Here, the parameter of the 5G system that can be set as the input data of the input layer is a parameter of the 5G system acquired from the NF of the control plane of the 5GC 30 (an example of a first parameter). In addition, the parameter of the 5G system that can be set as the input data of the input layer is a parameter of the 5G system acquired from the UPF 330 which is a network function that processes the user plane of the 5GC 30 (an example of a second parameter).

[0346] (First case)

[0347] As a first case, the AF 308 acquires information about the AI / ML model (an example of model information) from the NWDAF 313 through the NEF 302. After setting the format of the AI / ML model, the AF 308 can instruct the application server 40 to learn the AI / ML model together with the set format.

[0348] Here, the format of the AI / ML model is a setting of the type of AI / ML model, the kind of algorithm, the number of layers, or the number of nodes in each layer. In addition, if necessary, information about parameters of the 5G system input to the input layer (an example of parameter information) is included in the setting of the format.

[0349] When the parameters of the 5G system are necessary for the input of the AI / ML model, the application server 40 can obtain the necessary parameters of the 5G system through the NEF 302 and the AF 308 and generate a learning data set.

[0350] When the application server 40 completes the learning of the AI / ML model, the AF 308 creates setting information necessary for the format of the learned AI / ML model (hereinafter also referred to as a learned model).

[0351] Here, the setting information includes, for example, the type of learning model, the type of algorithm, the number of layers, the number of nodes in each layer, the weighting factors between nodes, and if necessary, also includes settings about the correspondence between the parameters of the 5G system input to the input layer and the corresponding nodes of the input layer.

[0352] It should be noted that when a plurality of AI / ML models are used in an application, the application server 40 sets the plurality of AI / ML models and performs learning of each model.

[0353] Fig. 22 3 is a diagram showing an example of a method for providing information of a 5G system through the NEF 302 and making a request to the 5G system according to an embodiment of the present disclosure. Fig. 22 In the example, AF 308 sends an AF request to NWDAF 313 via NEF 302.

[0354] The AF 308 transmits an AF request including setting information (an example of the second format information) necessary for the format (an example of the second format) of the learned model and for setting the quality of the learned model in the NWDAF 313 to the NWDAF 313 .

[0355] When setting of the requested learned model is permitted, the NWDAF 313 assigns an analysis ID (an example of model identification information) to the learned model, and returns a message including the analysis ID assigned to the learned model to the AF 308 as a response to the AF request.

[0356] When requesting the setting of a plurality of learned models, the AF 308 transmits an AF request including a plurality of items of setting information necessary for the format of the plurality of learned models and an instruction for setting the plurality of learned models in the NWDAF 313 to the NWDAF 313 .

[0357] The NWDAF 313 stores information about the learned AI / ML model assigned the analysis ID in the ADRF 313-4 (see Fig.13 , 14 wait).

[0358] (Second case)

[0359] Furthermore, as a second case, it is assumed that the application server 40 detects a degradation in the accuracy of the learned model in the application. In this case, the application server 40 can activate the process of relearning the AI / ML model based on the format set by the AF 308 in the first case.

[0360] The application server 40 may detect the degradation of the accuracy of the learned model in the application by the degradation of the QoE detected by the UE 10. Here, the application server 40 may obtain the QoE detected by the UE 10 from the UE 10 through the application layer. Alternatively, the application server 40 may obtain the QoE detected by the UE 10 as a parameter of the 5G system through the NEF 302. In the 5G system, the (R) AN 20 obtains the QoE from the UE 10, and the measurement collection entity (MCE), the trace collection entity (TCE), or the AMF 301 of the 5GC 30 collects the QoE from the (R) AN 20. Here, the QoE is collected for each application ID.

[0361] The 5GC 30 including operations, administration and maintenance (OAM) activates the measurement of QoE in the UE 10 on a signaling basis, for example through the AMF 301.

[0362] (R)AN 20 transmits a radio resource control (RRC) message including AppLayerMeasConfig as a setting related to measurement and reporting of QoE to UE 10. (R)AN 20 receives a report including the measurement result of QoE through an uplink RRC message. (R)AN 20 transmits a QoE reference ID corresponding to the received report including the measurement result of QoE to MCE / TCE.

[0363] The activation of QoE measurement can be implemented based on the use of a specific network slice or application (e.g., an application using a learned model). Subsequently, when the use of the specific network slice or application ends, the QoE measurement ends (deactivates).

[0364] Alternatively, the UE 10 that has received settings related to the measurement and reporting of QoE in advance from the 5GC 30 can determine the activation of the measurement of QoE according to the start of a specific network slice or application. Subsequently, when the use of the specific network slice or application ends, the UE 10 determines the end (deactivation) of the measurement of QoE.

[0365] In addition, the (R)AN 20 may release one or more application layer settings (eg, QoE measurements) configured in the UE 10 at any time using an RRCReconfiguration message.

[0366] An example in which the 5GC 30 including OAM detects the accuracy degradation of the learned model in the application through QoE is described here, but the method for detecting the degradation is not limited to this example. The 5GC 30 including OAM can set the measurement and report of another application layer in the UE 10, and detect the accuracy degradation of the learned model based on the measurement result of the application layer. An example of measurement of another application layer will be described later.

[0367] When the application server 40 completes the relearning of the AI / ML model, the AF 308 creates information necessary for the format of the relearned model based on the relearned AI / ML model (hereinafter also referred to as the relearned model).

[0368] The AF 308 transmits an AF request (an example of a second request) including setting information (an example of format information) necessary for the format of the relearned model corresponding to the analysis ID to be updated and an instruction to the NWDAF 313 to update the learned model to the NWDAF 313 .

[0369] When the requested update of the learned model is allowed, the NWDAF 313 returns a message including the analysis ID assigned to the updated learned model to the AF 308 as a response to the AF request.

[0370] (Third case)

[0371] In addition, as a third case, AF 308 Fig.21 The AF 308 can then obtain information about the AI / ML model from the NWDAF 313 through the NEF 302 and set the format of the AI / ML model. Fig. 22 Instructs NWDAF 313 to learn the AI / ML model along with the set format.

[0372] Here, the format of the AI / ML model is a setting of the type of AI / ML model, the kind of algorithm, the number of layers, or the number of nodes in each layer. In addition, if necessary, the setting of the correspondence between the parameters of the 5G system input to the input layer and the corresponding nodes of the input layer is included in the setting of the format. When the parameters of the 5G system are necessary for the input of the AI / ML model, the MTLF of the NWDAF 313 obtains the necessary parameters of the 5G system from each NF and performs learning of the set AI / ML model.

[0373] Here, when a plurality of AI / ML models are used in an application, the application server 40 instructs the NWDAF 313 to learn the plurality of AI / ML models together with a plurality of set formats.

[0374] When the learning of the set AI / ML model is completed, the MTLF of the NWDAF 313 assigns an analysis ID to the learned model. The NWDAF 313 sends a query to the AF 308 based on the Fig.21 The notification includes an analysis ID assigned to the learned model to provide notification of the completion of learning of the AI / ML model set by the NEF 302.

[0375] When an instruction on learning of multiple AI / ML models is given, the NWDAF 313 sends a notification including analysis IDs respectively assigned to the multiple learned models to the AF 308. Alternatively, the NWDAF 313 may send multiple notifications including analysis IDs assigned to the learned models to the AF 308 in the order of learning.

[0376] (The fourth case)

[0377] In addition, as a fourth case, the MTLF of NWDAF 313 notifies AF 308 of the Fig.21 The learning of the set AI / ML model is completed. Subsequently, the MTLF of the NWDAF 313 may store information about the learned model in the ADRF 313 - 4 and perform re-learning of the set AI / ML model.

[0378] When relearning of the set AI / ML model is completed, the MTLF of the NWDAF 313 updates the information about the learned model corresponding to the analysis ID stored in the ADRF 313-4 using the information about the relearned AI / ML model. The MTLF of the NWDAF 313 sends the AF 308 a message based on the AI / ML model. Fig.21 A notification including the analysis ID assigned to the learned model is provided to notify that the re-learning of the set AI / ML model is completed.

[0379] In addition, assume that the application server 40 detects that the accuracy of the learned model in the application has deteriorated. In this case, AF308 can Fig. 22 An analysis ID is specified, and a request is sent to the NWDAF 313, the request including an instruction for re-learning the AI / ML model corresponding to the analysis ID.

[0380] Furthermore, the application server 40 can detect the degradation of the accuracy of the learned model in the application through the degradation of the QoE detected by the UE 10 .

[0381] (Fifth case)

[0382] Furthermore, as a fifth case, the AF 308 requests settings necessary for implementing QoS control through an AF request in the established session to transmit and receive data input to and output from the learned model used in the application. Examples of this session include a PDU session.

[0383] In this example, an example of assigning an analysis ID to identify a learned model is described, but the method for identifying a learned model is not limited to this example. For example, NWDAF 313 assigns an AI / ML model ID to a learned model. The 5G system may use the AI / ML model ID as a substitute or supplement for the analysis ID to identify the learned model.

[0384] Although an example is shown here in which AF 308 provides necessary information to NWDAF 313 using AF request, AF request may be used to provide necessary information for various network functions of 5GC 30 through NEF 302 or directly.

[0385] <<10. Session management method in the first case>>

[0386] Fig.23 is a sequence diagram illustrating an example of a PDU session establishment process according to an embodiment of the present disclosure.

[0387] The AF 308 sends an AF request for setting a learned model used by an application to the NEF 302 according to the first case described in Section 9 (step S501). The NEF 302 sends the AF request to the NWDAF 313 (step S502). In this way, the AF 308 sends an AF request for setting a learned model used by an application to the NWDAF 313 through the NEF 302.

[0388] Here, an example of sending one AF request is described, but a plurality of AF requests each including different requests may be sent. Examples of different requests include the following AF requests:

[0389] -AF request for learning a previous AI / ML model or a learned model according to the requested format;

[0390] - an AF request for providing a notification regarding instructions for learning a set prior-to-learning AI / ML model; and

[0391] -AF request for notification providing instructions for updating the learned model.

[0392] The AF request includes an application ID for identifying the application.

[0393] When data input to or from the learned model processed by NWDAF 313 is routed from UPF 330 or is routed to UPF 330 through DN 340, the AF request includes information about N6 traffic routing requirements.

[0394] The information on the N6 traffic routing requirement includes N6 traffic routing information. The N6 traffic routing information includes information necessary for routing traffic to the DNAI specified by the application ID.

[0395] When the DNAI corresponds to user plane access to multiple QoS flows to the DN 340, the connection destination (eg, port number) to which each QoS flow is sent may be included in the N6 traffic routing information.

[0396] The NWDAF 313 sets the learned AI / ML model used by the application according to the AF request obtained from the AF 308 (step S503). Here, the settings of the learned AI / ML model are, for example, the type of AI / ML model, the type of algorithm, the number of layers, the number of nodes in each layer, the weighting factor between nodes, etc.

[0397] It should be noted that when the parameters of the 5G system are input to the input layer of the learned AI / ML model, the setting of the learned AI / ML model includes a setting (an example of corresponding information) about the correspondence between the corresponding node of the input layer and the parameters of the 5G system. The corresponding information may include first corresponding information in which the parameters and formats of the 5G system acquired from the NF of the control plane of the 5GC 30 are associated with each other. In addition, the corresponding information may include second corresponding information in which the parameters and formats of the 5G system acquired from the UPF 330, which is a network function that processes the user plane of the 5GC 30, are associated with each other.

[0398] When setting the learned AI / ML model, the NWDAF 313 assigns identification information so that the learned AI / ML model can be identified in the 5G system. The identification information is, for example, an analysis ID or an AI / ML model ID.

[0399] In addition, when data obtained from a network function (e.g., UPF 330) that processes a user plane is input to an input layer of a learned AI / ML model, the AF request includes a setting of the correspondence between a corresponding node of the input layer and a QoS parameter (e.g., QoS feature, 5QI, etc.). Here, when the input layer has a plurality of corresponding nodes and different QoS parameters are applied to the plurality of nodes, respectively, a plurality of QoS parameters may be set.

[0400] In addition, when the data output from the output layer of the learned AI / ML model is processed by the network function processing the user plane, the AF request includes a setting of the correspondence between the corresponding node of the output layer and the QoS parameter. Here, when the output layer has multiple corresponding nodes and different QoS parameters are applied to the multiple nodes respectively, multiple QoS parameters can be set.

[0401] Among the AF requests sent from the AF 308 , the NWDAF 313 sends the AF request including a request for applying QoS parameters to the input / output layer of the learned AI / ML model to the PCF 305 (step S504 ).

[0402] The PCF 305 creates policies / rules regarding QoS parameters to be applied to the input / output layer of the learned AI / ML model according to the received AF request (step S505 ).

[0403] When the setting of the learned AI / ML model is completed, the NWDAF 313 returns a response to the AF request to the NEF 302 (step S506). The NEF 302 sends a response to the AF request to the AF 308 (step S507). In this way, the NWDAF 313 returns a response to the AF request to the AF 308 through the NEF 302.

[0404] The response includes an analysis ID or an AI / ML model ID for identifying the learned AI / ML model. When multiple learned AI / ML models are set for the application, the response includes multiple analysis IDs or AI / ML model IDs.

[0405] The procedure from step S501 to step S507 described above is a procedure corresponding to the processing of the first case described in Section 9.

[0406] UE 10 sends a PDU Session Establishment Request as a NAS message to AMF 301 through (R)AN 20 (step S508).

[0407] Here, the 5G system may use S-NSSAI to inform the UE 10 that the NWDAF 313 may process learned AI / ML models used in applications activated by the UE 10 in the registration area.

[0408] For example, assume that a specific S-NSSAI corresponding to a network slice in which the NWDAF 313 processes the learned AI / ML model used in the application activated by the UE 10 is included in the allowed NSSAI. In this case, the UE 10 can send a PDU session establishment request including the specific S-NSSAI to the AMF 301.

[0409] like Fig.23 As shown in , AMF 301 selects SMF 306 based on the DNN and S-NSSAI requested by the UE included in the received message (step S509). Here, when the DNN requested by the UE is not included in the received message, AMF 301 selects the DNN for S-NSSAI and selects SMF 306 based on the selected DNN and S-NSSAI.

[0410] Here, the 5G system may specify any one of the DNN, S-NSSAI, and application ID, or NWDAF that processes the application and / or learning model used in the PDU session established through its set.

[0411] The AMF 301 sends a session management context generation request (Nsmf_PDUSession_CreateSMContext request) message for the PDU session to the selected SMF 306 (step S510).

[0412] The session management context generation request message includes a subscription permanent identifier (SUPI), an S-NSSAI, and a UE requested DNN or a selected DNN.

[0413] The SMF 306 that has received the session management context generation request message acquires the session management subscription data from the UDM 407 using the Nudm_SDM_Get message (step S511). This is implemented when the session management subscription data corresponding to the SUPI, S-NSSAI, and DNN are not available.

[0414] Furthermore, the SMF 306 registers the session management subscription information in the UDM 307 using the Nudm_SDM_Subscribe message, thereby causing a notification to be given when the session management subscription data is updated.

[0415] When SMF 306 can process the session management context generation request message received in the previous step S510, SMF 306 generates a session management context (SM context). SMF 306 returns a session management context generation response (Nsmf_PDUSession_CreateSMContext response) message including a session management context ID (SM context ID) to AMF301 (step S512).

[0416] When a second authentication and authorization process must be performed by the DN-AAA server during the PDU session establishment process, the SMF 306 activates the PDU session authentication / authorization (step S513).

[0417] When applying dynamic policy and charging control (PCC) to the PDU session to be established, SMF 306 performs PCF selection (step S514). Alternatively, SMF 306 can apply local policies.

[0418] In addition, SMF 306 establishes a session management policy association (SM policy association) with PCF 305. SMF 306 may perform a session management policy association establishment procedure (SM policy association establishment procedure) with PCF 305 to obtain a default PCC rule for the PDU session (step S515).

[0419] As a result, SMF 306 can obtain PCC rules before selecting UPF 330.

[0420] For example, PCC rules include policies / rules related to the learned AI / ML model used by the application corresponding to the application ID.

[0421] Examples of policies / rules related to learned AI / ML models include the following:

[0422] - Policies / rules for specifying the NWDAF 313 for processing learned AI / ML models used by applications;

[0423] - Policies / rules regarding the analytic ID or AI / ML model ID used to identify the learned AI / ML model;

[0424] - Strategies / rules regarding the correspondence between the corresponding nodes of the input layer of the learned AI / ML model and the parameters of the 5G system;

[0425] - policies / rules regarding routing between corresponding nodes of the input / output layer of the learned AI / ML model and the UPF 330; and

[0426] - Policies / rules regarding QoS parameters to be applied at the input / output layer of the learned AI / ML models.

[0427] SMF 306 specifies, based on the PCC rules acquired in the previous step S515, that the application uses one or more learned AI / ML models set in NWDAF 313. SMF 306 also performs UPF selection according to preset rules or acquired PCC rules, and selects one or more UPFs 330 (step S516).

[0428] The SMF 306 sends an N4 session establishment request message to the selected UPF 330 (step S517).

[0429] Through the N4 session establishment request message, the N4 rule for controlling uplink and downlink traffic in the UPF 330 is set.

[0430] For example, the N4 rules include information about packet detection rules (PDR), forwarding action rules (FAR), QoS enforcement rules (QER), usage reporting rules (URR) and buffer action rules (BAR). In addition, QoS monitoring can be set through the N4 session establishment request message.

[0431] The PDR includes information necessary for classifying packets in the UPF 330. For example, the information necessary for classifying packets is set according to the policy / rule regarding the QoS parameters applied to the input / output layer of the learned AI / ML model. The UPF 330 can classify packets of the PDU session by the QFI corresponding to the 5QI.

[0432] The FAR includes information about the method used to process a particular packet, such as forwarding, duplication, discarding, or buffering. The QER includes instructions about the QoS applied to the traffic. For example, information about instructions about the QoS applied to the traffic is set according to the policy / rules of the 5QI (QoS characteristics) for each QoS flow obtained from the PCF 305. The URR includes information necessary for measuring and reporting traffic. The BAR includes the time period and amount of data to be buffered, as well as information about the notification method in the control plane.

[0433] In addition, the N4 rules may include, for example, information about the routing of data input to or output from the learned AI / ML model processed by the NWDAF 313 in the UPF 330. This is implemented, for example, according to policies / rules regarding routing between corresponding nodes of the input / output layer of the learned AI / ML model and the UPF 330 (e.g., policies / rules regarding N6 traffic routing requirements).

[0434] When receiving the N4 session establishment request message, UPF 330 returns an N4 session establishment response message to SMF 306 (step S518).

[0435] It should be noted that when a plurality of UPFs 330 are selected for the PDU session in the foregoing step S516 , the N4 session establishment process is activated for each UPF 330 .

[0436] SMF 306 sends a Namf_Communication_N1N2MessageTransfer message to AMF 301 (step S519).

[0437] The Namf_Communication_N1N2MessageTransfer message includes a PDU session ID, N2 session management information (N2 SM information), CN tunnel information (CN tunnel information), S-NSSAI and an N1 session management container (N1 SM container).

[0438] The N2 session management information includes a PDU session ID, (multiple) QFIs, a QoS profile, etc. Here, the QoS profile may include information about a PDU set. For example, it may include information about a PDU set set for a QoS flow corresponding to a specific QFI.

[0439] In addition, when multiple QoS flows are assigned to one PDU session, a QoS profile set including multiple QoS profiles may be provided. For example, when multiple learned AI / ML models are processed in one PDU session, a QoS flow is set for each learned AI / ML model. Alternatively, when multiple learned AI / ML models are processed in one PDU session, for example, data input to and output from one learned AI / ML model is classified into multiple QoS flows.

[0440] Furthermore, data input to and output from the learned AI / ML models, as well as other user plane data, may be assigned to different QoS flows.

[0441] When multiple UPFs 330 are used for a PDU session, the CN tunnel information includes tunnel information about the multiple UPFs 330 terminating N3.

[0442] The N1 Session Management Container includes the PDU Session Establishment Accept and QoS rules that the AMF 301 should provide to the UE 10. The PDU Session Establishment Accept includes the S-NSSAI. The Namf_Communication_N1N2MessageTransfer message includes the PDU Session ID so that it can be found which access the AMF 301 uses for the UE 10.

[0443] The AMF 301 sends an N2 PDU Session Request message to the (R)AN 20 (step S520).

[0444] The N2 PDU Session Request message includes a PDU session ID having the UE 10 as a destination, a NAS message including an N1 session management container, and the N2 session management information received from the SMF 306 .

[0445] (R)AN 20 acquires PDU session ID, QFI, QoS profile (an example of the second QoS parameter), etc. from the N2 session management information included in the N2 PDU session request message. (R)AN 20 determines a processing method on a radio interface according to the QoS profile.

[0446] Furthermore, the (R)AN 20 transmits a NAS message included in the N2 PDU Session Request message to the UE 10 (step S521).

[0447] As described above, the NAS message includes the PDU session ID and the N1 session management container, and the N1 session management container includes the PDU session establishment acceptance and the QoS rule (an example of the first QoS parameter). The UE 10 determines the mapping between the traffic of the user plane in the uplink and the QoS flow according to the QoS rule.

[0448] In addition, the (R)AN 20 assigns AN tunnel information (AN Tunnel Information) to the PDU session. The AN Tunnel Information includes a tunnel endpoint of each participating (R)AN node and a QFI assigned to the tunnel endpoint.

[0449] (R)AN 20 updates the N2 session management information for notifying SMF 306. The N2 session management information includes information such as PDU session ID, AN tunnel information, a list of allowed or denied QFI(s), and user plane enforcement policy notification.

[0450] (R)AN 20 returns an N2 PDU Session Response message including N2 session management information to AMF 301 (step S522).

[0451] The AMF 301 acquires the N2SM information through the N2 PDU session response received from the (R)AN 20. The AMF 301 transmits a session management context update request message (Nsmf_PDUSession_UpdateSMContext Request) of the PDU session including the session management context ID and the acquired N2 session management information to the SMF 306 (step S523).

[0452] SMF 306 activates the N4 session modification procedure with UPF 330 and sends an N4 session modification request message to UPF 330 (step S524).

[0453] In addition to the transport rules, SMF 306 provides UPF 330 with AN tunnel information.

[0454] UPF 330 returns an N4 session modification response message to SMF 306 (step S525).

[0455] It should be noted that when multiple UPFs 330 are used in a PDU session, the preceding session modification procedure is implemented on all UPFs 330 terminating N3.

[0456] When receiving the N4 session modification response from the UPF 330, the SMF 306 returns a session management context update response message (Nsmf_PDUSession_UpdateSMContext response) to the AMF 301 (step S526).

[0457] Through the previous processing, a PDU session for connection with the application server 40 is established. One DNN and one S-NSSAI are associated with the PDU session. In the PDU session, the processing of the learned model among the various processes of the application processed by the application server 40 is processed by the NWDAF 313. In this way, the UE 10, the NWDAF 313 and the application server 40 cooperate to process the application.

[0458] When QoS monitoring is set, (R)AN 20 measures the delay of UL / DL packets in the (R)AN 20 part and provides the measured value to UPF 330 through reference point N3. UPF 330 calculates the delay of UL / DL packets at reference point N3 or N9. Here, when the QoS flow is a GBR QoS flow or a delay-critical GBR QoS flow, as QoS monitoring, (R)AN 20 can measure the amount of data processed during the period of 5G-AN PDB.

[0459] The UPF 330 may calculate the delay of the UL / DL packet in the 5GS using the measurement result of the UL / DL packet delay in the (R)AN 20 part and the measurement result of the UL / DL packet delay at the reference point N3 or N9.

[0460] In addition, the UPF 330 may measure jitter as a change in the measurement result of the UL / DL packet delay, that is, a difference between the measurement results of the UL / DL packet delay measured at two different timings.

[0461] The UPF 330 may report QoS monitoring results including measurement results of UL / DL packet delay for each QoS flow and jitter in the 5GS to the AF 308 or SMF 306 via the NEF 302.

[0462] In addition to QoS monitoring, AF 308 can set monitoring in the processing time of the learned AI / ML model in NWDAF 313 through AF request. For example, NWDAF 313 reports to AF 308 the measurement results of the time required for one attempt for one data set input to the input layer of the learned AI / ML model or the average time of one attempt for a preset number of attempts.

[0463] Based on the measurement results of QoS monitoring for each QoS flow, AF 308 can grasp the transmission delay associated with the data input to the input layer and the data of the output layer of the learned AI / ML model. In addition, by considering the measurement results of the processing time of the learned AI / ML model, AF 308 can grasp the processing time in the application including the processing time in the AI / ML model. Therefore, AF 308 can grasp the time required for the processing in the application including the transmission delay.

[0464] <<11. Session management method in the second case>>

[0465] In the session management method in the second case, the edge application server 41 is used.

[0466] Fig.24 is a sequence diagram illustrating another example of a PDU session establishment process according to an embodiment of the present disclosure.

[0467] Since the procedure from step S501 to step S507 is the same as Fig.23 It should be noted that the procedure from step S501 to step S507 is a procedure for the AF 308 to set the learned model used by the application in the NWDAF 313 through the NEF 302. It should be noted that in step S505, the PCF 305 creates the policy / rule necessary for setting the NWDAF 313 that processes the learned model used by the application in the edge application server 41 according to the AF request.

[0468] In addition, due to Fig.24 The procedure from step S508 to step S515 shown in Fig.23 , so their description is omitted here.

[0469] like Fig.24 As shown in FIG. 4 , based on the PCC rule acquired in step S515, the SMF 306 decides to set the NWDAF 313 that processes one or more learned AI / ML models used by the application in the edge application server 41. For example, when the network data analysis function is specified, the function of the NWDAF 313 is set in the edge application server 41. After the decision, the SMF 306 performs EASDF selection (step S601).

[0470] The SMF 306 sends a DNS context creation request (Neasdf_DNSContext_Create request) message to the selected EASDF 311 (step S602).

[0471] The EASDF 311 maintains the IP address, SUPI, Subscription Hidden Identifier (SUCI), etc. of the UE 10 included in the message, and creates a DNS context.

[0472] Subsequently, the EASDF 311 returns a response message (Neasdf_DNSContext_Create response) to the DNS context creation request message (step S603).

[0473] After this process, the SMF 306 adds the IP address of the EASDF 311 as a DNS server or resolver to the PDU Session Establishment Accept message to be sent to the UE 10 .

[0474] SMF 306 activates the EAS discovery procedure (step S604).

[0475] After the activation, the SMF 306 sends a DNS query including the IP address of the UE 10 to the EASDF 311. The EASDF 311 selects an edge application server 41 close to the UE 10 based on the IP address of the UE 10.

[0476] Here, the selection of the edge application server 41 may include a process of selecting the DN 340 to be connected to the edge application server 41 and setting the DNN included in the PDU session establishment request in the DN 340 .

[0477] The EASDF 311 sends a DNS response including the IP address of the edge application server 41 and / or the IP address of the DN 340 selected as a response to the DNS query.

[0478] In addition, the SMF 306 adds information about the location of the UE 10 to the DNS query. The EASDF 311 may select an edge application server 41 that is close to the UE 10 based on the information about the location of the UE 10.

[0479] Here, information about the location of UE 10 may be obtained from UE 10 or LMF 312 using, for example, the method described in Section 5, Location Information Management Function.

[0480] Furthermore, the EASDF 311 may select the edge application server 41 based on the load status of the edge application server 41 in addition to the information on the location of the UE 10 .

[0481] For example, when there are multiple candidates for edge application servers 41 within a preset range for the location of UE 10, EASDF 311 selects the edge application server 41 with the lowest load status. Alternatively, EASDF 311 selects an edge application server 41 close to UE 10 from among multiple candidates for edge application servers 41 whose load status is equal to or less than a preset threshold.

[0482] In addition, the EASDF 311 may select the edge application server 41 based on the capability / performance of the edge application server 41. For example, the edge application server 41 is classified based on the number of virtual central processing unit (CPU) / graphics processing unit (GPU) cores or memory capacity. The EASDF 311 selects the edge application server 41 close to the UE 10 from among a plurality of candidates for the edge application server 41 included in the classification of the capability / performance equal to or higher than that specified by the DNS context creation request or the DNS query.

[0483] The SMF 306 requests the NWDAF 313 to set, in the edge application server 41, a function of the NWDAF 313 that processes one or more learned AI / ML models used by the application (step S605).

[0484] The NWDAF 313 that receives the request from the SMF 306 executes the Fig.24 In the EAS41 in the process, the function of the NWDAF 313 including the AnLF is set (step S606).

[0485] The SMF 306 selects the PSAUPF 330 to be the PDU session anchor based on the selected DN 340 or edge application server 41 (step S607).

[0486] SMF 306 sends an N4 session establishment request message to the selected PSA UPF 330 (step S517), and then performs the Fig.23 The same processing as steps S518 to S526 in .

[0487] Through the previous processing, Fig.11 The distributed anchor point shown in FIG. 4 establishes a PDU session for connection with the edge application server 41. Here, the edge application server 41 acts as a computing node in the PDU session. One DNN and one S-NSSAI are associated with the PDU session.

[0488] In addition, in the edge application server 41, in addition to the function of the NWDAF 313 that processes one or more learned models used by the application, another process of the application performed by the application server 40 is set. In this PDU session, the UE 10, the edge application server 41, and the NWDAF 313 set in the edge application server 41 cooperate to process the application. In addition, the 5GC 30 can set the functions of the central unit (CU) and / or the distributed unit (DU) as the functions of the (R)AN20 in the edge application server 41.

[0489] In applications such as advanced driver assistance systems (ADAS) or assisting autonomous driving, it is assumed that the learned model is used to generate information to assist in vehicle control.

[0490] Acquiring information that assists vehicle control requires low latency. Therefore, the edge application server 41 handles the process of learning the model and also handles the process of generating information that assists vehicle control. As a result, the delay occurring between the vehicle in which the UE 10 is installed and the application server 40 is reduced.

[0491] Here, in order to generate information that assists vehicle control, information acquired by various sensors installed on the vehicle may be input into a learning model. That is, one or more learning models implement sensor fusion and generate information that assists vehicle control. Examples of various sensors include the following:

[0492] - GNSS receiver;

[0493] - Acceleration sensor;

[0494] - Gyro sensor;

[0495] - Camera (image sensor);

[0496] -Light Detection and Ranging (LiDAR); and

[0497] -Millimeter wave radar.

[0498] In addition, the vehicle may also be equipped with an inertial measurement unit (IMU) including an acceleration sensor, a gyro sensor, a magnetic field sensor, an atmospheric pressure sensor, a temperature sensor, etc. It should be noted that generating information to assist vehicle control includes generating a dynamic map.

[0499] Dynamic maps are map information obtained by adding various types of traffic information, etc. to highly accurate three-dimensional maps. Dynamic maps include the following layers:

[0500] - A static information layer including road surface information, lane information, etc. as a highly accurate 3D map;

[0501] -A quasi-static information layer including traffic regulations and road construction scheduling information;

[0502] - Quasi-dynamic information layer including accident information, congestion information, etc.; and

[0503] - Dynamic information layer including surrounding vehicles, pedestrians, etc.

[0504] (R)AN 20 can also implement communication with the vehicle in which UE 10 is installed, that is, vehicle-to-everything (V2X) communication through the Uu interface. The concept of V2X communication includes: vehicle-to-vehicle (V2V) which means communication between vehicles, vehicle-to-infrastructure (V2I) which means communication between vehicles and infrastructure equipment installed on the road, vehicle-to-pedestrian (V2P) which means communication between vehicles and pedestrians, and vehicle-to-network (V2N) which means communication between vehicles and networks such as the cloud.

[0505] (R)AN 20 may be a roadside unit (RSU) that provides V2N communication. The roadside unit may support sidelink communication through the PC5 interface to provide V2I communication.

[0506] The roadside unit is equipped with, for example, a camera (image sensor), LiDAR, millimeter wave radar, etc., and implements driving assistance or supports autonomous driving on dedicated roads or public roads. The roadside unit provides, for example, location information of the vehicle on which the UE 10 is installed, information on traffic lights, information on dynamic maps, etc.

[0507] The 5GC 30 may set measurements related to vehicle control accuracy in an ADAS or assisted autonomous driving application in a UE 10 installed on a vehicle as a measurement of QoE or another application layer.

[0508] The edge application server 41 may detect the accuracy degradation of the learned model in the application based on the measurement result of the application layer detected by the UE 10 (ie, the measurement result regarding the vehicle control accuracy).

[0509] In addition, the vehicle includes an unmanned aerial vehicle (UAV) represented by a drone.

[0510] Similarly, in applications that assist control of autonomous movement or manipulation of a robot, it is assumed that a learned model is used to assist in the generation of information for robot control.

[0511] Acquiring information that assists robot control requires low latency. Therefore, the edge application server 41 handles learning the model and also handles the generation of information that assists robot control. As a result, the delay occurring between the robot on which the UE 10 is installed and the application server 40 is reduced.

[0512] Here, in order to generate information to assist robot control, information obtained by various sensors installed on the robot can be input into the learning model. That is, one or more learning models implement sensor fusion and generate information to assist robot control. Depending on their application, the various sensors are, for example, various sensors such as inertial measurement units, voltage sensors, current sensors, travel time (ToF) sensors, and cameras (image sensors). Examples of the various sensors also include various sensors such as infrared sensors, ultrasonic sensors, pressure sensors, vibration sensors, humidity sensors, gas sensors, tactile sensors, olfactory sensors, taste sensors, and the like.

[0513] As a measurement of QoE or another application layer, the 5GC 30 may set measurements related to robot control accuracy in an application of controlling the autonomous movement or operation of the assisting robot in the UE 10 mounted on the robot.

[0514] The edge application server 41 may detect the accuracy degradation of the learned model in the application based on the measurement result of the application layer detected by the UE 10 (that is, the measurement result regarding the robot control accuracy).

[0515] Furthermore, in applications using extended reality (XR), it is assumed that the learned model is used for the generation of XR media and the encoding of XR media.

[0516] In the delivery of XR media, it is required to reduce the motion-to-screen delay that causes virtual reality (VR) sickness. Therefore, the edge application server 41 handles the generation of XR media and the processing of encoding XR media. As a result, the delay between the wearable device represented by the head-mounted display (HMD) installed on the UE 10 and the application server 40 is reduced.

[0517] Here, in the generation of XR media, tracking information and posture information (pose information) are used to specify the user's position, viewport, and display position of augmented reality (AR) content to be multiplexed.

[0518] In order to obtain such information, information obtained by various sensors installed on the wearable device may be input into the learning model set in the NWDAF 313. Examples of the various sensors include a GNSS receiver, an acceleration sensor, a gyro sensor, a camera (image sensor), a magnetic field sensor, an atmospheric pressure sensor, a temperature sensor, a tactile sensor, and an olfactory sensor.

[0519] It should be noted that the application using XR includes games. In this case, the various sensors include a game controller, a sensor (such as a gyro sensor) mounted on the game controller, and the like.

[0520] Here, the UE 10 is installed on a game controller. Information acquired by the game controller and the sensor installed on the game controller is sent to the edge application server 41 via the UE 10, and is input to the learning model set in the NWDAF 313.

[0521] In addition, the edge application server 41 that processes the generation of the XR media can adaptively control the format of the XR media based on the result of QoS monitoring. For example, when the measurement result of the bit rate of the GBR QoS flow assigned to the XR media does not meet the request value of the GFBR, the edge application server 41 changes the format to a format with a lower resolution than the current format.

[0522] Furthermore, when the measurement result of the aggregate bit rate of all non-GBR QoS flows of the PDU session sending XR media does not satisfy the requested value of the session AMBR, the edge application server 41 changes the format to a format having a lower resolution than the current format.

[0523] Furthermore, when the measurement result of the delay of the UL / DL packet does not satisfy the request value, the edge application server 41 changes the format to a format having a lower resolution than the current format.

[0524] Furthermore, when the measurement result of the jitter of the UL / DL packet does not satisfy the requested value, the edge application server 41 changes the format to a format having a lower resolution than the current format.

[0525] In addition, when the measurement result of the amount of data processed during the period of the 5G-AN PDB is lower than the amount of data processed in the current format, the edge application server 41 changes the format to a format having a lower resolution than the current format.

[0526] Alternatively, upon receiving a signal from the (R)AN 20 or SMF 306 indicating that the maximum data burst size cannot be met, the edge application server 41 changes the format to a format having a lower resolution than the current format.

[0527] Here, according to the request from the edge application server 41, the AF 308 sends an AF request to the 5GC 30, and the AF request includes the data size in frames with the desired format as the maximum data burst amount.

[0528] Furthermore, to ensure synchronization between the first PDU session and the second PDU session, the AF 308 may add a synchronization request for the two PDU sessions and / or an allowable delay amount between the two PDU sessions to the AF request.

[0529] Here, the first PDU session is, for example, a session established between the UE 101 installed on the game controller and the edge application server 41. The second PDU session is, for example, a session established between the UE 102 installed on the wearable device and the edge application server 41.

[0530] In addition, the AF 308 may request synchronization for the two PDU sessions at the granularity of the QoS flow.When a delay exceeding the allowable delay amount is emitted between the two PDU sessions, the perceived quality (eg, QoE) of the game degrades.

[0531] In addition, the first UE 101 and the second UE 102 can add information indicating the association between the two PDU sessions requiring synchronization to the PDU session establishment request. Based on the information indicating the association between the two PDU sessions, the 5GC 30 can identify the PDU sessions requiring synchronization and grasp their relationship.

[0532] The SMF 306 detects the delay amount in the two PDU sessions based on the QoS monitoring result, and implements control so that the relative delay amount between the two PDU sessions falls within the allowable delay amount.

[0533] For example, the SMF 306 instructs the PSA UPF 330 to add a first time offset to a PDU session with low latency (e.g., the first PDU session). Due to the first offset, the PSA UPF 330 can implement control so that the relative delay between the two PDU sessions falls within the allowable delay. The control of the delay can be set independently for the uplink and downlink.

[0534] The first offset is achieved, for example, by buffering packets of a PDU session with a small delay (e.g., the first PDU session). Alternatively, the first offset is achieved, for example, by preferentially processing packets of a PDU session with a large delay (e.g., the second PDU session) compared to packets of a PDU session with a small delay (e.g., the first PDU session).

[0535] Alternatively, the SMF 306 may set a synchronization request for two PDU sessions from the AF 308 and / or an allowable delay amount between the two PDU sessions in the PSA UPF 330 at the time of the N4 session establishment request.

[0536] In this case, the PSA UPF 330 may control the first time offset based on the QoS monitoring result without receiving a notification of the first time offset from the SMF 306.

[0537] Furthermore, when requesting synchronization for two PDU sessions, the AF 308 may add an allowable jitter to the AF request in addition to the allowable delay amount.

[0538] When it is determined based on the QoS monitoring result that the two PDU sessions for which synchronization is requested do not satisfy the allowable jitter, SMF 306 notifies (R) AN 20 of instructions for setting configuration grant (CG) and / or semi-persistent scheduling (SPS). Here, 5G-AN PDB may be included in the notification.

[0539] (R)AN 20 sets the CG for uplink QoS flow and the SPS for downlink QoS flow based on the 5G-AN PDB.

[0540] In addition, the first UE 101 can request the second UE 102 to establish a first PDU session through sidelink communication. At this time, the second UE 102 can operate the first UE 10 as a remote UE and as a UE to network relay. Here, in addition to the second PDU session in which the second UE 10 itself is the terminating destination, the second UE 10 also establishes a first PDU session for relaying.

[0541] Furthermore, in conjunction with the first UE 101 instructing the second UE 102 to operate as a UE-to-network relay, the second UE 102 may activate a PDU session update process and perform a process of integrating the first PDU session into the second PDU session.

[0542] Through this process, the first UE 101 and the second UE 102 may transmit and receive data to / from the edge application server 41 through the second PDU session.

[0543] Here, AF 308 requests 5GC 30 to implement synchronization of QoS flows in the second PDU session. SMF 306 sets the allowable delay amount between specific QoS flows in PSA UPF 330, for example, using N4 rules.

[0544] Here, an example of requesting synchronization for two PDU sessions is described. However, when an application provides services through two or more PDU sessions, the AF request may also include the number of PDU sessions for which synchronization is requested.

[0545] In addition, the game controller and the wearable device can be connected to each other via a wire. The UE 10 installed on the game controller or the wearable device establishes a PDU session with the edge application server 41.

[0546] Information acquired by the game controller and the sensor mounted on the game controller, information acquired by the sensor mounted on the wearable device, and information acquired by the display unit of the wearable device are sent and received through one PDU session.

[0547] Here, information acquired by the game controller and each sensor mounted on the game controller, information acquired by each sensor mounted on the wearable device, and information acquired by the display unit of the wearable device are mapped to one or more QoS flows.

[0548] AF 308 may request synchronization for two or more QoS flows to be processed in the one PDU session. The synchronization request for two or more QoS flows is, for example, a request to control the relative delay between QoS flows or the relative delay between QoS flows and the jitter between QoS flows within an allowable range.

[0549] In addition, when multiple users use an application (e.g., a game that allows multiplayer play), the AF 308 may request synchronization between PDU sessions established by the users. Here, the request for synchronization between PDU sessions is, for example, a request to control the relative delay between QoS flows or the relative delay between QoS flows and the jitter between QoS flows within an allowable range.

[0550] Depending on the location of the user, the 5GC 30 may establish a PDU session for the user with different edge application servers 41. Different PSA UPFs 330 may be assigned for PDU sessions with different edge application servers 41.

[0551] As an anchor point, 5GC 30 sets a first SMF 306-1 that manages one of the PDU sessions established by the user. The first SMF 306-1 instructs the second SMF 306-2 that manages another PDU session to control the previous time offset so that the relative delay between the PDU sessions falls within the allowable delay range. Alternatively, the first SMF 306-1 set as the anchor point can establish and directly manage the N4 session with the PSA UPF 330 that handles another PDU session.

[0552] Here, the 5G system may use S-NSSAI to inform the UE 10 that synchronization may be requested between two or more PDU sessions established for applications activated by the UE 10 in a registration area.

[0553] It is assumed that a specific S-NSSAI corresponding to a network slice that can control synchronization between two or more PDU sessions established for an application activated by the UE 10 is included in the allowed NSSAI. In this case, the UE 10 can send a PDU session establishment request including the specific S-NSSAI to the AMF 301.

[0554] Furthermore, the application separates the user plane processed by the edge application server 41 into data and other data that can be provided to multiple users as the same data.

[0555] 5GC 30 acquires data that can be provided to multiple users as the same data as multicast transmission data, and acquires other data as unicast transmission data.

[0556] Here, 5GC 30 can assign different QoS flows to multicast transmission data and unicast transmission data.

[0557] When an edge application server 41 sends user plane data of an application to multiple users, the 5GC 30 can instruct the edge application server 41 to implement the processing necessary to separate the multicast transmission data and the unicast transmission data from each other.

[0558] For example, the learned model set in the NWDAF 313 of the edge application server 41 determines the separation between the user plane data that can be provided as the same data to multiple users and other data.

[0559] In addition, when the number of users connected to one edge application server 41 exceeds a preset number, the 5GC 30 can instruct the edge application server 41 to implement the processing necessary to separate the multicast transmission data and the unicast transmission data from each other.

[0560] The Multicast / Broadcast Service (MBS) QoS flow is assigned to multicast transmission data. For example, a game that allows multiplayer play sends the same data as multicast transmission data to multiple users. On the other hand, a game that allows multiplayer play sends different data as unicast transmission data to multiple users respectively.

[0561] The 5GC 30 obtains a copy of the multicast transmission data and sends the copy of the multicast transmission data to the (R)AN 20 via shared transmission using the 5GC shared MBS traffic delivery method.

[0562] Alternatively, the 5GC 30 obtains a copy of the multicast transmission data. The 5GC 30 sends the copy of the multicast transmission data through a first PDU session with the first UE 101 using the 5GC separate MBS traffic delivery method. The 5GC 30 sends another copy of the multicast transmission data through a second PDU session with the second UE 102 using the 5GC separate MBS traffic delivery method.

[0563] For example, when (R)AN 20 supports MBS, 5GC 30 uses the 5GC shared MBS traffic delivery method. When (R)AN 20 does not support MBS, 5GC 30 uses the 5GC separate MBS traffic delivery method.

[0564] It should be noted that, needless to say, the method described herein for separating user plane processing into multicast transmission data and unicast transmission data and transmitting the separated data may be applied to applications other than games that allow multiplayer play.

[0565] For example, the method can be applied to the transmission of dynamic maps for multiple vehicles in ADAS or assisted autonomous driving applications. In this case, the 5GC 30 can send the information of the static information layer or the static information layer and the quasi-static information layer as multicast transmission data, and can send the information of other layers as unicast transmission data.

[0566] In addition, when multiple users use an application (such as a game that allows multiplayer play) through the 5G system of different PLMNs, AF 308 can obtain QoS monitoring results for each PLMN.

[0567] AF 308 determines a second time offset for the packet delay budget for each PLMN based on the QoS monitoring result obtained for each PLMN so as to ensure synchronization between PDU sessions across PLMNs. AF 308 provides the second time offset for the packet delay budget of the corresponding PLMN to the 5GC 30 of each PLMN through an AF request.

[0568] The SMF 306 of each PLMN sets packet processing according to the acquired second time offset by using the N4 rule, for example, in the PSA UPF 330. The packet processing is implemented by, for example, buffering or prioritizing packets of a specified PDU session.

[0569] Packet priority processing is achieved by changing 5QI, ARP, resource type representing 5QI, priority, packet delay budget, etc.

[0570] For example, the packet priority processing is implemented by changing the preemption capability and preemption vulnerability of ARP. Specifically, the packet priority processing is implemented by changing the preemption capability from "disable" to "enable" and / or changing the preemption vulnerability from "enable" to "disable".

[0571] Alternatively, packet priority processing is implemented by changing the resource type of non-GBR QoS flow to GBR QoS flow or changing the priority level from a low priority level to a high priority level. Alternatively, packet priority processing is implemented by changing the packet delay budget to a short time.

[0572] Here, PLMN can broadly include service providers that provide 5G services in the form of independent non-public networks (SNPNs) represented by local 5G.

[0573] In addition, for example, a personal area network (PAN) device for acquiring information detected by a sensor installed on a game controller is installed on the wearable device. The UE 10 installed on the wearable device can use the PAN device as one of the sensors installed on the wearable device to establish a PDU session with the edge application server 41.

[0574] Here, the PAN device is a device that complies with a wireless standard such as Infrared Data Association (IrDA), Bluetooth (registered trademark), ZigBee (registered trademark), or Ultra Wideband (UWB).

[0575] It should be noted that in applications other than gaming, similar to the game controller described previously, a wearable device (e.g., HMD) can be connected to a bio-embedded device for the Internet of Medical Things (IoMT) in addition to multiple other wearable devices (wristband type devices, smart watches, clip type devices, smart rings, etc.) via sidelink communication or a PAN device to obtain information on a sensor mounted on another wearable device or bio-embedded device.

[0576] The plurality of other wearable devices may request to establish each first PDU session with a second UE 10 mounted on a wearable device (eg, HMD) operating as a layer 2 UE-to-network relay via sidelink communication.

[0577] Alternatively, the second UE 10 mounted on a wearable device (eg, HMD) operating as a layer 3 UE-to-network relay may map bearers to one first PDU session for relaying via sidelinks with multiple other wearable devices or PAN devices.

[0578] When an application other than a game is, for example, a digital twin that reflects information processed by a virtual / cyber space in a real space or a metaverse that reflects real space information acquired from multiple users in a virtual space, the processing in the virtual space and the generation of the real space information to be reflected can be processed by an edge application server 41 assigned to each user. Therefore, in addition to the QoS monitoring result for each PLMN described above, the AF 308 can obtain the QoS monitoring result for each edge application server 41 or each PSA UPF 330. The AF 308 determines the third time offset of the packet delay budget for each edge application server 41 or each PSA UPF 330 based on the QoS monitoring result for each edge application server 41 or each PSA UPF 330 in the PLMN, so as to ensure synchronization between PDU sessions that transmit the real space or the information of the real space to be reflected in the information of the virtual space. The AF 308 provides the third time offset of the packet delay budget for each edge application server 41 or each PSA UPF 330 to the 5GC 30 of each PLMN through the AF request.

[0579] The SMF 306 of each PLMN sets packet processing according to the acquired third time offset by using the N4 rule in each PSA UPF 330. The packet processing is implemented by buffering or prioritizing packets of a specified PDU session, for example.

[0580] Furthermore, in the application of a surveillance camera / security camera system called a closed-circuit television (CCTV), it is assumed that the learned model is used for the analysis of moving image information acquired from the surveillance camera. In some cases, low latency is required for security purposes. Therefore, the edge application server 41 handles the analysis processing of the moving image information. As a result, the delay occurring between the surveillance camera on which the UE 10 is installed and the application server 40 is reduced.

[0581] Here, in order to analyze the moving image information, information acquired by various sensors installed on the surveillance camera may be input into the learning model. Examples of the various sensors include a camera (image sensor), an infrared sensor, an ultrasonic sensor, and a sound sensor.

[0582] Information acquired by sensors included in various hardware installed in the UE 10 is input to the input layer of the learning model processed by the NWDAF 313 set in the edge application server 41. The information is input to the input layer of the learning model by a network function (e.g., PSA UPF 330) that processes the user plane on the PDU session established or updated as described above.

[0583] In addition, in the UE-assisted mode, the LMF 312 may obtain measurement results on GNSS such as code phase, Doppler, or carrier phase, input the measurement results to the input layer of the learning model, and calculate information on the location of the UE 10 as its output. In this case, the LMF 312 may cause the NWDAF 313 set in the edge application server 41 to perform the processing in the learning model.

[0584] Alternatively, the LMF 312 may acquire a measurement result obtained by the (R)AN 20 measuring a signal transmitted from the UE 10 in a network-based mode, or a measurement result obtained by the UE 10 measuring a signal transmitted from the (R)AN 20 in a UE-assisted mode.

[0585] Here, examples of the location detection method using a signal transmitted from the UE 10 in the network-based mode include UL TDOA and UL AoA. Examples of the location detection method using a signal transmitted from the (R)AN 20 in the UE-assisted mode include OTDOA, multi-RTT, DL AoD, and DL TDOA.

[0586] For example, LMF 312 may input the acquired measurement results to the input layer of the learned model and calculate as its output information about the location of UE 10. In this case, LMF 312 may cause NWDAF 313 set in edge application server 41 to perform processing in the learned model.

[0587] Furthermore, in order to improve the position accuracy, the NWDAF 313 may input information acquired from sensors included in the device where the UE 10 is installed to the input layer of the learning model in addition to the measurement result of the RF sensing, and calculate information on the position of the UE 10 .

[0588] In addition, when the PDU session is a session through a TSN bridge, the NWDAF 313 can obtain the measurement result of the port-to-port delay from the TSN application function (AF) as a QoS monitoring result. The TSN AF can, for example, calculate the port-to-port delay as the difference between the ingress timestamp (TSi) and the egress timestamp (TSe) generated by the network-side TSN translator (NW-TT) or the device-side TSN translator (DS-TT), where the translator is a port of the TSN bridge from TSi and TSe.

[0589] In addition, when the resource type is a GBR QoS flow or a delay-critical GBR QoS flow, the SMF 306 may instruct the (R)AN 20 to set a configuration grant (CG) and / or a semi-persistent scheduling (SPS). The (R)AN 20 sets the CG for the uplink QoS flow and the SPS for the downlink QoS flow based on the 5G-AN PDB.

[0590] <<12. Session management method in the third case>>

[0591] In the third case, the edge application server 41 is used as in the second case.

[0592] Fig.25 is a sequence diagram illustrating another example of a PDU session establishment process according to an embodiment of the present disclosure.

[0593] Since the procedure from step S501 to step S507 is the same as Fig.23 It should be noted that the procedure from step S501 to step S507 is a procedure for the AF 308 to set the learned model used by the application in the NWDAF 313 through the NEF 302.

[0594] In addition, since the procedure from step S508 to step S515 in the PDU session establishment process is the same as Fig.24 The same as in , so the description is omitted here.

[0595] like Fig.25 As shown in FIG. 5 , the SMF 306 decides to set the NWDAF 313 that processes one or more learned AI / ML models used by the application in the edge application server 41 based on the PCC rule acquired in step S515. After the decision, the SMF 306 performs EASDF selection (step S601).

[0596] The SMF 306 also performs UPF selection according to preset rules or acquired PCC rules. The SMF 306 selects one or more UPFs (C-PSA) 330-1 that have user plane access with the central (C)-DN 340-1 (step S701).

[0597] The SMF 306 sends an N4 session establishment request message to the selected UPF (C-PSA) 330-1 (step S702).

[0598] Through the N4 session establishment request message, an N4 rule for controlling uplink and downlink traffic in the UPF (C-PSA) 330-1 is set.

[0599] When receiving the N4 session establishment request message, UPF (C-PSA) 330-1 sets the N4 rule and returns an N4 session establishment response message to SMF 306 (step S703).

[0600] It should be noted that when a plurality of UPF (C-PSA) 330 - 1 are selected for the PDU session in the previous step S701 , the N4 session establishment process is activated for each UPF (C-PSA) 330 - 1 .

[0601] The SMF 306 sends a DNS context creation request (Neasdf_DNSContext_Create request) message to the selected EASDF 311 (step S602).

[0602] EASDF 311 holds the IP address, SUPI, SUCI, etc. of UE 10 included in the message, and creates a DNS context. Subsequently, EASDF 311 returns a response message (Neasdf_DNSContext_Create response) to the DNS context creation request message (step S603).

[0603] After this process, the SMF 306 adds the IP address of the EASDF 311 as a DNS server or resolver to the PDU Session Establishment Accept message to be sent to the UE 10 .

[0604] SMF 306 activates the EAS discovery procedure (step S604).

[0605] The EASDF 311 sends a DNS query including the IP address of the UE 10 to the EASDF 311 , and selects an edge application server 41 close to the UE 10 based on the IP address of the UE 10 .

[0606] Here, the selection of the edge application server 41 may include a process of selecting the local (L)-DN 340-2 connected to the edge application server 41. In addition, the function of the L-DN 340-2 may be implemented in the physically same device as the edge application server 41 as a logically different function.

[0607] The EASDF 311 sends a DNS response including the IP address of the edge application server 41 and / or the IP address of the L-DN 340-2 selected as a response to the DNS query.

[0608] In addition, the SMF 306 may add information about the address of the UE 10 to the DNS query. The EASDF 311 may select an edge application server 41 that is close to the UE 10 and / or the L-DN 340-2 based on the information about the location of the UE 10.

[0609] Here, for example, the information about the location of the UE 10 may be acquired from the UE 10 or the LMF 312 using the method described in the location information management function in Section 5.

[0610] Furthermore, the EASDF 311 may select the edge application server 41 and / or the L-DN 340 - 2 based on the load status of the edge application server 41 in addition to the information on the location of the UE 10 .

[0611] For example, when there are multiple candidates for edge application servers 41 within a preset range for the location of UE 10, EASDF 311 selects the edge application server 41 with the lowest load status. Alternatively, EASDF 311 selects an edge application server 41 close to UE 10 from among multiple candidates for edge application servers 41 whose load status is equal to or lower than a preset threshold.

[0612] The SMF 306 requests the NWDAF 313 to set, in the edge application server 41, a function of the NWDAF 313 that processes one or more learned AI / ML models used by the application (step S605).

[0613] The NWDAF 313 that has received the request from the SMF 306 performs a process of setting the functions of the NWDAF 313 including the AnLF in the edge application server 41 (step S606 ).

[0614] SMF 306 sends a Namf_Communication_N1N2MessageTransfer message including an N2 PDU session request message to AMF 301 (step S519). Fig.23 , so their description is omitted here.

[0615] Based on the local (L)-DN 340-2 selected in step S604, the SMF 306 selects the UPF 330 that is connected to the edge application server 41 and acts as a PDU session anchor (L-PSA) in the 5G system (step S704).

[0616] Furthermore, the SMF 306 selects the UPF 330 that supports UL CL / BP (step S705).

[0617] The SMF 306 sends an N4 session management request message to the UPF (UL CL / BP) 330-2 (step S706).

[0618] When receiving the N4 Session Establishment Request message, the UPF (UL CL / BP) 330-2 sets the N4 rule and returns an N4 Session Establishment Response message to the SMF 306 (step S707).

[0619] The SMF 306 sends an N4 session establishment request message to the UPF (L-PSA) 330-3 (step S708).

[0620] When receiving the N4 session establishment request message, the UPF (L-PSA) 330-3 sets the N4 rule and returns an N4 session establishment response message to the SMF 306 (step S709).

[0621] Here, the N4 rule for UPF (L-PSA) 330-3 includes an instruction to assign the same DNN as C-DN 340-1 to L-DN 340-2 to which UPF (L-PSA) 330-3 is connected. According to the instruction from SMF 306, UPF (L-PSA) 330-3 sets the same DNN as C-DN 340-1 in L-DN 340-2.

[0622] The SMF 306 sends an N4 session modification request message to the UPF (C-PSA) 330-1 (step S710).

[0623] UPF (C-PSA) 330-1 implements the insertion of UPF (UL CL / BP) 330-2 and the reset of the updated N4 rules. The N4 rules are updated in association with the addition of a path to be locally routed by UPF (UL CL / BP) 330-2 and connected to UPF (L-PSA) 330-3. The N4 session modification response message is returned to SMF 306 (step S711).

[0624] SMF 306 returns a PDU session management context update response (Nsmf_PDUSession_Update SMContext Response) message to AMF 301 (step S712).

[0625] Here, the PDU session management context update response message includes an N2 message, and the N2 message includes N2 session management information, CN tunnel information (CN tunnel information), S-NSSAI and N1 session management container. It should be noted that the N2 session management information is updated by the insertion of UPF (UL CL / BP) 330-2 and the addition of the path to be locally guided by UPF (UL CL / BP) 330-2.

[0626] The N1 Session Management Container includes the PDU Session Establishment Accept and QoS rules that the AMF 301 should provide to the UE 10. The PDU Session Establishment Accept includes the S-NSSAI and the IP address of the EASDF 311-2.

[0627] The AMF 301 sends the N2 message included in the PDU Session Management Context Update Response message to the (R)AN 20 (step S713).

[0628] (R)AN 20 sends the PDU Session Establishment Accept and QoS Rules included in the N2 message to UE 10 using the N1 message (step S714), and returns a response to the N2 message (step S715).

[0629] Through the previous processing, Fig.10 The form of session breakout shown in FIG. 4 establishes a PDU session for connecting the application server 40 and the edge application server 41 to the UE 10. One DNN and one S-NSSAI are associated with the PDU session.

[0630] Furthermore, the function of the NWDAF 313 that processes one or more learned models used by the application is set in the edge application server 41. Another process of the application is set in the application server 40.

[0631] For example, information acquired by sensors included in various hardware installed in the UE 10 is input to the input layer of the learning model processed by the NWDAF 313 set in the edge application server 41. The information is input to the input layer by a network function (e.g., UPF (L-PSA) 330-3) that processes the user plane on the PDU session established or updated as described above.

[0632] In this PDU session, the UE 10, the NWDAF 313 set in the edge application server 41, and the application server 40 cooperate to process the application.

[0633] In addition, as another example of this embodiment, AF 308 requests the 5G system to set one or more QoS flows for the second data transmission other than the first data transmission. The request is sent to the 5G system through the AF request.

[0634] The first data transmission is, for example, data transmission between UE 10 and edge application server 41 caused by processing of the learned model of the application. The second data transmission is, for example, data transmission between UE 10 and application server 40 caused by processing other than processing of the learned model.

[0635] The PCF 305 generates policies / rules regarding the second data transmission based on the AF request including settings of one or more QoS flows for the second data transmission.

[0636] The SMF 306 may find the 5QI for each flow of the traffic applied to the second data transmission based on the PCC rule acquired in step S515. In addition, the SMF 306 may find the QoS characteristics from the 5QI applied to each flow.

[0637] The SMF 306 determines whether one or more processes other than the process of the learned model will be processed by the application server 40 or by the edge application server 41 from the QoS characteristics of the traffic applied to the second data transmission.

[0638] For example, when the resource type is a GBR QoS flow, SMF 306 determines that one or more processes other than the learned model process will be processed by edge application server 41. When the resource type is a non-GBR QoS flow, SMF 306 determines that one or more processes other than the learned model process will be processed by application server 40.

[0639] In addition, when the resource type is delay critical GBR, SMF 306 may determine that one or more processes other than the process of the learned model will be processed by the edge application server 41. When the resource type is a GBR QoS flow other than delay critical GBR, SMF 306 may determine that one or more processes other than the process of the learned model will be processed by the application server 40. Here, the maximum data burst size is set for delay critical GBR.

[0640] Furthermore, the SMF 306 determines whether processing other than the processing of the learned model is to be processed by the application server 40 or by the edge application server 41 based on the priority.

[0641] For example, when the priority level is 30 or lower, SMF 306 determines that processing other than the learned model processing will be processed by edge application server 41. When the priority level exceeds 30, SMF 306 determines that processing other than the learned model processing will be processed by application server 40.

[0642] Furthermore, the SMF 306 determines whether processing other than the processing of the learned model will be processed by the application server 40 or by the edge application server 41 based on the packet delay budget.

[0643] For example, when the packet delay budget is 50ms or less, SMF 306 determines that processing other than that of the learned model will be processed by edge application server 41. When the packet delay budget exceeds 50ms, SMF 306 determines that processing other than that of the learned model will be processed by application server 40.

[0644] Furthermore, the SMF 306 determines whether processing other than the processing of learning the AI / ML model will be processed by the application server 40 or by the edge application server 41 based on the allowable jitter of the packet.

[0645] For example, when the packet allowable jitter is 10ms or less, SMF 306 determines that processing other than processing of learned AI / ML models will be processed by edge application server 41. When the packet allowable jitter exceeds 10ms, SMF 306 determines that processing other than processing of learned AI / ML models will be processed by application server 40.

[0646] In addition, the SMF 306 determines whether processing other than processing of the learned AI / ML model will be processed by the application server 40 or by the edge application server 41 based on a 5QI set including one or more 5QIs. The 5QI set includes a 5QI assigned to a QoS flow requiring a low packet delay budget, and the 5QI set is set based on a PCC rule.

[0647] For example, when the 5QI assigned to the traffic for the second data transmission is included in the set 5QI set, the SMF 306 determines that the processing other than the processing of the learned AI / ML model will be processed by the edge application server 41. When the 5QI assigned to the traffic for the second data transmission is not included in the set 5QI set, the SMF 306 determines that the processing other than the processing of the learned AI / ML model will be processed by the application server 40.

[0648] Here, the request to the 5G system through the AF request can be dynamically executed based on the results of the QoS monitoring for each QoS flow described above.

[0649] Furthermore, as in the session management method in the second case, the AF 308 can request synchronization between two or more PDU sessions that are locally routed and established with the edge application server 41. This improves QoE in XR and gaming applications.

[0650] In step S519, the Namf_Communication_N1N2MessageTransfer message sent from SMF 306 to AMF 301 includes N2 session management information. The N2 session management information includes (multiple) QFIs and QoS profiles for the traffic of the first data transmission, and (multiple) QFIs and QoS profiles for the traffic of the second data transmission.

[0651] In a PDU session according to another example, the UE 10, the edge application server 41, the NWDAF 313 set in the edge application server 41, and the application server 40 cooperate to process an application.

[0652] As another example of this example, SMF 306 can activate delay measurement of E2E UL / DL packets between UE 10 and UPF (C-PSA) 330-1 for QoS flows through processing of one or more learned models used by the application. This is implemented based on the QoS monitoring policy included in the PCC rules.

[0653] The (R)AN 20 measures UL / DL packet delay in the (R)AN 20 portion and provides the measured value to the UPF (C-PSA) 330-1 through the reference point N3.

[0654] The UPF (C-PSA) 330-1 calculates the UL / DL packet delay at the reference point N3 or N9. The UPF (C-PSA) 330-1 transmits the QoS monitoring result to the SMF 306 based on a predetermined condition.

[0655] Here, the predetermined condition is, for example, sending only once, sending periodically, or event triggering. The event is, for example, that the E2E delay of the UL packet does not meet the first QoS parameter or the E2E delay of the DL packet does not meet the second QoS parameter.

[0656] When an event is detected (based on the QoS monitoring result, the E2E delay of the UL packet does not meet the first QoS parameter or the E2E delay of the DL packet does not meet the second QoS parameter), the SMF 306 determines that the learned model will be processed by the edge application server 41. The SMF 306 can activate Fig.25 The processing after S604 in order to update the Fig.23 The PDU session established by the session management method in the first case shown in .

[0657] It should be noted that the QoS monitoring described above can be set not only for the QoS flow processed by the learned AI / ML model, but also for another QoS flow processed by the application.

[0658] also, Fig.25 An example is shown in which one edge application server 41 is locally routed in one PDU session, but the number of edge application servers 41 to be locally routed is not limited to one. For example, when multiple candidates for edge application servers 41 that meet the set allowable delay time are found through the AF request, the SMF 306 can establish a PDU session including paths with two or more edge application servers 41. For example, the function of the NWDAF 313 for learning an AI / ML model for application processing can be set in one edge application server 41-1 (not shown), and other processing of the application can be set in another edge application server 41-2 (not shown).

[0659] Here, SMF 306 sets UPF (L-PSA) 330-3 for each edge application server 41. SMF 306 sets a transmission destination for each QoS flow in UPF (UL CL / BP) 330-2, so that each QoS flow processed by the application server 40, the edge application server 41-1 or the edge application server 41-2 is transmitted to the correct transmission destination.

[0660] <<13. Session management method in the fourth case>>

[0661] The session management approach in the fourth case involves distributed processing of AI / ML models.

[0662] When the UE 10 has a certain level of operating capability, such as Figure 4 As shown in , the operation of the learned model of the application can be processed in a distributed manner through the client (ie, UE 10) and the cloud (ie, the application server 40 or the edge application server 41).

[0663] The 5G system may manage the processing capabilities of the UE 10 as a UE radio capability.

[0664] Fig.26 : is a flowchart showing a processing example when the operation of the learning model of the application according to the embodiment of the present disclosure is distributed through the client and the cloud.

[0665] The network function of the 5G system (eg, NWDAF 313) confirms the UE radio capability of the UE 10 to the UCMF 310 in response to the PDU session establishment request of the UE 10 (step S801).

[0666] The NWDAF 313 acquires information on the processing capability of the UE 10 (step S802 ).

[0667] Here, the information about the processing capability is information about the processing capability of the operating unit. The processing capability is, for example, the number of clocks (base clocks or maximum clocks), the number of cores, the number of threads, the memory bandwidth, or the memory capacity of the MPU, CPU, or GPU. Alternatively, the processing capability may be the number of clocks (base clocks or maximum clocks), the number of cores, the number of threads, the memory bandwidth, or the memory capacity of the system-level chip (SoC) on which the MPU, CPU, and GPU are installed. In addition, the processing capability may be a classification (e.g., a category) set by a combination thereof.

[0668] The NWDAF 313 determines a division point of the learned AI / ML model used in the application activated by the UE 10 based on the acquired information about the processing capability (step S803 ).

[0669] Here, the NWDAF 313 determines the layers to be divided so that more hidden layers including the input layer are included in the first AI / ML model as the processing capability of the UE 10 increases. The second AI / ML model includes the remaining hidden layers including the division point layer and the output layer.

[0670] The NWDAF 313 instructs the SMF 306 to set the first AI / ML model in the UE 10 .

[0671] For example, the SMF 306 adds information necessary for setting the first AI / ML model in the UE 10 (e.g., the format of the first AI / ML model) to the N1 session management container (N1 SM container), and Fig.23 The processing from S519 to S521 in FIG. 1 sets the first AI / ML model in UE 10 (step S804 ).

[0672] like Fig.26 As shown in , when the setting of the first AI / ML model in the UE 10 is completed, the NWDAF 313 sets the second AI / ML model in the AnLF (step S805).

[0673] Through the previous processing, the 5G system can implement the processing of using the learned model in the application in a distributed manner through the UE 10 and the NWDAF 313 in the PDU session established for the application activated by the UE 10.

[0674] Through this process, data that is input to the learned model and is related to the privacy of the UE 10 and the individual using the UE 10 is not directly transmitted to devices other than the UE 10. The UE 10 transmits the data of the intermediate layer processed by the first AI / ML model to devices other than the UE 10, so that security and personal privacy can be improved.

[0675] For example, when the application is extended reality (XR), as information that is assumed to be input to the AI / ML model and processed, there is tracking information for indicating the user's position, viewport, and display position of the AR content to be multiplexed. In addition, examples of this information include information obtained from various sensors (GNSS receivers, acceleration sensors, gyroscope sensors, cameras (image sensors), magnetic field sensors, atmospheric pressure sensors, temperature sensors, etc.) installed on wearable devices represented by HMDs in order to obtain posture information (posture information). A lot of this information involves privacy.

[0676] In addition, when the application is a surveillance camera / security camera system, it is assumed that information acquired from various sensors (camera (image sensor), infrared sensor, ultrasonic sensor, sound sensor, etc.) installed on the surveillance camera is input into the AI / ML model and processed to analyze the moving image information. A lot of this information involves privacy.

[0677] Furthermore, the information on the processing capability may be information on the transmission efficiency of the radio unit of the UE 10. Examples of the information on the transmission efficiency of the radio unit include the following information:

[0678] - UE category;

[0679] - Maximum UL / DL data rate;

[0680] - Maximum number of UL / DL component carriers supported;

[0681] - Maximum number of multiple-input multiple-output (MIMO) layers in UL / DL;

[0682] - Maximum modulation order for UL / DL;

[0683] - Maximum bandwidth in UL / DL multi-connectivity; and

[0684] - Maximum bandwidth in carrier aggregation.

[0685] In addition, as another example of this embodiment, the network function of the 5G system (e.g., NWDAF 313) can determine the division point of the learned model used in the application activated by UE 10. In this case, NWDAF 313 determines the division point based on, for example, information about the communication quality between UE 10 and the base station device 20 to which UE 10 is connected.

[0686] Here, the information on the communication quality is various parameters included in the measurement report reported from the UE 10 to the (R)AN 20. Alternatively, the information may be a measurement result on the delay amount of the UL / DL packet in the (R)AN 20 part, which is measured by the (R)AN 20 and reported to the UPF 330 through the reference point N3.

[0687] Examples of the various parameters included in the measurement report include the following parameters:

[0688] - Channel State Information (CSI) - RSRP;

[0689] -CSI-RSRQ;

[0690] -CSI signal-to-interference-and-noise ratio (SINR);

[0691] - Synchronization Signal (SS) - RSRP;

[0692] -SS-RSRQ; and

[0693] -SS-SINR.

[0694] As the communication quality between UE 10 and base station device 20 to which UE 10 is connected deteriorates, NWDAF 313 determines the layers to be divided so that more hidden layers including the input layer are included in the third AI / ML model. The fourth AI / ML model includes the remaining hidden layers, including the division point layer and the output layer.

[0695] SMF 306 adds information necessary for setting the third AI / ML model in UE 10 (e.g., the format of the third AI / ML model) to the N1 session management container (N1 SM container), and sets the third AI / ML model in UE 10. This is, for example, based on Fig.23 It is implemented by processing from step S519 to step S521 in .

[0696] Furthermore, the communication quality between the UE 10 and the base station device 20 to which the UE 10 is connected may change dynamically. Therefore, the NWDAF 313 may dynamically control the division point of the AI / ML model according to the communication quality.

[0697] In addition, the UE 10 requests a change in the division point of the AI / ML model according to a resource usage state in the MPU, the CPU, the GPU, or the SoC on which the MPU, the CPU, and the GPU are mounted.

[0698] For example, the UE 10 sends a PDU session update request including a notification about a partition point change of the AI / ML model.

[0699] When receiving a notification about a change in a division point of the AI / ML model through the AMF 301 or the SMF 306, the NWDAF 313 determines a different division point of the AI / ML model.

[0700] The NWDAF 313 determines the layers to be split so that fewer hidden layers including the input layer are included in the third AI / ML model as the usage ratio of computing resources used for operations in the UE 10 increases. The fourth AI / ML model includes the remaining hidden layers including the split point layer and the output layer.

[0701] Here, the UE 10 may add the resource usage ratio to the notification regarding the division point change of the AI / ML model.

[0702] The foregoing describes a network function, session management method, policy management method, and system for a 5G system for an application including processing of one or more AI / ML models. It goes without saying that the present embodiment is not limited to a 5G system and can be widely applied to next generation and subsequent systems including Beyond 5G (B5G).

[0703] In addition, the PDU session used above is an example of a path for transmitting user plane data and is not limited thereto. The technology disclosed in this specification can be applied to a broad concept of setting a termination destination and forming a path for transmitting user plane data in any unit such as traffic, packet, and flow.

[0704] <<14. Other embodiments>>

[0705] The embodiments described above are by way of example, and various modifications and applications are possible.

[0706] For example, the control device that controls the information processing device 300, the base station device 20, and the wireless communication device 10 of the above-described embodiments may be implemented by a dedicated computer system or a general-purpose computer system.

[0707] For example, a communication program for performing the operations described above is stored in a computer-readable recording medium such as an optical disk, a semiconductor memory, a magnetic tape, or a flexible disk and distributed. Subsequently, the program is installed in a computer, for example, and the processing described above is performed to configure the control device. At this time, the control device may be a device (such as a personal computer) outside the information processing device 300, the base station device 20, and the wireless communication device 10. In addition, the control device may be a device (such as a control unit 33, 240, or 150) inside the information processing device 300, the base station device 20, and the wireless communication device 10.

[0708] In addition, the communication program may be stored in a disk device included in a server device on a network such as the Internet, so that, for example, the communication program can be downloaded to a computer. In addition, the functions described above may be implemented by the cooperation of an operating system (OS) and application software. In this case, the portion other than the OS may be stored in a medium and distributed, or the portion other than the OS may be stored in a server device, so that, for example, the portion other than the OS can be downloaded to a computer.

[0709] Among the processes described in the foregoing embodiments, all or part of the processes described as being automatically implemented may be implemented manually, or all or part of the processes described as being manually implemented may be automatically implemented by known methods. In addition, unless otherwise indicated, the processing procedures, specific names, and information including various types of data and parameters shown in the foregoing description and the accompanying drawings may be arbitrarily changed. For example, the various types of information shown in the accompanying drawings are not limited to the information shown.

[0710] In addition, each component of each device shown in the drawings is a functional concept and does not necessarily need to be physically configured as shown in the drawings. That is, the specific form of distribution and integration of each device is not limited to the form shown, and all or part of it can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc. It should be noted that this configuration through distribution and integration can be implemented dynamically.

[0711] Furthermore, the embodiments described above may be appropriately combined in the case where the processing contents do not contradict each other. Furthermore, the order of the respective steps shown in the sequence diagrams and flowcharts of the embodiments described above may be appropriately changed.

[0712] In addition, the present embodiment can be implemented as any configuration constituting a device or system, such as a processor as a system large-scale integration (LSI), a module using multiple processors, a unit using multiple modules, a set obtained by further adding other functions to the unit, etc. (that is, a partial configuration of the device).

[0713] It should be noted that in the present embodiment, a system means a collection of a plurality of components (devices, modules (components), etc.), and it is not important whether all the components are in the same housing. Therefore, a plurality of devices housed in separate housings and connected to each other via a network and one device housing a plurality of modules in one housing are both systems.

[0714] In addition, the present embodiment may, for example, adopt a cloud computing configuration, in which a function is collaboratively shared and processed by multiple devices via a network.

[0715] <<15. Conclusion>>

[0716] Although the embodiments of the present disclosure are described above, the technical scope of the present disclosure is not limited to the embodiments described above, and various modifications may be made without departing from the gist of the present disclosure. In addition, components of different embodiments and modifications may be appropriately combined with each other.

[0717] Furthermore, the effects of the embodiments described here are merely examples and not limitations, and other effects may be provided.

[0718] It should be noted that the present technology may also have the following configurations.

[0719] (1) An information processing device for executing an application, the information processing device comprising:

[0720] The control unit issues a request for processing of the learning model used in the application to a device having a network data analysis function of the mobile network through a network exposure function and / or directly.

[0721] (2) The information processing device according to (1), wherein the control unit acquires model information about the learning model supported by the mobile network through at least one of a network data analysis function and a network exposure function.

[0722] (3) The information processing device according to (2), wherein the control unit sets a format of at least one of the learning models to be processed by the device having a network data analysis function based on the model information.

[0723] (4) The information processing device according to (3), wherein the control unit issues a request to the device, the request including application identification information for identifying the application and format information on a format of the set learning model.

[0724] (5) The information processing device according to (3) or (4), wherein

[0725] The model information includes at least information about the type of the learning model, and

[0726] The control unit sets the type of the learning model to a format for the application.

[0727] (6) An information processing device according to (5), wherein the type includes at least one of a neural network model and a deep neural network model.

[0728] (7) The information processing device according to any one of (3) to (6), wherein

[0729] The model information includes at least information about an algorithm for learning the model, and

[0730] The control unit formats the algorithm of the learning model into a format for the application.

[0731] (8) An information processing device according to (7), wherein the algorithm includes at least one of a convolutional neural network, a recurrent neural network, a fully connected neural network, a long short-term memory (LSTM), and an autoencoder.

[0732] (9) The information processing device according to any one of (3) to (8), wherein

[0733] The model information includes information about at least one of a range of the number of layers and a range of the number of nodes per layer that can be set in the learning model, and

[0734] The control unit sets at least one of the number of layers and the number of nodes of the learning model to a format for the application.

[0735] (10) The information processing device according to any one of (3) to (9), wherein

[0736] The model information includes information about the type of the learning model, and

[0737] The control unit sets the type of the learning model to a format for the application.

[0738] (11) The information processing device according to (10), wherein

[0739] The information about the type includes type information about the type of the learning model and algorithm information about the algorithm of the learning model.

[0740] The type information includes information about at least one of a neural network model and a deep neural network model, and

[0741] The algorithm information includes information about at least one of a convolutional neural network, a recurrent neural network, a fully connected neural network, a long short-term memory (LSTM), and an autoencoder.

[0742] (12) The information processing device according to any one of (3) to (11), wherein

[0743] The control unit acquires, as the model information, parameter information about parameters that can be acquired from a network function of the mobile network, and

[0744] A request is issued to the device, the request including corresponding information in which the parameter and input data of the learning model in the format are associated with each other.

[0745] (13) The information processing device according to (12), wherein

[0746] The parameter information includes first parameter information about a first parameter that can be obtained from a network function for a control plane of the mobile network, and second parameter information about a second parameter that can be obtained from a network function for processing a user plane, and

[0747] The correspondence information includes first correspondence information in which a first parameter and the format are associated with each other, and second correspondence information in which a second parameter and the format are associated with each other.

[0748] (14) The information processing device according to any one of (1) to (13), wherein the control unit issues a request for processing of the learning model to the device, the request including a notification giving an instruction on learning of the learning model.

[0749] (15) The information processing device according to (14), wherein the device performs learning of the learning model according to the notification included in the request.

[0750] (16) An information processing device according to (14) or (15), wherein, in response to a request to the device, a control unit obtains model identification information for identifying the learned learning model through at least one of a network data analysis function and a network exposure function.

[0751] (17) An information processing device according to (16), wherein the control unit specifies model identification information and issues a second request for processing the learning model to the device, the second request including a relearning notification giving instructions for relearning the acquired learning model.

[0752] (18) The information processing device according to any one of (1) to (13), wherein

[0753] The control unit performs the following steps:

[0754] obtaining parameter information about parameters that may be obtained from a network function of a mobile network;

[0755] Based on the correspondence between the parameters and the format of the learning model, implementing learning of the learning model by using the parameters as input data of the learning model; and

[0756] A request is issued to the device, the request including second format information about the acquired learning model in a second format.

[0757] (19) The information processing device according to (18), wherein the second format information includes information on settings of a format of the learning model and weighting factors between nodes of the learning model.

[0758] (20) The information processing device according to (18) or (19), wherein:

[0759] The control unit performs the following steps:

[0760] In response to the request to the device, obtaining model identification information for identifying a learning model through at least one of a network data analysis function and a network exposure function; and

[0761] A second request is issued to the device through a network exposure function or directly, wherein the second request includes an instruction to update the learning model indicating the model identification information.

[0762] (21) The information processing device according to (20), wherein

[0763] The control unit performs the following steps:

[0764] generating a third format of an updated learning model; and

[0765] A second request for giving an instruction to update the learning model in the set third format is issued to the device.

[0766] (22) The information processing device according to (21), wherein

[0767] The control unit performs the following steps:

[0768] Relearning the learning model based on the quality of experience (QoE) of the application; and

[0769] A third format is generated based on the learning model for which re-learning is performed.

[0770] (23) The information processing device according to any one of (14) to (22), wherein

[0771] a control unit issuing a request to the device, the request comprising information about a first quality of service (QoS) parameter and information about a second QoS parameter,

[0772] The first QoS parameter and the second QoS parameter are parameters applied in a session for transmitting at least one of input data input to the learned learning model used in the application and output data output from the learning model,

[0773] The first QoS parameter is applied to a network function processing a user plane of a mobile network so as to obtain a second parameter obtainable from the network function processing the user plane among various input data, and

[0774] The second QoS parameter is applied to a network function handling a user plane in order to transmit the outgoing data.

[0775] (24) An information processing device for establishing or updating a session for processing an application, the information processing device comprising:

[0776] A control unit obtains, based on the application, a strategy applied to one or more learning models processed by the network data analysis function for the application.

[0777] (25) The information processing device according to (24), wherein the application is specified based on at least one of a data network name and an application ID.

[0778] (26) The information processing device according to (24) or (25), wherein

[0779] The strategies include:

[0780] Setting of correspondence between parameters obtained from network functions of the mobile network and input data of the learning model;

[0781] Setting of correspondence between a second parameter acquired from a function for processing a user plane among the parameters and the input data;

[0782] a first QoS parameter applied to a second parameter; and

[0783] A second QoS parameter is applied to the output data of the learned model.

[0784] (27) The information processing device according to (26), wherein

[0785] The control unit performs the following steps:

[0786] sending the first QoS parameter and the second QoS parameter to a first network function having a first interface with the base station device in a control plane;

[0787] Sending the first QoS parameter to the base station device through the first interface; and

[0788] The second QoS parameter is sent to the communication device using a non-access stratum (NAS) message via a second interface with the base station device.

[0789] (28) The information processing device according to any one of (24) to (27), wherein

[0790] The control unit receives a message requesting to establish the one session,

[0791] The message includes a data network name and slice identification information identifying a network slice, and

[0792] The control unit indicates a network data analysis function for processing the one or more learning models based on at least one of the data network name, application ID and slice identification information.

[0793] (29) The information processing device according to (28), wherein, when the network data analysis function is specified, the control unit sets the specified network data analysis function in the edge application server.

[0794] (30) The information processing device according to (29), wherein the control unit selects the edge application server based on location information about the communication device acquired from at least one of the communication device and the location management function.

[0795] (31) An information processing device according to any one of (24) to (30), wherein, when first network slice selection assistance information corresponding to a network slice in which a network data analysis function processes one or more learning models used in the application is included in allowed NSSAI, a control unit requests establishment or update of the session including the first network slice selection assistance information.

[0796] (32) An information processing device according to (31), wherein, when the first network slice selection auxiliary information is included in a request for establishment or update of the one session, the control unit indicates one or more learning models to be processed by the network data analysis function for the application.

[0797] (33) An information processing device according to any one of (24) to (32), wherein

[0798] The control unit sets one or more QoS flows for traffic generated by a second process different from the first process among the processes executed by the specified application and transmitted to / from the communication device, and

[0799] A first process is implemented using the one or more learning models in a network data analysis function.

[0800] (34) An information processing device according to (33), wherein:

[0801] The control unit performs the following steps:

[0802] When the resource type of the QoS flow is a GBR QoS flow, setting a second process corresponding to the QoS flow in the edge application server; and

[0803] When the resource type of the QoS flow is a non-GBR QoS flow, a second process corresponding to the QoS flow is set in the application server.

[0804] (35) The information processing device according to (33) or (34), wherein

[0805] The control unit performs the following steps:

[0806] When the priority level of the QoS flow is equal to or lower than the set threshold, setting a second process corresponding to the QoS flow in the edge application server; and

[0807] When the priority level of the QoS flow exceeds the set threshold, a second process corresponding to the QoS flow is set in the application server.

[0808] (36) The information processing device according to (33) or (34), wherein

[0809] The control unit performs the following steps:

[0810] When the packet delay budget of the QoS flow is equal to or lower than the set threshold, setting a second process corresponding to the QoS flow in the edge application server; and

[0811] When the packet delay budget of the QoS flow exceeds the set threshold, a second process corresponding to the QoS flow is set in the application server.

[0812] (37) The information processing device according to (33) or (34), wherein

[0813] The control unit performs the following steps:

[0814] When the 5QI of the QoS flow is included in the set set, setting a second process corresponding to the QoS flow in the edge application server; and

[0815] When the 5QI of the QoS flow is not included in the set set, a second process corresponding to the QoS flow is set in the application server.

[0816] (38) The information processing device according to (29) or (30), wherein:

[0817] The control unit performs the following steps:

[0818] determining whether a first QoS parameter and a second QoS parameter are satisfied;

[0819] As a result of the determining, when the first QoS parameter and the second QoS parameter are not satisfied, setting a specified network data analysis function in the edge application server;

[0820] The first QoS parameter is applied to a second parameter acquired from a function processing a user plane among parameters acquired from a network function of the mobile network; and

[0821] The second QoS parameter is applied to the output data of the learned model.

[0822] (39) An information processing device according to any one of (24) to (38), wherein

[0823] The control unit performs the following steps:

[0824] obtaining at least one of capability and communication quality of a communication device;

[0825] Based on at least one of a capability of the communication device and a communication quality, determining that the network data analysis function divides the one or more learning models into a first learning model and a second learning model;

[0826] sending first format information about the format of the first learning model to the communication device;

[0827] The communication device sets a first learning model using the received first format information; and

[0828] The network data analysis function sets the second learning model.

[0829] (40) The information processing device according to any one of (24) to (39), wherein

[0830] The control unit performs the following steps:

[0831] acquiring, from a base station device connected to the communication device, quality information on the quality of communication between the communication device and the base station device;

[0832] determining, based on the quality information, that the network data analysis function divides the one or more learning models into a third learning model and a fourth learning model; and

[0833] transmitting third format information regarding the format of the third learning model to the communication device;

[0834] The communication device sets a third learning model using the received third format information; and

[0835] The network data analysis function sets the fourth learning model.

[0836] (41) An information processing device having a function of managing a policy applied to a session, the information processing device comprising:

[0837] a control unit that receives a request through a network exposure function or directly from an application function, the request including application identification information for identifying an application and format information about a format of a learning model, and

[0838] A policy is generated based on the application identification information and the format information included in the request.

[0839] (42) An information processing device according to (41), wherein the format information includes at least one of model information about the learning model and correspondence information about the correspondence between parameters of the mobile network system input to the input layer of the learning model and input data of the learning model.

[0840] (43) An information processing device according to (42), wherein the model information includes information about any one of the type of learning model, the type of algorithm, the number of layers, the number of nodes in each layer, and the weighting factors between nodes.

[0841] (44) A communication device for establishing a session with an edge application server for processing a user plane, the communication device comprising a control unit, the control unit implementing the following steps:

[0842] sending input data to be input into the learned learning model to be processed by the network data analysis function set in the edge application server;

[0843] obtaining QoS rules to be applied to the input data; and

[0844] Flow identification information for identifying a QoS flow is assigned to the input data based on a QoS rule.

[0845] (45) The communication device according to (44), wherein

[0846] The control unit performs the following steps:

[0847] acquiring first setting information for setting a first learning model among a first learning model and a second learning model obtained by dividing the learning model; and

[0848] The values ​​of the nodes of the output layer of the first learning model are sent as the input data.

[0849] (46) The communication device according to (44) or (45), wherein

[0850] Identify the learning model through the model identification information, and

[0851] Flow identification information for identifying a QoS flow is assigned to each item of model identification information.

[0852] (47) The communication device according to (45), wherein

[0853] identifying a first learning model through first model identification information, and

[0854] Flow identification information for identifying a QoS flow is assigned to each item of first model identification information.

[0855] (48) A communication device according to any one of (44) to (47), wherein

[0856] The session adds a function of an uplink classifier (UL CL) or a function of a branch point (BP) to a second user plane with the application server, and connects the user plane with the application server in the form of session grooming,

[0857] The application server performs a second process using the result of the learned learning model, and

[0858] The control unit transmits the second data for the second process to the application server through the function of UL CL or BP.

[0859] (49) A communication device according to any one of (44) to (48), wherein

[0860] The control unit performs the following steps:

[0861] Send input data via base station equipment;

[0862] obtaining measurement information and reporting settings regarding measurements of signals received from a base station device;

[0863] reporting a measurement result of the signal based on the measurement information; and

[0864] First setting information for setting a first learning model among a first learning model and a second learning model obtained by dividing the learning model is acquired based on the measurement result.

[0865] (50) A communication device according to any one of (44) to (49), wherein

[0866] The control unit performs the following steps:

[0867] sending a session update request including a request to change a partition point of the learning model; and

[0868] As a response to the update request, first setting information on settings of a first learning model obtained by dividing the learning model is acquired.

[0869] (51) The communication device according to (50), wherein

[0870] The control unit performs the following steps:

[0871] obtaining a resource usage ratio of at least one of a central processing unit (CPU), a graphics processing unit (GPU), and a system-on-chip (SoC) on which the CPU and the GPU are mounted; and

[0872] Sending a session update request is determined based on the usage ratio.

[0873] (52) The communication device according to any one of (44) to (51), wherein the control unit transmits the second session update request including a request for re-learning of the learning model.

[0874] (53) The communication device according to (52), wherein the control unit measures quality of experience (QoE) for the application using the learning model, and determines the sending of the second session update request based on the measured QoE.

[0875] (54) A communication device according to any one of (44) to (53), wherein

[0876] The control unit obtains the allowed NSSAI in response to the registration request including the first network slice selection assistance information,

[0877] The first network slice selection assistance information corresponds to a network slice in which the network data analysis function processes the learning model, and

[0878] When the first network slice selection assistance information is included in the allowed NSSAI, the control unit requests establishment of the session.

[0879] (55) A communication device according to (54), wherein, when the first network slice selection assistance information is not included in the allowed NSSAI, the control unit stops the request for session establishment.

[0880] (56) The communication device according to any one of (44) to (55), further comprising:

[0881] Sensors, including

[0882] The control unit sends sensor data acquired by the sensor as the input data.

[0883] (57) An information processing device comprising a control unit, the control unit implementing the following steps:

[0884] Establishing multiple sessions with multiple communication devices using one application; and

[0885] Synchronization between the plurality of sessions is set in a network function that manages the plurality of sessions.

[0886] (58) An information processing device according to (57), wherein

[0887] The control unit performs the following steps:

[0888] obtaining relevant information indicating associations between the plurality of sessions; and

[0889] A session for which synchronization is to be set among the plurality of sessions is specified based on the relevant information.

[0890] (59) The information processing device according to (57) or (58), wherein

[0891] The network function sets an allowable delay amount associated with the synchronization in a second network function processing user plane data of the plurality of sessions.

[0892] (60) The information processing device according to (59), wherein the allowable delay amount is set independently for uplink and downlink.

[0893] (61) An information processing device according to any one of (57) to (60), wherein

[0894] The network function provides, based on the QoS monitoring result of each of the plurality of sessions, first offset information about the first offset for the synchronization to a second network function processing user plane data of the plurality of sessions, and

[0895] The second network function controls buffering based on a first offset and an allowable delay associated with synchronization for one of the plurality of sessions.

[0896] (62) An information processing device according to any one of (57) to (60), wherein

[0897] The network function provides, based on the QoS monitoring result of each of the plurality of sessions, first offset information about the first offset for the synchronization to a second network function processing user plane data of the plurality of sessions, and

[0898] The second network function controls prioritized packet processing based on a first offset and an allowable delay associated with synchronization for one of the plurality of sessions.

[0899] (63) An information processing device according to any one of (57) to (60), wherein

[0900] The second network function processing the user plane data of the plurality of sessions implements the following steps:

[0901] detecting a first offset for the synchronization based on the QoS monitoring result of each of the plurality of sessions; and

[0902] Buffering is controlled based on a first offset and an allowable delay associated with synchronization for one of the plurality of sessions.

[0903] (64) An information processing device according to any one of (57) to (60), wherein

[0904] The second network function processing the user plane data of the plurality of sessions implements the following steps:

[0905] detecting a first offset for the synchronization based on the QoS monitoring result of each of the plurality of sessions; and

[0906] Prioritized packet processing is controlled based on a first offset and an allowable delay associated with synchronization for one of the plurality of sessions.

[0907] (65) The information processing device according to any one of (57) to (64), wherein the network function acquires a request for synchronization between a plurality of sessions established for the application through an application function (AF).

[0908] (66) The information processing device according to (65), wherein the network function obtains an allowable delay amount related to the synchronization through an AF request.

[0909] (67) The information processing device according to (65) or (66), wherein the network function acquires, through an AF request, designation information for designating a QoS flow to be synchronized.

[0910] (68) An information processing device according to any one of (65) to (67), wherein

[0911] The control unit generates a policy for managing the session based on the AF request, and

[0912] The network function acquires the created policy and controls the session according to the acquired policy.

[0913] (69) An information processing device according to any one of (57) to (68), wherein

[0914] The application is processed by an edge application server, and

[0915] The control unit establishes the multiple sessions with the edge application server in the form of a distributed anchor point.

[0916] (70) An information processing device according to any one of (57) to (68), wherein

[0917] The application is processed by an edge application server, and

[0918] The control unit establishes the multiple sessions with the edge application server in the form of session grooming.

[0919] (71) An information processing device according to any one of (57) to (70), wherein

[0920] The plurality of communication devices include a control device and a display device equipped with a sensor, and

[0921] The application implements the following steps:

[0922] receiving, through a first session of the plurality of sessions, sensor information detected by the sensor;

[0923] generating moving image data based on the sensor information; and

[0924] The moving image data is transmitted to a display device through a second session of the plurality of sessions.

[0925] (72) The information processing device according to (71), wherein the control unit requests the display device to establish the first session for relaying to the control device through side link communication with the display device.

[0926] (73) The information processing device according to (72), wherein the control unit notifies the display device to operate as a UE to network relay.

[0927] (74) An information processing device according to any one of (71) to (73), wherein

[0928] The control unit performs the following steps:

[0929] The session update process is activated, and the first session and the second session are integrated.

[0930] (75) An information processing device according to (57), wherein

[0931] A first session included in the plurality of sessions is a session between a first communication device included in the plurality of communication devices and a first edge application server,

[0932] A second session included in the plurality of sessions is a session between a second communication device included in the plurality of communication devices and a second edge application server,

[0933] The control unit selects a third network function as an anchor point from the first network function managing the first session to the second network function managing the second session, and

[0934] The third network function controls synchronization between the first session and the second session.

[0935] (76) An information processing device according to (75), wherein

[0936] The third network function implements the following steps:

[0937] detecting a first offset for the synchronization based on a QoS monitoring result of each of the first session and the second session; and

[0938] First offset information about the first offset is provided to a fourth network function that processes user plane data for the first session.

[0939] (77) The information processing device according to (76), wherein the fourth network function controls buffering for the first session based on the first offset information.

[0940] (78) The information processing device according to (76), wherein the fourth network function controls priority packet processing for the first session based on the first offset information.

[0941] (79) An information processing device according to any one of (75) to (78), wherein

[0942] a third session included in the plurality of sessions is a session in a first public land mobile network (PLMN) between a first communication device among the plurality of communication devices and a first edge application server,

[0943] A fourth session included in the plurality of sessions is a session between a second communication device included in the plurality of communication devices and a second edge application server, and

[0944] The network function of the first PLMN obtains synchronization information about synchronization between the third session and the fourth session through the AF request, and

[0945] A second offset is set in a fourth network function of the first PLMN based on the synchronization information, and the fourth network function processes user plane data of the first session.

[0946] (80) The information processing device according to (79), wherein the fourth session is a session of the first PLMN.

[0947] (81) The information processing device according to (79), wherein the fourth session is a session of the second PLMN.

[0948] (82) The information processing device according to any one of (79) to (81), wherein the synchronization information is set based on a first QoS monitoring result regarding a third session and a second QoS monitoring result regarding a fourth session.

[0949] (83) The information processing device according to any one of (79) to (82), wherein the fourth network function controls buffering for the third session based on second offset information about the second offset.

[0950] (84) The information processing device according to any one of (79) to (82), wherein the fourth network function controls the priority packet processing for the third session based on the second offset information about the second offset.

[0951] (85) An information processing device according to (57), wherein

[0952] A first session included in the plurality of sessions is a session between a first communication device included in the plurality of communication devices and an edge application server,

[0953] A second session included in the plurality of sessions is a session between a second communication device included in the plurality of communication devices and the edge application server, and

[0954] The control unit performs the following steps:

[0955] Separating user plane data processed by the edge application server into data for multicast transmission and data for unicast transmission;

[0956] acquiring multicast data for the same multicast transmission, and transmitting the multicast data to the first communication device and the second communication device; and

[0957] Different unicast data for unicast transmission is acquired, and the unicast data is transmitted through the first session and the second session.

[0958] (86) The information processing device according to (85), wherein the control unit sends a copy of the multicast data via shared transmission.

[0959] (87) An information processing device according to (85), wherein

[0960] The control unit performs the following steps:

[0961] Sending a copy of the multicast data via the first session; and

[0962] Another copy of the multicast data is sent via the second session.

[0963] (88) An information processing device according to any one of (85) to (87), wherein when the base station device to which the first communication device and the second communication device are connected supports multicast / broadcast service (MBS), the control unit sends a copy of the multicast data via shared transmission.

[0964] (89) An information processing device according to any one of (85) to (88), wherein, when a base station device to which the first communication device and the second communication device are connected does not support MBS, a control unit sends one copy of the multicast data through a first session and sends another copy of the multicast data through a second session.

[0965] (90) The information processing device according to any one of (85) to (89), wherein when the number of communication devices connected to the edge application server exceeds a preset number, the control unit separates the multicast data and the unicast data from each other.

[0966] (91) The information processing device according to any one of (85) to (90), wherein the control unit assigns the MBS QoS flow to the multicast data.

[0967] Reference numerals list

[0968] 10——Wireless communication equipment

[0969] 20——Base station equipment

[0970] 31, 110, 210 - Communication unit

[0971] 32, 120, 220 - storage unit

[0972] 33, 150, 240 - Control unit

[0973] 40 - Application Server

[0974] 41 - Edge Application Server

[0975] 111, 211 - Receiving and processing unit

[0976] 112, 212——Sending processing unit

[0977] 113, 213, 214——Antenna

[0978] 130, 230——Network communication unit

[0979] 140 - Input / output unit

[0980] 300——Information processing equipment

Claims

1. An information processing device for executing an application, the information processing device comprising: The control unit issues a request for processing of the learning model used in the application to a device having a network data analysis function of the mobile network through a network exposure function and / or directly.

2. The information processing device according to claim 1, wherein: The control unit acquires model information about a learning model supported by the mobile network through at least one of a network data analysis function and a network exposure function.

3. The information processing device according to claim 2, wherein: The control unit sets a format of at least one of the learning models to be processed by the device having a network data analysis function based on the model information.

4. The information processing device according to claim 3, wherein: The control unit issues a request to the device, the request including application identification information for identifying the application and format information about the format of the set learning model.

5. The information processing device according to claim 3, wherein The control unit acquires, as the model information, parameter information about parameters that can be acquired from a network function of the mobile network, and A request is issued to the device, the request including corresponding information in which the parameter and input data of the learning model in the format are associated with each other.

6. The information processing device according to claim 5, wherein The parameter information includes first parameter information about a first parameter that can be obtained from a network function for a control plane of the mobile network, and second parameter information about a second parameter that can be obtained from a network function for processing a user plane, and The correspondence information includes first correspondence information in which a first parameter and the format are associated with each other, and second correspondence information in which a second parameter and the format are associated with each other.

7. The information processing device according to claim 1, wherein: The control unit issues a request for processing of a learning model to the device, the request including a notification giving an instruction on learning of the learning model.

8. The information processing device according to claim 1, wherein The control unit performs the following steps: obtaining parameter information about parameters that may be obtained from a network function of a mobile network; Based on the correspondence between the parameters and the format of the learning model, implementing learning of the learning model by using the parameters as input data of the learning model; and A request is issued to the device, the request including second format information about the acquired learning model in a second format.

9. The information processing device according to claim 8, wherein The control unit performs the following steps: In response to the request to the device, obtaining model identification information for identifying a learning model through at least one of a network data analysis function and a network exposure function; and A second request is issued to the device through a network exposure function or directly, wherein the second request includes an instruction to update the learning model indicating the model identification information.

10. The information processing device according to claim 7, wherein The control unit sends a request to the device, the request including information about the first quality of service QoS parameter and information about the second QoS parameter, The first QoS parameter and the second QoS parameter are parameters applied in a session for transmitting at least one of input data input to the learned learning model used in the application and output data output from the learning model, The first QoS parameter is applied to a network function processing a user plane of a mobile network so as to obtain a second parameter obtainable from the network function processing the user plane among various input data, and The second QoS parameter is applied to a network function handling a user plane in order to transmit the outgoing data.

11. An information processing device for establishing or updating a session for processing an application, the information processing device comprising: A control unit obtains, based on the application, a strategy applied to one or more learning models processed by the network data analysis function for the application.

12. The information processing device according to claim 11, wherein: The application is specified based on at least one of a data network name and an application ID.

13. The information processing device according to claim 11, wherein The strategies include: Setting of correspondence between parameters obtained from network functions of the mobile network and input data of the learning model; Setting of correspondence between a second parameter acquired from a function for processing a user plane among the parameters and the input data; a first QoS parameter applied to a second parameter; as well as A second QoS parameter is applied to the output data of the learned model.

14. The information processing device according to claim 13, wherein The control unit performs the following steps: sending the first QoS parameter and the second QoS parameter to a first network function having a first interface with the base station device in a control plane; Sending the first QoS parameter to the base station device through the first interface; as well as The second QoS parameter is sent to the communication device using a non-access stratum (NAS) message via a second interface with the base station device.

15. The information processing device according to claim 11, wherein The control unit receives a message requesting to establish the one session, The message includes a data network name and slice identification information identifying a network slice, and The control unit indicates a network data analysis function for processing the one or more learning models based on at least one of the data network name, application ID and slice identification information.

16. The information processing device according to claim 15, wherein: When the network data analysis function is specified, the control unit sets the specified network data analysis function in the edge application server.

17. The information processing device according to claim 16, wherein The control unit performs the following steps: determining whether a first QoS parameter and a second QoS parameter are satisfied; As a result of the determining, when the first QoS parameter and the second QoS parameter are not satisfied, setting a specified network data analysis function in the edge application server; The first QoS parameter is applied to a second parameter acquired from a function processing a user plane among parameters acquired from a network function of the mobile network; and The second QoS parameter is applied to the output data of the learned model.

18. The information processing device according to claim 11, wherein The control unit performs the following steps: obtaining at least one of capability and communication quality of a communication device; Based on at least one of a capability of the communication device and a communication quality, determining that the network data analysis function divides the one or more learning models into a first learning model and a second learning model; sending first format information about the format of the first learning model to the communication device; The communication device sets a first learning model using the received first format information; The network data analysis function sets the second learning model; as well as The communication quality is the communication quality between the communication device and the base station device, and a notification of the communication quality is provided by the base station device connected to the communication device.

19. An information processing device having a function of managing a policy applied to a session, the information processing device comprising: a control unit that receives a request through a network exposure function or directly from an application function, the request including application identification information for identifying an application and format information about a format of a learning model, and A policy is generated based on the application identification information and the format information included in the request.

20. The information processing device according to claim 19, wherein: The format information includes at least one of model information about the learning model and correspondence information about the correspondence between parameters of the mobile network system input to an input layer of the learning model and input data of the learning model.

Images