WO2024018586A1 - System for predicting terminal communication quality - Google Patents

System for predicting terminal communication quality Download PDF

Info

Publication number
WO2024018586A1
WO2024018586A1 PCT/JP2022/028329 JP2022028329W WO2024018586A1 WO 2024018586 A1 WO2024018586 A1 WO 2024018586A1 JP 2022028329 W JP2022028329 W JP 2022028329W WO 2024018586 A1 WO2024018586 A1 WO 2024018586A1
Authority
WO
WIPO (PCT)
Prior art keywords
terminal
information
unmanaged
management
communication quality
Prior art date
Application number
PCT/JP2022/028329
Other languages
French (fr)
Japanese (ja)
Inventor
央也 小野
裕希 坂上
友宏 谷口
Original Assignee
日本電信電話株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電信電話株式会社 filed Critical 日本電信電話株式会社
Priority to PCT/JP2022/028329 priority Critical patent/WO2024018586A1/en
Publication of WO2024018586A1 publication Critical patent/WO2024018586A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements

Definitions

  • the present disclosure relates to a system for predicting communication quality of a terminal.
  • wireless communication lines When using network services, users use communication lines provided by communication carriers. For example, in the case of wireless communication lines, there are standards such as 3GPP 5G NR, LTE, IEEE 802.11 (wireless LAN), and IEEE 802.16 (WiMAX).
  • 3GPP 5G NR Long Term Evolution
  • LTE Long Term Evolution
  • IEEE 802.11 wireless LAN
  • WiMAX IEEE 802.16
  • User terminals can communicate using multiple communication standards. For example, a smartphone can select whether to use LTE, wireless LAN, or Bluetooth. It is also possible to use different access means of different carriers that use the same communication standard. Each of these access methods has different communication quality such as bandwidth and delay, so it is possible to maximize the user's quality of experience (QoE) by appropriately using them depending on the purpose. Become.
  • the access means determination function used by the terminal does not necessarily need to be present in the terminal itself, and can be held by a router or server on the network, or by an external terminal.
  • a message method called ANDSF Access Network Discovery and Selection Function
  • 3GPP 3GPP
  • this type of external control method optimization is performed using the combination of connection destinations of a group of terminals within the control area as a variable, thereby suppressing quality deterioration due to competition in access methods between terminals, and improving quality. This makes it possible to control connection destinations, which leads to improved fairness.
  • Non-Patent Documents 1 to 3 There are several techniques for predicting the communication quality of each terminal using the connection pattern between the terminal and the network as input (see, for example, Non-Patent Documents 1 to 3).
  • the first method is to use a mathematical model.
  • Throughput and the like are analytically predicted from the behavior of the communication protocol based on the SNR (Signal-Noise Ratio), the allocated frequency width, the number of devices connected to the base station device, and the like.
  • the second method is to use network simulation.
  • [Non-Patent Document 3] predicts communication quality by simulating application communication from terminals and servers and simulating the behavior of each packet within a virtual network topology.
  • the third method is based on machine learning regression.
  • Patent Document 4 Patent Document 1] a prediction model of communication quality is learned from simulation results and communication results at an actual terminal, and communication quality for an unknown connection pattern is predicted.
  • the connected control device When using these quality prediction methods, the connected control device performs prediction after collecting information necessary for prediction from terminals and network devices in advance through TCP/IP communication. For example, when making predictions using a mathematical model, it is possible to estimate the upper bound of the realized throughput from Shannon's equation by collecting information on the radio field strength that each terminal receives from each wireless base station. At this time, it is expected that the more detailed the terminal information and network information within the system is acquired, the more accurately the communication results at the actual terminals can be predicted.
  • the destination control device When applying these technologies, it is ideal for the destination control device to manage information on all terminals within the system, but in an actual communication environment, terminals that are not under the control of the destination control device, Using the same radio frequency band may affect the quality of managed terminals. At this time, the connection destination control device cannot grasp the number and location of terminals that are not under management, or the traffic patterns generated, at least using the same method as the terminals under management.
  • the quality prediction method described above is based on the assumption that all terminals in the system are under management, and if interference with unmanaged terminals occurs, the prediction accuracy will deteriorate. Although it is possible to obtain the traffic generated by unmanaged terminals using packet capture, etc., it is only possible to obtain throughput etc. at the moment of capture, and the degree of interference when changing the connection pattern of managed terminals cannot be determined. Expected to be different. For example, even if there is an unmanaged terminal that is generating 4 Mbps of traffic at a certain time based on packet capture, depending on the connection pattern of the management terminal, the unmanaged traffic may be about 3 Mbps depending on the degree of network congestion.
  • communication information at a certain moment through packet capture alone is insufficient as a method to obtain information on unmanaged terminals, and information on the maximum data rate at which traffic can be generated from which position on unmanaged terminals can be obtained. It is desirable to be able to do so.
  • 3GPP TS 24.312 V15.0.0 (2018-06), Access Network Discovery and Selection Function (ANDSF) Management t Object (MO), https://portal. 3gpp. org/desktopmodules/Specifications/SpecificationDetails. aspx? specificationId 1077 I. B. Dhia et al. , “Optimization of access points selection and resource allocation in heterogeneous wireless net ork,”2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communica tions (PIMRC), Montreal, QC, 2017, pp. 1-7. “ns-3
  • the present disclosure aims to make it possible to predict the communication quality obtained by a managed terminal even if there is an unmanaged terminal.
  • the system of the present disclosure includes a management terminal connectable to a network and a device of the present disclosure that acquires communication quality information from the management terminal, and is a system that predicts the communication quality of the management terminal.
  • the apparatus and method of the present disclosure include: Obtaining management terminal information including the geographical location of the management terminal and communication quality information obtained from the management terminal, predicting unmanaged terminal information including the geographical location of the unmanaged terminal based on the management terminal information and the communication quality information; The communication quality of the management terminal is predicted using the predicted unmanaged terminal information.
  • the device of the present disclosure may predict the unmanaged terminal information using an unmanaged terminal prediction model that has previously learned the communication quality obtained by the management terminal when the unmanaged terminal exists.
  • the management terminal information can be represented by a two-dimensional array or a three-dimensional array of pixels corresponding to the geographical location of the management terminal.
  • the unmanaged terminal information is represented by a two-dimensional array or three-dimensional array of pixels corresponding to the geographical location of the unmanaged terminal, and the pixel positions of the managed terminal information and the unmanaged terminal information are The positions may match.
  • the device of the present disclosure may predict the communication quality obtained by the management terminal for each network when the management terminal connects to the network, and determine the network to which the management terminal connects based on the prediction. .
  • the device of the present disclosure acquires the management terminal information and the communication quality information from a plurality of the management terminals, extracts a connection pattern when the plurality of management terminals connect to a network, and selects a connection pattern from among the extracted connection patterns.
  • An arbitrary connection pattern may be determined, and the plurality of management terminals may be connected to the network using the determined connection pattern.
  • the program of the present disclosure is a program for realizing a computer as each functional unit included in the device of the present disclosure, and is a program for causing the computer to execute each step of a method executed by the device of the present disclosure. .
  • the present disclosure even if an unmanaged terminal exists, the communication quality obtained by the managed terminal can be predicted. Therefore, even if there are unmanaged terminals, the present disclosure makes it possible to accurately predict the communication quality obtained by each management terminal in a virtual combination of connection destinations, and to select the optimal combination of connection destinations. It becomes possible.
  • An example of the network configuration of this embodiment is shown.
  • An example of the operation of the connected control device during learning is shown.
  • An example of the operation of the connection destination control device at the time of prediction is shown.
  • An example of the architecture of Conditional GAN is shown.
  • An example of the configuration of a terminal and a connected control device is shown.
  • An example of the learning flow of the unmanaged terminal prediction model is shown.
  • An example of the learning flow of the management terminal quality prediction model is shown.
  • An example of a prediction flow is shown.
  • An example of a dataset generation method is shown.
  • FIG. 1 shows an example of the network configuration of this embodiment.
  • an example is an environment in which the terminals 91#1 to 91#4 can connect to the Internet 93 via two types of networks (sometimes abbreviated as NW) 92#1 and 92#2. shows.
  • the system of the present disclosure includes a connection destination control device 94 that functions as a device according to the present disclosure.
  • the terminals 91#1 to 91#4 there are terminals 91#1 to 91#3 under the control of the destination control device 94 and a terminal 91#4 not under the control of the destination control device 94.
  • the terminals 91#1 to 91#3 are referred to as “management terminals", and the terminal 91#4 is referred to as an "unmanaged terminal".
  • the connection destination control device 94 controls the communication quality obtained by the management terminals 91#1 to 91#3 when the management terminals 91#1 to 91#3 connect to the network 92#1 or 92#2. 92#2, and based on the prediction, determine the network to which the management terminals 91#1 to 91#3 will connect, and connect the plurality of management terminals 91#1 to 91#3 to the network using the determined connection pattern. Connect to. At this time, the connection destination control device 94 lists the patterns in which each of the management terminals 91#1 to 91#3 connects to the network 92#1 or 92#2, and performs each management for any connection pattern extracted from the list. The communication quality achieved by terminals 91#1 to 91#3 is predicted.
  • connection destination control device 94 connects the terminals 91#1 and 91#2 to the network 92#1, connects the terminal 91#3 to the network 92#2, and connects the terminal to the first connection pattern.
  • a second connection pattern can be extracted in which terminal 91#1 is connected to network 92#1 and terminals 91#2 and 91#3 are connected to network 92#2.
  • the connection destination control device 94 derives the communication qualities ⁇ and ⁇ of the first and second connection patterns, respectively, and determines the optimal connection pattern based on the magnitude of the communication qualities ⁇ and ⁇ . Thereby, the system of this embodiment can optimize the connection pattern of each management terminal 91#1 to 91#3.
  • the connection destination control device 94 determines a connection pattern based on the following information. ⁇ Management terminal information of management terminals 91#1 to 91#3 ⁇ Communication quality information of management terminals 91#1 to 91#3 ⁇ Network information of networks 92#1 and 92#2
  • the management terminal information includes the geographical locations of the management terminals 91#1 to 91#3.
  • the management terminal information may include any information regarding the communication quality of the management terminals 91#1 to 91#3, such as data rate, traffic pattern, and relative coordinates with the connected base station.
  • the communication quality information is information regarding the communication quality of the management terminals 91#1 to 91#3, and includes, for example, upstream/downstream throughput, delay, jitter, etc. of the management terminals 91#1 to 91#3.
  • the network information includes the geographic locations of the base stations of networks 92#1 and 92#2.
  • the network information may include the frequency channel and bandwidth used by the base station.
  • the connection destination control device 94 further determines a connection pattern based on the unmanaged terminal information of the unmanaged terminal 91#4.
  • the unmanaged terminal information includes the position of the unmanaged terminal 91#4.
  • the unmanaged terminal information includes any information that affects the communication quality of the unmanaged terminal 91#4 to the managed terminals 91#1 to 91#3, such as data rate, traffic pattern, and relative coordinates with the connected base station. You can stay there.
  • unmanaged terminal information including the geographical location of unmanaged terminal 91#4 is collected by prediction, and the same type of parameters as when collecting management terminal information from management terminals 91#1 to 91#3 are used. obtain.
  • highly accurate quality prediction results for the management terminals 91#1 to 91#3 are obtained, taking into consideration the behavior of the unmanaged terminal 91#4.
  • FIG. 2 shows an example of the operation of the connection destination control device 94 during learning.
  • the connection destination control device 94 inputs certain management terminal information, certain network information, and communication quality information of the management terminal information realized under certain network information, and calculates information such as the geographical location and traffic pattern of the unmanaged terminal 91#4. Uses an unmanaged terminal prediction model that outputs information.
  • the management terminal information and network information used for input are represented by a two-dimensional array or a three-dimensional array of pixels in which geographical locations and pixel locations correspond to each other.
  • the number of arrays matches the number of input parameters used for prediction.
  • the output is represented by a two-dimensional or three-dimensional array of pixels in which the geographic location and pixel location correspond.
  • the number of arrays matches the predicted number of parameters of the unmanaged terminal 91#4.
  • the managed terminal information and the unmanaged terminal information match in pixel position and geographical position.
  • the two-dimensional array includes latitude and longitude dimensions.
  • the three-dimensional array includes latitude, longitude, and height dimensions.
  • a two-dimensional array or three-dimensional array of pixels may be prepared for each type of parameter. In this case, a multidimensional array of four or more dimensions may be used.
  • pixel positions will be represented as a two-dimensional matrix for easy understanding.
  • the management terminal information can be expressed by equation (1).
  • the usage frequency channel of the base station of network 92#1 located at the pixel position of the first row and first column is ⁇ 1
  • the usage frequency channel of the base station of network 92#2 located at the pixel position of the third row and fourth column can be expressed by equation (2).
  • the unmanaged terminal 91#4 with a data rate ⁇ 1 exists at the pixel position of the 1st row and 4th column
  • the unmanaged terminal 91 #4 with a data rate ⁇ 2 exists at the pixel position of the 1st row and 4th column. If it is found that the unmanaged terminal information exists at the pixel position of the fourth row and second column, the unmanaged terminal information can be expressed by equation (3).
  • FIG. 3 shows an example of the operation of the connection destination control device 94 at the time of prediction. Since the unmanaged terminal information obtained by the unmanaged terminal prediction model can be expressed using the same parameters as the managed terminal information, it can be treated in the same way as the managed terminal information, and various communication patterns and connection patterns can be expressed using an arbitrary method. It is possible to predict the communication quality of a terminal. Then, the prediction result becomes the communication quality with interference of the unmanaged terminal 91#4 taken into account.
  • conditional generative adversarial network which is a type of supervised machine learning model.
  • CVAE Conditional Variable Autoencoder
  • Conditional GAN will be described below as an example.
  • FIG. 4 shows an example of the architecture of Conditional GAN.
  • the Conditional GAN two learning devices, a generator 81 and a discriminator 82, coexist.
  • the generator 81 is a neural network that generates a multidimensional array of unmanaged terminal information from management terminal information, network information, and labels.
  • communication quality information obtained by the management terminals 91#1 to 91#3 is used as a label.
  • the discriminator 82 is a neural network that determines whether the input multidimensional array is true data derived from a dataset or false data generated by the generator 81.
  • an unmanaged terminal prediction model that generates unmanaged terminal information is learned so that the unmanaged terminal information generated by the generator 81 cannot be distinguished from real data.
  • the generator 81 can generate more accurate unmanaged terminal information.
  • the learning results of the generator 81 that are determined to be true data by the discriminator 82 are used as an unmanaged terminal prediction model.
  • the learned unmanaged terminal prediction model may be installed in the connection destination control device 94, the connection destination control device 94 itself may be provided with learning functions such as the generator 81 and the discriminator 82.
  • FIG. 5 shows a configuration example of the terminal 91 and the connection destination control device 94.
  • the terminal 91 includes a communication interface (IF) 11 , an application 12 , a location information acquisition section 13 , a terminal information acquisition/notification section 14 , and a communication result notification section 15 .
  • the connection destination control device 94 includes a communication interface (IF) 41, a communication information aggregation unit 42, a data set storage unit 43, an unmanaged terminal prediction model 44, a management terminal quality prediction model 45, a connection pattern optimization program 46, and connection destination control. 47.
  • IF communication interface
  • the terminal 91 and the connection destination control device 94 of the present disclosure can also be realized by a computer and a program, and the program can be recorded on a recording medium or provided through a network.
  • - Communication interface 11 An interface that communicates with the destination control device 94 directly or over a network. The same interface that communicates with the base station of network 92#1 or 92#2 may be used.
  • -Application 12 An arbitrary application provided in the terminal 91.
  • - Location information acquisition unit 13 Acquires the geographical position of the terminal 91 in coordinates using GPS or the like.
  • - Terminal information acquisition/notification unit 14 Acquires location information from the location information acquisition unit 13, acquires the average data rate generated by the application 12, and acquires management terminal information necessary for quality prediction of the management terminal.
  • - Communication result notification unit 15 Acquires communication quality information including what kind of communication quality was obtained (throughput and delay) as a result of the acquired management terminal information, and notifies it to the connection destination control device 94.
  • - Communication interface 41 An interface for communicating with the terminal 91. It may also include an interface for communicating with base stations of networks 92#1 and 92#2.
  • - Communication information aggregation unit 42 Processes the management terminal information and communication quality information acquired from the terminal group and stores it as a data set. At the time of prediction, the processed management terminal information and communication quality information are output to the unmanaged terminal prediction model 44 and the management terminal quality prediction model 45.
  • - Data set storage unit 43 Stores the data set used for learning of Conditional GAN, etc., and passes the data to the unmanaged terminal prediction model 44 during learning.
  • - Unmanaged terminal prediction model 44 Predicts unmanaged terminal information using a learning model such as Conditional GAN.
  • - Management terminal quality prediction model 45 Predicts the communication quality of management terminals 91#1 to 91#3 by inputting management terminal information and predicted unmanaged terminal information. At this time, the management terminal quality prediction model 45 may use network information in addition to the management terminal information and unmanaged terminal information.
  • - Connection pattern optimization program 46 Using the prediction results of the management terminal quality prediction model 45, discovers the optimal connection pattern for the management terminals 91#1 to 91#3. For example, the connection pattern optimization program 46 extracts connection patterns when each management terminal connects to each network 92#1 and 92#2, and determines an arbitrary connection pattern from among the extracted connection patterns.
  • - Connection destination control unit 47 Instructs the management terminals 91#1 to 91#3 to switch to the connection pattern determined by the connection pattern optimization program 46.
  • FIG. 6 shows an example of the learning flow of the unmanaged terminal prediction model 44.
  • the data set required for learning the unmanaged terminal prediction model 44 can be generated by, for example, a network simulator.
  • the network simulator simulates a virtual network in which multiple terminals exist (S11), and the communication information aggregation unit 42 of the connection destination control device 94 aggregates the simulation results (S12) and converts them into a multidimensional array format of data sets. After that (S13), it is stored in the data set storage unit 43 (S14).
  • the unmanaged terminal prediction model 44 learns unmanaged terminal information for each data set stored in the data set storage unit 43 (S15) (S16).
  • S15 data set storage unit 43
  • S16 When using a network simulator, it is possible to generate data on a plurality of unmanaged terminal information using one simulation data, and the generation method will be described later.
  • FIG. 7 shows an example of a learning flow of the management terminal quality prediction model 45.
  • the data set required for learning the management terminal quality prediction model 45 can be generated from the management terminal information and communication quality information of the actual terminal. Learning may be performed by regarding some of the management terminals 91#1 to 91#3 as unmanaged terminals.
  • the management terminals 91#1 to 91#3 communicate with the peripheral terminals at the same time (S21), and the connection destination control device 94 aggregates the management terminal information and communication quality information at that time (S22).
  • the connection destination control device 94 converts the aggregated information into a data set multidimensional array format (S23), and then stores it in the data set storage unit 43 (S24).
  • the management terminal quality prediction model 45 performs learning for each data set stored in the data set storage unit 43 (S25) (S26).
  • FIG. 8 shows an example of a prediction flow.
  • the management terminals 91#1 to 91#3 communicate with peripheral terminals at the same time (S31), and the communication information aggregation unit 42 aggregates the management terminal information and communication quality information at that time (S32).
  • the unmanaged terminal prediction model 44 inputs the management terminal information and communication quality information of the immediately preceding management terminals 91#1 to 91#3, and calculates unmanaged terminals such as the position of the unmanaged terminal 91#4 and the communication traffic pattern to be generated. Predict information (S33).
  • the output of the unmanaged terminal prediction model 44 becomes the input of the managed terminal quality prediction model 45.
  • the management terminal quality prediction model 45 combines the aggregation results of the communication information aggregation unit 42 and the prediction results of the unmanaged terminal prediction model 44 (S34), and uses the same method as the quality prediction of the management terminal when there is no unmanaged terminal,
  • the management terminal/unmanaged terminal is collectively regarded as a management terminal and quality prediction is performed (S35).
  • connection pattern optimization program 46 input data related to the management terminals 91#1 to 91#3 out of the output of the management terminal quality prediction model 45. This is because the connection pattern optimization program 46 can only control the network connection state of the management terminal.
  • FIG. 9 shows an example of a data set generation method.
  • a conceptual diagram of a method for generating a dataset from data that aggregates communication results from a network simulation or an actual terminal is shown.
  • the aggregated data is all arbitrarily set simulation parameters or management terminal information, but this data is divided into a pseudo multidimensional array consisting of management terminals and a multidimensional array consisting of unmanaged terminals,
  • the set of multidimensional arrays is regarded as one piece of data.
  • the multidimensional array of management terminal information is used as a label in the Conditional GAN, and that of unmanaged terminals is used as an input array to the discriminator 82.
  • the management terminal information is expressed by equation (4) and the communication quality information is expressed by equation (5)
  • the first data set expressed by equations (6) to (8) and the equation ( 9) to a second data set expressed by equation (11) are generated.
  • the first data set includes management terminal information represented by formula (6), unmanaged terminal information represented by formula (7), communication quality information of the management terminal represented by formula (8), including.
  • the second data set includes management terminal information represented by formula (9), unmanaged terminal information represented by formula (10), communication quality information of the management terminal represented by formula (11), including.
  • the present disclosure learns in advance unmanaged terminal information including the location of unmanaged terminal 91#4, and when predicting the communication quality of management terminals 91#1 to 91#3, management terminal information Obtain unmanaged terminal information with the same type of parameters as . As a result, the present disclosure can obtain highly accurate quality prediction results for the management terminals 91#1 to 91#3 in consideration of the behavior of the unmanaged terminal 91#4.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The purpose of the present disclosure is to make it possible to predict the communication quality that can be obtained at a managed terminal, even if there is a non-managed terminal.　The device and method of the present disclosure: acquire managed terminal information relating to a managed terminal and communication quality information that can be obtained at the managed terminal; predict non-managed terminal information on the basis of the managed terminal information and the communication quality information; and use the predicted non-managed terminal information to predict the communication quality for the managed terminal.

Description

端末の通信品質を予測するシステムSystem to predict terminal communication quality
　本開示は、端末の通信品質を予測するシステムに関する。 The present disclosure relates to a system for predicting communication quality of a terminal.
　ユーザはネットワークサービスを利用するとき、通信キャリアの提供する通信回線を利用する。例えば、無線通信回線の場合、３ＧＰＰ　５Ｇ　ＮＲやＬＴＥ、ＩＥＥＥ８０２．１１（無線ＬＡＮ）、ＩＥＥＥ　８０２．１６（ＷｉＭＡＸ）等の規格が存在する。 When using network services, users use communication lines provided by communication carriers. For example, in the case of wireless communication lines, there are standards such as 3GPP 5G NR, LTE, IEEE 802.11 (wireless LAN), and IEEE 802.16 (WiMAX).
　ユーザー端末は、複数の通信規格を使い分けて通信を行うことができる。例えば、スマートホンは、ＬＴＥと無線ＬＡＮ、Ｂｌｕｅｔｏｏｔｈのどれを利用するか選択することができる。また、同一の通信規格を利用した異種キャリアのアクセス手段を使い分けることも可能である。このような各アクセス手段は帯域や遅延等の通信品質が異なっているため、これらを用途に応じて適切に使い分けることでユーザの体感品質（ＱｏＥ：　Ｑｕａｌｉｔｙ　ｏｆ　Ｅｘｐｅｒｉｅｎｃｅ）を最大化することが可能になる。 User terminals can communicate using multiple communication standards. For example, a smartphone can select whether to use LTE, wireless LAN, or Bluetooth. It is also possible to use different access means of different carriers that use the same communication standard. Each of these access methods has different communication quality such as bandwidth and delay, so it is possible to maximize the user's quality of experience (QoE) by appropriately using them depending on the purpose. Become.
　また、端末の利用するアクセス手段決定機能は、必ずしも端末本体に存在する必要はなく、ネットワーク上のルータやサーバ、もしくは外部端末が保持することも可能である。実際、そのようなネットワーク上の装置から端末に対して接続先の指示を行う方式を想定したメッセージ方式　ＡＮＤＳＦ（Ａｃｃｅｓｓ　Ｎｅｔｗｏｒｋ　Ｄｉｓｃｏｖｅｒｙ　ａｎｄ　Ｓｅｌｅｃｔｉｏｎ　Ｆｕｎｃｔｉｏｎ）が、３ＧＰＰにて標準化されている（非特許文献１を参照）。このような外部制御の方式を適用することで、制御エリア内の端末群の接続先組み合わせを変数とした最適化が実行することにより、端末間のアクセス手段の競合などによる品質低下抑制や、品質の公平性向上につながる接続先制御が可能となる。 Furthermore, the access means determination function used by the terminal does not necessarily need to be present in the terminal itself, and can be held by a router or server on the network, or by an external terminal. In fact, a message method called ANDSF (Access Network Discovery and Selection Function), which assumes a method for instructing a terminal to connect from a device on such a network, has been standardized by 3GPP (see Non-Patent Document 1). reference). By applying this type of external control method, optimization is performed using the combination of connection destinations of a group of terminals within the control area as a variable, thereby suppressing quality deterioration due to competition in access methods between terminals, and improving quality. This makes it possible to control connection destinations, which leads to improved fairness.
　前記のように、ある装置が対象エリア内の端末群の接続先組み合わせを制御可能であるとき、その装置は最適な接続先組み合わせを導出する処理を必要とする。その際、解の候補となる様々な接続先組み合わせによって実現される通信品質を把握する必要があるが、通信品質の把握は実際の接続を伴わず、予測によって実現できることが望ましい。なぜなら、ある接続パターンでの通信品質を予測できると、切り替え処理に伴う一時的なＮＷ切断や低品質ネットワーク接続中の影響を排除できるとともに、接続パターンあたりの品質確認時間を抑制できるためである。 As mentioned above, when a device can control the combination of connection destinations of a group of terminals within a target area, that device requires processing to derive the optimal combination of connection destinations. In this case, it is necessary to understand the communication quality achieved by various connection destination combinations that are candidate solutions, but it is desirable that communication quality can be realized by prediction without involving actual connections. This is because if the communication quality for a certain connection pattern can be predicted, it is possible to eliminate the effects of temporary NW disconnection due to switching processing and low-quality network connection, and it is also possible to suppress the quality confirmation time per connection pattern.
　以上のことから、端末群の接続先組み合わせを最適化するためには、現在の接続状態以外のパターンも含めた仮想の接続先組み合わせに対して、各端末で得られる通信品質を予測する機能が必要となる。 Based on the above, in order to optimize the connection destination combinations of a group of terminals, it is necessary to have a function that predicts the communication quality obtained by each terminal for virtual connection destination combinations that include patterns other than the current connection state. It becomes necessary.
　端末とネットワーク間の接続パターンを入力として、各端末の通信品質を予測する技術はいくつか存在する（例えば、非特許文献１～３参照。）。
　一つ目に、数理モデルを用いる方法がある。［非特許文献２］では、ＳＮＲ（Ｓｉｇｎａｌ－Ｎｏｉｓｅ　Ｒａｔｉｏ）や割り当て周波数幅、その基地局装置への接続台数などを基に、通信プロトコルの振る舞いからスループットなどを解析的に予測している。
　二つ目に、ネットワークシミュレーションを用いる方法がある。［非特許文献３］では、端末やサーバーからのアプリケーション通信を模擬し、パケットごとの振る舞いを仮想のネットワークトポロジー内でシミュレーションすることで通信品質を予測している。
　三つめに、機械学習回帰による方法がある。［非特許文献４，特許文献１］では、シミュレーション結果や実端末での通信結果から通信品質の予測モデルを学習しておき、未知の接続パターンに対しての通信品質を予測している。 There are several techniques for predicting the communication quality of each terminal using the connection pattern between the terminal and the network as input (see, for example, Non-Patent Documents 1 to 3).
The first method is to use a mathematical model. In [Non-Patent Document 2], throughput and the like are analytically predicted from the behavior of the communication protocol based on the SNR (Signal-Noise Ratio), the allocated frequency width, the number of devices connected to the base station device, and the like.
The second method is to use network simulation. [Non-Patent Document 3] predicts communication quality by simulating application communication from terminals and servers and simulating the behavior of each packet within a virtual network topology.
The third method is based on machine learning regression. In [Non-Patent Document 4, Patent Document 1], a prediction model of communication quality is learned from simulation results and communication results at an actual terminal, and communication quality for an unknown connection pattern is predicted.
　これらの品質予測手法を用いる時、接続先制御装置は端末やネットワーク装置から予測に必要な情報をＴＣＰ／ＩＰ通信などによってあらかじめ収集したうえで予測を行う。例えば数理モデルによって予測する場合、各端末が各無線基地局から受信する電波強度情報を収集することで、シャノンの式から実現スループットの上界を推定するなどのことができる。この時、系内の端末情報やネットワーク情報をより詳細に取得するほど、現実の端末における通信結果を精度よく予測できることが期待される。 When using these quality prediction methods, the connected control device performs prediction after collecting information necessary for prediction from terminals and network devices in advance through TCP/IP communication. For example, when making predictions using a mathematical model, it is possible to estimate the upper bound of the realized throughput from Shannon's equation by collecting information on the radio field strength that each terminal receives from each wireless base station. At this time, it is expected that the more detailed the terminal information and network information within the system is acquired, the more accurately the communication results at the actual terminals can be predicted.
　これらの技術を適用するとき、接続先制御装置が系内の全ての端末の情報を管理することが理想的であるが、実際の通信環境では、接続先制御装置の管理下にない端末が、同一無線周波数帯を利用するなどして、管理下の端末の品質に影響を与えることがある。このとき接続先制御装置は、少なくとも管理下端末と同じ手法では、管理下にない端末の台数や位置、発生させるトラヒックパターンを把握することができない。 When applying these technologies, it is ideal for the destination control device to manage information on all terminals within the system, but in an actual communication environment, terminals that are not under the control of the destination control device, Using the same radio frequency band may affect the quality of managed terminals. At this time, the connection destination control device cannot grasp the number and location of terminals that are not under management, or the traffic patterns generated, at least using the same method as the terminals under management.
　前述の品質予測手法は、系内の端末がすべて管理下にある状態を前提としており、管理外端末との干渉が発生する場合、その予測精度が劣化する。管理外端末が発生させているトラヒックはパケットキャプチャなどによって取得することも可能だが、あくまで取得した瞬間のスループットなどが取得できるのみであり、管理端末の接続パターンを変化させた場合の干渉の程度は異なることが予想される。例えば、パケットキャプチャによってある時刻に４Ｍｂｐｓのトラヒックを発生させている管理外端末がいたとしても、管理端末の接続パターンによっては、ネットワークの混雑具合に応じ３Ｍｂｐｓ程度の管理外トラヒックになる場合がある。そのため、パケットキャプチャによるある瞬間の通信情報のみでは管理外端末の情報取得手法としては不十分であり、管理外端末がどの位置から最大のどれほどのデータレートでトラヒックを発生させ得るかの情報を取得できることが望ましい。 The quality prediction method described above is based on the assumption that all terminals in the system are under management, and if interference with unmanaged terminals occurs, the prediction accuracy will deteriorate. Although it is possible to obtain the traffic generated by unmanaged terminals using packet capture, etc., it is only possible to obtain throughput etc. at the moment of capture, and the degree of interference when changing the connection pattern of managed terminals cannot be determined. Expected to be different. For example, even if there is an unmanaged terminal that is generating 4 Mbps of traffic at a certain time based on packet capture, depending on the connection pattern of the management terminal, the unmanaged traffic may be about 3 Mbps depending on the degree of network congestion. Therefore, communication information at a certain moment through packet capture alone is insufficient as a method to obtain information on unmanaged terminals, and information on the maximum data rate at which traffic can be generated from which position on unmanaged terminals can be obtained. It is desirable to be able to do so.
ＷＯ２０２２／０３８７６０号公報WO2022/038760 publication
　本開示は、管理外端末が存在する場合であっても、管理端末で得られる通信品質を予測可能にすることを目的とする。 The present disclosure aims to make it possible to predict the communication quality obtained by a managed terminal even if there is an unmanaged terminal.
　本開示のシステムは、ネットワークに接続可能な管理端末と、前記管理端末から通信品質情報を取得する、本開示の装置と、を備え、前記管理端末の通信品質を予測するシステムである。 The system of the present disclosure includes a management terminal connectable to a network and a device of the present disclosure that acquires communication quality information from the management terminal, and is a system that predicts the communication quality of the management terminal.
　本開示の装置及び方法は、
　管理端末の地理的位置を含む管理端末情報と当該管理端末で得られた通信品質情報とを取得し、
　前記管理端末情報及び前記通信品質情報に基づいて、管理外端末の地理的位置を含む管理外端末情報を予測し、
　予測した前記管理外端末情報を用いて、前記管理端末の通信品質を予測する。 The apparatus and method of the present disclosure include:
Obtaining management terminal information including the geographical location of the management terminal and communication quality information obtained from the management terminal,
predicting unmanaged terminal information including the geographical location of the unmanaged terminal based on the management terminal information and the communication quality information;
The communication quality of the management terminal is predicted using the predicted unmanaged terminal information.
　本開示の装置は、前記管理外端末が存在する場合に前記管理端末が得られる通信品質を予め学習した管理外端末予測モデルを用いて、前記管理外端末情報を予測してもよい。 The device of the present disclosure may predict the unmanaged terminal information using an unmanaged terminal prediction model that has previously learned the communication quality obtained by the management terminal when the unmanaged terminal exists.
　前記管理端末情報は、前記管理端末の地理的位置に対応する２次元配列又は３次元配列のピクセルで表わすことができる。この場合、前記管理外端末情報は、前記管理外端末の地理的位置に対応する２次元配列又は３次元配列のピクセルで表され、前記管理端末情報及び前記管理外端末情報のピクセル位置と地理的位置が一致してもよい。 The management terminal information can be represented by a two-dimensional array or a three-dimensional array of pixels corresponding to the geographical location of the management terminal. In this case, the unmanaged terminal information is represented by a two-dimensional array or three-dimensional array of pixels corresponding to the geographical location of the unmanaged terminal, and the pixel positions of the managed terminal information and the unmanaged terminal information are The positions may match.
　本開示の装置は、前記管理端末がネットワークに接続した場合に前記管理端末で得られる通信品質をネットワークごとに予測し、前記予測に基づいて、前記管理端末が接続するネットワークを決定してもよい。 The device of the present disclosure may predict the communication quality obtained by the management terminal for each network when the management terminal connects to the network, and determine the network to which the management terminal connects based on the prediction. .
　本開示の装置は、複数の前記管理端末から前記管理端末情報及び前記通信品質情報を取得し、複数の前記管理端末がネットワークに接続した場合の接続パターンを抽出し、抽出した接続パターンのなかから任意の接続パターンを決定し、決定した接続パターンで、複数の前記管理端末をネットワークに接続させてもよい。 The device of the present disclosure acquires the management terminal information and the communication quality information from a plurality of the management terminals, extracts a connection pattern when the plurality of management terminals connect to a network, and selects a connection pattern from among the extracted connection patterns. An arbitrary connection pattern may be determined, and the plurality of management terminals may be connected to the network using the determined connection pattern.
　本開示のプログラムは、本開示に係る装置に備わる各機能部としてコンピュータを実現させるためのプログラムであり、本開示に係る装置が実行する方法に備わる各ステップをコンピュータに実行させるためのプログラムである。 The program of the present disclosure is a program for realizing a computer as each functional unit included in the device of the present disclosure, and is a program for causing the computer to execute each step of a method executed by the device of the present disclosure. .
　なお、上記各開示は、可能な限り組み合わせることができる。 Note that the above disclosures can be combined as much as possible.
　本開示によれば、管理外端末が存在する場合であっても、管理端末で得られる通信品質を予測できる。このため、本開示は、管理外端末が存在する場合であっても、仮想の接続先組み合わせにおいて各管理端末で得られる通信品質を精度良く予測することができ、最適な接続先組み合わせの選択が可能となる。 According to the present disclosure, even if an unmanaged terminal exists, the communication quality obtained by the managed terminal can be predicted. Therefore, even if there are unmanaged terminals, the present disclosure makes it possible to accurately predict the communication quality obtained by each management terminal in a virtual combination of connection destinations, and to select the optimal combination of connection destinations. It becomes possible.
本実施形態のネットワーク構成の一例を示す。An example of the network configuration of this embodiment is shown. 接続先制御装置の学習時の動作の一例を示す。An example of the operation of the connected control device during learning is shown. 接続先制御装置の予測時の動作の一例を示す。An example of the operation of the connection destination control device at the time of prediction is shown. Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮのアーキテクチャの一例を示す。An example of the architecture of Conditional GAN is shown. 端末及び接続先制御装置の構成例を示す。An example of the configuration of a terminal and a connected control device is shown. 管理外端末予測モデルの学習フローの一例を示す。An example of the learning flow of the unmanaged terminal prediction model is shown. 管理端末品質予測モデルの学習フローの一例を示す。An example of the learning flow of the management terminal quality prediction model is shown. 予測フローの一例を示す。An example of a prediction flow is shown. データセット生成方法の一例を示す。An example of a dataset generation method is shown.
　以下、本開示の実施形態について、図面を参照しながら詳細に説明する。なお、本開示は、以下に示す実施形態に限定されるものではない。これらの実施の例は例示に過ぎず、本開示は当業者の知識に基づいて種々の変更、改良を施した形態で実施することができる。なお、本明細書及び図面において符号が同じ構成要素は、相互に同一のものを示すものとする。 Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the present disclosure is not limited to the embodiments shown below. These implementation examples are merely illustrative, and the present disclosure can be implemented with various changes and improvements based on the knowledge of those skilled in the art. Note that components with the same reference numerals in this specification and the drawings indicate the same components.
（本開示の概要）
　図１に、本実施形態のネットワーク構成の一例を示す。本実施形態では、端末９１＃１～９１＃４がネットワーク（ＮＷと略記する場合がある。）９２＃１及び９２＃２の２種類のネットワークを介してインターネット９３に接続可能な環境である例を示す。本開示のシステムは、本開示に係る装置として機能する接続先制御装置９４を備える。端末９１＃１～９１＃４のなかには、接続先制御装置９４の管理下の端末９１＃１～９１＃３と、接続先制御装置９４の管理外の端末９１＃４と、が存在する。本開示では、端末９１＃１～９１＃３を「管理端末」と称し、端末９１＃４を「管理外端末」と称する。 (Summary of this disclosure)
FIG. 1 shows an example of the network configuration of this embodiment. In this embodiment, an example is an environment in which the terminals 91#1 to 91#4 can connect to the Internet 93 via two types of networks (sometimes abbreviated as NW) 92#1 and 92#2. shows. The system of the present disclosure includes a connection destination control device 94 that functions as a device according to the present disclosure. Among the terminals 91#1 to 91#4, there are terminals 91#1 to 91#3 under the control of the destination control device 94 and a terminal 91#4 not under the control of the destination control device 94. In this disclosure, the terminals 91#1 to 91#3 are referred to as "management terminals", and the terminal 91#4 is referred to as an "unmanaged terminal".
　接続先制御装置９４は、管理端末９１＃１～９１＃３がネットワーク９２＃１又は９２＃２に接続した場合に管理端末９１＃１～９１＃３で得られる通信品質をネットワーク９２＃１及び９２＃２ごとに予測し、前記予測に基づいて、管理端末９１＃１～９１＃３が接続するネットワークを決定し、決定した接続パターンで、複数の管理端末９１＃１～９１＃３をネットワークに接続させる。このとき、接続先制御装置９４は、各管理端末９１＃１～９１＃３がネットワーク９２＃１又は９２＃２に接続するパターンをリストアップし、そのなかから抽出した任意の接続パターンについて各管理端末９１＃１～９１＃３で実現される通信品質を予測する。 The connection destination control device 94 controls the communication quality obtained by the management terminals 91#1 to 91#3 when the management terminals 91#1 to 91#3 connect to the network 92#1 or 92#2. 92#2, and based on the prediction, determine the network to which the management terminals 91#1 to 91#3 will connect, and connect the plurality of management terminals 91#1 to 91#3 to the network using the determined connection pattern. Connect to. At this time, the connection destination control device 94 lists the patterns in which each of the management terminals 91#1 to 91#3 connects to the network 92#1 or 92#2, and performs each management for any connection pattern extracted from the list. The communication quality achieved by terminals 91#1 to 91#3 is predicted.
　本実施形態の場合、接続先制御装置９４は、端末９１＃１及び９１＃２をネットワーク９２＃１に接続し、端末９１＃３をネットワーク９２＃２に接続した第１の接続パターンと、端末９１＃１をネットワーク９２＃１に接続し、端末９１＃２及び９１＃３をネットワーク９２＃２に接続した第２の接続パターンが抽出され得る。接続先制御装置９４は、第１及び第２の接続パターンの通信品質α及びβをそれぞれ導出し、通信品質α及びβの大小等に基づいて最適接続パターンを決定する。これにより、本実施形態のシステムは、各管理端末９１＃１～９１＃３の接続パターンを最適にすることができる。 In the case of this embodiment, the connection destination control device 94 connects the terminals 91#1 and 91#2 to the network 92#1, connects the terminal 91#3 to the network 92#2, and connects the terminal to the first connection pattern. A second connection pattern can be extracted in which terminal 91#1 is connected to network 92#1 and terminals 91#2 and 91#3 are connected to network 92#2. The connection destination control device 94 derives the communication qualities α and β of the first and second connection patterns, respectively, and determines the optimal connection pattern based on the magnitude of the communication qualities α and β. Thereby, the system of this embodiment can optimize the connection pattern of each management terminal 91#1 to 91#3.
　接続先制御装置９４は、以下の情報に基づいて、接続パターンを決定する。
　・管理端末９１＃１～９１＃３の管理端末情報
　・管理端末９１＃１～９１＃３の通信品質情報
　・ネットワーク９２＃１及び９２＃２のネットワーク情報 The connection destination control device 94 determines a connection pattern based on the following information.
・Management terminal information of management terminals 91#1 to 91#3 ・Communication quality information of management terminals 91#1 to 91#3 ・Network information of networks 92#1 and 92#2
　管理端末情報は、管理端末９１＃１～９１＃３の地理的位置を含む。管理端末情報は、データレート、トラヒックパターン、接続基地局との相対座標など、管理端末９１＃１～９１＃３の通信品質に関する任意の情報を含んでいてもよい。
　通信品質情報は、管理端末９１＃１～９１＃３の通信品質に関する情報であり、例えば、管理端末９１＃１～９１＃３の上り・下りスループット、遅延、ジッタ等を含む。
　ネットワーク情報は、ネットワーク９２＃１及び９２＃２の基地局の地理的位置を含む。ネットワーク情報は、基地局の利用周波数チャネル、帯域幅を含んでいてもよい。 The management terminal information includes the geographical locations of the management terminals 91#1 to 91#3. The management terminal information may include any information regarding the communication quality of the management terminals 91#1 to 91#3, such as data rate, traffic pattern, and relative coordinates with the connected base station.
The communication quality information is information regarding the communication quality of the management terminals 91#1 to 91#3, and includes, for example, upstream/downstream throughput, delay, jitter, etc. of the management terminals 91#1 to 91#3.
The network information includes the geographic locations of the base stations of networks 92#1 and 92#2. The network information may include the frequency channel and bandwidth used by the base station.
　接続先制御装置９４は、さらに、管理外端末９１＃４の管理外端末情報に基づいて、接続パターンを決定する。管理外端末情報は、管理外端末９１＃４の位置を含む。管理外端末情報は、データレート、トラヒックパターン、接続基地局との相対座標など、管理外端末９１＃４が管理端末９１＃１～９１＃３の通信品質に影響を与える任意の情報を含んでいてもよい。 The connection destination control device 94 further determines a connection pattern based on the unmanaged terminal information of the unmanaged terminal 91#4. The unmanaged terminal information includes the position of the unmanaged terminal 91#4. The unmanaged terminal information includes any information that affects the communication quality of the unmanaged terminal 91#4 to the managed terminals 91#1 to 91#3, such as data rate, traffic pattern, and relative coordinates with the connected base station. You can stay there.
　管理外端末９１＃４は管理外であるため、管理外端末９１＃４から管理外端末情報を収集することはできない。そこで、本開示では、管理外端末９１＃４の地理的位置を含む管理外端末情報を予測によって収集し、管理端末９１＃１～９１＃３から管理端末情報を収集する場合と同種のパラメータを得る。これによって、管理外端末９１＃４の振る舞いを考慮した、管理端末９１＃１～９１＃３の高精度な品質予測結果を得る。 Since the unmanaged terminal 91#4 is unmanaged, unmanaged terminal information cannot be collected from the unmanaged terminal 91#4. Therefore, in the present disclosure, unmanaged terminal information including the geographical location of unmanaged terminal 91#4 is collected by prediction, and the same type of parameters as when collecting management terminal information from management terminals 91#1 to 91#3 are used. obtain. As a result, highly accurate quality prediction results for the management terminals 91#1 to 91#3 are obtained, taking into consideration the behavior of the unmanaged terminal 91#4.
　図２に、接続先制御装置９４の学習時の動作の一例を示す。接続先制御装置９４はある管理端末情報、あるネットワーク情報、あるネットワーク情報下において実現された管理端末情報の通信品質情報、を入力として、管理外端末９１＃４の地理的位置やトラヒックパターンなどの情報を出力する管理外端末予測モデルを用いる。 FIG. 2 shows an example of the operation of the connection destination control device 94 during learning. The connection destination control device 94 inputs certain management terminal information, certain network information, and communication quality information of the management terminal information realized under certain network information, and calculates information such as the geographical location and traffic pattern of the unmanaged terminal 91#4. Uses an unmanaged terminal prediction model that outputs information.
　入力に使用する管理端末情報とネットワーク情報は、地理的位置とピクセル位置が対応関係を持つ２次元配列もしくは３次元配列のピクセルで表す。配列の個数は、予測に用いる入力パラメータ数と一致する。
　出力は、地理的位置とピクセル位置が対応関係を持つ２次元配列もしくは３次元配列のピクセルで表す。配列の個数は、予測する管理外端末９１＃４のパラメータ数と一致する。
　管理端末情報及び管理外端末情報は、ピクセル位置及び地理的位置が一致する。２次元配列は、緯度及び経度の次元を含む。３次元配列は、緯度、経度及び高さの次元を含む。パラメータの種類ごとに２次元配列もしくは３次元配列のピクセルを用意してもよい。この場合、４次元以上の多次元配列を用いてもよい。以下、理解が容易になるよう、ピクセル位置を２次元行列で表して説明する。 The management terminal information and network information used for input are represented by a two-dimensional array or a three-dimensional array of pixels in which geographical locations and pixel locations correspond to each other. The number of arrays matches the number of input parameters used for prediction.
The output is represented by a two-dimensional or three-dimensional array of pixels in which the geographic location and pixel location correspond. The number of arrays matches the predicted number of parameters of the unmanaged terminal 91#4.
The managed terminal information and the unmanaged terminal information match in pixel position and geographical position. The two-dimensional array includes latitude and longitude dimensions. The three-dimensional array includes latitude, longitude, and height dimensions. A two-dimensional array or three-dimensional array of pixels may be prepared for each type of parameter. In this case, a multidimensional array of four or more dimensions may be used. In the following, pixel positions will be represented as a two-dimensional matrix for easy understanding.
　例えば、地理的位置を４×４の領域に分割した場合、第２行第２列のピクセル位置に位置する端末９１＃１の要求データレートがα１であり、第３行第３列のピクセル位置に位置する端末９１＃２の要求データレートがα２である場合、管理端末情報は式（１）で表すことができる。
Figure JPOXMLDOC01-appb-M000001
 For example, if a geographical location is divided into 4×4 areas, the requested data rate of terminal 91#1 located at the pixel position of the second row and second column is α 1, and the requested data rate of the terminal 91#1 located at the pixel position of the second row and second column is α 1 , When the requested data rate of the terminal 91#2 located at the position is α2 , the management terminal information can be expressed by equation (1).
Figure JPOXMLDOC01-appb-M000001
　第１行第１列のピクセル位置に位置するネットワーク９２＃１の基地局の利用周波数チャネルがβ１であり、第３行第４列のピクセル位置に位置するネットワーク９２＃２の基地局の利用周波数チャネルがβ２である場合、ネットワーク情報は式（２）で表すことができる。
Figure JPOXMLDOC01-appb-M000002
 The usage frequency channel of the base station of network 92#1 located at the pixel position of the first row and first column is β 1 , and the usage frequency channel of the base station of network 92#2 located at the pixel position of the third row and fourth column When the frequency channel is β 2 , the network information can be expressed by equation (2).
Figure JPOXMLDOC01-appb-M000002
　管理外端末予測モデルを用いた予測に基づいて、データレートγ１の管理外端末９１＃４が第１行第４列のピクセル位置に存在し、データレートγ２の管理外端末９１＃４が第４行第２列のピクセル位置に存在することが得られた場合、管理外端末情報は式（３）で表すことができる。
Figure JPOXMLDOC01-appb-M000003
 Based on the prediction using the unmanaged terminal prediction model, the unmanaged terminal 91#4 with a data rate γ 1 exists at the pixel position of the 1st row and 4th column, and the unmanaged terminal 91 #4 with a data rate γ 2 exists at the pixel position of the 1st row and 4th column. If it is found that the unmanaged terminal information exists at the pixel position of the fourth row and second column, the unmanaged terminal information can be expressed by equation (3).
Figure JPOXMLDOC01-appb-M000003
　図３に、接続先制御装置９４の予測時の動作の一例を示す。管理外端末予測モデルにより得た管理外端末情報は管理端末情報と同じパラメータで表現可能であるから、それらを管理端末情報と同列に扱って任意の方法によってその他の通信パターンや接続パターンについての各端末の通信品質予測を行うことができる。そして、その予測結果は管理外端末９１＃４の干渉考慮済みの通信品質となる。 FIG. 3 shows an example of the operation of the connection destination control device 94 at the time of prediction. Since the unmanaged terminal information obtained by the unmanaged terminal prediction model can be expressed using the same parameters as the managed terminal information, it can be treated in the same way as the managed terminal information, and various communication patterns and connection patterns can be expressed using an arbitrary method. It is possible to predict the communication quality of a terminal. Then, the prediction result becomes the communication quality with interference of the unmanaged terminal 91#4 taken into account.
　管理外端末予測モデルの実装例の一つとして、教師有り機械学習モデルの一種である条件付き敵対的生成ネットワーク（Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮ：　Ｃｏｎｄｉｔｉｏｎａｌ　Ｇｅｎｅｒａｔｉｖｅ　Ａｄｖｅｒｓａｒｉａｌ　Ｎｅｔｗｏｒｋ）を使用する方法が考えられる。その他、本手法の実装に利用可能なものとして、多次元配列の生成モデルであるＣＶＡＥ（Ｃｏｎｄｉｔｉｏｎａｌ　Ｖａｒｉａｂｌｅ　Ａｕｔｏｅｎｃｏｄｅｒ）等が考えられるが、以降はＣｏｎｄｉｔｉｏｎａｌＧＡＮ　を例に挙げて説明する。 One possible implementation example of the unmanaged terminal prediction model is to use a conditional generative adversarial network (GAN), which is a type of supervised machine learning model. In addition, CVAE (Conditional Variable Autoencoder), which is a generation model for multidimensional arrays, etc. can be considered as something that can be used to implement the present method, but Conditional GAN will be described below as an example.
　図４に、Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮのアーキテクチャの一例を示す。Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮでは、ジェネレータ８１とディスクリミネータ８２という二つの学習器が共存する。 FIG. 4 shows an example of the architecture of Conditional GAN. In the Conditional GAN, two learning devices, a generator 81 and a discriminator 82, coexist.
　ジェネレータ８１は、管理端末情報とネットワーク情報とラベルから、管理外端末情報の多次元配列を生成するニューラルネットワークである。本実施形態では、管理端末９１＃１～９１＃３で得られた通信品質情報がラベルとして用いられる。 The generator 81 is a neural network that generates a multidimensional array of unmanaged terminal information from management terminal information, network information, and labels. In this embodiment, communication quality information obtained by the management terminals 91#1 to 91#3 is used as a label.
　ディスクリミネータ８２は、入力された多次元配列が、データセット由来の真のデータか、ジェネレータ８１が生成した偽のデータかを判別するニューラルネットワークである。本実施形態では、ジェネレータ８１によって生成された管理外端末情報が真のデータと判別がつかないよう、管理外端末情報を生成する管理外端末予測モデルを学習する。一方で、ディスクリミネータ８２が真偽を精度よく判別できるよう学習を行うことで、ジェネレータ８１はより精度の良い管理外端末情報を生成できるようになる。 The discriminator 82 is a neural network that determines whether the input multidimensional array is true data derived from a dataset or false data generated by the generator 81. In this embodiment, an unmanaged terminal prediction model that generates unmanaged terminal information is learned so that the unmanaged terminal information generated by the generator 81 cannot be distinguished from real data. On the other hand, by learning so that the discriminator 82 can accurately determine authenticity, the generator 81 can generate more accurate unmanaged terminal information.
　ディスクリミネータ８２で真のデータと判別されるようになったジェネレータ８１での学習結果が管理外端末予測モデルとして用いられる。学習済みの管理外端末予測モデルが接続先制御装置９４に搭載されていてもよいが、接続先制御装置９４自身がジェネレータ８１及びディスクリミネータ８２のような学習機能を備えていてもよい。 The learning results of the generator 81 that are determined to be true data by the discriminator 82 are used as an unmanaged terminal prediction model. Although the learned unmanaged terminal prediction model may be installed in the connection destination control device 94, the connection destination control device 94 itself may be provided with learning functions such as the generator 81 and the discriminator 82.
　図５に、端末９１及び接続先制御装置９４の構成例を示す。端末９１は、通信インタフェース（ＩＦ）１１、アプリケーション１２、位置情報取得部１３、端末情報取得・通知部１４、通信結果通知部１５を備える。接続先制御装置９４は、通信インタフェース（ＩＦ）４１、通信情報集約部４２、データセット格納部４３、管理外端末予測モデル４４、管理端末品質予測モデル４５、接続パターン最適化プログラム４６、接続先制御部４７を備える。 FIG. 5 shows a configuration example of the terminal 91 and the connection destination control device 94. The terminal 91 includes a communication interface (IF) 11 , an application 12 , a location information acquisition section 13 , a terminal information acquisition/notification section 14 , and a communication result notification section 15 . The connection destination control device 94 includes a communication interface (IF) 41, a communication information aggregation unit 42, a data set storage unit 43, an unmanaged terminal prediction model 44, a management terminal quality prediction model 45, a connection pattern optimization program 46, and connection destination control. 47.
　本開示の端末９１及び接続先制御装置９４は、コンピュータとプログラムによっても実現でき、プログラムを記録媒体に記録することも、ネットワークを通して提供することも可能である。 The terminal 91 and the connection destination control device 94 of the present disclosure can also be realized by a computer and a program, and the program can be recorded on a recording medium or provided through a network.
（端末９１）
・通信インタフェース１１：直接もしくはネットワーク越しに、接続先制御装置９４と通信を行うインタフェースである。ネットワーク９２＃１又は９２＃２の基地局と通信するインタフェースと同じものを用いても良い。
・アプリケーション１２：端末９１に備わる任意のアプリケーションである。
・位置情報取得部１３：ＧＰＳなどによって、端末９１の地理的位置を座標で取得する。
・端末情報取得・通知部１４：位置情報取得部１３から位置情報を取得し、アプリケーション１２の発生させる平均のデータレートを取得するなど、管理端末の品質予測に必要な管理端末情報を取得する。
・通信結果通知部１５：取得した管理端末情報の結果、どのような通信品質が得られたか（スループットや遅延）を含む通信品質情報を取得し、接続先制御装置９４に通知する。 (terminal 91)
- Communication interface 11: An interface that communicates with the destination control device 94 directly or over a network. The same interface that communicates with the base station of network 92#1 or 92#2 may be used.
-Application 12: An arbitrary application provided in the terminal 91.
- Location information acquisition unit 13: Acquires the geographical position of the terminal 91 in coordinates using GPS or the like.
- Terminal information acquisition/notification unit 14: Acquires location information from the location information acquisition unit 13, acquires the average data rate generated by the application 12, and acquires management terminal information necessary for quality prediction of the management terminal.
- Communication result notification unit 15: Acquires communication quality information including what kind of communication quality was obtained (throughput and delay) as a result of the acquired management terminal information, and notifies it to the connection destination control device 94.
（接続先制御装置９４）
・通信インタフェース４１：端末９１と通信を行うインタフェースである。ネットワーク９２＃１及び９２＃２の基地局と通信を行うインタフェースを備えていてもよい。
・通信情報集約部４２：端末群から取得した管理端末情報及び通信品質情報を加工し、データセットとして格納する。予測時にはその加工した管理端末情報及び通信品質情報を、管理外端末予測モデル４４と管理端末品質予測モデル４５へ出力する。
・データセット格納部４３：Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮなどの学習に用いるデータセットを格納し、学習時には管理外端末予測モデル４４にデータを渡す。
・管理外端末予測モデル４４：Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮなどの学習モデルを用いて、管理外端末情報の予測を行う。
・管理端末品質予測モデル４５：管理端末情報と、予測された管理外端末情報を入力として、管理端末９１＃１～９１＃３の通信品質を予測する。このとき、管理端末品質予測モデル４５は、管理端末情報及び管理外端末情報に加え、ネットワーク情報を用いてもよい。
・接続パターン最適化プログラム４６：管理端末品質予測モデル４５の予測結果を用いて、管理端末９１＃１～９１＃３の最適な接続パターンを発見する。例えば、接続パターン最適化プログラム４６は、各管理端末が各ネットワーク９２＃１及び９２＃２に接続した場合の接続パターンを抽出し、抽出した接続パターンのなかから任意の接続パターンを決定する。
・接続先制御部４７：接続パターン最適化プログラム４６が決定した接続パターンとなるよう、管理端末９１＃１～９１＃３に対して切り替え指示を行う。 (Connection destination control device 94)
- Communication interface 41: An interface for communicating with the terminal 91. It may also include an interface for communicating with base stations of networks 92#1 and 92#2.
- Communication information aggregation unit 42: Processes the management terminal information and communication quality information acquired from the terminal group and stores it as a data set. At the time of prediction, the processed management terminal information and communication quality information are output to the unmanaged terminal prediction model 44 and the management terminal quality prediction model 45.
- Data set storage unit 43: Stores the data set used for learning of Conditional GAN, etc., and passes the data to the unmanaged terminal prediction model 44 during learning.
- Unmanaged terminal prediction model 44: Predicts unmanaged terminal information using a learning model such as Conditional GAN.
- Management terminal quality prediction model 45: Predicts the communication quality of management terminals 91#1 to 91#3 by inputting management terminal information and predicted unmanaged terminal information. At this time, the management terminal quality prediction model 45 may use network information in addition to the management terminal information and unmanaged terminal information.
- Connection pattern optimization program 46: Using the prediction results of the management terminal quality prediction model 45, discovers the optimal connection pattern for the management terminals 91#1 to 91#3. For example, the connection pattern optimization program 46 extracts connection patterns when each management terminal connects to each network 92#1 and 92#2, and determines an arbitrary connection pattern from among the extracted connection patterns.
- Connection destination control unit 47: Instructs the management terminals 91#1 to 91#3 to switch to the connection pattern determined by the connection pattern optimization program 46.
（管理外端末予測モデルの学習）
　図６に、管理外端末予測モデル４４の学習フローの一例を示す。
管理外端末予測モデル４４の学習にあたって必要となるデータセットは、例えばネットワークシミュレータによって生成することができる。ネットワークシミュレータは複数端末が存在する仮想のネットワークについてシミュレーションを行い（Ｓ１１）、接続先制御装置９４の通信情報集約部４２がシミュレーション結果を集約し（Ｓ１２）、データセットの多次元配列の形式に変換した後（Ｓ１３）、データセット格納部４３に保存する（Ｓ１４）。管理外端末予測モデル４４は、データセット格納部４３に保存されているデータセットごとに（Ｓ１５）、管理外端末情報の学習を行う（Ｓ１６）。ネットワークシミュレータの時１回のシミュレーションデータを用いて複数の管理外端末情報のデータが生成可能であるが、その生成方法については後述する。 (Learning of unmanaged terminal prediction model)
FIG. 6 shows an example of the learning flow of the unmanaged terminal prediction model 44.
The data set required for learning the unmanaged terminal prediction model 44 can be generated by, for example, a network simulator. The network simulator simulates a virtual network in which multiple terminals exist (S11), and the communication information aggregation unit 42 of the connection destination control device 94 aggregates the simulation results (S12) and converts them into a multidimensional array format of data sets. After that (S13), it is stored in the data set storage unit 43 (S14). The unmanaged terminal prediction model 44 learns unmanaged terminal information for each data set stored in the data set storage unit 43 (S15) (S16). When using a network simulator, it is possible to generate data on a plurality of unmanaged terminal information using one simulation data, and the generation method will be described later.
（管理端末品質予測モデルの学習）
　図７に、管理端末品質予測モデル４５の学習フローの一例を示す。管理端末品質予測モデル４５の学習にあたって必要となるデータセットは、実端末の管理端末情報及び通信品質情報によって生成することができる。管理端末９１＃１～９１＃３であるもののうちいくつかを管理外端末と見なして学習を行ってもよい。
　管理端末９１＃１～９１＃３は周辺端末と同時に通信を行い（Ｓ２１）、そのときの管理端末情報及び通信品質情報を接続先制御装置９４が集約する（Ｓ２２）。接続先制御装置９４は、集約した情報をデータセットの多次元配列の形式に変換した後（Ｓ２３）、データセット格納部４３に保存する（Ｓ２４）。管理端末品質予測モデル４５は、データセット格納部４３に保存されているデータセットごとに（Ｓ２５）、管理端末品質予測モデル４５の学習を行う（Ｓ２６）。 (Learning of management terminal quality prediction model)
FIG. 7 shows an example of a learning flow of the management terminal quality prediction model 45. The data set required for learning the management terminal quality prediction model 45 can be generated from the management terminal information and communication quality information of the actual terminal. Learning may be performed by regarding some of the management terminals 91#1 to 91#3 as unmanaged terminals.
The management terminals 91#1 to 91#3 communicate with the peripheral terminals at the same time (S21), and the connection destination control device 94 aggregates the management terminal information and communication quality information at that time (S22). The connection destination control device 94 converts the aggregated information into a data set multidimensional array format (S23), and then stores it in the data set storage unit 43 (S24). The management terminal quality prediction model 45 performs learning for each data set stored in the data set storage unit 43 (S25) (S26).
（学習後の実運用）
　図８に、予測フローの一例を示す。管理端末９１＃１～９１＃３が周辺端末と同時に通信を行い（Ｓ３１）、そのときの管理端末情報及び通信品質情報を通信情報集約部４２が集約する（Ｓ３２）。管理外端末予測モデル４４は、直前の管理端末９１＃１～９１＃３の管理端末情報及び通信品質情報を入力として、管理外端末９１＃４の位置や発生させる通信トラヒックパターンなどの管理外端末情報を予測する（Ｓ３３）。 (Actual operation after learning)
FIG. 8 shows an example of a prediction flow. The management terminals 91#1 to 91#3 communicate with peripheral terminals at the same time (S31), and the communication information aggregation unit 42 aggregates the management terminal information and communication quality information at that time (S32). The unmanaged terminal prediction model 44 inputs the management terminal information and communication quality information of the immediately preceding management terminals 91#1 to 91#3, and calculates unmanaged terminals such as the position of the unmanaged terminal 91#4 and the communication traffic pattern to be generated. Predict information (S33).
　管理外端末予測モデル４４の出力は管理端末品質予測モデル４５の入力となる。管理端末品質予測モデル４５は、通信情報集約部４２の集約結果と管理外端末予測モデル４４の予測結果を結合し（Ｓ３４）、管理外端末無し時の管理端末の品質予測と同様の手法で、管理端末／管理外端末をまとめて管理端末とみなし品質予測を行う（Ｓ３５）。 The output of the unmanaged terminal prediction model 44 becomes the input of the managed terminal quality prediction model 45. The management terminal quality prediction model 45 combines the aggregation results of the communication information aggregation unit 42 and the prediction results of the unmanaged terminal prediction model 44 (S34), and uses the same method as the quality prediction of the management terminal when there is no unmanaged terminal, The management terminal/unmanaged terminal is collectively regarded as a management terminal and quality prediction is performed (S35).
　ステップＳ３５の予測手法は非特許文献［２］［３］［４］などのいかなる手法を用いても良い。接続パターン最適化プログラム４６には、管理端末品質予測モデル４５の出力のうち、管理端末９１＃１～９１＃３に関するデータが入力されれば十分である。なぜなら接続パターン最適化プログラム４６は、管理端末のネットワーク接続状態のみ制御可能であるからである。 As the prediction method in step S35, any method such as those described in non-patent documents [2], [3], and [4] may be used. It is sufficient for the connection pattern optimization program 46 to input data related to the management terminals 91#1 to 91#3 out of the output of the management terminal quality prediction model 45. This is because the connection pattern optimization program 46 can only control the network connection state of the management terminal.
（データセット生成方法）
　図９に、データセット生成方法の一例を示す。あるネットワークシミュレーションもしくは実端末での通信結果を集約したデータから、データセットを生成する方法の概念図を示す。集約されたデータは、全て恣意的に設定したシミュレーションパラメータもしくは管理端末の情報であるが、このデータを疑似的に管理端末からなる多次元配列と、管理外端末からなる多次元配列に分割し、その多次元配列の集合を一つのデータとみなす。 (Dataset generation method)
FIG. 9 shows an example of a data set generation method. A conceptual diagram of a method for generating a dataset from data that aggregates communication results from a network simulation or an actual terminal is shown. The aggregated data is all arbitrarily set simulation parameters or management terminal information, but this data is divided into a pseudo multidimensional array consisting of management terminals and a multidimensional array consisting of unmanaged terminals, The set of multidimensional arrays is regarded as one piece of data.
　管理端末情報の多次元配列は、Ｃｏｎｄｉｔｉｏｎａｌ　ＧＡＮにおけるラベルとして用いられ、管理外端末のそれはディスクリミネーター８２への入力配列として用いられる。管理端末と管理外端末の分割方法を無作為に抽出することで、一つの集約データから多数のデータをデータセットとして生成することができる。 The multidimensional array of management terminal information is used as a label in the Conditional GAN, and that of unmanaged terminals is used as an input array to the discriminator 82. By randomly selecting a method for dividing managed terminals and unmanaged terminals, it is possible to generate a large amount of data as a data set from one aggregated data.
　例えば、管理端末情報が式（４）で表され、通信品質情報が式（５）で表される場合、式（６）～式（８）で表される第１のデータセットと、式（９）～式（１１）で表される第２のデータセットと、を生成する。
Figure JPOXMLDOC01-appb-M000004
Figure JPOXMLDOC01-appb-M000005
 For example, if the management terminal information is expressed by equation (4) and the communication quality information is expressed by equation (5), the first data set expressed by equations (6) to (8) and the equation ( 9) to a second data set expressed by equation (11) are generated.
Figure JPOXMLDOC01-appb-M000004
Figure JPOXMLDOC01-appb-M000005
（第１のデータセット）
　第１のデータセットは、式（６）で表される管理端末情報と、式（７）で表される管理外端末情報と、式（８）で表される管理端末の通信品質情報と、を含む。
Figure JPOXMLDOC01-appb-M000006
Figure JPOXMLDOC01-appb-M000007
Figure JPOXMLDOC01-appb-M000008
 (first data set)
The first data set includes management terminal information represented by formula (6), unmanaged terminal information represented by formula (7), communication quality information of the management terminal represented by formula (8), including.
Figure JPOXMLDOC01-appb-M000006
Figure JPOXMLDOC01-appb-M000007
Figure JPOXMLDOC01-appb-M000008
（第２のデータセット）
　第２のデータセットは、式（９）で表される管理端末情報と、式（１０）で表される管理外端末情報と、式（１１）で表される管理端末の通信品質情報と、を含む。
Figure JPOXMLDOC01-appb-M000009
Figure JPOXMLDOC01-appb-M000010
Figure JPOXMLDOC01-appb-M000011
 (Second dataset)
The second data set includes management terminal information represented by formula (9), unmanaged terminal information represented by formula (10), communication quality information of the management terminal represented by formula (11), including.
Figure JPOXMLDOC01-appb-M000009
Figure JPOXMLDOC01-appb-M000010
Figure JPOXMLDOC01-appb-M000011
　以上説明したように、本開示は、管理外端末９１＃４の位置を含む管理外端末情報を予め学習し、管理端末９１＃１～９１＃３の通信品質の予測の際に、管理端末情報と同種のパラメータの管理外端末情報を得る。これによって、本開示は、管理外端末９１＃４の振る舞いを考慮した、管理端末９１＃１～９１＃３の高精度な品質予測結果を得ることができる。 As explained above, the present disclosure learns in advance unmanaged terminal information including the location of unmanaged terminal 91#4, and when predicting the communication quality of management terminals 91#1 to 91#3, management terminal information Obtain unmanaged terminal information with the same type of parameters as . As a result, the present disclosure can obtain highly accurate quality prediction results for the management terminals 91#1 to 91#3 in consideration of the behavior of the unmanaged terminal 91#4.
１１、４１：通信インタフェース（ＩＦ）
１２：アプリケーション
１３：位置情報取得部
１４：端末情報取得・通知部
１５：通信結果通知部
４２：通信情報集約部
４３：データセット格納部
４４：管理外端末予測モデル
４５：管理端末品質予測モデル
４６：接続パターン最適化プログラム
４７：接続先制御部
８１：ジェネレータ
８２：ディスクリミネータ
９１：端末
９２：ネットワーク
９３：インターネット
９４：接続先制御装置 11, 41: Communication interface (IF)
12: Application 13: Location information acquisition unit 14: Terminal information acquisition/notification unit 15: Communication result notification unit 42: Communication information aggregation unit 43: Data set storage unit 44: Unmanaged terminal prediction model 45: Management terminal quality prediction model 46 : Connection pattern optimization program 47: Connection destination control unit 81: Generator 82: Discriminator 91: Terminal 92: Network 93: Internet 94: Connection destination control device

Claims (8)

  1. 　管理端末の地理的位置を含む管理端末情報と当該管理端末で得られた通信品質情報とを取得し、
    　前記管理端末情報及び前記通信品質情報に基づいて、管理外端末の地理的位置を含む管理外端末情報を予測し、
    　予測した前記管理外端末情報を用いて、前記管理端末の通信品質を予測する、
    　装置。     Obtaining management terminal information including the geographical location of the management terminal and communication quality information obtained from the management terminal,
    predicting unmanaged terminal information including the geographical location of the unmanaged terminal based on the management terminal information and the communication quality information;
    predicting communication quality of the management terminal using the predicted unmanaged terminal information;
    Device.
  2. 　前記管理外端末が存在する場合に前記管理端末が得られる通信品質を予め学習した管理外端末予測モデルを用いて、前記管理外端末情報を予測する、
    　請求項１に記載の装置。     predicting the unmanaged terminal information using an unmanaged terminal prediction model that has learned in advance the communication quality obtained by the management terminal when the unmanaged terminal exists;
    The device according to claim 1.
  3. 　前記管理端末情報は、前記管理端末の地理的位置に対応する２次元配列又は３次元配列のピクセルで表され、
    　前記管理外端末情報は、前記管理外端末の地理的位置に対応する２次元配列又は３次元配列のピクセルで表され、
    　前記管理端末情報及び前記管理外端末情報のピクセル位置と地理的位置が一致する、
    　請求項１に記載の装置。     The management terminal information is represented by a two-dimensional array or a three-dimensional array of pixels corresponding to the geographical location of the management terminal,
    The unmanaged terminal information is represented by a two-dimensional array or three-dimensional array of pixels corresponding to the geographical location of the unmanaged terminal,
    The pixel position of the managed terminal information and the unmanaged terminal information match the geographical position,
    A device according to claim 1.
  4. 　前記管理端末がネットワークに接続した場合に前記管理端末で得られる通信品質をネットワークごとに予測し、
    　前記予測に基づいて、前記管理端末が接続するネットワークを決定する、
    　請求項１に記載の装置。     predicting the communication quality obtained by the management terminal for each network when the management terminal connects to the network;
    determining a network to which the management terminal connects based on the prediction;
    The device according to claim 1.
  5. 　複数の前記管理端末から前記管理端末情報及び前記通信品質情報を取得し、
    　複数の前記管理端末がネットワークに接続した場合の接続パターンを抽出し、
    　抽出した接続パターンのなかから任意の接続パターンを決定し、
    　決定した接続パターンで、複数の前記管理端末をネットワークに接続させる、
    　請求項１に記載の装置。     acquiring the management terminal information and the communication quality information from a plurality of the management terminals;
    extracting a connection pattern when a plurality of said management terminals connect to the network;
    Determine an arbitrary connection pattern from the extracted connection patterns,
    connecting the plurality of management terminals to the network using the determined connection pattern;
    A device according to claim 1.
  6. 　ネットワークに接続可能な管理端末と、
    　前記管理端末から通信品質情報を取得する、請求項１から５のいずれかに記載の装置と、
    　を備えるシステム。     A management terminal that can connect to the network,
    The device according to any one of claims 1 to 5, which acquires communication quality information from the management terminal;
    A system equipped with
  7. 　管理端末の地理的位置を含む管理端末情報と当該管理端末で得られた通信品質情報とを取得し、
    　前記管理端末情報及び前記通信品質情報に基づいて、管理外端末の地理的位置を含む管理外端末情報を予測し、
    　予測した前記管理外端末情報を用いて、前記管理端末の通信品質を予測する、
    　方法。     Obtain management terminal information including the geographical location of the management terminal and communication quality information obtained from the management terminal,
    predicting unmanaged terminal information including the geographical location of the unmanaged terminal based on the management terminal information and the communication quality information;
    predicting communication quality of the management terminal using the predicted unmanaged terminal information;
    Method.
  8. 　請求項１から５のいずれかに記載の装置に備わる各機能部としてコンピュータを実現させるためのプログラム。 A program for realizing a computer as each functional unit included in the device according to any one of claims 1 to 5.
PCT/JP2022/028329 2022-07-21 2022-07-21 System for predicting terminal communication quality WO2024018586A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/028329 WO2024018586A1 (en) 2022-07-21 2022-07-21 System for predicting terminal communication quality

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2022/028329 WO2024018586A1 (en) 2022-07-21 2022-07-21 System for predicting terminal communication quality

Publications (1)

Publication Number Publication Date
WO2024018586A1 true WO2024018586A1 (en) 2024-01-25

Family

ID=89617549

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2022/028329 WO2024018586A1 (en) 2022-07-21 2022-07-21 System for predicting terminal communication quality

Country Status (1)

Country Link
WO (1) WO2024018586A1 (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2017028340A (en) * 2015-07-15 2017-02-02 日本電信電話株式会社 Csma/ca communication quality management system and method
JP2021022913A (en) * 2019-07-25 2021-02-18 パナソニック株式会社 Controller and control method
WO2022038760A1 (en) * 2020-08-21 2022-02-24 日本電信電話株式会社 Device, method, and program for predicting communication quality

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2017028340A (en) * 2015-07-15 2017-02-02 日本電信電話株式会社 Csma/ca communication quality management system and method
JP2021022913A (en) * 2019-07-25 2021-02-18 パナソニック株式会社 Controller and control method
WO2022038760A1 (en) * 2020-08-21 2022-02-24 日本電信電話株式会社 Device, method, and program for predicting communication quality

Similar Documents

Publication Publication Date Title
Qin et al. A software defined networking architecture for the internet-of-things
Li et al. Toward software-defined cellular networks
AU2019272327B2 (en) Method for determining background traffic transfer policy and apparatus
Amani et al. Programmable policies for data offloading in LTE network
Chung et al. Machine learning based path management for mobile devices over MPTCP
CN109905280B (en) Simulation method and system for mobile satellite network
Barmpounakis et al. Data analytics for 5G networks: A complete framework for network access selection and traffic steering
Alshaer et al. Software-defined networking-enabled heterogeneous wireless networks and applications convergence
Mouawad et al. Inter-slice mobility management solution in V2X environment
Laselva et al. Advancements of QoE assessment and optimization in mobile networks in the machine era
WO2022038760A1 (en) Device, method, and program for predicting communication quality
Sánchez-Mahecha et al. QoS evaluation of the future high-efficiency IEEE 802.11 ax WLAN standard
Lu et al. An access selection mechanism in 5G network slicing
CN111669785B (en) Mobile data distribution method of heterogeneous wireless network based on campus network
De Schepper et al. Flow management and load balancing in dynamic heterogeneous LANs
Kassir et al. Service placement for real-time applications: Rate-adaptation and load-balancing at the network edge
WO2024018586A1 (en) System for predicting terminal communication quality
Hoque et al. SDN-DTN Combined Architecture in Post Disaster Scenario—A new way to start
Groen et al. TRACTOR: Traffic Analysis and Classification Tool for Open RAN
Labraoui et al. Opportunistic SDN-controlled wireless mesh network for mobile traffic offloading
Dababneh Lte traffic generation and evolved packet core (epc) network planning
Lee et al. A practical SDN-based data offloading framework
Anand et al. A machine learning-based xApp for 5G O-RAN to mitigate co-tier interference and improve qoe for various services in a hetnet environment
Baldesi et al. Optimized cooperative streaming in wireless mesh networks
US11722930B2 (en) Systems and methods for dynamic rule determination for user plane data in a wireless network

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22951972

Country of ref document: EP

Kind code of ref document: A1