JP7367257B1 - Communication management device and communication management method - Google Patents

Abstract

An object of the present invention is to allocate radio resources according to requests regarding communication quality of network slices using a simpler configuration. A communication management device 1 acquires information regarding a plurality of network slices, and the information regarding the plurality of network slices includes request information regarding communication quality associated with each of the plurality of network slices. An acquisition unit 10, a second acquisition unit 11 that acquires the initial value of the number of resource blocks, and the initial value is given as an unknown input to the learned first machine learning model, and the learned first machine learning model performs calculations. a calculation unit 12 that outputs the number of optimized resource blocks to be allocated to each of the plurality of network slices associated with the request information; and a notification unit 15 that notifies the base station 30 of the number. [Selection diagram] Figure 1

Description

The present invention relates to a communication management device and a communication management method.

Network slicing has been known as a technology that has attracted attention in SA (Stand Alone) 5G (fifth generation mobile communication system). In network slicing, one physical network from a radio access network (RAN) such as a base station to a core network is virtually divided to provide networks according to usage purposes.

Each network slice that is virtually divided by network slicing exists on the same physical network, and wireless resources are allocated based on the required bandwidth and communication delay requirements according to service requests and the like. As shown in FIG. 16, radio resources are composed of a plurality of resource blocks (RBs) defined on a two-dimensional plane consisting of a time axis and a frequency axis. A resource block is a block in which a plane consisting of a time axis and a frequency axis is divided into a matrix of minimum units of frequency x time. Further, the total number of resource blocks is given by frequency (number of subcarriers) x time (number of time slots) x space (number of spatial streams).

In the conventional network slicing technique, as shown in FIG. 16, a plurality of divided network slices #1 to #N share the entire radio resource. That is, each of the plurality of network slices has a configuration that allows it to access and use all of the resource blocks on each plane. As described above, in the conventional technology, each network slice does not have a dedicated resource block. Therefore, for example, even though a high wireless communication speed is not required in a certain network slice, a resource block with a communication speed higher than the requested communication speed may be allocated.

Therefore, for example, Patent Document 1 discloses a communication management system that determines the amount of resources to be allocated to communication for each communication service in a base station. In the communication management system according to Patent Document 1, information regarding the communication quality of one or more communication terminals communicating with the base station is obtained from a base station, which is aggregated for each value indicating a communication service. Furthermore, based on the acquired information, the base station determines the amount of resources to be allocated to communication for each communication service, and notifies the base station of the determined amount of resources.

However, since Patent Document 1 does not have a configuration in which a dedicated resource block is provided for each network slice, processing for appropriate radio resource allocation based on aggregation of communication quality, requested throughput, etc. becomes complicated.

JP2022-117825A

As described above, with the conventional technology, it has not been possible to allocate radio resources in accordance with the requirements regarding the communication quality of network slices using a simpler configuration.

The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to allocate radio resources according to requests regarding communication quality of network slices with a simpler configuration.

In order to solve the above problems, a communication management device according to the present invention is configured to acquire information regarding a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated, and information regarding the plurality of network slices. a first acquisition unit, the information including request information regarding communication quality associated with each of the plurality of network slices; and an initial number of the resource blocks of the radio resource to be allocated to each of the plurality of network slices. a second acquisition unit configured to acquire a value; and a second acquisition unit configured to provide the initial value acquired by the second acquisition unit as an unknown input to the trained first machine learning model; a first calculation unit configured to perform calculations on a learning model and output an optimized number of resource blocks to be allocated to each of the plurality of network slices associated with the request information; and a notification unit configured to notify a base station of the optimized number of resource blocks allocated to each network slice.

Further, in the communication management device according to the present invention, when the initial value of the number of resource blocks to be allocated to each of the plurality of network slices is given as an input, the plurality of networks with which the request information is associated a first storage unit configured to store a first machine learning model that outputs a predicted value of the number of resource blocks to be allocated to each of the slices; The first machine learning model is configured to learn parameters such that the sum of the predicted values of the number of resource blocks output from the learning model becomes the total number of the resource blocks constituting the radio resource. a first learning unit, the first storage unit further stores the learned first machine learning model constructed by the first learning unit, and the first calculation unit further stores the first machine learning model constructed by the first learning unit; The learned first machine learning model may be read from the storage unit and the learned first machine learning model may be calculated.

In order to solve the above problems, a communication management device according to the present invention is configured to acquire information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated, and the information regarding the network slice includes: a third acquisition unit that includes request information regarding communication quality associated with the network slice and a set value of the number of resource blocks allocated to the network slice; and the request information associated with the network slice; Information about the resource blocks to be assigned to the network slice by giving the set value of the number of resource blocks to the learned second machine learning model as an unknown input, and performing calculations on the learned second machine learning model. and a notification unit configured to notify a base station of information regarding the resource block for each resource element to be allocated to the network slice. , the resource elements include a frequency element, a time element, and a spatial element.

Furthermore, in the communication management device according to the present invention, element information including a set number of elements for any one or two of the frequency element, the time element, and the spatial element is provided. a fourth acquisition unit configured to acquire the request information associated with the network slice, the setting value of the number of resource blocks, and the third machine learning that has been trained using the element information as unknown inputs; information about the resource block to be assigned to the network slice by performing calculations on the learned third machine learning model, from among the frequency element, the time element, and the spatial element. and a third calculation unit configured to output for each resource element other than the resource element related to the element information acquired by the notification unit, the notification unit is configured to output information for each resource element related to the element information to be allocated to the network slice. Information regarding the resource block for each resource element other than the resource element may be notified to the base station.

Furthermore, in the communication management device according to the present invention, the request information associated with the network slice, the set value of the number of resource blocks, and the resource blocks for each of the resource elements allocated to the network slice are further provided. a second learning unit configured to learn a relationship with information using a second machine learning model; and storing the learned second machine learning model constructed by the second learning unit. and a second storage unit configured to read out the learned second machine learning model from the second storage unit, and perform calculations on the learned second machine learning model. It's okay.

Further, in the communication management device according to the present invention, the request information associated with the network slice, the setting value of the number of resource blocks, the element information, the frequency element, the time element, and the It is configured to learn, using a third machine learning model, a relationship between each of the resource elements other than the resource element related to the element information among the spatial elements and information regarding the resource block to be allocated to the network slice. and a third storage section configured to store the trained third machine learning model constructed by the third learning section, and the third calculation section The learned third machine learning model may be read out from the third storage unit and the learned third machine learning model may be calculated.

In order to solve the above-mentioned problems, a communication management method according to the present invention includes a first acquisition step of acquiring information regarding a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated. a first acquisition step in which the information regarding network slices includes request information regarding communication quality associated with each of the plurality of network slices; and the resource of the radio resource to be allocated to each of the plurality of network slices. a second acquisition step of acquiring an initial value of the number of blocks, and supplying the initial value acquired in the second acquisition step to the trained first machine learning model as an unknown input; a first calculation step of calculating the learning model and outputting an optimized number of resource blocks to be allocated to each of the plurality of network slices associated with the request information; and a notification step configured to notify a base station of the optimized number of resource blocks to be allocated.

Further, in the communication management method according to the present invention, when the initial value of the number of resource blocks to be allocated to each of the plurality of network slices is given as an input, the plurality of networks with which the request information is associated a first storing step of storing a first machine learning model that outputs a predicted value of the number of resource blocks to be allocated to each slice in a first storage unit; a first learning step of learning parameters of the first machine learning model such that the sum of the predicted values of the number of resource blocks output from the learning model becomes the total number of the resource blocks constituting the radio resource; The first storage step further stores the trained first machine learning model constructed in the first learning step in the first storage unit, and the first calculation step further comprises: The learned first machine learning model may be read out from the first storage unit and the learned first machine learning model may be calculated.

Further, in the communication management method according to the present invention, a third acquisition step of acquiring information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated, the information regarding the network slice including the information regarding the network slice. a third acquisition step including request information regarding the associated communication quality and a set value of the number of the resource blocks allocated to the network slice; The set value of the number is given to the trained second machine learning model as an unknown input, and the trained second machine learning model is operated to obtain information regarding resource blocks to be allocated to the network slice for each resource element. a second calculation step of outputting, and a notification step of notifying a base station of information regarding the resource block for each of the resource elements allocated to the network slice, the resource elements include a frequency element, a time element, and a spatial element. including.

Furthermore, in the communication management method according to the present invention, element information including a set number of elements for any one or two of the frequency element, the time element, and the spatial element is provided. a fourth acquisition step of acquiring the request information associated with the network slice, the setting value of the number of resource blocks, and the element information as unknown inputs to a trained third machine learning model; The learned third machine learning model is computed to obtain information regarding the resource block to be allocated to the network slice from among the frequency element, the time element, and the spatial element acquired in the fourth acquisition step. a third calculation step configured to output for each resource element other than the resource element related to the element information, and the notification step is configured to output the resource element other than the resource element related to the element information to be allocated to the network slice. Information regarding the resource blocks for each resource element may be notified to the base station.

Further, in the communication management method according to the present invention, the request information associated with the network slice, the set value of the number of resource blocks, and the resource blocks for each of the resource elements allocated to the network slice are further provided. a second learning step of learning a relationship with information using a second machine learning model; and a second learning step of storing the learned second machine learning model constructed in the second learning step in a second storage unit. The second calculation step may include a storage step, and the second calculation step may read the learned second machine learning model from the second storage unit and perform calculations on the learned second machine learning model.

Further, in the communication management method according to the present invention, the request information associated with the network slice, the setting value of the number of resource blocks, and the element information, the frequency element, the time element, and the A third learning step of learning, using a third machine learning model, a relationship between each of the resource elements other than the resource element related to the element information among spatial elements and information regarding the resource block to be allocated to the network slice. and a third storage step of storing the learned third machine learning model constructed in the third learning step in a third storage section, and the third calculation step includes the step of storing the learned third machine learning model constructed in the third learning step in the third storage section. The learned third machine learning model may be read and the learned third machine learning model may be calculated.

According to the present invention, an initial value of the number of resource blocks of radio resources to be allocated to each of a plurality of network slices is given as an unknown input to a trained first machine learning model, and the trained first machine learning model is operated. and outputs the optimized number of resource blocks to be allocated to each of the plurality of network slices associated with the request information. Therefore, with a simpler configuration, it is possible to allocate radio resources in accordance with requests regarding communication quality of network slices.

FIG. 1 is a block diagram showing the configuration of a communication management system including a communication management device according to a first embodiment of the present invention. FIG. 2 is a diagram for explaining an overview of the communication management device according to the first embodiment. FIG. 3 is a block diagram showing the hardware configuration of the communication management device according to the first embodiment. FIG. 4 is a diagram for explaining learning processing by the learning section according to the first embodiment. FIG. 5 is a flowchart showing the operation of the communication management device according to the first embodiment. FIG. 6 is a flowchart showing the operation of the communication management device according to the first embodiment. FIG. 7 is a block diagram showing the configuration of a communication management system including a communication management device according to the second embodiment. FIG. 8 is a diagram for explaining learning processing by the learning section according to the second embodiment. FIG. 9 is a diagram for explaining an overview of the communication management device according to the second embodiment. FIG. 10 is a flowchart showing the operation of the communication management device according to the second embodiment. FIG. 11 is a flowchart showing the operation of the communication management device according to the second embodiment. FIG. 12 is a block diagram showing the configuration of a communication management system including a communication management device according to the third embodiment. FIG. 13 is a diagram for explaining learning processing by the learning section according to the third embodiment. FIG. 14 is a flowchart showing the operation of the communication management device according to the third embodiment. FIG. 15 is a flowchart showing the operation of the communication management device according to the third embodiment. FIG. 16 is a diagram for explaining network slicing according to a conventional example.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to FIGS. 1 to 15.

[First embodiment]
FIG. 1 is a block diagram showing the configuration of a communication management system including a communication management device 1 according to the first embodiment of the present invention. The communication management device 1 according to the present embodiment calculates the optimized number of resource blocks to be allocated to each of the network slices associated with request information regarding communication quality, and notifies the base station 30 of the resource block allocation information. .

[Communication management system configuration]
First, an overview of a communication management system including a communication management device 1 according to an embodiment of the present invention will be explained. As shown in FIG. 1, the communication management system includes, for example, a communication management device 1 compatible with an SA type 5G wireless communication system, a communication terminal 2, a radio access network 3 including a base station 30, and a core network 4.

The communication management device 1 and the core network 4 are connected via a network NW such as a LAN or WAN. Furthermore, the base station 30 included in the radio access network 3 and the core network 4 are connected via a network L such as a backhaul link. Further, in FIG. 1, as an example, the communication management device 1 and the base station 30 are connected via the network NW, but the communication management device 1 and the base station 30 are connected via a dedicated communication network. It may also be a configuration. Alternatively, the connection between the communication management device 1 and the base station 30 via the network NW can be omitted.

The communication terminal 2 is realized by a mobile communication terminal such as a smartphone equipped with a SIM, a PDA (Personal Digital Assistant), a tablet computer, or a laptop computer (so-called notebook computer). Furthermore, the communication terminal 2 is also realized by an IoT device that has a unique IP address and can be connected to the Internet.

The communication terminal 2 communicates with the core network 4 via the radio access network 3 and requests a service from the core network 4. Specifically, the communication terminal 2 sets an identifier indicating service characteristics according to the service to be used, and transmits it to the core network 4 via the radio access network 3. The communication terminal 2 can communicate with the core network 4 by using resource blocks in the network slice according to the requested service, connect to an appropriate data network, and perform data communication related to the service.

A different network slice is set for each service requested by the communication terminal 2 depending on its purpose. For example, a network slice that supports high speed and large capacity, a network slice that supports multiple simultaneous connections, a network slice that supports high reliability and low delay, etc. are provided. That is, each network slice requires a certain communication quality such as an appropriate communication speed and communication delay depending on the application. In this embodiment, it is assumed that a preset number of network slices are set depending on the service type and characteristics.

Additionally, each network slice is assigned a unique network slice ID, and each network slice ID is associated with requirements regarding communication quality such as communication delay, bandwidth, and reliability, as well as information regarding security policies and authentication requirements. ing.

The radio access network 3 includes a plurality of base stations 30 and enables communication between the communication terminals 2 located in each base station 30 and the core network 4 .

The base station 30 is configured of a wireless base station compatible with the 5G system, and relays communication between the communication terminal 2 located within the area and the core network 4. The base station 30 may be realized by, for example, a computer connected via a bus, including a processor, a main storage device, a communication interface, an auxiliary storage device, and an input/output I/O, and a program that controls these hardware resources. I can do it.

The base station 30 receives from the communication terminal 2 an identifier indicating the service type and characteristics of the service used by the communication terminal 2, and transmits it to the core network 4. Furthermore, the base station 30 according to the present embodiment receives notification of allocation information of resource blocks to be allocated to each network slice from the communication management device 1. Furthermore, the base station 30 according to the present embodiment allocates resource blocks to network slices based on the allocation information notified from the communication management device 1. Furthermore, the base station 30 allocates resource blocks within the network slice according to the service to the communication terminal 2 based on the service type and the identifier indicating the characteristics.

The core network 4 is configured of, for example, an SA type 5G core network. The core network 4 shown in FIG. 1 includes an NSSF (Network Slice Selection Function) 40 and an AMF (Access and Mobility Management Function) 41, and other devices and functions are omitted from illustration.

The NSSF 40 is a network function that selects a network slice to be used by the communication terminal 2 when the communication terminal 2 connects to the core network 4 via the base station 30. The NSSF 40 selects available network slice candidates based on the type and characteristics of the service requested by the communication terminal 2, contract information, location information, terminal information, etc., and notifies the AMF 41 of the candidates.

The AMF 41 is a network function that performs registration and updating of location information and connection status of the communication terminal 2, and movement management. Further, the AMF 41 selects and manages a network slice to be used by the communication terminal 2 based on the available network slice candidates notified from the NSSF 40. Information on the network slice selected by the AMF 41 is notified to the base station 30.

[Functional block of communication management device]
As shown in FIG. 1, the communication management device 1 includes a first acquisition unit 10, a second acquisition unit 11, a learning unit (first learning unit) 12, a calculation unit (first calculation unit) 13, a storage unit (first storage unit) 14, and a notification unit 15.

The first acquisition unit 10 acquires information regarding a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated. The information regarding the plurality of network slices acquired by the first acquisition unit 10 includes request information regarding communication quality associated with each network slice. The request information includes requirements regarding communication quality including communication delay, bandwidth, reliability, and the like.

The second acquisition unit 11 acquires an initial value of the number of resource blocks of radio resources to be allocated to each of a plurality of network slices. The second acquisition unit 11 can acquire the initial value of the number of resource blocks from an external server or the like via the network NW, for example. Alternatively, the second acquisition unit 11 can acquire the initial value of the number of resource blocks by accepting an input operation performed by the user. The initial value of the number of resource blocks of radio resources to be allocated to each of a plurality of network slices is, for example, the approximate number of resource blocks that a user who manages and operates a wireless communication network desires to allocate to each network slice. can do. Further, the initial value of the number of resource blocks can be set, for example, based on the total number of resource blocks set in the radio resource.

The learning unit 12 determines that when an initial value of the number of resource blocks to be allocated to a plurality of network slices is given as an input, the sum of predicted values of the number of resource blocks output from the first machine learning model constitutes a radio resource. Parameters of the first machine learning model are learned so that the total number of resource blocks is obtained. Further, the learning unit 12 reads out the first machine learning model before learning, which is stored in the first model storage unit 140 of the storage unit 14, and performs a learning process. As described above, the total number of resource blocks is a value given by frequency (number of subcarriers) x time (number of time slots) x space (number of spatial streams).

The first machine learning model trained by the learning unit 12 allocates resource blocks to each of a plurality of network slices associated with request information when an initial value of the number of resource blocks to be allocated to each of a plurality of network slices is given as an input. Output the predicted value of the number of resource blocks. The learning unit 12 can use the initial value of the number of resource blocks acquired by the second acquisition unit 11 as learning data.

FIG. 4 shows a neural network structure adopted as an example of the first machine learning model for learning by the learning unit 12. The neural network comprises an input layer x, a hidden layer h, and an output layer y. In the example of FIG. 4, the number of input nodes and output nodes is the same, and these correspond to N network slices that are preset according to the number of services and the like.

The learning unit 12 provides the number of resource blocks to be allocated to each network slice to the input layer of the neural network, applies an activation function to the weighted sum of inputs, and passes an output determined by threshold processing to the output layer. As shown in FIG. 4, initial values of the number of resource blocks (RB) for network slices #1 to #N are input to each input node x ₁ to x _N of the input layer. Each output node y ₁ to y _N of the output layer is a predicted value of the neural network for the input value.

The size and elements of the machine learning model, including the number of layers in the hidden layer h and the sparseness and density of the connections between the nodes of the neural network, are not limited as long as the design provides sufficient inference accuracy. It may be a fully connected or sparse structure.

The learning unit 12 further introduces an objective function that evaluates the error between the total number of resource blocks of radio resources and the predicted value y of the neural network, thereby calculating the total value of the predicted value of the number of resource blocks of each network slice. The parameters of the neural network are learned so that the total number of resource blocks constituting the wireless resource is the same.

The learning unit 12 adjusts the weight parameters of the neural network so that the objective function E expressed by the following equation (1) becomes minimum, that is, zero.
E={(y ₁ +y ₂ +y ₃ +...+y _N-1 +y _N )-total number of RB} ² ...(1)
Note that the mean square error, cross entropy, cross covariance, structural similarity, clustering loss, etc. can be employed as the objective function. The learning unit 12 can optimize the objective function using a gradient method using an error backpropagation method, a stochastic gradient descent method, or the like.

Returning to FIG. 1, the calculation unit 13 supplies the initial value of the number of resource blocks to be allocated to each network slice, acquired by the second acquisition unit 11, as an unknown input to the trained first machine learning model. The first machine learning model is calculated to output an optimized number of resource blocks to be allocated to each of a plurality of network slices associated with request information regarding communication quality.

As described above, the initial value of the number of resource blocks allocated to each network slice acquired by the second acquisition unit 11 is determined based on the relationship between the total number of resource blocks and the communication quality that each network slice should satisfy. It may not be the number of resource blocks.

For example, if the total number of resource blocks is 300, the initial value of the number of resource blocks for network slice #1 shown in FIG. 4 is 30, and the initial value of the number of resource blocks for network slice #2 is 40. do. By giving such an initial value of the number of resource blocks as input to the trained first machine learning model and performing calculations on the trained first machine learning model, the total number of resource blocks and each network slice can be satisfied. The optimal number of resource blocks for each network slice from the viewpoint of the desired communication quality is output. For example, the optimized number of resource blocks for network slice #1 is output as 32, the initial value of the optimized number of resource blocks for network slice #2 is output as 38, and so on.

FIG. 2 is a schematic diagram showing resource blocks allocated to each network slice. FIG. 2 shows a two-dimensional plane in which radio resources are defined by a time axis and a frequency axis, and a spatial axis. As shown in FIG. 2, radio resources are divided into resource blocks composed of time slots and subcarriers. In this embodiment, as shown in each hatched area from "NW slice #1" to "NW slice #N" in FIG. 2, a dedicated resource block is allocated in units of network slices. Furthermore, an optimized number of resource blocks are allocated to each network slice through the learning process by the learning unit 12 and the calculation of the learned first machine learning model by the calculation unit 13.

More specifically, for example, when 2×2 MIMO consisting of two transmitting antennas and two receiving antennas is set, spatial streams s ₁ , s ₂ are generated along the spatial axis in FIG. is provided. Therefore, the number of optimized resource blocks in spatial stream s ₂ in addition to spatial stream s ₁ is determined.

Furthermore, in the present embodiment, when allocating the optimized number of resource blocks determined by the calculation unit 13 to network slices, the range of allocatable resource blocks can be set in advance for each network slice. For example, as shown in FIG. 2, a prior setting is provided such that network slices from "NW slice #1" to "NW slice #N" can each use time slots t ₁ to t _k . be able to. Furthermore, a setting can be made in which resource blocks are allocated in order from "NW slice #1" to "NW slice #N" and from subcarriers f ₁ to f _k .

In this manner, in this embodiment, it is possible to set in advance the range of resource blocks to be allocated to each network slice. Therefore, based on the optimal number of resource blocks output by the calculation unit 13, for example, the optimized number of resource blocks is selected in the order of "NW slice #1" to "NW slice #N" in the predetermined frequency range. Resource blocks can be allocated.

Returning to FIG. 1, the storage unit 14 includes a first model storage unit 140. The storage unit 14 stores information regarding the network slice acquired by the first acquisition unit 10. The storage unit 14 also stores request information regarding communication quality associated with the network slice ID of the network slice.

The first model storage unit 140 stores a first machine learning model before learning and a trained first machine learning model.

The notification unit 15 notifies the base station 30 of the optimized number of resource blocks to be allocated to each of the plurality of network slices. Specifically, the notification unit 15 can associate the optimized number of resource blocks output by the calculation unit 13 with the network slice ID and notify the base station 30 via the network NW.

[Hardware configuration of communication management device]
Next, an example of a hardware configuration that implements the communication management device 1 having the above-described functions will be described using FIG. 3.

As shown in FIG. 3, the communication management device 1 includes, for example, a computer connected via a bus 101, including a processor 102, a main storage device 103, a communication interface 104, an auxiliary storage device 105, and an input/output I/O 106; This can be realized by a program that controls these hardware resources.

The main storage device 103 stores in advance programs for the processor 102 to perform various controls and calculations. The processor 102 and the main storage device 103 realize each function of the communication management device 1, such as the first acquisition section 10, the second acquisition section 11, the learning section 12, and the calculation section 13 shown in FIG.

The communication interface 104 is an interface circuit for establishing a network connection between the communication management device 1 and various external electronic devices.

The auxiliary storage device 105 includes a readable and writable storage medium and a drive device for reading and writing various information such as programs and data to and from the storage medium. For the auxiliary storage device 105, a semiconductor memory such as a hard disk or a flash memory can be used as a storage medium.

The auxiliary storage device 105 has a program storage area that stores a communication management program executed by the communication management device 1 . Further, the auxiliary storage device 105 has an area for storing a learning program for learning the first machine learning model. The auxiliary storage device 105 implements the storage unit 14 described in FIG. Furthermore, for example, it may have a backup area for backing up the above-mentioned data, programs, and the like.

The input/output I/O 106 is an input/output device that inputs signals from external devices and outputs signals to external devices.

[Operation of communication management device]
Next, the operation of the communication management device 1 having the above-described configuration will be explained with reference to the flowcharts of FIGS. 5 and 6.

First, with reference to FIG. 5, learning processing by the communication management device 1 will be described. First, the first acquisition unit 10 acquires information regarding a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated (step S1). The information regarding the plurality of network slices acquired in step S1 includes request information regarding communication quality associated with each network slice.

Next, the second acquisition unit 11 acquires an initial value of the number of resource blocks to be allocated to each of the plurality of network slices (step S2). Subsequently, the learning unit 12 determines whether the sum of the predicted values of the number of resource blocks output from the first machine learning model is the resource blocks constituting the radio resource when the initial value acquired in step S2 is given as an input. The parameters of the first machine learning model are learned so that the total number becomes the same (step S3). Specifically, the learning unit 12 reads the first machine learning model stored in the first model storage unit 140 of the storage unit 14 and performs the learning process.

The learning unit 12 employs, for example, the model with the neural network structure shown in FIG. Adjust network weight parameters. The learning unit 12 can optimize the objective function using a gradient method using an error backpropagation method, a stochastic gradient descent method, or the like. The trained first machine learning model constructed by the learning unit 12 is stored in the first model storage unit 140 (step S4).

Next, calculation processing by the communication management device 1 will be explained with reference to FIG. First, the second acquisition unit 11 acquires an initial value of the number of resource blocks to be allocated to each network slice (step S10). Next, the calculation unit 13 loads the learned first machine learning model from the first model storage unit 140 (step S11).

Next, the calculation unit 13 supplies the initial value of the number of resource blocks to be allocated to each network slice, which was acquired by the second acquisition unit 11 in step S10, as an unknown input to the trained first machine learning model. The first machine learning model is calculated to output the optimized number of resource blocks to be allocated to each of the plurality of network slices associated with the communication quality request information (step S12).

Thereafter, the notification unit 15 notifies the base station 30 of the optimized number of resource blocks allocated to each network slice, which was output in step S12 (step S13).

The base station 30 that has received the notification sets and registers the optimized number of resource blocks in association with the network slice ID of each network slice. Based on the number of resource blocks associated with the network slice ID, the base station 30 allocates resource blocks within the network slice corresponding to the communication quality request to the communication terminal 2 that uses a predetermined service. Furthermore, the notification unit 15 can allocate resource blocks for each network slice even for each base station 30.

As explained above, according to the communication management device 1 according to the first embodiment, each of the plurality of network slices associated with request information regarding communication quality is Since the optimized number of resource blocks to be allocated is output, it is possible to allocate radio resources according to requests regarding communication quality of network slices with a simpler configuration.

Furthermore, according to the communication management device 1 according to the first embodiment, the optimized number of resource blocks to be allocated to a plurality of network slices associated with communication quality request information is output, so that the number of resource blocks is optimized for each network slice. Dedicated resource blocks can be allocated according to QoS (Quality of Service) requests. Furthermore, it is possible to allocate resource blocks for each network slice in accordance with QoS requests for each base station 30.

Further, according to the communication management device 1 according to the first embodiment, the first machine learning model is trained, and the total number of resource blocks and the number of requests for each network slice are calculated based on the initial value of the number of resource blocks. The number of resource blocks optimized for the communication quality is learned. Therefore, it is possible to reduce the database requirements without requiring a huge database capacity by mapping the resource parameters of radio resources such as the number of subcarriers, the number of time slots, and the number of spatial streams using the network slice ID as a key. Become.

[Second embodiment]
Next, a second embodiment of the present invention will be described. In the following description, the same components as those in the first embodiment described above will be denoted by the same reference numerals, and the description thereof will be omitted.

In the first embodiment, a case has been described in which the optimized number of resource blocks for the total number of resource blocks is calculated for the initial value of the number of resource blocks to be allocated to each of a plurality of network slices. On the other hand, in the second embodiment, when the number of resource blocks to be allocated to each network slice is allocated as a fixed value, the network slice for each resource element including the frequency element, time element, and space element of radio resources is Obtain resource block allocation information for.

[Functional block of communication management device]
FIG. 7 is a block diagram showing the configuration of a communication management device 1A according to the second embodiment. The communication management device 1A includes a learning section (second learning section) 12A, a calculation section (second calculation section) 13A, a storage section (second storage section) 14A, a notification section 15, and a third acquisition section 16.

First, the third acquisition unit 16 acquires information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated. The information regarding the network slice acquired by the third acquisition unit 16 includes request information regarding communication quality associated with the network slice. Further, the information regarding the network slice acquired by the third acquisition unit 16 includes a set value of the number of resource blocks allocated to the network slice. The setting value for the number of resource blocks is the number of resource blocks to be allocated to a network slice, which is set as a fixed value in advance.

The learning unit 13A uses a second machine learning model to determine the relationship between the request information and the set value of the number of resource blocks associated with the network slice, and the information regarding the resource block for each resource element of the radio resource allocated to the network slice. Learn by using

The resource elements include subcarriers (frequency elements) on the frequency axis, time slots (time elements) on the time axis, and spatial streams (spatial elements) on the spatial axis that define radio resources. To explain using the example of FIG. 2, resource elements include subcarriers f ₁ to f _k on the frequency axis and time slots t ₁ to t _k on the time axis in a two-dimensional plane of radio resources. Furthermore, spatial streams s ₁ and s ₂ on the spatial axis shown in FIG. 2 are included in the resource element.

In this way, the number and value of each resource element corresponds to the number and value of subcarriers f ₁ to f _k in the frequency axis, time slots t ₁ to t _k in the time axis, and spatial streams s ₁ , s ₂ in the spatial axis. It can be expressed as

FIG. 8 shows a neural network structure adopted as an example of the second machine learning model for learning by the learning unit 12A. The neural network comprises an input layer x, a hidden layer h, and an output layer y. In the example of FIG. 8, request information regarding the communication quality associated with "NW slice ID" and "NW slice ID" which are the network slice IDs of the network slices are input to the input node. The output nodes y ₁ to y _m correspond to the values of "frequency element f,""time element t," and "spatial element s."

For example, as an output node of "frequency element f", an output node corresponding to subcarriers f ₁ to f _k on the frequency axis, and as an output node of "time element t", an output node corresponding to time slots t ₁ to t _k , Furthermore, output nodes corresponding to the spatial streams s ₁ and s ₂ can be provided as output nodes of the "spatial element s". Note that the number of each resource element of the output node can be determined in advance.

The learning unit 12A provides the network slice ID of the network slice, the request information regarding the communication quality associated with the network slice ID, and the setting value of the number of resource blocks to the input layer of the neural network, and applies an activation function to the weighted sum of the inputs. and pass the output determined by threshold processing to the output layer.

The learning unit 12A further introduces an objective function that evaluates the error between the correct label value and the output value y of the neural network, thereby improving the network slice ID of the network slice and the communication associated with the network slice ID. The parameters of the neural network are learned so that the set values for the quality-related request information and the number of resource blocks become the values of the frequency element f, the time element t, and the spatial element s of the correct label.

The learning unit 12A calculates the value of the frequency element f, the value of the time element t, and the spatial element of the correct label for the input values of the request information regarding communication quality and the setting value of the number of resource blocks associated with the network slice ID. A neural network is trained using the training data given the value of s.

The learning unit 12A adjusts the weight parameters of the neural network so that the objective function becomes minimum, that is, zero. The learning unit 12A can optimize the objective function using a gradient method using an error backpropagation method, a stochastic gradient descent method, or the like. The trained second machine learning model constructed by the learning unit 12A is stored in the second model storage unit 141, which will be described later.

Returning to FIG. 7, the calculation unit 13A supplies the request information regarding the communication quality associated with the network slice and the set value of the number of resource blocks as unknown inputs to the trained second machine learning model, and Computes the learning model and outputs information regarding resource blocks to be allocated to network slices for each resource element. The calculation unit 13A reads the learned second machine learning model from the second model storage unit 141 and performs calculation. Information regarding resource blocks for each resource element output by the calculation unit 13A is passed to the notification unit 15.

As shown in FIG. 9, in this embodiment, in a two-dimensional plane of wireless resources consisting of a time axis and a frequency axis, resource blocks with non-consecutive frequencies and time slots can be allocated to each network slice. can. In FIG. 9, resource blocks in each hatched area indicate resource blocks allocated to each network slice. Note that in FIG. 9, illustration of the spatial streams s ₁ and s ₂ on the spatial axis is omitted.

The storage unit 14A includes a second model storage unit 141. The storage unit 14A stores information regarding the resource blocks acquired by the third acquisition unit 16, request information regarding communication quality associated with the network slice, and setting values for the number of resource blocks allocated to the network slice.

The second model storage unit 141 stores the trained second machine learning model constructed by the learning unit 12A.

The notification unit 15 notifies the base station 30 of information regarding resource blocks for each resource element to be allocated to a network slice. More specifically, the notification unit 15 notifies the base station 30 of information regarding the resource block specified by the frequency element, time element, and space element determined by the calculation unit 13A.

[Operation of communication management device]
Next, the operation of the communication management device 1A having the above-described configuration will be explained with reference to the flowcharts of FIGS. 10 and 11. FIG. 10 is a flowchart showing learning processing by the communication management device 1A. FIG. 11 is a flowchart showing arithmetic processing by the communication management device 1A.

First, as shown in FIG. 10, the learning unit 12A prepares teacher data (step S20). More specifically, the learning unit 12A calculates frequency element values, time element values, and spatial Learning is performed using training data in which information about resource blocks represented by element values is given as correct labels.

Next, the learning unit 12A determines the relationship between the request information regarding the communication quality associated with the network slice and the setting value of the number of resource blocks, and the information regarding the resource block for each resource element allocated to the network slice. Learning is performed using the learning model (step S21). The learning unit 12A learns the second machine learning model using the teacher data prepared in step S20.

Next, the second model storage unit 141 stores the learned second machine learning model constructed in step S21 (step S22).

Next, as shown in FIG. 11, the third acquisition unit 16 acquires information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated (step S30). The information regarding the network slice that the third acquisition unit 16 acquires in step S30 includes request information regarding communication quality associated with the network slice and a setting value for the number of resource blocks allocated to the network slice.

Next, the calculation unit 13A loads the learned second machine learning model from the second model storage unit 141 (step S31). Subsequently, the calculation unit 13A supplies the request information regarding the communication quality associated with the network slice and the setting value of the number of resource blocks as unknown inputs to the trained second machine learning model, and and outputs information regarding resource blocks to be allocated to network slices for each resource element (step S32).

Next, the notification unit 15 notifies the base station 30 of the information obtained in step S32 regarding the resource block for each resource element to be allocated to the network slice (step S33). The notification unit 15 can notify the base station 30 of information on resource blocks identified by the frequency element value, time element value, and spatial element value associated with the network slice ID. The base station 30 that has received the notification allocates and registers the specific resource block related to the notification to the network slice related to the network slice ID.

As explained above, according to the communication management device 1A according to the second embodiment, the request information regarding the communication quality associated with the network slice ID and the set value of the number of resource blocks are used as unknown inputs for the learned communication management device 1A. 2 machine learning model, the learned second machine learning model performs calculations, and information regarding resource blocks to be allocated to the network slice is output for each resource element including a frequency element, a time element, and a spatial element. Therefore, a resource block specified by a frequency element value, a time element value, and a spatial element value can be allocated to a network slice to which a set value of the number of resource blocks is fixedly allocated.

Furthermore, according to the communication management device 1A according to the second embodiment, information regarding resource blocks to be allocated to network slices is output for each resource element including frequency elements, time elements, and space elements, so that requests regarding communication quality can be satisfied. , more appropriate resource blocks can be allocated to network slices.

[Third embodiment]
Next, a third embodiment of the present invention will be described. In the following description, the same components as those in the first and second embodiments described above will be denoted by the same reference numerals, and the description thereof will be omitted.

In the second embodiment, a case has been described in which information regarding resource blocks for each resource element including a frequency element, a time element, and a spatial element is obtained when the number of resource blocks is fixedly allocated to a network slice. . On the other hand, in the third embodiment, when information regarding any one or two resource elements among the frequency element, time element, and spatial element is set to a fixed value, the information is allocated to a network slice. , obtain information about resource blocks for each other resource element.

[Functional block of communication management device]
As shown in FIG. 12, the communication management device 1B according to the present embodiment includes a learning unit (third learning unit) 12B, a calculation unit (third calculation unit) 13B, a storage unit (third storage unit) 14B, a notification 15, a third acquisition section 16, and a fourth acquisition section 17. The communication management device 1B according to the third embodiment differs in configuration from the communication management device 1A according to the second embodiment in that it further includes a fourth acquisition unit 17. Hereinafter, a description will be given focusing on the configuration different from the second embodiment.

First, the fourth acquisition unit 17 acquires element information including the set number of elements for any one or two of the frequency element, time element, and spatial element. For example, the fourth acquisition unit 17 can acquire the number of subcarriers on the frequency axis when setting the frequency element as a fixed value. Alternatively, when setting the time element as a fixed value, the number of time slots on the time axis can be obtained. Similarly, the fourth acquisition unit 17 can acquire the number of spatial streams on the spatial axis, which is set as a fixed value.

The fourth acquisition unit 17 also acquires not only the number of frequency elements, time elements, and spatial elements, but also the values of subcarriers f ₁ to f _k that are specific frequency elements, and the time as shown in the example of FIG. Values of time slots t ₁ to t _k that are elements and values of spatial streams s ₁ , s _{2 ,} etc. that are spatial elements may be obtained.

The learning unit 12B acquires request information regarding communication quality associated with the network slice, a set value for the number of resource blocks to be allocated to the network slice, element information acquired by the fourth acquisition unit 17, a frequency element, a time element, and A third machine learning model is used to learn the relationship between each resource element other than the resource element related to the element information acquired by the fourth acquisition unit 17 among the spatial elements and the information regarding the resource block to be allocated to the network slice. .

FIG. 13 shows a neural network structure adopted as an example of the third machine learning model for learning by the learning unit 12B. The neural network comprises an input layer x, a hidden layer h, and an output layer y. In the example of FIG. ₁₃ , the input node The number of resource blocks is input.

In the example of FIG. 13, the value of the “spatial element s” acquired by the fourth acquisition unit 17 is input to the input nodes x ₂ to x _l . The output nodes y ₁ to y _m correspond to the values of "frequency element f" and "time element t." For example, as an output node of "frequency element f", an output node corresponding to subcarriers f ₁ to f _k on the frequency axis, and as an output node of "time element t", an output node corresponding to time slots t ₁ to t _k can be provided. Note that the number of each resource element of the output node can be determined in advance.

The learning unit 12B acquires the network slice ID of the network slice, the request information regarding communication quality associated with the network slice ID, the set value of the number of resource blocks, and the element information of the resource element acquired by the fourth acquisition unit 17 through neural processing. It is applied to the input layer of the network, an activation function is applied to the weighted sum of the inputs, and the output determined by threshold processing is passed to the output layer.

The learning unit 12B further introduces an objective function that evaluates the error between the correct label value and the output value y of the neural network, thereby improving the network slice ID of the network slice and the communication associated with the network slice ID. The parameters of the neural network are learned so that the request information regarding quality, the set value of the number of resource blocks, and the element information of the spatial element become the value of the frequency element f and the value of the time element t of the correct label.

The learning unit 12B determines a correct label based on the request information regarding communication quality associated with the network slice ID, the setting value of the number of resource blocks, and the input values of the spatial element s s ₁ and s ₂ which are element information. The neural network is trained using teacher data to which the value of the frequency element f and the value of the time element t are given.

The learning unit 12B adjusts the weight parameters of the neural network so that the objective function becomes minimum, that is, zero. The learning unit 12B can optimize the objective function using a gradient method using an error backpropagation method, a stochastic gradient descent method, or the like. The trained third machine learning model constructed by the learning unit 12B is stored in the third model storage unit 142, which will be described later.

Returning to FIG. 12, the calculation unit 13B has learned request information regarding communication quality associated with a network slice, a setting value of the number of resource blocks to be assigned as a fixed value to a network slice, and element information of a resource element as unknown inputs. The fourth acquisition unit 17 performs calculations on the learned third machine learning model to obtain information regarding resource blocks to be allocated to the network slice among the frequency element, time element, and spatial element. Output for each resource element other than the resource element related to the element information obtained by.

The storage unit 14B includes a third model storage unit 142. The storage unit 14B stores element information of one or two of the frequency element, time element, and spatial element acquired by the fourth acquisition unit 17.

The third model storage unit 142 stores the trained third machine learning model constructed by the learning unit 12B.

The notification unit 15 notifies the base station 30 of information regarding resource blocks for each resource element to be allocated to the network slice. More specifically, the notification unit 15 notifies the base station 30 of information regarding the resource block indicated by the value of the resource element determined by the calculation unit 13B.

[Operation of communication management device]
Next, the operation of the communication management device 1B having the above-described configuration will be explained with reference to the flowcharts of FIGS. 14 and 15. FIG. 14 is a flowchart showing learning processing by the communication management device 1B. FIG. 15 is a flowchart showing arithmetic processing by the communication management device 1B.

First, as shown in FIG. 14, the learning unit 12B acquires element information (step S200). More specifically, in step S200, element information regarding any one or two resource elements among the frequency element, time element, and space element to be assigned as a fixed value to the network slice is acquired. The learning unit 12B can acquire the element information acquired by the fourth acquisition unit 17.

Next, the learning unit 12B prepares teacher data used in the learning process (step S201). Specifically, the learning unit 12B uses input values including request information regarding communication quality associated with a network slice, a setting value for the number of resource blocks to be set as a fixed value, and element information of a resource element to be set as a fixed value. In contrast, teacher data is prepared in which the values of resource elements other than the resource elements set as fixed values are given as correct labels.

Next, the learning unit 12B obtains request information regarding communication quality associated with the network slice, a setting value for the number of resource blocks to be set as a fixed value, element information of a resource element to be set as a fixed value, a frequency element, a time element. , and among the spatial elements, the relationship between each resource element other than the resource element related to the element information set as a fixed value and the information regarding the resource block to be allocated to the network slice is learned using the third machine learning model (step S202).

The third model storage unit 142 stores the learned third machine learning model constructed in step S202 (step S203).

Next, as shown in FIG. 15, the third acquisition unit 16 acquires information regarding the network slice to which the resource block is allocated (step S300). The information regarding the network slice that the third acquisition unit 16 acquires in step S300 includes request information regarding the communication quality associated with the network slice and a setting value for the number of resource blocks allocated to the network slice.

Next, the fourth acquisition unit 17 acquires element information including the set number of elements for any one or two of the frequency element, time element, and spatial element (step S301). Using the example of FIG. 13, in step S301, the fourth acquisition unit 17 can acquire the number of spatial streams s ₁ and s ₂ or the values of spatial streams s ₁ and s ₂ .

Next, the calculation unit 13B loads the trained third machine learning model from the third model storage unit 142 (step S302). Next, the arithmetic unit 13B uses as unknown inputs the requested information regarding the communication quality associated with the network slice, the setting value of the number of resource blocks to be assigned as a fixed value to the network slice, and the element information of the resource element. In step S301, the fourth acquisition unit calculates the learned third machine learning model and acquires information regarding the resource block to be assigned to the network slice among the frequency element, time element, and spatial element. Each resource element other than the resource element related to the element information acquired in Step 17 is output (step S303).

Using the example of FIG. 13, in step S303, an element including request information regarding communication quality associated with the network slice ID, the number of resource blocks set as a fixed value, and the number of spatial elements s set as a fixed value. The information is given as unknown input to the trained third machine learning model. The calculation unit 13B calculates the learned third machine learning model and outputs information regarding the resource block given by the value of the frequency element f and the value of the time element t.

Next, the notification unit 15 notifies the base station 30 of the information regarding the resource block for each resource element obtained in step S303 (step S304). The base station 30 that receives the notification allocates and registers resource blocks to the network slice. According to the example of FIG. 13, the notification unit 15 can notify the base station 30 of the value of the frequency element f and the value of the time element t associated with the network slice ID.

As described above, according to the communication management device 1B according to the third embodiment, request information regarding communication quality associated with a network slice ID, a set value of the number of resource blocks, and a resource element set as a fixed value The element information is given to the trained third machine learning model as an unknown input, and the trained third machine learning model is operated to obtain information about the resource blocks to be allocated to the network slice in the frequency element, time element, and space. Among the elements, output for each resource element other than the resource element set as a fixed value. Therefore, to a network slice to which the set value of the number of resource blocks and a specific resource element are fixedly allocated, it is not possible to allocate a resource block specified by the value of a resource element other than the resource element set as a fixed value. can.

In addition, in the third embodiment described above, as illustrated in FIG. 13, among the frequency element, time element, and spatial element, the element information regarding the spatial element is acquired as a fixed value as one resource element, and the remaining The case of finding values regarding the frequency element and time element of is explained. However, as described above, the resource element set as the input node of the third machine learning model may be a frequency element or a time element. Furthermore, the resource elements set as input nodes of the third machine learning model can be not only one type of resource element, but also two resource elements selected from three resource elements as fixed values. In this case, what is output by the calculation unit 13B is the value of the remaining one resource element.

Furthermore, in the embodiments described above, a case has been described in which a neural network model including an input layer, a hidden layer, and an output layer is used as the first machine learning model, the second machine learning model, and the third machine learning model. As the neural network model employed as the second machine learning model and the third machine learning model, a recurrent neural network (RNN) or a deep multimodal neural network can be used.

In addition to the above-mentioned neural network model, the first machine learning model, second machine learning model, and third machine learning model are based on deep learning with multiple layers of neural networks, such as logistic regression, random forest, decision tree, etc. May be used. In addition to these supervised learnings, generative models such as generative adversarial networks and variational orthogonal encoders may be used as machine learning models for unsupervised learning.

In addition, in the above-mentioned embodiment, although the case where it is a communication management system based on 5G was illustrated, the communication management system based on LTE or 6G may be sufficient.

Although the embodiments of the communication management device and the communication management method of the present invention have been described above, the present invention is not limited to the described embodiments, and those skilled in the art can imagine it within the scope of the invention described in the claims. Various possible modifications can be made.

DESCRIPTION OF SYMBOLS 1... Communication management device, 10... First acquisition part, 11... Second acquisition part, 12... Learning part, 13... Arithmetic part, 14... Storage part, 15... Notification part, 2... Communication terminal, 3... Wireless access network , 30... Base station, 4... Core network, NSSF... 40, AMF... 41, 101... Bus, 102... Processor, 103... Main storage device, 104... Communication interface, 105... Auxiliary storage device, 106... Input/output I/ O, 140...first model storage unit, L, NW...network.

Claims (12)

The configuration is configured to acquire information regarding a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated, and the information regarding the plurality of network slices includes information regarding communication quality associated with each of the plurality of network slices. a first acquisition part that includes request information;
a second acquisition unit configured to acquire an initial value of the number of resource blocks of the radio resource to be allocated to each of the plurality of network slices;
The initial value acquired by the second acquisition unit is given to the learned first machine learning model as an unknown input, and the learned first machine learning model is operated to associate the request information. a first calculation unit configured to output an optimized number of resource blocks to be allocated to each of the plurality of network slices;
and a notification unit configured to notify a base station of the optimized number of resource blocks to be allocated to each of the plurality of network slices. The communication management device according to claim 1,
Furthermore, when the initial value of the number of resource blocks to be allocated to each of the plurality of network slices is given as an input, the number of resource blocks to be allocated to each of the plurality of network slices associated with the request information is a first storage unit configured to store a first machine learning model that outputs a predicted value;
When the initial value is given as an input, the sum of the predicted values of the number of resource blocks output from the first machine learning model is the total number of the resource blocks constituting the radio resource, a first learning unit configured to learn parameters of the first machine learning model;
The first storage unit further stores the trained first machine learning model constructed by the first learning unit,
The communication management device, wherein the first calculation unit reads the learned first machine learning model from the first storage unit and performs calculation of the learned first machine learning model. The configuration is configured to acquire information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated, and the information regarding the network slice includes request information regarding communication quality associated with the network slice, and request information regarding the communication quality associated with the network slice. a third acquisition unit that includes a setting value for the number of allocated resource blocks;
The request information associated with the network slice and the set value of the number of resource blocks are given to the trained second machine learning model as unknown inputs, and the trained second machine learning model is operated. , a second calculation unit configured to output information regarding the resource block allocated to the network slice for each resource element;
a notification unit configured to notify a base station of information regarding the resource block for each of the resource elements allocated to the network slice;
The communication management device, wherein the resource element includes a frequency element, a time element, and a space element. The communication management device according to claim 3,
Furthermore, a fourth acquisition unit configured to acquire element information including a set number of elements for any one or two of the resource elements of the frequency element, the time element, and the spatial element. and,
The request information associated with the network slice, the setting value of the number of resource blocks, and the element information are provided as unknown inputs to the trained third machine learning model, and the trained third machine learning model The information regarding the resource block to be allocated to the network slice is calculated based on the resource element related to the element information acquired by the fourth acquisition unit among the frequency element, the time element, and the spatial element. a third calculation unit configured to output for each resource element other than;
The communication management device is characterized in that the notification unit notifies the base station of information regarding the resource block for each of the resource elements other than the resource element related to the element information to be allocated to the network slice. The communication management device according to claim 3,
Furthermore, a second machine learning model calculates the relationship between the request information and the set value of the number of resource blocks associated with the network slice, and information regarding the resource block for each resource element allocated to the network slice. a second learning section configured to learn using the
a second storage unit configured to store the trained second machine learning model constructed by the second learning unit;
The communication management device, wherein the second calculation unit reads the learned second machine learning model from the second storage unit and performs calculation of the learned second machine learning model. The communication management device according to claim 4,
Furthermore, the request information associated with the network slice, the set value of the number of resource blocks, and the element information, and the element information of the frequency element, the time element, and the spatial element. a third learning unit configured to learn, using a third machine learning model, a relationship between each of the resource elements other than the resource element and the information regarding the resource block to be allocated to the network slice;
a third storage unit configured to store the trained third machine learning model constructed by the third learning unit;
The communication management device, wherein the third calculation unit reads the learned third machine learning model from the third storage unit and performs calculation of the learned third machine learning model. A communication management method executed by a communication management device, the method comprising:
A first acquisition step of acquiring information about a plurality of network slices to which resource blocks, which are units of division of radio resources, are allocated, the information about the plurality of network slices including the information associated with each of the plurality of network slices. , a first acquisition step including request information regarding communication quality;
a second obtaining step of obtaining an initial value of the number of resource blocks of the radio resource to be allocated to each of the plurality of network slices;
The initial value acquired in the second acquisition step is given to the learned first machine learning model as an unknown input, and the learned first machine learning model is operated to associate the request information. a first calculation step of outputting an optimized number of resource blocks to be allocated to each of the plurality of network slices;
a notification step configured to notify a base station of the optimized number of resource blocks to be allocated to each of the plurality of network slices. The communication management method according to claim 7,
Furthermore, when the initial value of the number of resource blocks to be allocated to each of the plurality of network slices is given as an input, the number of resource blocks to be allocated to each of the plurality of network slices associated with the request information is a first storage step of storing a first machine learning model that outputs a predicted value in a first storage unit;
When the initial value is given as an input, the sum of the predicted values of the number of resource blocks output from the first machine learning model is the total number of the resource blocks constituting the radio resource, a first learning step of learning parameters of the first machine learning model;
The first storage step further stores the trained first machine learning model constructed in the first learning step in the first storage unit,
The communication management method, wherein the first calculation step reads the learned first machine learning model from the first storage unit and performs calculation of the learned first machine learning model. A communication management method executed by a communication management device, the method comprising:
a third acquisition step of acquiring information regarding a network slice to which a resource block, which is a unit of division of radio resources, is allocated, the information regarding the network slice including request information regarding communication quality associated with the network slice; a third obtaining step, comprising a set value for the number of resource blocks allocated to a network slice;
The request information associated with the network slice and the set value of the number of resource blocks are given to the trained second machine learning model as unknown inputs, and the trained second machine learning model is operated. , a second calculation step of outputting information regarding the resource block allocated to the network slice for each resource element;
a notification step of notifying a base station of information regarding the resource block for each of the resource elements allocated to the network slice;
The communication management method, wherein the resource element includes a frequency element, a time element, and a space element. The communication management method according to claim 9,
Further, a fourth acquisition step of acquiring element information including a set number of elements for any one or two of the resource elements of the frequency element, the time element, and the spatial element;
The request information associated with the network slice, the setting value of the number of resource blocks, and the element information are provided as unknown inputs to the trained third machine learning model, and the trained third machine learning model The information regarding the resource block to be allocated to the network slice is calculated based on the resource element related to the element information acquired in the fourth acquisition step among the frequency element, the time element, and the spatial element. a third calculation step configured to output for each resource element other than;
The communication management method is characterized in that the notification step notifies the base station of information regarding the resource block for each of the resource elements other than the resource element related to the element information to be allocated to the network slice. The communication management method according to claim 9,
Furthermore, a second machine learning model calculates the relationship between the request information and the set value of the number of resource blocks associated with the network slice, and information regarding the resource block for each resource element allocated to the network slice. a second learning step of learning using
a second storage step of storing the learned second machine learning model constructed in the second learning step in a second storage unit;
The communication management method, wherein the second calculation step reads the learned second machine learning model from the second storage unit and performs calculation of the learned second machine learning model. The communication management method according to claim 10,
Furthermore, the request information associated with the network slice, the set value of the number of resource blocks, and the element information, and the element information of the frequency element, the time element, and the spatial element. a third learning step of learning, using a third machine learning model, a relationship between each of the resource elements other than the resource element and the information regarding the resource block to be allocated to the network slice;
a third storage step of storing the learned third machine learning model constructed in the third learning step in a third storage unit;
The communication management method, wherein the third calculation step reads the learned third machine learning model from the third storage unit and performs calculation of the learned third machine learning model.