Communication management device, communication management method, and communication management system

Abstract

The objective is to improve the throughput of each of the multiple communication terminals under management. [Solution] The communication management device 1 includes a generator 131 that generates pseudo-communication volume time-series data similar to the true communication volume time-series data, using the time-series data of the communication terminal with the largest total communication volume for each selected time period as the true communication volume time-series data; a discriminator 132 that distinguishes between the pseudo-communication volume time-series data generated by the generator 131 and the true communication volume time-series data; and a notification unit 16 that notifies a plurality of communication terminals 2 of the trained generator 131' constructed by the learning unit 13 as communication management information.

Description

Technical Field

[0001] The present invention relates to a communication management device, a communication management method, and a communication management system.

Background Art

[0002] In recent years, with the increase in the number of communication terminals connected to a mobile communication network, there has been a demand for more efficient data communication. For example, Patent Document 1 collects information related to wireless communication such as the received signal strength of a communication terminal, determines the congestion level of a wireless line based on information such as the throughput of the communication terminal and the received signal strength, and determines a schedule for data communication performed with the communication terminal according to the congestion level. A communication management system is disclosed.

[0003] However, among IoT terminals and the like, there are cases where it is necessary to transmit data at a predetermined cycle, so communication to avoid congestion may be difficult. Therefore, it is difficult to adopt the off-peak communication described in Patent Document 1 for IoT terminals that need to transmit data at a predetermined cycle. In addition, in the communication management technology described in Patent Document 1, traffic adjustment is performed throughout the mobile communication network. Therefore, it is also difficult to improve the throughput of each of a specific plurality of communication terminals selected as management targets.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In the conventional technology, it has been difficult to improve the throughput of each of a plurality of communication terminals to be managed.

[0006] This invention was made to solve the above-mentioned problems and aims to improve the throughput of each of the multiple communication terminals under management. [Means for solving the problem]

[0007] To solve the above-mentioned problems, the communication management device according to the present invention comprises: a collection unit configured to collect time-series data of communication volume for each time period of a plurality of communication terminals, each identified by a subscriber identifier; a selection unit configured to select the time-series data of communication volume of the communication terminal with the largest total communication volume for each time period based on the collected time-series data of communication volume; a generator configured to perform adversarial learning of a generative model having a generator that generates pseudo-communication volume time-series data similar to the true communication volume time-series data, using the selected time-series data of communication volume of the communication terminal with the largest total communication volume for each time period as the true communication volume time-series data; and a discriminator that distinguishes between the pseudo-communication volume time-series data generated by the generator and the true communication volume time-series data; and a notification unit configured to notify the plurality of communication terminals of the trained generator constructed by the learning unit as communication management information.

[0008] Furthermore, in the communication management device according to the present invention, the plurality of communication terminals may be a plurality of communication terminals to which common identification information has been assigned.

[0009] Furthermore, in the communication management device according to the present invention, the collection unit may collect time-series data of the communication volume for each time period of the plurality of communication terminals to which the common identification information has been assigned, from the user plane function of the core network.

[0010] Furthermore, in the communication management device according to the present invention, the plurality of communication terminals may be a plurality of communication terminals located in the same communication area.

[0011] Furthermore, in the communication management device according to the present invention, the time-series data of the pseudo-communication volume may have a data distribution that minimizes the statistical distance from the data distribution of the time-series data of the true communication volume.

[0012] To solve the above-mentioned problems, the communication management method according to the present invention comprises: a collection step of collecting time-series data of communication volume for each time period of a plurality of communication terminals, each identified by a subscriber identifier; a selection step of selecting time-series data of communication volume of the communication terminal with the largest total communication volume for each time period based on the collected time-series data of communication volume; a learning step of performing adversarial learning on a generative model having a generator that generates pseudo-communication volume time-series data similar to the true communication volume time-series data, using the selected time-series data of communication volume of the communication terminal with the largest total communication volume for each time period as true communication volume time-series data; and a discriminator that distinguishes between the pseudo-communication volume time-series data generated by the generator and the true communication volume time-series data; and a notification step of notifying the plurality of communication terminals of the trained generator constructed in the learning step as communication management information.

[0013] Furthermore, in the communication management method according to the present invention, the plurality of communication terminals may be a plurality of communication terminals to which common identification information has been assigned.

[0014] Furthermore, in the communication management method according to the present invention, the collection step may collect time-series data of the communication volume for each time period of the plurality of communication terminals to which the common identification information has been assigned, from the user plane function of the core network.

[0015] Furthermore, in the communication management method according to the present invention, the plurality of communication terminals may be a plurality of communication terminals located in the same communication area.

[0016] Furthermore, in the communication management method according to the present invention, each of the plurality of communication terminals may further include an acquisition step of acquiring the learned generator of the communication management information notified in the notification step, a generation step of generating pseudo-communication volume time-series data similar to the true communication volume time-series data using the learned generator, and a communication step of controlling the transmission of data in each time period based on the generated pseudo-communication volume time-series data.

[0017] To solve the above-mentioned problems, the communication management system according to the present invention is a communication management system comprising the above-mentioned communication management device and the plurality of communication terminals, wherein each of the plurality of communication terminals comprises an acquisition unit configured to acquire the learned generator of the communication management information notified from the communication management device, a generation unit configured to generate pseudo-communication volume time-series data similar to the true communication volume time-series data using the learned generator, and a communication unit configured to control the transmission of data in each time period based on the generated pseudo-communication volume time-series data. [Effects of the Invention]

[0018] According to the present invention, adversarial learning is performed on a generative model having a generator that generates pseudo-communication volume time-series data similar to the true communication volume time-series data, using the time-series data of the communication terminal with the largest total communication volume for each selected time period as the true communication volume time-series data, and a discriminator that distinguishes between the pseudo-communication volume time-series data generated by the generator and the true communication volume time-series data. As a result, the throughput of each of the multiple communication terminals under management can be improved. [Brief explanation of the drawing]

[0019] [Figure 1] Figure 1 is a block diagram showing the configuration of a communication management system including a communication management device according to an embodiment of the present invention. [Figure 2] Figure 2 is a block diagram showing the configuration of a communication terminal according to this embodiment. [Figure 3]Figure 3 is a diagram illustrating the time-series data of communication volume collected by the communication management device according to this embodiment. [Figure 4] Figure 4 is a diagram illustrating the structure of the configuration information table provided by the communication management system according to this embodiment. [Figure 5] Figure 5 is a block diagram showing the configuration of the learning unit included in the communication management device according to this embodiment. [Figure 6] Figure 6 is a diagram illustrating the learning unit included in the communication management device according to this embodiment. [Figure 7] Figure 7 is a diagram illustrating the learning unit included in the communication management device according to this embodiment. [Figure 8] Figure 8 is a block diagram showing the hardware configuration of the communication management device according to this embodiment. [Figure 9] Figure 9 is a block diagram showing the hardware configuration of the communication management device according to this embodiment. [Figure 10] Figure 10 is a sequence diagram showing an overview of the operation of the communication management system according to this embodiment. [Figure 11] Figure 11 is a sequence diagram showing an overview of the operation of the communication management system according to this embodiment. [Figure 12] Figure 12 is a flowchart showing the learning process by the communication management device according to this embodiment. [Modes for carrying out the invention]

[0020] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to Figures 1 to 12.

[0021] [Configuration of the communication management system] First, an overview of a communication management system comprising a communication management device 1 and a communication terminal 2 according to an embodiment of the present invention will be described. Figure 1 is a block diagram showing the configuration of a communication management system according to an embodiment of the present invention.

[0022] The communication management system according to this embodiment is installed, for example, in a mobile communication network compatible with the 5G communication standard. The communication management system comprises a communication management device 1, communication terminals 2, a base station 3, a core network 4, and a data network 5. The communication management system manages the optimal data transmission timing for each of the multiple communication terminals 2 under its management.

[0023] Communication terminal 2 is implemented by mobile communication terminals such as smartphones, tablet computers, and laptop computers, as well as IoT terminals that use the 5G mobile communication network, such as smart meters, wearable devices, and industrial sensors. Communication terminal 2 is equipped with SIM2a, and the contract profile of SIM2a stores the user's subscriber identification information, including identifier information such as the subscriber identifier (IMSI: International Mobile Subscriber Identity) assigned to the mobile phone line contract, the subscriber user's telephone number (MSISDN: Mobile Subscriber International Subscriber Directory Number), and the SIM card number (ICCID: Integrated Circuit Card Identifier). Communication terminal 2 is uniquely identified by the IMSI.

[0024] Each communication terminal 2 is also assigned a terminal IP address that uniquely identifies the terminal. The IP address is assigned to the communication terminal 2 via SMF42 after the session is established. In this embodiment, there are multiple communication terminals 2, and multiple communication terminals 2 located in the same communication area 3A are assigned a group ID (common identification information). The group ID can be assigned to terminals among the multiple communication terminals 2 located in the same communication area 3A that share predetermined attributes related to communication and services. The communication management processing in this embodiment is performed for each of the multiple communication terminals 2 with the same group ID.

[0025] For example, if communication terminal 2 consists of industrial sensors, a common group ID can be assigned to multiple industrial sensors that are located in the same communication area 3A, used for similar purposes, and that need to transmit similar amounts of data at regular intervals, thus having similar communication patterns and requirements.

[0026] Base station 3 is a wireless base station compliant with the 5G communication standard and relays communication between communication terminals 2 located in communication area 3A and the core network 4. Base station 3 is connected to the core network 4 via a backhaul link. Base station 3 receives data from communication terminals 2 using predefined wireless resources.

[0027] Core network 4 is connected to communication management device 1 via a network NW such as LAN, WAN, or the Internet. Core network 4 includes nodes within the C-plane: AMF (Access and Mobility Management Function) 40, UDM (Unified Data Management) / UDR (Unified Data Repository) 41, SMF (Session Management Function) 42, and PCF (Policy Control Function) 43. Core network 4 also includes multiple UPF (User Plane Function) 44 within the U-plane. Functional nodes within the U-plane and C-plane that are included in core network 4 are not shown in the diagram.

[0028] The AMF40 is a node that provides mobility control functions and performs movement control such as location registration, paging, and handover.

[0029] The UDM / UDR41 manages subscriber profiles, performs authentication, and manages mobility. In this embodiment, the UDM / UDR41, in response to instructions from the communication management device 1, creates and stores a configuration information table T2, as shown in Figure 4, which associates the IMSI, group ID, base station ID of multiple communication terminals 2, and the communication paths of the UPF44 set in the data communication of multiple communication terminals 2. The instructions from the communication management device 1 include grouping information that associates multiple IMSIs and group IDs of the managed devices specified in advance. The IP address value of the UPF44 in the configuration information table T2 is registered after the communication paths are set by the SMF42 and PCF43 described later.

[0030] The UDM / UDR41 can configure the configuration information table T2 by adding a group ID field to the subscriber profile. The UDM / UDR41 according to this embodiment is equipped with a communication interface 41a for communicating with the communication management device 1. The configuration information table T2 managed by the UDM / UDR41 is synchronized with the communication management device 1, and the same contents are stored there (first storage unit 14). In addition, the UDM / UDR41 may be configured as a single device with the UDM and UDR, or it may be a device in which the UDM and UDR are arranged separately.

[0031] SMF42 is a session management function that establishes, modifies, and releases PDU (Packet Data Unit) sessions between communication terminal 2 and data networks such as the Internet. Based on the PCC (Policy and Charging Control) policy from PCF43, SMF42 sets the appropriate communication path for data communication between communication terminal 2 and UPF44.

[0032] PCF43 determines QoS and policies and provides them to SMF42. PCF43 applies PCC rules according to the 3GPP (registered trademark) specification and creates PCC policies for configuring the communication path of UPF44 through which communication terminal 2 communicates, in response to instructions from communication management device 1.

[0033] The UPF44 is a user plane function that processes data between the base station 3 and a data network 5 such as the internet. Multiple UPF44s are provided in the core network 4. In this embodiment, the UPF44 records time-series data of the communication volume of the communication terminal 2 for each time period in memory. In this embodiment, the UPF44 is equipped with a communication interface 44a for communicating with the communication management device 1.

[0034] Figure 3 is a diagram illustrating the time-series data of communication volume recorded by each UPF44 and collected by the collection unit 11 of the communication management device 1, which will be described later. As shown in Table T1 of Figure 3, each UPF44 records a log of time-series data of communication volume for each time period in memory, for each IMSI and group ID. i (i=1, 2, ..., N) are time slots set for the communication management processing in this embodiment.

[0035] Furthermore, regarding data usage [bps], the time zone t i By setting a threshold for the data transfer rate [bps] value for each time period, the data is converted to binary data. i The UPF44 takes the value "1" if the data traffic [bps] exceeds the threshold (data traffic is present), and "0" if it does not exceed the threshold (no data traffic). Note that the process of converting the time-series data of data traffic into binary data by thresholding the data traffic [bps] may be performed by the data collection unit 11 of the communication management device 1, rather than by the UPF44.

[0036] Data network 5 is an external network such as the internet or a cloud network. Communication terminal 2 communicates data with data network 5 via base station 3 and the UPF44 of core network 4.

[0037] [Functional blocks of the communication management device] As shown in Figure 1, the communication management device 1 comprises a setting unit 10, a collection unit 11, a selection unit 12, a learning unit 13, a first storage unit 14, a second storage unit 15, and a notification unit 16. The communication management device 1 learns a generative model through adversarial learning using the timing of data transmission from the communication terminal 2 with the best throughput among a plurality of communication terminals 2 as true communication volume time series data, and notifies the plurality of communication terminals 2 under management of the trained generator constructed through learning as communication management information.

[0038] The configuration unit 10 specifies the IMSI and group ID based on the location registration request signal from each communication terminal 2 and instructs the UPF44 settings of each communication terminal 2. Specifically, the configuration unit 10 transmits grouping information to the UDM / UDR41, which associates the previously registered managed IMSI and group ID. Based on the grouping information from the configuration unit 10, the UDM / UDR41 creates a configuration information table T2 (Figure 4). Furthermore, the configuration unit 10 instructs the user plane function settings for data communication of each communication terminal 2 based on the group ID associated with the IMSI included in the location registration request signal from each communication terminal 2.

[0039] The configuration unit 10 requests the creation of a PCC policy from the PCF43 of the core network 4, specifying the group ID, IMSI, and the IP address of the communication management device 1. In response to the creation request, the PCF43, SMF42, and UDM / UDR41 of the core network 4 cooperate to set the appropriate communication path for UPF44 for the data communication of the communication terminal 2. Once the UPF44 communication path is set for the data communication of the communication terminal 2, the configuration information table T2 updated by the UDM / UDR41 is synchronized with, for example, the first storage unit 14, and the updated configuration information table T2 is stored in the first storage unit 14.

[0040] The collection unit 11 collects time-series data of the traffic volume of each of the plurality of communication terminals 2 identified by IMSI for each time period. More specifically, the collection unit 11 collects time-series data of the traffic volume of each of the plurality of communication terminals 2 to which a group ID is assigned from the UPF 44 included in the core network 4. The collection unit 11 collects time-series data of the traffic volume of each of the plurality of communication terminals 2 for each time period for each group ID via the network NW from the UPF 44 related to the setting of the communication path of the data communication of the communication terminal 2 by the setting unit 10. As shown in FIG. 3, for each of IMSI_1 to IMSI_N to which "Group 1" is assigned, the collection unit 11 collects time-series data (t N at which takes a value of "1" for data communication or "0" for no data communication (t i , x i ).

[0041] The selection unit 12 selects the time-series data of the traffic volume of the communication terminal 2 with the largest total traffic volume for each time period based on the collected time-series data of the traffic volume. More specifically, the selection unit 12, based on the time-series data of the traffic volume (t i , x i ) collected by the collection unit 11, for each IMSI with a common group ID, selects the time-series data of the traffic volume of the communication terminal 2 with the largest value (Σx N ) obtained by summing the traffic volumes for each time period (t1 to t i ). The communication terminal 2 with the largest total traffic volume is the communication terminal 2 with the highest throughput among the plurality of communication terminals 2 included in the group ID.

[0042] More specifically, the selection unit 12 compares the total values of the traffic volumes of all IMSIs with a common group ID, and can select N (N is a positive integer of 2 or more) time-series data of the traffic volumes in descending order of the total values. The selection of the N data can be set to the number required to obtain sufficient accuracy in adversarial learning by the learning unit 13 described later, such as the top 4%.

[0043] The learning unit 13 performs adversarial learning of a generative model that includes a generator 131 that generates pseudo-data series data similar to the true data series data, using the time series data of the selected communication terminal 2 with the largest total data volume for each time period as the true data series data, and a discriminator 132 that distinguishes between the pseudo-data series data generated by the generator 131 and the true data series data.

[0044] As shown in Figure 5, the learning unit 13 trains a Generative Adversarial Network (GAN) having a generator 131 and a discriminator 132 in an adversarial manner. The learning unit 13 constructs a trained generator 131' through its training.

[0045] Figures 6 and 7 schematically represent the neural network configurations of the GAN generator 131 and discriminator 132 used by the learning unit 13. As shown in Figure 6, the generator 131 consists of a neural network having an input layer, a hidden layer, and an output layer. The hidden layer can consist of a fully connected layer and a deconvolution layer. The generator 131 is a model that generates time-series data of pseudo-communication volume from random noise. For example, m randomly sampled Gaussian noise vectors are input to the input node of the generator 131 (z1~z m The generator 131 performs sum-of-products operations on the input and weight parameters, and threshold processing using an activation function to produce the output G(z1)~G(z N Outputs ).

[0046] The classifier 132 shown in Figure 7 consists of a neural network having an input layer, a hidden layer, and an output layer. The hidden layer can be a convolutional layer, and the output layer can be a fully connected layer. In the example in Figure 7, the training data consists of N samples of time-series data of communication volume x1~x N The following is given as input. As mentioned above, the time-series data of communication volume is represented as binary data. Therefore, each input node in the input layer of the neural network constituting the classifier 132 has a time period t iA value of either 0 or 1 for the amount of traffic is input. In the example in Figure 5, the true traffic time series data is input to the classifier 132 as training data 134.

[0047] The classifier 132 performs a sum-of-products operation on the input and weight parameters, and thresholding using an activation function, to produce an output of, for example, 1 or 0. When the classifier 132 correctly identifies the training data relating to the input true traffic volume time series data as true traffic volume time series data, it outputs y=1. On the other hand, when it correctly identifies the training data relating to the input pseudo-traffic volume time series data as pseudo-traffic volume time series data, it outputs y=0. In this way, the classifier 132 is a model that distinguishes the model distribution generated by the generator 131 from the data distribution of the training data, which is the true distribution. The pseudo-traffic volume time series data has a data distribution that minimizes the statistical distance from the data distribution of the true traffic volume time series data.

[0048] Figure 5 is a block diagram illustrating the adversarial learning of the GAN by the learning unit 13. The generator 131 of the GAN adopted by the learning unit 13 is denoted as function G, and the discriminator 132 is denoted as function D. The time series data of true traffic volume is denoted as x, the predicted value output by the discriminator 132 is denoted as y, and the correct label is denoted as t. The correct label t is set to 1 for the time series data of true traffic volume and to 0 for the time series data of pseudo-traffic volume generated by the generator 131. In this case, the discriminator 132 performs a binary classification problem with the cross-entropy E given by equation (1) below. CE It can be expressed as follows.

[0049]

number

[0050] The first term inside the brace in equation (1) above represents t n lny n In this case, the predicted value y of the classifier 132 n However, the true labels for the time-series data of the actual traffic volume are t n It is desirable to approach the value of =1. On the other hand, the second term inside the brace represents (1-t n)ln(1-y n In this case, the predicted value y of the classifier 132 n However, the value of the correct label (1-t) used to distinguish the time-series data of pseudo-traffic volume from the actual data is different. n It is desirable for cross-entropy E to approach 0. CE This value is maximized when the predicted value matches the correct label value.

[0051] Here, the generator 131 that constitutes the GAN has parameter w G ,θ G It has a function G(w G ,θ G ) is expressed as. Also, the classifier 132 has parameter w D ,θ D It has a function D(w D ,θ D This is expressed as ). The cross-entropy E in equation (1) above CE The objective function E of a GAN comprising a generator 131 and a discriminator 132 based on the above can be expressed by the following equation (2).

number

[0052] The first term of equation (2) above represents E D(x)=1 lnD(w D ,θ D ) is the expected value that the classifier 132 identifies as true traffic volume time series data. The second term of equation (2) above represents E D(x)=0 ln(1-D(G(w G ,θ G ),w D ,θ D)) is the expected value at which the discriminator 132 identifies the pseudo-communication volume time series data generated by the generator 131 as pseudo-communication volume time series data. In GAN training, the generator 131 and the discriminator 132 are trained adversarially by min-max optimization of the objective function E. Therefore, the generator 131 is trained to generate pseudo-communication volume time series data that deceives the discriminator 132, and the discriminator 132 is trained to identify the pseudo-communication volume time series data generated by the generator 131 as pseudo-communication volume time series data.

[0053] During the training of classifier 132, when given time-series data of true traffic volume, classifier 132 is trained to maximize the first term of the objective function E in equation (2) above by producing an output close to y=1. On the other hand, when given time-series data of pseudo-traffic volume, classifier 132 is trained to maximize the second term of the objective function E by producing an output close to y=0.

[0054] In the learning of generator 131, D(G(w) in equation (2) above G ,θ G ),w D ,θ D )(D(G(z)) in Figure 5) is close to 1 G(w G ,θ G The objective function E is minimized by outputting (G(z) in Figure 5). The learning unit 13 uses a learning procedure that alternately updates the parameters of the generator 131 and the discriminator 132. Details of the learning procedure for the generator 131 and discriminator 132 by the learning unit 13 will be described later.

[0055] When the objective function E of the GAN is optimized, the learning unit 13 stores the trained generator 131' in the second storage unit 15.

[0056] The first storage unit 14 stores a configuration information table T2 (Figure 4) in which the IP address, IMSI, group ID, and base station ID of the UPF44 related to the communication path setting by the setting unit 10 are associated with each other. The first storage unit 14 can synchronize with the configuration information table T2 created and updated by the UDM / UDR41. Alternatively, the configuration may involve periodically acquiring the configuration information table T2 created and updated by the UDM / UDR41, or referring to and reflecting the updated configuration information table T2.

[0057] The second memory unit 15 stores the trained generators 131' constructed by the learning unit 13. Since the trained generators 131' are generators constructed for each group ID, each trained generator 131' is associated with a group ID.

[0058] The notification unit 16 notifies multiple communication terminals 2 of the trained generator 131' constructed by the learning unit 13 as communication management information. More specifically, the notification unit 16 notifies multiple communication terminals 2 having the same group ID of the trained generator 131' constructed for each group ID via the network NW.

[0059] [Communication terminal function blocks] Next, the functional blocks of the communication terminal 2 according to this embodiment will be described with reference to Figure 2. Note that multiple communication terminals 2 are composed of the same functional blocks.

[0060] The third memory unit 20 stores the learned generator 131' notified by the notification unit 16 of the communication management device 1 as communication management information.

[0061] The acquisition unit 21 acquires the learned generator 131' of communication management information notified from the communication management device 1. Specifically, the acquisition unit 21 reads the learned generator 131' from the third storage unit 20.

[0062] The generation unit 22 uses the trained generator 131' to generate pseudo-communication volume time series data that is similar to the true communication volume time series data. The generated pseudo-communication volume time series data is information that indicates the data transmission timing that is very similar to the data transmission timing of the communication terminal 2 with the best throughput within the group ID. For example, the pseudo-communication volume time series data (t i ,x i ) for each time period (t1, t2, t3, t4, ..., t N In ) (x1,x2,x3,x4,···,x N Let's consider the case where ) = (0,1,1,0,···,1). In this case, the timing of data transmission is indicated as follows: "0" for time zone t1, "1" for time zone t2, ... for data transmission.

[0063] The communication unit 23 controls data transmission in each time period based on the generated pseudo-communication volume time series data. For example, the communication unit 23 can take into account the threshold used when converting the communication volume time series data to a binary representation, and control the transmission of data in the "0" time period t1, where no data is transmitted, to either not transmit data or transmit data with a communication volume that does not exceed the threshold. On the other hand, in the "1" time period t2, where data is transmitted, it can transmit data with a communication volume that exceeds the threshold. The communication unit 23 can control communication, for example, by a grant-free method that transmits data at predetermined timings without requiring additional permission from the base station 3.

[0064] [Hardware configuration of the communication management device] Next, an example of a hardware configuration for realizing the communication management device 1 having the functions described above will be explained using Figure 8.

[0065] As shown in Figure 8, the communication management device 1 can be implemented, for example, by a computer equipped with a processor 102, main memory 103, communication interface 104, auxiliary storage 105, and input / output I / O 106 connected via a bus 101, and a program to control these hardware resources. The communication management device 1 can also include a display device 107 connected via the bus 101.

[0066] Processor 102 is implemented using CPUs, GPUs, FPGAs, ASICs, etc.

[0067] The main memory 103 contains pre-stored programs for the processor 102 to perform various controls and calculations. The processor 102 and the main memory 103 work together to realize the various functions of the communication management device 1, such as the setting unit 10, collection unit 11, selection unit 12, learning unit 13, and notification unit 16 shown in Figure 1.

[0068] The communication interface 104 is an interface circuit for networking the communication management device 1 with various external electronic devices.

[0069] The auxiliary storage device 105 consists of a read / write storage medium and a drive device for reading and writing various information such as programs and data to the storage medium. The auxiliary storage device 105 can use semiconductor memory such as a hard disk or flash memory as the storage medium.

[0070] The auxiliary storage device 105 has a program storage area for storing the GAN learning program and communication management program executed by the communication management device 1. The auxiliary storage device 105 also has an area for storing grouping information where group IDs and IMSIs are associated. The auxiliary storage device 105 enables the implementation of the first storage unit 14 and the second storage unit 15 described in Figure 1. Furthermore, it may also have, for example, a backup area for backing up the aforementioned data and programs.

[0071] The I / O106 is an input / output device that accepts signals from external devices and outputs signals to external devices.

[0072] The display device 107 is composed of an organic EL display, a liquid crystal display, and the like.

[0073] [Hardware configuration of communication terminals] Next, an example of a hardware configuration for realizing the communication terminal 2 having the functions described above will be explained using Figure 9.

[0074] As shown in Figure 9, the communication terminal 2 can be realized by a computer equipped with a processor 202, main memory 203, communication interface 204, auxiliary storage 205, and input / output I / O 206, all connected via a bus 201, and a program that controls these hardware resources. Furthermore, the communication terminal 2 may include a display device 207 connected via the bus 201.

[0075] The main memory 203 contains pre-stored programs for the processor 202 to perform various controls and calculations. The processor 202 and the main memory 203 work together to realize the various functions of the communication terminal 2, such as the acquisition unit 21, generation unit 22, and communication unit 23 shown in Figure 2.

[0076] The auxiliary storage device 205 has a program storage area for storing the arithmetic program of the learned generator 131' executed by the communication terminal 2. It also has a program storage area for storing the communication program executed by the communication terminal 2. The auxiliary storage device 205 realizes the third storage unit 20 described in Figure 2. Furthermore, it may have, for example, a backup area for backing up the aforementioned data and programs.

[0077] [Operation of the communication management device] Next, the operation of the communication management device 1 having the above-described configuration will be explained with reference to the sequence diagrams in Figures 10 and 11, and the flowchart in Figure 12.

[0078] Figure 10 shows a sequence illustrating the grouping of communication terminals 2 by the communication management system and the setting of communication paths for data communication. First, the communication management device 1 transmits grouping information, which includes the group ID and IMSI, to the UDM / UDR 41 (step S100). The grouping information is stored in the communication management device 1 in advance. Next, the UDM / UDR 41 creates a configuration information table T2 (Figure 4) based on the received grouping information (step S101). In step S101, the values ​​for "group ID" and "IMSI" in the configuration information table T2 are registered, while the values ​​for the other items are not yet entered.

[0079] Subsequently, the communication terminal 2 located in communication area 3A transmits a location registration request signal from the base station 3 to the core network 4 (step S102). The location registration request signal includes the IMSI of the communication terminal 2 and the base station ID of the base station 3 where it is located. When the UDM / UDR 41 receives the location registration request signal via the AMF 40, the UDM / UDR 41 associates the group ID with the base station ID and IMSI included in the received location registration request signal using the information in the configuration information table T2 (Figure 4) created in step S101, and then forwards the location registration request signal to the communication management device 1 (step S103). Note that in step S103, the value of the base station ID in the configuration information table T2 is set.

[0080] Next, the setting unit 10 of the communication management device 1 requests the PCF43 to create a PCC policy by specifying the IMSI, base station ID, group ID, and its own IP address included in the location registration request signal received in step S103 (step S104). Subsequently, the PCF43 creates a PCC policy based on the specified requirements (step S105). The PCF43 creates a PCC policy that includes information on which UPF44 the data communication from the communication terminal 2 will pass through. The PCC policy specifies the optimal UPF44 as the communication path for data communication from the communication terminal 2 to the data network 5.

[0081] Next, PCF43 sends the created PCC policy to SMF42 (step S106). Then, SMF42 sends data communication path configuration information related to user plane functions such as setting communication paths defined in the PCC policy to UPF44, which is specified by the PCC policy (step S107). Next, UPF44 registers the received data communication path configuration information in memory (step S108).

[0082] Next, UPF44 sends an ACK to SMF42, notifying it of its IP address (step S109). Furthermore, SMF42 sends an ACK to the communication management device 1, notifying it of UPF44's IP address (step S110). Subsequently, the communication management device 1 notifies UDM / UDR41 of UPF44's IP address and also sends an ACK to communication terminal 2 (step S111). At this time, the IP address of UPF44 communicating with communication terminal 2 is set in the configuration information table T2. The configuration information table T2 updated by UDM / UDR41 is also stored in the first storage unit 14 of the communication management device 1. After that, a data communication path is established between communication terminal 2 and UPF44 (step S112). Communication terminal 2 then performs data communication with the data network 5.

[0083] Subsequently, UPF44 records logs of the communication volume for each time period of multiple communication terminals 2 (step S113). The communication volume logs recorded by UPF44 are time-series data of the communication volume recorded for each IMSI of the group ID (t i ,x i ) Furthermore, the communication management device 1 then performs communication management processing (step S114).

[0084] Next, referring to the sequence in Figure 11, the communication management process (step S114) described in Figure 10 will be explained in more detail. First, the collection unit 11 of the communication management device collects time-series data of communication volume from UPF44 for IMSIs (communication terminals 2) with a common group ID (step S1). The collection unit 11 refers to the setting information table T2 stored in the first storage unit 14 and, for example, identifies the IP address "IP11" of the UPF44 that IMSI_1 to IMSI_N of group ID "1" communicate with, and collects time-series data of communication volume from the UPF44 of the identified IP address "IP11" (table T1 in Figure 3).

[0085] Next, the selection unit 12 of the communication management device 1 selects the time-series data of the communication terminal 2 with the largest total communication volume for each time period, based on the time-series data of the collected communication volume (step S2). In step S2, the selection unit 12 selects the time-series data of the communication volume for each time period (t i ,x i Based on this, for each IMSI with a common group ID, for each time period (t1~t N The sum of the data usage (Σx i Select the time-series data of communication terminal 2, which has the largest ) and the best throughput.

[0086] In step S2, the time-series data of the N data volumes with the highest throughput are obtained. For example, the total data volumes of all IMSIs can be compared, and the N data points with the largest values ​​can be collected. The time-series data of the data volumes selected in step S2 are used as training data in adversarial learning, i.e., time-series data of true data volumes.

[0087] Next, the learning unit 13 of the communication management device 1 performs a learning process for a GAN comprising a generator 131 and a discriminator 132 (step S3). More specifically, in step S3, the learning unit 13 performs adversarial learning of a generative model having a generator 131 that generates pseudo-communication volume time series data similar to the true communication volume time series data, using the time series data of the communication terminal 2 with the largest total communication volume per time period, selected in step S2, as the true communication volume time series data, and a discriminator 132 that distinguishes between the pseudo-communication volume time series data generated by the generator 131 and the true communication volume time series data (step S3).

[0088] Figure 12 is a flowchart illustrating the GAN learning process performed by the learning unit 13 of the communication management device 1. First, the learning unit 13 acquires true communication volume time series data to be used as training data (step S50). In step S50, a set of N true communication volume time series data is acquired. The true communication volume time series data represents the timing of data transmission from the communication terminal 2 with high throughput among multiple communication terminals 2 that share a common group ID.

[0089] Here, as shown in the GAN learning process by the learning unit 13 in Figure 5, the true traffic volume time series data acquired in step S50 is used as training data 134 to be input when learning the classifier 132.

[0090] Next, the learning unit 13 inputs the true traffic volume time series data as training data 134 into the classifier 132, and sets the parameter w of the classifier 132 so that it can distinguish the true traffic volume time series data from the true traffic volume time series data (y=1). D ,θ D The system learns and updates the data (step S51). In step S51, the learning unit 13 can train the classifier 132 on true traffic time series data using, for example, the backpropagation method. Step S51 pre-constructs a classifier 132 that can distinguish true traffic time series data from true traffic time series data.

[0091] Next, the learning unit 13 generates Gaussian noise 130 and provides a random vector of the generated Gaussian noise 130 as input to the generator 131 (step S52). Subsequently, the generator 131 uses the input z and weight parameter w based on the given Gaussian noise 130. G ,θ G A sum-of-products operation and threshold processing using an activation function are performed to generate time-series data G(z) of pseudo-communication volume (step S53).

[0092] Next, we train the classifier 132. The training of the classifier 132 involves the parameters w of the generator 131. D ,θ D This is done with the parameters fixed. First, the learning unit 13 provides the training data 134 of the true traffic volume time series data collected in step S50 as input to the classifier 132, and uses backpropagation or the like to maximize the objective function E in equation (2) above, thereby controlling the parameters w D ,θ D Update (step S54). Note that the label for training data 134 is set to 1 (time series data of true traffic volume).

[0093] Next, the learning unit 13 provides the time-series data of the pseudo-communication volume generated by the generator 131 in step S53 as input to the discriminator 132, and uses methods such as backpropagation to optimize the parameters w so that the objective function E in equation (2) above is maximized. D ,θ D Update (step S55). That is, in steps S54 and S55, in order to maximize the objective function E in equation (2) above, the first term is D(w D ,θ D )=1 is output, and the second term is D(G(w G ,θ G ),w D ,θ D The optimization is performed so that ) = 0. Note that the training data 134 has a label of 0 (time series data of pseudo-communication volume).

[0094] The training of the classifier 132 in steps S54 and S55 corresponds to the dashed arrow ("training") which indicates that the classifier error is calculated in block 135 of the objective function E based on the output 133 from the classifier 132, and then the error is backpropagated to the classifier 132, as shown in Figure 5.

[0095] Next, the generator 131 is trained. In the training of the generator 131, the parameters of the discriminator 132 are fixed. The learning unit 13 trains the generator 131 so that time-series data of pseudo-communication volume is generated when random Gaussian noise 130 is given to the generator 131. Specifically, the learning unit 13 calculates the error in order to minimize the objective function E in equation (2) above, and adjusts the parameters w using methods such as backpropagation. G ,θ G Update (step S56).

[0096] The learning in step S56 corresponds to the dashed arrow in Figure 5, which indicates that the error is backpropagated to the generator 131. In other words, in step S56, the time-series data of the pseudo-communication volume generated by the generator 131 in Figure 5 is input to the discriminator 132, the generator error is calculated from its output 133 in block 135 of the objective function E, and then the dashed arrow ("training") indicates that the error is backpropagated to the generator 131.

[0097] Subsequently, the learning of the classifier 132 and generator 131 from steps S53 to S56 is repeated until the value of the objective function E reaches a Nash equilibrium and converges (step S57: NO). On the other hand, if the value of the objective function E converges (step S57: YES), the processing from steps S51 to S57 is repeated until the learning of the generator 131 and classifier 132 is performed using the remaining N-1 time series data of true communication volumes out of the N time series data of true communication volumes (step S58: NO).

[0098] Subsequently, if the generator 131 and discriminator 132 are trained using the remaining N-1 true communication volume time series data (step S58: YES), the learning unit 13 stores the trained generator 131' in the second storage unit 15 (step S4). The trained generator 131' is constructed through the above process.

[0099] Next, returning to Figure 11, the notification unit 16 notifies multiple communication terminals 2 having the same group ID via the network NW of the trained generator 131' constructed for each group ID as communication management information (step S5). Subsequently, the acquisition unit 21 of the communication terminal 2 acquires the trained generator 131' (step S6). Next, the generation unit 22 of the communication terminal 2 uses the trained generator 131' to generate pseudo-communication volume time-series data similar to the time-series data of true communication volume (step S7). Subsequently, the communication unit 23 of the communication terminal 2 controls the timing of data transmission of its own terminal based on the pseudo-communication volume time-series data generated in step S7 and transmits the data (step S8).

[0100] Similarly, steps S6 to S8 are performed for multiple other managed communication terminals 2 (IMSI_2 to IMSI_N) belonging to the same group ID (group 1). Furthermore, steps S1 to S8 are also performed for communication management processes for multiple communication terminals 2 belonging to different group IDs.

[0101] As described above, the communication management device 1 according to this embodiment selects time-series data of the communication volume of a high-throughput communication terminal 2 as true data in adversarial learning, and learns a generator 131 that generates pseudo-communication volume time-series data similar to the data transmission timing of the high-throughput communication terminal 2 through GAN learning. Then, the learned generator 131' is notified to multiple communication terminals 2 as communication management information. As a result, the throughput of each of the multiple communication terminals 2 under management can be improved.

[0102] Furthermore, according to the communication management device 1 of this embodiment, multiple communication terminals 2 to be managed are grouped in advance, and the optimal data transmission timing is learned for each group. Therefore, it is possible to improve the throughput of each of the communication terminals 2 that need to perform data communication even during times when communication bandwidth is congested.

[0103] In the above embodiment, an example was given of a communication management system compliant with 5G, but a communication management system compliant with 4G / LTE, 6G, etc., may also be used.

[0104] Furthermore, while the embodiment described above focuses on the use of GANs as the generative model, the generative model is not limited to GANs. For example, VAEs (Variational Autoencoders), Autoregressive Models, Energy-Based Models, Transformer-Based Generative Models, etc., can be used.

[0105] While embodiments of the communication management device, communication management method, and communication management system of the present invention have been described above, the present invention is not limited to the embodiments described above, and various modifications that a person skilled in the art can envision are possible within the scope of the invention described in the claims. [Explanation of Symbols]

[0106] 1...Communication management device, 10...Setting unit, 11...Collection unit, 12...Selection unit, 13...Learning unit, 14...First memory unit, 15...Second memory unit, 16...Notification unit, 2...Communication terminal, 2a...SIM, 3...Base station, 4...Core network, 40...AMF, 41...UDM / UDR, 42...SMF, 43...PCF, 44...UPF, 5...Data network, 101, 201...Bus, 102, 202...Processor, 103, 203... Main memory; 41a, 44a, 104, 204... Communication interface; 105, 205... Secondary memory; 106, 206... Input / Output I / O; 107, 207... Display device; 130... Noise; 131... Generator; 132... Discriminator; 131'... Trained generator; 133... Output; 134... Training data; 135... Block of objective function E; T1... Table; NW... Network.

Claims

1. A collection unit configured to collect time-series data of communication volume for each time period from multiple communication terminals, each identified by a subscriber identifier, A selection unit configured to select the time-series data of the communication terminal with the largest total communication volume for each time period, based on the time-series data of the collected communication volume, A learning unit configured to perform adversarial learning of a generative model having: a generator that generates pseudo-data series data similar to the true data series data, using the time series data of the data volume of the selected communication terminal with the largest total data volume for each time period as the true data volume time series data; and a discriminator that distinguishes between the pseudo-data volume time series data generated by the generator and the true data volume time series data. A notification unit configured to notify the plurality of communication terminals of the learned generator constructed by the learning unit as communication management information, A communication management device equipped with the following features.

2. In the communication management device according to claim 1, The aforementioned plurality of communication terminals are a plurality of communication terminals to which common identification information has been assigned. A communication management device characterized by the following features.

3. In the communication management device according to claim 2, The collection unit collects time-series data of the communication volume for each time period from the user plane function of the core network, for the multiple communication terminals to which the common identification information has been assigned. A communication management device characterized by the following features.

4. In the communication management device according to claim 1, The aforementioned multiple communication terminals are multiple communication terminals located within the same communication area. A communication management device characterized by the following features.

5. In the communication management device according to claim 1, The time-series data of the simulated traffic volume has a data distribution that minimizes the statistical distance from the data distribution of the time-series data of the true traffic volume. A communication management device characterized by the following features.

6. A collection step involves collecting time-series data on the amount of data transmitted over time periods from multiple communication terminals, each identified by a subscriber identifier. A selection step involves selecting the time-series data of the communication terminal with the largest total communication volume for each time period, based on the time-series data of the collected communication volume. A learning step in which a generative model is performed, having a generator that generates pseudo-data series data similar to the time series data of the data volume of a selected communication terminal with the largest total data volume for each time period, using the time series data of the data volume of the selected communication terminal as the true data volume time series data, and a discriminator that distinguishes between the pseudo-data volume time series data generated by the generator and the true data volume time series data, A notification step in which the trained generator constructed in the learning step is notified to the multiple communication terminals as communication management information; A communication management method comprising the following features.

7. In the communication management method described in claim 6, The aforementioned plurality of communication terminals are a plurality of communication terminals to which common identification information has been assigned. A communication management method characterized by the following features.

8. In the communication management method described in claim 7, The collection step involves collecting time-series data of the communication volume for each time period from the user plane function of the core network, for the multiple communication terminals to which the common identification information has been assigned. A communication management method characterized by the following features.

9. In the communication management method described in claim 6, The aforementioned multiple communication terminals are multiple communication terminals located within the same communication area. A communication management method characterized by the following features.

10. In the communication management method described in claim 6, Furthermore, each of the aforementioned communication terminals, An acquisition step to acquire the learned generator of the communication management information notified in the notification step, A generation step of generating pseudo-communication volume time series data similar to the true communication volume time series data using the trained generator, A communication step that controls the transmission of data at each time period based on the generated pseudo-communication volume time-series data. A communication management method comprising the following features.

11. A communication management device according to any one of claims 1 to 5, The aforementioned multiple communication terminals and In a communication management system equipped with, Each of the aforementioned communication terminals is An acquisition unit configured to acquire the learned generator of the communication management information notified from the communication management device, A generation unit configured to generate pseudo-communication volume time-series data similar to the true communication volume time-series data using the aforementioned trained generator, A communication unit configured to control the transmission of data during each time period based on the generated pseudo-communication volume time-series data, A communication management system equipped with the following features.

Images