Network slice management device and management method

Abstract

A network slice management device according to one aspect manages network slices in a cellular network. The network slice management device has a control unit. The control unit uses a trained AI model for application identification to input packet information indicating the operation of a packet of an application in the cellular network and output an application name. The control unit uses a trained AI model for communication control to input the application name and output communication control information indicating information for communication control in the cellular network of the packet. Further, when a packet including the name of a new application has been received, the control unit causes the trained AI model for application identification and the trained AI model for communication control to be retrained using at least the name of the new application.

Description

Network slice management device and management method

[0001] The present invention relates to a network slice management device and a management method.

[0002] Network slicing has been introduced in mobile communication systems including cellular networks. Network slicing is a technology that virtually divides a physical network constructed by a carrier to create multiple virtual networks. Each virtual network is called a network slice. Network slicing allows carriers to create network slices according to the service requirements of different service types, such as enhanced Mobile Broadband (eMBB), ultra-reliable and low latency communications (URLLC), and massive machine type communications (mMTC). Network slicing can, for example, optimize network resources.

[0003] On the other hand, users may use terminal devices such as smartphones to execute various applications. In particular, there are now use cases in which new applications not seen before, such as remote medical care or robot control, are executed.

[0004] 3GPP TS 38.300 V17.5.0 (2023-06)3GPP TS 23.501 V17.9.0 (2023-06)

[0005] A network slice management device according to one embodiment is a network slice management device that manages network slices in a cellular network. The network slice management device has a control unit. The control unit uses a trained AI model for application identification to input packet information indicating the behavior of packets of an application in the cellular network and output an application name. The control unit also uses a trained AI model for communication control to input the application name and output communication control information indicating information for communication control of the packet in the cellular network. Furthermore, when the control unit receives a packet including the name of a new application, it re-trains the trained AI model for communication control and the trained AI model for application identification using at least the name of the new application.

[0006] Also, one aspect of the management method is a management method in a network slice management device that manages network slices in a cellular network. The management method includes a step of inputting packet information indicating the behavior of packets of an application in the cellular network using a trained AI model for application identification and outputting an application name. The management method also includes a step of inputting the application name using a trained AI model for communication control and outputting communication control information indicating information for communication control of the packet in the cellular network. Furthermore, when a packet including the name of a new application is received, the management method includes a step of relearning the trained AI model for application identification and the trained AI model for communication control using at least the name of the new application.

[0007] FIG. 1 is a diagram illustrating an example of the configuration of a communication system according to the first embodiment. FIG. 2 is a diagram illustrating an example of the configuration of a cellular network according to the first embodiment. FIG. 3 is a diagram illustrating an example of a network slice according to the first embodiment. FIG. 4 is a diagram illustrating an example of a network slice according to the first embodiment. FIG. 5 is a diagram illustrating an example of network slice management according to the first embodiment. FIG. 6 is a diagram illustrating an example of the configuration of a UE according to the first embodiment. FIG. 7 is a diagram illustrating an example of the configuration of a node according to the first embodiment. FIG. 8 is a diagram illustrating an example of the configuration of a CN device. FIG. 9 is a diagram illustrating an example of the configuration of a network slice management device according to the first embodiment. FIG. 10 is a diagram illustrating an example of operation according to the first embodiment. FIG. 11 is a diagram illustrating an example of generation of a trained model according to the first embodiment. FIGS. 12(A) and 12(B) are diagrams illustrating an example of generation of a trained model for application identification according to the first embodiment. FIGS. 13(A) and 13(B) are diagrams illustrating an example of generation of a trained model for network slice design according to the first embodiment. FIGS. 14(A) and 14(B) are diagrams illustrating an example of generation of a trained model for communication control according to the first embodiment. FIG. 15 is a diagram illustrating an example of the configuration of a packet according to the first embodiment.

[0008] [First Embodiment] As described above, network slicing has been introduced in mobile communication systems. Meanwhile, new applications not previously seen in mobile communication systems, such as telemedicine or robot control, may be implemented in mobile communication systems. For example, telemedicine requires not only high-speed, large-capacity communication but also low-latency communication. Furthermore, for example, robot control requires both low-latency communication and multi-connection communication.

[0009] Operators of mobile communication system infrastructure manually decide which network slice to apply for each application, from among high-speed, large-capacity (eMBB), ultra-low latency (URLLC), and multiple connections (mMTC), depending on the communication requirements of each application.

[0010] However, given the current situation in which an increase in new applications is expected in the future, it may take a huge amount of time and effort for operators to manually decide which network slice to apply to each application.

[0011] Here, a network slice refers to each virtual network obtained by virtually dividing a physical network constructed by a carrier in a cellular network. A network slice may be a logical network in a cellular network, which allows dynamic allocation of resources in the cellular network. Alternatively, a network slice may be a network obtained by virtually dividing a cellular network according to services to be provided to users.

[0012] A network slice includes a RAN slice, a transport slice, and a core slice. The RAN slice provides, for example, resource control and priority control for each network slice in a RAN (Radio Access Network) that performs radio access control. The RAN slice may also be called a RAN slice subnet. Furthermore, the transport slice provides network slice functions for each network, for example, a fronthaul (a network between a Radio Unit (RU), a Distributed Unit (DU), and a Central Unit (CU) when the RAN is separated into an RU, a DU, and a CU), a middlehaul (a network between a DU and a CU), and a backhaul (a network between a CU and a core network (CN)). The transport slice may also be called a transport slice subnet. Furthermore, the core slice provides, for example, core network functions in each network slice. The core slice may also be called a core slice subnet.

[0013] The network slice that the operator manually applies is intended to be applied to the entire network slice, including the RAN slice, transport slice, and core slice, for each application.

[0014] Therefore, the first embodiment aims to set optimal communication control within a network.

[0015] The first embodiment will be specifically described below with reference to the drawings. In the description of the drawings, the same or similar parts are denoted by the same or similar reference numerals.

[0016] (1) Example of the Configuration of a Communication System First, an example of the configuration of a communication system according to the first embodiment will be described.

[0017] FIG. 1 is a diagram illustrating an example of the configuration of a communication system 1.

[0018] 1, the communication system 1 includes a user equipment (UE) 100, a node 110, a CN (Core Network) device 120, and a network slice management device 130. The communication system 1 may include a cellular network 10. The cellular network 10 includes the UE 100, the node 110, the CN device 120, and the network slice management device 130.

[0019] The cellular network 10 is a network capable of wireless communication with a mobile UE 100. The cellular network 10 is also a network to which a 3GPP-standard mobile communication system is applied. The cellular network 10 is, for example, a network compliant with the 3GPP-standard 5th Generation System (5GS). The cellular network 10 may partially adopt a network compliant with the 3GPP-standard Long Term Evolution (LTE) system, or may at least partially adopt a 6th Generation (6G) system that is scheduled to be standardized in the future. The cellular network 10 may also be a mobile communication system. A configuration example of the cellular network 10 will be described later.

[0020] The UE 100 is a mobile wireless communication device. The UE 100 may be any device used by a user, but for example, the UE 100 may be a mobile phone terminal (including a smartphone), a tablet terminal, a notebook PC, a communication module (including a communication card or a chipset), a sensor or a device provided in a sensor, a vehicle or a device provided in a vehicle (Vehicle UE), an aircraft or a device provided in an aircraft (Aerial UE, or UAV (Unmanned Aerial Vehicle)). Alternatively, the UE 100 may be an IoT (Internet of Things) device, an IoT sensor, or the like.

[0021] The node 110 may be connected as one node or as multiple nodes in the cellular network 10. The node 110 may function as a base station in the cellular network 10. Details of the node 110 will be described later. The node 110 may also be called a network node.

[0022] The CN device 120 functions as a communication device in the cellular network 10. The CN device 120 may be a device that connects the cellular network 10 to an external network. Alternatively, the CN device 120 may be a device having a gateway function that connects the cellular network 10 to an external network. The CN device 120 may be an Access and Mobility Management Function (AMF) that manages access and mobility control of the UE 100. Alternatively, the CN device 120 may be a Session Management Function (SMF) that manages a communication session of the UE 100 in the cellular network 10. Alternatively, the CN device 120 may be a User Plane Function (UPF) that functions as a termination point of a communication session (e.g., a Protocol Data Unit (PDU) session) with the UE 100 in the cellular network 10 and exchanges user data with the UE 100.

[0023] The network slice management device 130 manages one or more network slices in the cellular network 10. As a network slice orchestrator, the network slice management device 130 manages the RAN slice, transport slice, and core slice included in the network slice. The network slice management device 130 may manage the generation of network slices. The network slice management device 130 may also configure network slices in the node 110 and the CN device 120 by sending configuration information to the node 110 and the CN device 120. Note that "management" of a network slice may include "control" of the network slice. Alternatively, "management" of a network slice and "control" of a network slice may be used to have the same meaning.

[0024] In the first embodiment, the network slice management device 130 can also manage network slices using a learning function based on AI (Artificial Intelligence). Details will be described later. Note that, hereinafter, the network slice management device 130 may be referred to as a "network slice orchestrator" (or simply as an "orchestrator").

[0025] (1.1) Example of Cellular Network Configuration Next, an example of the configuration of the cellular network 10 will be described.

[0026] FIG. 2 is a diagram illustrating an example of the configuration of the cellular network 10 according to the first embodiment.

[0027] As shown in FIG. 2 , the cellular network 10 includes a UE 100 , a Radio Access Network (RAN) 30 , a Core Network (CN) 40 , and an orchestrator 130 .

[0028] The RAN 30 includes one or more nodes 110 (110-1 to 110-3 in the example of FIG. 2) described above. The nodes 110 are connected to each other via inter-node interfaces. The nodes 110 may also be referred to as base stations. When the cellular network 10 is 5GS, the nodes 110 are also referred to as gNBs. The nodes 110 may be configured with an RU, a CU, and a DU (i.e., functionally divided).

[0029] The node 110 manages one or more cells. The node 110 performs wireless communication with the UE 100 that has established a connection with the node 110's own cell. The node 110 has a radio resource management (RRM) function, a routing function for user data (hereinafter simply referred to as "data"), a measurement control function for mobility control and scheduling, and the like. Note that the term "cell" is used to indicate the smallest unit of a wireless communication area. The term "cell" is also used to indicate a function or resource for performing wireless communication with the UE 100. One cell belongs to one carrier frequency (hereinafter simply referred to as "frequency").

[0030] The CN 40 includes the CN device 120 described above.

[0031] (1.1.1) Network Slice The network slice according to the first embodiment is constructed in the cellular network 10. Here, the network slice will be described.

[0032] As described above, a network slice is a virtual network created by virtually dividing a physical network built by an operator. By building a network slice, services such as high-speed, large-capacity (eMBB), ultra-low latency (URLLC), and multiple connections (mMTC) can be provided to the UE 100 by occupying resources within the cellular network 10.

[0033] FIG. 3 is a diagram illustrating an example of a network slice. The network slice is configured on a network consisting of a RAN 30 and a CN 40. One or more network slices can be configured on the network. Each network slice is associated with one service type. Examples of service types include eMBB, URLLC, Massive Internet of Things (MIoT), Vehicle to Everything (V2X), and High Performance Machine Type Communication (HMTC). For example, as shown in FIG. 3, network slice #1 is associated with the service type eMBB, network slice #2 is associated with the service type MIoT, and network slice #3 is associated with the service type HMTC.

[0034] Each network slice is provided with a slice identifier that identifies the network slice. An example of a slice identifier is S-NSSAI (Single Network Slicing Selection Assistance Information). The S-NSSAI includes an 8-bit SST (slice / service type). The S-NSSAI may further include a 24-bit SD (slice differentiator).

[0035] SST is an example of service type information indicating a service type to which a network slice is associated. For example, SST=1 indicates eMBB, SST=2 indicates URLLC, SST=3 indicates MIoT (Massive Internet of Things), SST=4 indicates V2X (Vehicle to everything), and SST=5 indicates HMTC (High Performance Machine Type Communication). However, these are merely examples, and SST values ​​may be associated with other service types, or other service types may be associated with other SST values.

[0036] The SD is information for differentiating multiple network slices associated with the same service type.

[0037] Information including multiple S-NSSAIs is called NSSAI (Network Slice Selection Assistance Information). One or more network slices may be grouped to form a slice group. A slice group is a group including one or more network slices, and a slice group identifier is assigned to the slice group.

[0038] As shown in FIG. 3, multiple slices may be configured for one UE 100, in which case the UE 100 can simultaneously receive multiple services via each slice.

[0039] As shown in FIG. 3 , the network slices include hardware and functional blocks in the RAN 30 and the CN 40. In the example of FIG. 3 , network slice #1 includes an RU, a DU, a CU #1, and a UPF #1. Network slice #2 includes an RU, a DU, a CU #2, and a UPF #2. Network slice #3 includes an RU, a DU, a CU #3, and a UPF #3. Each hardware and logical functional block in the RAN 30 and the CN 40 may be shared by multiple network slices. Each network slice can provide services such as eMBB and URLLC to the UE 100 by using the hardware and functional blocks. Hereinafter, the hardware and functional blocks in the cellular network 10 may be simply referred to as "resources."

[0040] As described above, a network slice includes a RAN slice, a transport slice, and a core slice. Figure 4 is a diagram illustrating an example of a network slice including a RAN slice, a transport slice, and a core slice. In the example of Figure 4, network slice #1 includes RAN slice #1, transport slice #1, and core slice #1. Network slice #2 includes RAN slice #2, transport slice #2, and core slice #2. Network slice #3 includes RAN slice #3, transport slice #3, and core slice #3. In this way, by including a RAN slice, a transport slice, and a core slice in a network slice, it is possible to dynamically change the resources used in the RAN slice, for example, and to flexibly respond to the requirements of each network slice for each RAN slice, transport slice, and core slice.

[0041] The orchestrator 130 (network slice management device 130) includes a RAN slice controller 131, a transport slice controller 132, and a core slice controller 133 to manage the RAN slice, transport slice, and core slice included in the network slice, respectively. The orchestrator 130 also includes a network slice controller 135 to manage the entire network slice.

[0042] The RAN slice controller 131 controls RAN slices in the RAN 30. For example, the RAN slice controller 131 controls radio resources used in wireless communication with the UE 100 and which RAN slices resources (e.g., RU, DU, CU, etc.) in the RAN 30 used in communication with the UE 100 correspond to. The RAN slice controller 131 may control each RAN slice in accordance with instructions from the network slice controller 135. Furthermore, when a packet is transmitted in the RAN 30, the RAN slice controller 131 may obtain RAN packet information indicating how the packet operated in the RAN 30. The RAN slice controller 131 outputs the RAN packet information to the network slice controller 135.

[0043] The transport slice controller 132 controls transport slices in the fronthaul, middlehaul, and backhaul. For example, the transport slice controller 132 controls route information indicating the route from the fronthaul to the backhaul for each transport slice. The transport slice controller 132 may also control each transport slice in accordance with instructions from the network slice controller 135. Furthermore, when a packet is transmitted in the backhaul, middlehaul, and fronthaul, the transport slice controller 132 may obtain transport packet information indicating how the packet operated in the backhaul, middlehaul, and fronthaul. The transport slice controller 132 outputs the transport packet information to the network slice controller 135.

[0044] The core slice controller 133 controls which CN device 120 is applied to each network slice. The core slice controller 133 may control each core slice according to instructions from the network slice controller 135. The core slice controller 133 also outputs core packet information indicating how packets transmitted in the CN 40 behaved to the network slice controller 135.

[0045] The network slice controller 135 manages the entire network slice in the cellular network. The network slice controller 135 may instruct the RAN slice controller 131, the transport slice controller 132, and the core slice controller 133 to configure a RAN slice, a transport slice, and a core slice, respectively. The network slice controller 135 may configure a network slice as shown in FIG. 4 in advance and transmit network slice configuration information indicating the configured network slice to the AMF (or SMF) which is the CN device 120. The AMF (or SMF) may select an allowed network slice (specifically, an Allowed NSSAI) for a network slice (specifically, a Requested NSSAI) requested by the UE 100 based on the network slice configuration information.

[0046] 5(A) and 5(B) are diagrams showing examples of network slice configuration information according to the first embodiment. The network slice configuration information is managed by the orchestrator 130. As shown in FIGS. 5(A) and 5(B), each network slice is identified by an S-NSSAI, and the RAN slice, transport slice, and core slice included in the network slice are linked to the S-NSSAI.

[0047] In the example of Figure 5(A), the network slice #1 shown in Figure 4 is represented by S-NSSAI #1, and S-NSSAI #1 is linked to the identification information (ID) of RAN slice #1, the identification information (ID) of transport slice #1, and the identification information (ID) of core slice #1. This allows the network slice controller 135 to manage the RAN slice #1, transport slice #1, and core slice #1 in the network slice #1. The ID of the RAN slice #1 then links the radio resources used in the RAN slice #1 to the identification information of the RU, DU, and CU #1 in the RAN 30. This allows the network slice controller 135 to manage the radio resources and hardware used in the RAN slice #1.

[0048] Furthermore, the network slice controller 135 associates route information #1 with the ID of the transport slice #1 and manages it, thereby enabling management of which backhaul, which middlehaul, and which fronthaul are used for the transport slice #1. Furthermore, the network slice controller 135 associates UPF #1 with the ID of the core slice #1 and manages it, thereby enabling management of which core network function is included in the core slice #1 (UPF #1 in FIG. 5A). Note that, regarding route information, a route table (or routing table) may be prepared that stores route information representing each route from the backhaul to the fronthaul. The network slice controller 135 may acquire route information using the route table.

[0049] In addition, the network slice configuration information and route table shown in Figures 5 (A) and 5 (B) may be stored in memory within the orchestrator 130.

[0050] Returning to FIG. 4, the network slice controller 135 may obtain packet information indicating the behavior of packets transmitted in the cellular network 10 based on the RAN packet information, the transport packet information, and the core packet information.

[0051] (1.1.2) Example of Configuration of Each Device in the Cellular Network Next, an example of the configuration of each device in the cellular network 10 will be described.

[0052] First, a configuration example of the UE 100 will be described.

[0053] 6 is a diagram illustrating an example of the configuration of a UE 100 (user equipment) according to the first embodiment. As illustrated in FIG. 6, the UE 100 includes a receiving unit 101, a transmitting unit 102, and a control unit 103. The receiving unit 101 and the transmitting unit 102 configure a wireless communication unit that performs wireless communication with a node 110.

[0054] The receiving unit 101 performs various reception operations under the control of the control unit 103. The receiving unit 101 includes an antenna and a receiver. The receiver converts a radio signal received by the antenna into a baseband signal (received signal) and outputs the baseband signal to the control unit 103.

[0055] The transmitting unit 102 performs various transmissions under the control of the control unit 103. The transmitting unit 102 includes an antenna and a transmitter. The transmitter converts a baseband signal (transmission signal) output by the control unit 103 into a radio signal and transmits it from the antenna.

[0056] The control unit 103 performs various controls and processes in the UE 100. Such processes include processes of each layer described below. The control unit 103 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in the processes by the processor. The processor may include a baseband processor and a CPU (Central Processing Unit). The baseband processor performs modulation / demodulation, encoding / decoding, etc. of baseband signals. The CPU executes programs stored in the memory to perform various processes. Note that the processes or operations in the UE 100 described below may be performed by the control unit 103. Furthermore, transmission of messages in the UE 100 described below may be performed by the transmitting unit 102, and reception of messages in the UE 100 may be performed by the receiving unit 101.

[0057] Next, an example of the configuration of the node 110 will be described.

[0058] 7 is a diagram showing the configuration of a node 110 (base station) according to the first embodiment. As shown in FIG. 7, the node 110 includes a transmitting unit 111, a receiving unit 112, a control unit 113, and a network communication unit 115. The transmitting unit 111 and the receiving unit 112 constitute a wireless communication unit that performs wireless communication with the UE 100. The network communication unit 115 constitutes a network communication unit that performs communication with the CN device 120.

[0059] The transmitting unit 111 performs various transmissions under the control of the control unit 113. The transmitting unit 111 includes an antenna and a transmitter. The transmitter converts a baseband signal (transmission signal) output by the control unit 113 into a radio signal and transmits it from the antenna.

[0060] The receiving unit 112 performs various types of reception under the control of the control unit 113. The receiving unit 112 includes an antenna and a receiver. The receiver converts a radio signal received by the antenna into a baseband signal (received signal) and outputs the baseband signal to the control unit 113.

[0061] The control unit 113 performs various controls and processes in the node 110. Such processes include processes in each layer described below. The control unit 113 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in the processes by the processor. The processor may include a baseband processor and a CPU. The baseband processor performs modulation / demodulation and encoding / decoding of baseband signals. The CPU executes programs stored in the memory to perform various processes. Note that the processes or operations in the node 110 described below may be performed by the control unit 113. Furthermore, message transmission in the node 110 described below may be performed by the transmitter 111 and the network communication unit 115, and message reception in the node 110 may be performed by the receiver 112 and the network communication unit 115.

[0062] The network communication unit 115 is connected to the CN device 120 via an NG interface, which is an interface between a base station and a core network. The network communication unit 115 is also connected to other nodes via an Xn interface. The network communication unit 115 may also perform various processes under the control of the control unit 113.

[0063] Next, an example of the configuration of the CN device 120 will be described.

[0064] 8 is a diagram illustrating an example of the configuration of the CN device 120 according to the first embodiment. As shown in FIG. 8, the CN device 120 includes a receiving unit 121, a transmitting unit 122, and a control unit 123.

[0065] The receiving unit 121 receives various types of messages under the control of the control unit 123. The receiving unit 121 receives messages transmitted from the node 110 (e.g., messages transmitted via the N2 interface). The receiving unit 121 may also receive messages transmitted from other CN devices (e.g., messages transmitted via the N11 interface). The receiving unit 121 outputs the received messages to the control unit 123.

[0066] The transmitting unit 122 performs various transmissions under the control of the control unit 123. In accordance with instructions from the control unit 123, the transmitting unit 122 transmits messages received from the control unit 123 (e.g., messages via the N2 interface) to the node 110. In addition, in accordance with instructions from the control unit 123, the transmitting unit 122 transmits messages received from the control unit 123 (e.g., messages via the N11 interface) to other CN devices.

[0067] The control unit 123 performs various controls and processes in the CN device 120. The control unit 123 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in the processing by the processor. The processor may include a CPU. The CPU executes programs stored in the memory to perform various processes. Note that the processing or operations in the CN device 120, which will be described later, may be performed by the control unit 123. Furthermore, message transmission in the CN device 120, which will be described later, may be performed by the transmitting unit 122, and message reception in the CN device 120 may be performed by the receiving unit 121.

[0068] Next, an example configuration of the network slice management device (or orchestrator) 130 will be described.

[0069] 9 is a diagram illustrating an example of the configuration of the orchestrator 130. As illustrated in FIG. 9, the orchestrator 130 includes a receiving unit 136, a transmitting unit 137, and a control unit 138.

[0070] The receiving unit 136 performs various types of reception under the control of the control unit 138. For example, the receiving unit 136 receives RAN packet information from the RAN 30, core packet information from the CN 40, and transport packet information from the RAN 30 and the CN 40. The receiving unit 136 outputs the RAN packet information, core packet information, and transport packet information to the control unit 138.

[0071] The transmitting unit 137 performs various transmissions under the control of the control unit 138. For example, the transmitting unit 137 transmits a message or the like to the RAN 30 and / or the CN 40 in accordance with an instruction from the control unit 138.

[0072] The control unit 138 performs various controls and processes in the orchestrator 130. The control unit 138 may be a block that realizes the functions of the RAN slice controller 131, the transport slice controller 132, the core slice controller 133, and the network slice controller 135. The control unit 138 may be divided into four control units 138 to realize the functions of the RAN slice controller 131, the transport slice controller 132, the core slice controller 133, and the network slice controller 135, respectively. The control unit 138 includes at least one processor and at least one memory. The memory stores programs executed by the processor and information used in processing by the processor. The processor may include a CPU. The CPU executes programs stored in the memory to perform various processes. Note that the processing or operations in the orchestrator 130 described below may be performed by the control unit 138.

[0073] (2) Operation Example According to First Embodiment Next, an operation example according to the first embodiment will be described.

[0074] FIG. 10 is a diagram illustrating an example of operation according to the first embodiment.

[0075] In the first embodiment, when an application is executed in UE 100, orchestrator 130 creates a learning model (or AI model; hereinafter, sometimes referred to as a "learning model") based on packet information.

[0076] Generally, when creating a learning model, learning is performed by inputting learning data into the learning model. The state in which learning data is input into the learning model and the learning model is learning is sometimes referred to as a "training mode." Furthermore, a learning model learning in learning mode is sometimes referred to as a "training model" (or a learning AI model). On the other hand, inference result data can be obtained by inputting inference data into the "training model" after the "training model" has undergone a certain amount of learning. The learning model after the "training model" has undergone a certain amount of learning is sometimes referred to as a "trained model" (or a trained AI model). The state in which inference result data is obtained from inference data using a trained model is sometimes referred to as an "inference mode." The inference mode may also be referred to as an execution mode or an operation mode.

[0077] In general, in a learning mode, learning is performed by providing correct answer data (or teacher data) as learning data, which is sometimes referred to as supervised learning. On the other hand, in a learning mode, learning is performed without providing correct answer data as learning data, which is sometimes referred to as unsupervised learning. In the first embodiment, a case where supervised learning is used will be mainly described, but the present invention is not limited to this, and unsupervised learning may also be used.

[0078] 11 is a diagram illustrating an example of generation of a trained model according to the first embodiment. As illustrated in FIG. 11 , the control unit 138 of the orchestrator 130 includes a model training unit 1380 and a model inference unit 1381.

[0079] The model learning unit 1380 generates a model under training from training data, and also generates a trained model by inputting the training data into the model under training. On the other hand, the model inference unit 1381 uses the trained model to output inference result data from inference data.

[0080] 10 , in the learning mode, the orchestrator 130 performs learning on each model under training in each phase from step S2 to step S4, and generates a trained model. Then, in the inference mode, the orchestrator 130 outputs inference result data using the trained model in each phase from step S5 to step S8.

[0081] In the following, first, an example of operation in the learning mode will be described, and then an example of operation in the inference mode will be described.

[0082] (2.1) Example of Operation in Learning Mode In the learning mode, a trained model is generated in each of the following three phases.

[0083] (2.1.1) Application Identification (Step S2)

[0084] (2.1.2) Network slice design (step S3)

[0085] (2.1.3) Communication Control (Step S4) The following will be described in order. Note that the operation example shown in Fig. 10 will be described assuming that it is performed under the following assumptions.

[0086] That is, the UE 100 executes an application and generates a packet. The packet includes an application name. The application name may be included in the trailer portion of the packet or in the payload portion of the packet (e.g., FIG. 15 ). For example, the network slice controller 135 can acquire the application name from a packet transmitted over the cellular network 10 in the RAN 30 or the CN 40. The network slice controller 135 can also acquire the application name from an external network. Note that each slice controller (the RAN slice controller 131, the transport slice controller 132, and / or the core slice controller 133) may acquire the application name from a packet transmitted over the cellular network 10 and output the acquired application name to the network slice controller 135.

[0087] (2.1.1) Application Identification (Step S2) When an application is executed in the UE 100, a packet related to the application is transmitted from the UE 100. The packet is input to the RAN 30 and the CN 40 of the cellular network 10. At this time, the orchestrator 130 collects communication path information of the packet from the node 110 included in the RAN 30 and / or the CN device 120 included in the CN 40 (Step S1).

[0088] First, the communication path information may include packet information indicating the operation of the packet data in the cellular network. The packet information may be packet information of one packet transmitted from the UE 100 itself. Alternatively, the packet information may be packet information of multiple packets transmitted from the UE 100 itself. Alternatively, the packet information may be packet information of not only the UE 100 itself but also other UEs 100. The packet information includes, for example, at least one of the following pieces of information:

[0089] (2.1.1.1) Data length of data included in the payload portion of packet data

[0090] (2.1.1.2) Packet Data Transmission Interval

[0091] (2.1.1.3) The number of UEs 100 (or terminal devices) connected to the cellular network when executing the application, and

[0092] (2.1.1.4) In the data volume learning mode accumulated in the transmission queue of the CN device 120 in the cellular network, the orchestrator 130 can obtain the actual behavior of packets in the cellular network 10 from packet information.

[0093] Second, the communication path information may include network slice configuration information indicating information for configuring a network slice. The network slice configuration information is, for example, configuration information related to a network slice configured in the cellular network 10 when the UE 100 executes an application. The network slice configuration information is, for example, as shown in FIG. 5 . The network slice configuration information is configured in the CN device 120 (e.g., the AMF) when the UE 100 performs a registration procedure with the cellular network 10. The network slice configuration information may be generated by the CN device 120 and transmitted from the CN device 120 to the orchestrator 130. Alternatively, the orchestrator 130 may generate the network slice configuration information and transmit it to the AMF, and the AMF may select any one of the network slice configuration information received from the orchestrator 130 during the registration procedure.

[0094] Third, the communication path information may include communication control information. The communication control information indicates information for controlling communication of a packet in the cellular network 10. In other words, the communication control information indicates information for controlling communication of the packet in the cellular network 10 when the packet is transmitted in the cellular network 10. The communication control information may include, for example, information about a QoS flow established between the UE 100 and a UPF (an example of the CN device 120). The information about the QoS flow may include a QoS flow identifier (QFI: QoS Flow ID) that distinguishes each QoS flow from other QoS flows, a 5QI (5G QoS Identifier) ​​that indicates the characteristics of each QoS flow, and / or a priority (ARP: Allocation and Retention Priority) of each QoS flow. The information about the QoS flow is set in the SMF (an example of the CN device 120) when the UE 100 establishes a PDU session with the UPF. Information about the QoS flow (communication control information) may be generated by the SMF and transmitted from the SMF to the orchestrator 130. The communication control information may also include, for example, information about radio resources allocated by the node 110 to the UE 100. The information about the radio resources may include, for example, time resources, frequency resources, and / or modulation levels. The information about the radio resources is set in the node 110 when the node 110 performs wireless communication with the UE 100. The information about the radio resources (communication control information) may also be generated by the node 110 and transmitted from the node 110 to the orchestrator 130. Alternatively, the communication control information may include information about computing resources in the node 110. The computing resources may be expressed by the number of CPUs, the number of CPU clocks per unit time, and / or the processing speed of the CPU per unit time. Alternatively, the computing resources may be expressed by the number of memories and / or memory capacity. Alternatively, the communication control information may include information about routing. The information about routing is information related to the selection of a communication path for a packet.Alternatively, the information regarding routing may be information indicating which node among the plurality of nodes 110 a packet is to be transmitted through.

[0095] In the application identification (step S2), the orchestrator 130 performs learning using the learning model using packet information from the collected communication path information.

[0096] FIG. 12A is a diagram illustrating an example of learning using a model under training for application identification according to the first embodiment.

[0097] As shown in Fig. 12A, in the learning mode, packet information and application names are input to a learning model for application identification, and the learning model for application identification is trained. The learning data is the packet information and the application name. The application name may be information that represents a specific application, such as "video distribution service," "telemedicine," or "robot control."

[0098] There is a certain relationship between application names and packet information. For example, in the case of a "video distribution service," large volumes of data are transmitted, so the data length of the user data included in the packets may be greater than a certain value. Also, in the case of "telemedicine," video data and the like is transmitted in real time, so the data length may be greater than a certain value and the packet transmission interval may be less than a certain value. By having a training model learn application names and packet information that have such a relationship with each other, it is possible to generate a trained model that outputs application names when packet information is input.

[0099] The control unit 138 (model learning unit 1380) acquires packet information from the cellular network 10, acquires the application name from the packet, and learns a model under training for application identification. After completing the learning, the control unit 138 (model learning unit 1380) generates a trained model for application identification (or a trained AI model for application identification) as the trained trained model for application identification.

[0100] (2.1.2) Network slice design (step S3) Next, the network slice design (step S3) in FIG. 10 will be described.

[0101] In the network slice design (step S3), network slice setting information is used from the communication path information collected by the orchestrator 130. Also, in the network slice design (step S3), learning is performed in a learning mode using a learning model for network slice design.

[0102] FIG. 13(A) is a diagram showing a learning example using a model under training for network slice design according to the first embodiment.

[0103] As shown in Figure 13 (A), in the learning mode, learning is performed on a learning model for network slice design using network slice setting information and application name as learning data.

[0104] There is a certain relationship between the application name and the network slice setting information. For example, in the case of a "video distribution service," since a large amount of data is transmitted, network slice setting information corresponding to a network slice used as high-speed, large-capacity (eMBB) may be used. Furthermore, for example, in the case of "telemedicine," since video data and the like are transmitted in real time, network slice setting information corresponding to a network slice used as ultra-low latency (URLLC) may be used. Furthermore, for example, in the case of "robot control," since control is performed using a large number of cameras, network slice setting information corresponding to a network slice used as multi-connection (mMTC) may be used. By training a model under training using application names and network slice setting information that have a certain relationship with each other in this way, it is possible to generate a trained model that outputs network slice setting information when an application name is input.

[0105] That is, the control unit 138 (model learning unit 1380) acquires the application name from the packet, acquires network slice setting information from the orchestrator 130 or the CN device 120, etc., and trains a training model for network slice design. After completing the training, the control unit 138 (model learning unit 1380) generates a trained model for network slice design (or a trained AI model for network slice design) from the trained training model for network slice design.

[0106] The application name to be input into the training model for network slice design may be the application name obtained from the training model for application identification (or the trained AI model for application identification).

[0107] (2.1.3) Communication Control (Step S4) Next, the communication control (step S4) shown in FIG. 10 will be described.

[0108] In the communication control (step S4), communication control information is used from the communication path information collected by the orchestrator 130. Also, in the communication control (step S4), in the learning mode, a learning model for communication control is trained. In the first embodiment, in a situation where a network slice is set between the RAN 30 and the CN 40 of the cellular network 10, information set in the communication path (e.g., QoS flow) between the UE 100 and the UPF is described as the target of the communication control information.

[0109] FIG. 14A is a diagram illustrating an example of learning using a learning model for communication control according to the first embodiment.

[0110] As shown in FIG. 14A, in the learning mode, the application name and communication control information are used as learning data to train a learning model for communication control.

[0111] There is a certain relationship between the application name and the communication control information. For example, in the case of a "video distribution service," large volumes of data are transmitted, so communication control information that guarantees a certain bit rate or higher may be set. Furthermore, in the case of "telemedicine," video data and the like are transmitted in real time, so communication control information that guarantees a certain upper limit on packet data delay (packet delay budget) and a certain packet error rate or lower may be set. Furthermore, in the case of "robot control," video data is acquired using many cameras, so communication control information that guarantees a certain bit rate or higher and a certain packet error rate or lower may be set. By training the application name and communication control information, which have a certain relationship between them, using a training model for communication control, it is possible to generate a trained model that outputs communication control information when an application name is input.

[0112] That is, the control unit 138 (model learning unit 1380) acquires the application name from the packet data, acquires communication control information from the orchestrator 130 or the CN device 120 (or the orchestrator 130 or the node 110), etc., and trains a training model for communication control. After completing the training, the control unit 138 (model learning unit 1380) generates a trained model for communication control (or a trained AI model for communication control) using the training model for communication control as a trained model.

[0113] The application name to be input to the training model for communication control may be the application name obtained from the training model for application identification (or the trained model for application identification).

[0114] 11 , after a trained model is created in each phase (steps S2 to S4), processing using the trained model is executed in each phase from step S5 to step S7 in inference mode. Next, an example of operation in inference mode (steps S5 to S7) will be described. However, in the following example of operation, it is assumed that the same application as that executed in learning mode is executed in UE 100 in inference mode.

[0115] (2.2) Example of operation in inference mode

[0116] (2.2.1) Application Identification (Step S5) FIG. 12(B) is a diagram illustrating an example of inference using a trained model for application identification according to the first embodiment. As shown in FIG. 12(B), in inference mode, packet information is input to a trained model for application identification to infer an application name. The inference data is packet information, and the inference result data is the application name. The packet information input to the trained model for application identification is operation information of packets transmitted in association with an application executed in the UE 100 in inference mode. The packet information may be acquired by the orchestrator 130 from the node 110 and / or the CN device 120, as in the learning mode. Then, the control unit 138 (or the model inference unit 1381) of the orchestrator 130 infers an application name from the packet information using the trained model for application identification.

[0117] (2.2.2) Network Slice Design (Step S6) FIG. 13(B) is a diagram showing an example of inference using the trained model for network slice design according to the first embodiment. As shown in FIG. 13(B), in inference mode, an application name is input to the trained model for network slice design to infer network slice setting information. The inference data is the application name, and the inference result data is the network slice setting information. The application name inferred from the trained model for application identification (Step S5, FIG. 12(B)) is used as the application name input to the trained model for network slice design. The control unit 138 (or the model inference unit 1381) of the orchestrator 130 infers network slice setting information from the application name using the trained model for network slice design.

[0118] (2.2.3) Communication Control (Step S7) FIG. 14(B) is a diagram showing an example of inference using the trained model for communication control according to the first embodiment. As shown in FIG. 14(B), in inference mode, an application name is input to the trained model for communication control to infer communication control information. The inference data is the application name, and the inference result data is communication control information. The application name inferred from the trained model for application identification (Step S5, FIG. 12(B)) is used as the application name input to the trained model for communication control. In the control unit 138 (or the model inference unit 1381) of the orchestrator 130, the trained model for communication control is used to infer communication control information from the application name.

[0119] 10 , in step S8, the orchestrator 130 uses the network slice setting information (step S6) to set the network slice of the cellular network 10, and uses the communication control information (step S7) to control packet communication (or communication paths) in the cellular network 10. Thereafter, the orchestrator 130 uses packets transmitted via the set communication paths within the cellular network 10 to collect communication path information (step S1), and repeats the above-described steps S5 to S8.

[0120] (2.3) When a new application is executed Next, an example of operation when an application that has never been executed before, that is, an application that has not been generated as a learned model (hereinafter, sometimes referred to as a “new application”), is executed in UE 100 will be described.

[0121] When a new application is executed on UE 100, orchestrator 130 retrains the trained model for application identification, the trained model for network slice design, and the trained model for communication control. Specifically, when an application that has never been executed on UE 100 (i.e., a new application) is executed on UE 100, and the control unit 138 receives a packet (e.g., a second packet) including the application name of the application, orchestrator 130 retrains the trained model for application identification and the trained model for network slice design using at least the application name. Furthermore, when the control unit 138 receives a packet including the application name of the new application, orchestrator 138 retrains the trained model for communication control using at least the application name.

[0122] By retraining the trained model for application identification, the trained model for network slice design, and the trained model for communication control, it is possible to generate a trained model for application identification, a trained model for network slice design, and a trained model for communication control corresponding to a new application. Even when packet data is received by executing a new application, the control unit 138 can acquire the application name, network slice setting information, and communication control information corresponding to the new application by using the retrained trained model for application identification, the trained model for network slice design, and the trained model for communication control. For example, by applying network slice setting information corresponding to the new application to the cellular network 10, it is possible to automatically set a network slice corresponding to the new application in the cellular network 10. Therefore, it is possible to automatically apply an optimal network slice to an application (especially a new application). Furthermore, for example, by applying communication control information corresponding to the new application to the cellular network 10, it is also possible to optimize the transmission path of packets transmitted by the new application. Therefore, it is possible to set optimal communication control within a network (e.g., the cellular network 10).

[0123] The specific details of re-learning are as follows:

[0124] (2.3.1) Retraining the Trained Model for Application Identification When a new application is executed, the control unit 138 acquires the name of the new application from the trailer or payload of a packet (second packet) received over the cellular network 10. Specifically, the control unit 138 acquires the application name from the node 110 or the CN device 120. Alternatively, the control unit 138 may acquire the application name from an external network. For example, the control unit 138 stores previously acquired application names in memory and determines that the acquired application name is the name of a new application if it differs from the stored application name. The control unit 138 (model training unit 1380) then acquires packet information about the packet from the CN device 120 and retrains the trained model for application identification using the packet information about the packet and the name of the new application ( FIG. 12(A) ). Through retraining, the control unit 138 (model training unit 1380) generates a trained model for application identification that also supports the new application. Then, the control unit 138 (model inference unit 1381) uses the learned model for identifying the application to infer the name of the new application from the packet information (FIG. 12(B)).

[0125] (2.3.2) Retraining the Trained Model for Network Slice Design Retraining the trained model for network slice design is also performed based on packets received in the cellular network 10 upon execution of a new application. That is, the control unit 138 (model training unit 1380) retrains the trained model for network slice design using the network slice setting information (second network slice setting information) set in the cellular network 10 upon execution of the new application and the name of the new application included in the packet ( FIG. 13(A) ). The control unit 138 (model training unit 1380) generates a retrained trained model for network slice design.

[0126] Then, the control unit 138 (model inference unit 1381) uses the re-learned trained model for network slice design to obtain network slice setting information for the new application from the application name (Figure 13 (B)).

[0127] The name of the new application to be input into the retrained trained model for network slice design may be the application name inferred from the retrained trained model for application identification.

[0128] (2.3.3) Re-learning of the trained model for communication control Re-learning of the trained model for communication control is also performed based on packets received in the cellular network 10 upon execution of a new application. That is, the control unit 138 (model training unit 1380) re-trains the trained model for communication control using communication control information (second communication control information) set in the cellular network 10 upon execution of the new application and the name of the new application included in the packet ( FIG. 14(A) ). The control unit 138 (model training unit 1380) generates a re-trained trained model for communication control.

[0129] Then, the control unit 138 (model inference unit 1381) uses the re-learned trained model for communication control to acquire communication control information of the new application from the application name (FIG. 14(B)).

[0130] The name of the new application to be input into the re-trained trained model for communication control may be the application name obtained from the re-trained trained model for application identification.

[0131] [Other Embodiments] The above-described operational flows are not limited to being implemented independently, but can be implemented by combining two or more operational flows. For example, some steps of one operational flow may be added to another operational flow, or some steps of one operational flow may be replaced with some steps of another operational flow. In each flow, it is not necessary to execute all steps, and only some steps may be executed.

[0132] In the above-described embodiments and examples, an example in which the base station is an NR base station (gNB) is described, but the base station may also be an LTE base station (eNB) or a 6G base station.

[0133] Also, the term "network node" or "node" primarily refers to a base station, but may also refer to a device in the core network or part of a base station (CU, DU, or RU).

[0134] A program may be provided that causes a computer to execute each process performed by the UE 100, the node 110, the CN device 120, or the network slice management device 130. The program may be recorded on a computer-readable medium. Using a computer-readable medium, the program can be installed on a computer. Here, the computer-readable medium on which the program is recorded may be a non-transitory recording medium. The non-transitory recording medium is not particularly limited, and may be, for example, a recording medium such as a CD-ROM or a DVD-ROM. Furthermore, circuits that execute each process performed by the UE 100, the node 110, the CN device 120, or the network slice management device 130 may be integrated, and at least a portion of the UE 100, the node 110, the CN device 120, or the network slice management device 130 may be configured as a semiconductor integrated circuit (chipset, SoC).

[0135] As used in this disclosure, the terms "based on" and "depending on / in response to" do not mean "based only on" or "depending only on," unless expressly stated otherwise. The term "based on" means both "based only on" and "based at least in part on." Similarly, the term "depending on" means both "depending only on" and "depending at least in part on." The terms "include," "comprise," and variations thereof do not mean including only the listed items, but may mean including only the listed items or may include additional items in addition to the listed items. Additionally, the term "or," as used in this disclosure, is not intended to mean an exclusive or. Furthermore, any reference to elements using designations such as "first," "second," etc., as used in this disclosure does not generally limit the quantity or order of those elements. These designations may be used herein as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed therein or that the first element must precede the second element in some way. In this disclosure, where articles are added by translation, such as a, an, and the in English, these articles shall include the plural unless the context clearly indicates otherwise.

[0136] Additionally, the functions performed by the UE 100 or the base station 200 (network node) may be implemented in circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions. A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may also be a programmed processor that executes a program stored in a memory.

[0137] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0138] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of the hardware and software used to configure the hardware and / or processor.

[0139] Although one embodiment has been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design changes can be made within the scope of the gist. Furthermore, the embodiments, operation examples, and processes can be appropriately combined within the scope of not being inconsistent.

[0140] (Additional Note)

[0141] (Supplementary Note 1) A network slice management device that manages network slices in a cellular network, comprising: a control unit that uses a trained AI model for application identification to input packet information indicating the behavior of packets of an application in the cellular network and outputs an application name; and a control unit that uses a trained AI model for communication control to input the application name and output communication control information indicating information for communication control of the packet in the cellular network, wherein when the control unit receives a packet including the name of a new application, the network slice management device re-trains the trained AI model for application identification and the trained AI model for communication control using at least the name of the new application.

[0142] (Supplementary Note 2) The network slice management device according to Supplementary Note 1, wherein when the control unit receives the packet, it retrains the trained AI model for application identification using the name of the new application and packet information of the packet.

[0143] (Supplementary Note 3) The network slice management device according to Supplementary Note 1, wherein when the control unit receives the packet, it retrains the trained AI model for communication control using the name of the new application and the communication control information set when the new application is executed.

[0144] (Supplementary Note 4) The network slice management device according to Supplementary Note 1, wherein the communication control information includes information about a QoS flow set between a user device and a core network device. (Supplementary Note 5) The network slice management device according to Supplementary Note 1, wherein the packet information represents at least one of the data length of the data included in the payload portion of the packet, the transmission interval of the packet, the number of the user devices connected to the cellular network when executing the application, and the amount of data accumulated in the transmission queue of a core network device in the cellular network.

[0145] (Supplementary Note 6) A management method in a network slice management device that manages network slices in a cellular network, comprising: a step of using a trained AI model for application identification to input packet information indicating the behavior of packets of an application in the cellular network and outputting an application name; a step of using a trained AI model for communication control to input the application name and output communication control information indicating information for communication control of the packet in the cellular network; and a step of, when a packet including the name of a new application is received, re-learning the trained AI model for application identification and the trained AI model for communication control using at least the name of the new application.

[0146] 1: Communication system 10: Cellular network 20: SDN 30: RAN 40: CN 100: UE 102: Transmitter 103: Controller 110: Node 111: Transmitter 113: Controller 120: CN device 121: Receiver 122: Transmitter 123: Controller 130: Network slice management device 131: RAN slice controller 132: Transport slice controller 133: Core slice controller 136: Receiver 137: Transmitter 138: Controller 1380: Model learning unit 1381: Model inference unit

Claims

1. A network slice management device that manages network slices in a cellular network, comprising: a control unit that uses a trained AI model for application identification to input packet information indicating the behavior of packets of an application in the cellular network and outputs an application name; and a control unit that uses a trained AI model for communication control to input the application name and output communication control information indicating information for communication control of the packet in the cellular network, wherein when the control unit receives a packet including the name of a new application, the control unit re-trains the trained AI model for application identification and the trained AI model for communication control using at least the name of the new application.

2. A network slice management device as described in claim 1, wherein when the control unit receives the packet, it retrains the trained AI model for application identification using the name of the new application and the packet information of the packet.

3. A network slice management device as described in claim 1, wherein when the control unit receives the packet, it retrains the trained AI model for communication control using the name of the new application and the communication control information set when the new application is executed.

4. A network slice management device as described in claim 1, wherein the communication control information includes information regarding a QoS flow established between a user device and a core network device.

5. A network slice management device as described in claim 1, wherein the packet information represents at least one of the data length of the data contained in the payload portion of the packet, the transmission interval of the packet, the number of user devices connected to the cellular network when executing the application, and the amount of data accumulated in the transmission queue of a core network device in the cellular network.

6. A management method in a network slice management device that manages network slices in a cellular network, comprising: a step of using a trained AI model for application identification to input packet information indicating the behavior of packets of an application in the cellular network and outputting an application name; a step of using a trained AI model for communication control to input the application name and output communication control information indicating information for communication control of the packet in the cellular network; and a step of, when a packet including the name of a new application is received, re-training the trained AI model for application identification and the trained AI model for communication control using at least the name of the new application.

Images