Route control device

Abstract

A route control device (100) comprises: an inference unit (118) that uses a route calculation rule inference model to infer, on the basis of link information between nodes forming a dynamic network, a route calculation rule to use in the computation of a route between two nodes included in the dynamic network; and a route computation unit (113) that uses the inferred route calculation rule to compute a route between two nodes. The route control device (100) may also comprise a link analysis unit (111) that computes link information on the basis of the position of a node and information about the environment of said node at a future time, and a route setting unit (119) that creates a routing table for a node of the dynamic network on the basis of the computed route and transmits the routing table to said node.

Description

Route control device

[0001] The present invention relates to a route control device for controlling routes in a dynamic network.

[0002] Non-Terrestrial Networks (NTNs), a type of dynamic network, are networks consisting of satellites, stratospheric platforms, unmanned aerial vehicles (drones), ships, smartphones, etc. As the fifth-generation mobile communication system (5G) becomes more widespread, non-terrestrial networks that connect various nodes from space to the air and ground are attracting attention as next-generation communication networks known as Beyond 5G and sixth-generation mobile communication systems (6G).

[0003] In non-terrestrial networks, the topology changes dynamically because some nodes move. In conventional routing control methods, the topology is detected by exchanging control messages on the network, and communication routes are calculated using a route calculation rule based on a shortest path algorithm (see Non-Patent Document 1).

[0004] C. Perkins, et al., "Ad hoc On-Demand Distance Vector (AODV) Routing," The Internet Society, Request for Comments 3561, July 2003.

[0005] In networks with dynamically changing environments and topologies, such as non-terrestrial networks, the time period during which links between nodes can communicate is limited. For example, a low-earth orbit satellite and a ground station can communicate for a maximum of about 10 minutes. Furthermore, link bandwidth, speed, and connection availability vary depending on environmental factors such as weather. Therefore, routing control is required to determine communication routes that are fast enough to adapt to network changes in order to increase packet arrival rates during limited time periods.

[0006] When conventional routing control methods are applied to dynamic networks, the faster and more drastic the environmental changes become, the more frequent the control message exchanges and route calculations that are required to keep up with the changes, resulting in increased network bandwidth overhead and a lack of convergence in route calculations. This leads to an increase in the number of cases where communication routes cannot be established in time, and a decrease in packet arrival rate.

[0007] In addition, route calculation rules based on a fixed, single metric (e.g., the inverse of bandwidth)-based shortest-path algorithm make it difficult to select routes (route control) that increase packet arrival rates in response to environmental changes. For example, it is difficult to adaptively calculate routes (route control) that increase packet arrival rates by adopting the geographically shortest route in good weather and bypassing areas with bad weather in bad weather, even though this is a longer route.

[0008] The present invention has been made in view of the above background, and an object of the present invention is to enable route calculation in a dynamic network.

[0009] In order to solve the above-mentioned problems, the route control device of the present invention comprises an inference unit that infers a route calculation rule to be used when calculating a route between two nodes included in a dynamic network based on link information between the nodes that make up the dynamic network, using a route calculation rule inference model, which is a machine learning model, and a route calculation unit that calculates a route between the two nodes using the inferred route calculation rule.

[0010] According to the present invention, it is possible to enable route calculation in dynamic networks.

[0011] FIG. 1 is a diagram showing the configuration of a non-terrestrial network according to the present embodiment; FIG. 2 is a functional block diagram of a route control device according to the present embodiment; FIG. 3 is a data configuration diagram of a scenario database according to the present embodiment; FIG. 4 is a flowchart of a learning data generation process according to the present embodiment; FIG. 5 is a flowchart of a route calculation rule inference model generation process according to the present embodiment; FIG. 6 is a flowchart of a route setting process according to the present embodiment; and FIG. 7 is a hardware configuration diagram showing an example of a computer that realizes the functions of a route control device according to the present embodiment.

[0012] A route control device for a dynamic network in a mode (embodiment) for carrying out the present invention will be described below. The following description will be given taking a non-terrestrial network as an example of the dynamic network, but the present invention is not limited to this. For example, the dynamic network may include a robot or relay station that moves on the floor or ground surface indoors or outdoors, or may include a balloon or drone that is near the ground surface (for example, up to an altitude of about 100 meters).

[0013] The route control device generates a network model of the non-terrestrial network at each time by referring to a scenario containing time-series information on link information created based on assumed changes in the positions of mobile nodes in the non-terrestrial network and expected environmental changes. Link information includes, for example, propagation distance, received signal strength, location (in a vacuum, in the atmosphere, near the Earth's surface, etc.), propagation loss, and rain attenuation. Location refers to the location of the link; for example, a link between satellites is in a vacuum, while a link between a low-altitude aircraft and a ground terminal is near the Earth's surface.

[0014] Next, the route control device uses the network model to calculate a route between two nodes according to each of the multiple route calculation rules. If there are multiple calculated routes, the route control device then selects one route that satisfies a predetermined evaluation condition, and generates learning data in which link information is used as an explanatory variable and the route calculation rule used to calculate the selected route is used as a target variable. Furthermore, the route control device uses the learning data to generate a route calculation rule inference model that infers a route calculation rule based on the link information.

[0015] The route control device calculates link information between nodes based on environmental information such as the future positions and weather of the nodes that make up the non-terrestrial network. Next, the route control device infers a route calculation rule using a route calculation rule inference model based on the link information, and calculates a route between two nodes using the route calculation rule. Next, the route control device generates a route table for each node based on the route and sends it to the node. Each node forwards communication packets based on the route table.

[0016] By having each node forward communication packets based on the routing table generated by such a routing control device, communications that satisfy predetermined evaluation criteria are possible in non-terrestrial networks. Route calculation rules include selecting a route with the shortest transmission distance or selecting a route with the largest minimum value for the power density-to-noise power density ratio per bit of each link. By calculating a route using these route calculation rules, the routing control device can calculate a route that satisfies predetermined evaluation criteria more quickly than by calculating and selecting a route using all route calculation rules. Examples of predetermined evaluation criteria include packet arrival rate, bandwidth, and delay. The routing control device also generates a routing table by calculating routes in advance based on the node's future position and environment. This enables stable communications even in non-terrestrial networks, where topology and environment change rapidly.

[0017] <<Configuration of Non-Terrestrial Network>> Fig. 1 is a diagram showing the configuration of a non-terrestrial network 500 according to this embodiment. The non-terrestrial network 500 includes satellites 511 to 514, ground stations 521 to 524, a ground node 531, and a ground network 580. The satellites 511 to 514 can communicate with the ground stations 521 to 524 and with the satellites 511 to 514. The ground stations 521 to 524 and the ground node 531 can communicate with each other via the ground network 580.

[0018] The non-terrestrial network 500 may be configured to include mobile nodes such as stratospheric platforms and unmanned aerial vehicles instead of the satellites 511 to 514. The non-terrestrial network 500 may also be a vertically multi-layered non-terrestrial network including mobile nodes such as geostationary satellites, low-earth orbit satellites, stratospheric platforms, and unmanned aerial vehicles.

[0019] Here, we consider the communication path (route) between the satellite 514 and the terrestrial node 531. It is considered desirable to shorten the wireless network route, which is inferior to the terrestrial network 580 in terms of bandwidth, delay, and packet arrival rate, and to route via the terrestrial station 524 and the terrestrial network 580. However, if the weather conditions at the terrestrial station 524 are poor and the weather conditions at the terrestrial station 523 are also poor, a route via the satellite 513, the satellite 512, the terrestrial station 522, and the terrestrial network 580 is desirable. This is because poor weather conditions cause the link between the satellite and the terrestrial station to become unstable, resulting in a low packet arrival rate. Therefore, a higher packet arrival rate provides more stable communication, even if the route is longer and the delay increases. When a wireless network is included, the route with the shortest length (the fewest number of nodes passed through) is not necessarily the desirable route. In other words, there is not necessarily a single route calculation rule for calculating a route.

[0020] 2 is a functional block diagram of the route control device 100 according to this embodiment. The route control device 100 is a computer, and includes a control unit 110, a storage unit 130, and a communication unit 180. The communication unit 180 includes a communication device, and is capable of transmitting and receiving data to and from other nodes in the non-terrestrial network 500 via the terrestrial network 580.

[0021] <Route Control Device: Storage Unit> The storage unit 130 is configured to include storage devices such as a read-only memory (ROM), a random access memory (RAM), a solid-state drive (SSD), etc. The storage unit 130 stores a scenario database 140, a propagation model 131, a route calculation rule 132, learning data 150, a route calculation rule inference model database 160, trajectory information 133, environmental information 134, and a program 138.

[0022] <Storage Unit: Propagation Model and Route Calculation Rule> The propagation model 131 is a propagation model related to radio wave propagation in wireless communication between the satellites 511 to 514 and between the satellites 511 to 514 and the terrestrial stations 521 to 524. The route calculation rule 132 is a rule for calculating a route in the non-terrestrial network 500, and includes, for example, the following (1) to (4): (1) Select a route that maximizes the minimum value of the ratio of power density per bit to noise power density among the links that make up the route in the wireless section. (2) Select a route that maximizes the sum of the ratios of power density per bit to noise power density among the links that make up the route in the wireless section. (3) Select a route that minimizes the sum of the transmission distances in the wireless section. (4) Select a route that minimizes the maximum value of the bit error rate among the links that make up the route in the wireless section.

[0023] <<Storage Unit: Learning Data>> The learning data 150 is learning data used to generate a route calculation rule inference model. The explanatory variables of the learning data 150 are link information, which will be described later. The objective variable is identification information (a label, which will be described later) that identifies one route calculation rule included in the route calculation rules 132. Details of the generation of the learning data 150 will be described later.

[0024] <Storage Unit: Route Computation Rule Inference Model Database> The route computation rule inference model database 160 stores a route computation rule inference model, which is a machine learning model that infers a route computation rule used when calculating a route between two nodes included in the non-terrestrial network 500. The machine learning model is, for example, a neural network, but is not limited to this.

[0025] <Storage Unit: Orbit Information and Environmental Information> The orbit information 133 includes the future positions and velocities of the satellites 511-514. The environmental information 134 includes environmental information including future meteorological conditions around the ground stations 521-524. The environmental information 134 may also include future environmental information (space weather forecast) around the satellites 511-514. Note that if the non-terrestrial network 500 includes an unmanned aerial vehicle (drone), the environmental information includes future meteorological conditions around the unmanned aerial vehicle. The program 138 includes descriptions of the learning data generation process (see FIG. 4), the route calculation rule inference model generation process (see FIG. 5), and the route setting process (see FIG. 6), which will be described later.

[0026] 3 is a data configuration diagram of the scenario database 140 according to this embodiment. The scenario database 140 includes one or more scenarios 141. The scenario 141 is information related to a path between two nodes indicated by a target node 142, and includes node information and environment information 143, link information 144, a communication determination result 145, and a label 146 within a predetermined period of time.

[0027] The node information and environment information 143 includes the positions and velocities of the satellites 511 to 514 within a predetermined period of time. The node information and environment information 143 also includes environmental information including meteorological conditions around the ground stations 521 to 524 within a predetermined period of time. The positions and velocities of the satellites 511 to 514 included in the node information and environment information 143 are not limited to information that actually occurred in the past, but may also be positions and velocities that may occur in the future. Furthermore, the environmental information including meteorological conditions around the ground stations 521 to 524 included in the node information and environment information 143 is not limited to information that actually occurred in the past, but may also be assumed environmental information.

[0028] The link information 144 is information related to links that can be paths between two nodes indicated by the target node 142, and includes any of the propagation distance, received signal strength (shown as "received strength" in FIG. 3), location, propagation loss, etc. of the link at the scenario time. Note that the link is a link between satellites 511 to 514, or a link between a satellite 511 to 514 and a ground station 521 to 524. As will be described later, the link information 144 is calculated based on the node information and environment information 143.

[0029] A scenario time is a time within the scenario 141. The interval between scenario times may be a fixed interval in seconds or minutes, or may be a variable interval. Between one scenario time and the next, changes in link information such as propagation distance and received signal strength are small and can be considered constant.

[0030] The communication determination result 145 indicates whether or not a route calculated using each route calculation rule included in the route calculation rule 132 is communicable in a network model configured including the links indicated in the link information 144 at each scenario time. Note that a route is a route between two nodes indicated in the target node 142. The network model will be described later.

[0031] The label 146 is identification information of a route calculation rule used when calculating a route selected from among routes that have communication and that the communication determination unit 114 has determined to have communication using a predetermined method. The predetermined method will be described later.

[0032] <Route Control Device: Control Unit> Returning to Fig. 2, the control unit 110 will be described. The control unit 110 is configured to include a CPU (Central Processing Unit) and is equipped with a link analysis unit 111, a network model generation unit 112, a route calculation unit 113, a communication determination unit 114, a learning data generation unit 115, a learning unit 116, a trajectory and environment information acquisition unit 117, an inference unit 118, and a route setting unit 119. The control unit 110 may be configured to include a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), etc.

[0033] <<Control Unit: Link Analysis Unit>> The link analysis unit 111 calculates link information such as propagation distance, received signal strength, location, propagation loss, and rain attenuation as information related to the links of the non-terrestrial network 500 based on the positions and speeds of the satellites 511 to 514 and environmental information. More specifically, the link analysis unit 111 calculates link information 144 based on node information and environmental information 143 included in the scenario 141 (see FIG. 3 ). The link analysis unit 111 also calculates future link information based on the orbit information 133 and environmental information 134.

[0034] <<Control Unit: Network Model Generation Unit>> The network model generation unit 112 generates a network model of the non-terrestrial network 500 based on the link information 144 (see FIG. 3 ) at each scenario time. This network model is not limited to a network model of the entire non-terrestrial network 500, but may be a network model showing a part of the non-terrestrial network 500 used to calculate a route between two nodes in the target node 142. The attributes (performance, quality) of the links in the network model are indicated in the link information 144. The network model generation unit 112 also generates a network model based on the link information for the future calculated by the link analysis unit 111. Note that, hereinafter, this network model will also be referred to as a network model for the future.

[0035] If the attributes of a link indicated in the link information do not satisfy a predetermined condition, the network model generation unit 112 does not include the link in the network model. For example, if the propagation distance of the link is equal to or greater than a predetermined threshold, the network model generation unit 112 does not include the link in the network model.

[0036] <<Control Unit: Path Calculation Unit>> The path calculation unit 113 calculates a path between two nodes that are target nodes in the network model generated by the network model generation unit 112. To explain in detail, the path calculation unit 113 calculates a path in the network model at each scenario time according to each path calculation rule in the path calculation rule 132. Furthermore, the path calculation unit 113 calculates a path in the future network model generated by the network model generation unit 112 according to a path calculation rule inferred by the inference unit 118, which will be described later.

[0037] <<Control unit: Communication determination unit>> The communication determination unit 114 determines whether the route calculated by the route calculation unit 113 is connected in the network model generated based on the link information 144 at each scenario time, and stores the result in the communication determination result 145 (see Figure 3).

[0038] <<Control Unit: Learning Data Generation Unit>> The learning data generation unit 115 generates learning data 150 by referring to the label 146 (see FIG. 3 ) from among the routes calculated by the route calculation unit 113. More specifically, the learning data generation unit 115 uses a predetermined method to select one route from among the routes determined to be connected by the communication determination unit 114 at each scenario time. Next, the learning data generation unit 115 acquires a label, which is identification information for the route calculation rule used when calculating the selected route, and stores the label in the label 146.

[0039] The predetermined method includes a random selection method, a method of treating labels (identification information of empirical route rules) as numerical values ​​and selecting the minimum / maximum label, and a method of selecting a label that is identification information of the route calculation rule used when calculating the route that best satisfies the predetermined evaluation conditions. The predetermined evaluation conditions are, for example, packet arrival rate, bandwidth, delay, or a combination of these, predicted based on the link information that makes up the route.

[0040] Next, the learning data generation unit 115 generates learning data that is identification information of a path calculation rule whose explanatory variable is the link information 144 and whose objective variable is the label 146 at each scenario time, and stores the generated learning data in the learning data 150.

[0041] <<Control Unit: Learning Unit>> The learning unit 116 generates a path calculation rule inference model, which is a machine learning model, using the learning data 150, and stores the model in a path calculation rule inference model database 160. A path calculation rule inference model is generated for each target node 142 (see FIG. 3 ).

[0042] <<Control Unit: Orbit and Environment Information Acquisition Unit>> The orbit and environment information acquisition unit 117 acquires information about the future orbits of the satellites 511 to 514 and stores it in orbit information 133. The orbit and environment information acquisition unit 117 also acquires future weather conditions and the like and stores them in environment information 134.

[0043] <<Control Unit: Inference Unit>> The inference unit 118 infers a route calculation rule to be used for calculating a route between two nodes in a network model in the future. More specifically, the inference unit 118 infers a route calculation rule (identification information (label) of the route calculation rule) based on link information using a route calculation rule inference model.

[0044] <<Control Unit: Route Setting Unit>> The route setting unit 119 sets a route table that reflects routes calculated using a future network model in the nodes of the non-terrestrial network 500. More specifically, the route calculation unit 113 calculates routes for two nodes of the non-terrestrial network 500 in the future network model according to the route calculation rule inferred by the inference unit 118. Next, the route setting unit 119 generates a route table for each node included in the non-terrestrial network 500 based on all routes between the two nodes, and transmits the route table to the node. Each node forwards communication packets according to the route table.

[0045] <Learning Data Generation Process> Figure 4 is a flowchart of the learning data generation process according to this embodiment. The learning data generation process is a process for generating learning data for generating a machine learning model that infers a route calculation rule used when calculating a route between two nodes in the non-terrestrial network 500. At the start of the learning data generation process, the scenario database 140 (see Figure 3) is assumed to store a scenario 141 in which node information and environment information 143 for the two nodes is stored. The learning data generation process is a process that is repeated for each pair of two nodes included in the non-terrestrial network 500.

[0046] In step S11, the learning data generation unit 115 starts the process of repeating steps S12 to S19 for each scenario 141. In step S12, the link analysis unit 111 calculates a scenario time. More specifically, the link analysis unit 111 calculates a scenario time with a fixed-length interval and stores it in the scenario time of the link information 144 (see FIG. 3). The link analysis unit 111 may calculate the scenario time with a variable-length interval so that the amount of change in the node information / environment information 143 falls within a predetermined range. In step S13, the learning data generation unit 115 starts the process of repeating steps S14 to S19 for each scenario time calculated in step S12.

[0047] In step S14, the link analysis unit 111 calculates link information at the scenario time. In step S15, the network model generation unit 112 generates a network model based on the link information calculated in step S14.

[0048] In step S16, the route calculation unit 113 starts the process of repeating steps S17 to S18 for each route calculation rule included in the route calculation rule 132. Hereinafter, the route calculation rule that is the target of this process will be referred to as the target route calculation rule. In step S17, the route calculation unit 113 calculates a route between two nodes in the network model generated in step S15 according to the target route calculation rule. In step S18, the communication determination unit 114 determines whether the route calculated in step S17 is connected, and stores the determination result (yes / no) in the communication determination result 145 (see FIG. 3 ).

[0049] In step S19, the learning data generation unit 115 uses a predetermined method to select one route from the routes determined to be connected in step S18, acquires a label that is identification information of the route calculation rule (the route calculation rule to be processed) used when calculating the selected route, and stores the label in label 146 (see FIG. 3 ). Next, the learning data generation unit 115 generates learning data in which the explanatory variable is link information 144 and the objective variable is the acquired label, and stores the learning data in learning data 150.

[0050] <<Route Calculation Rule Inference Model Generation Process>> Fig. 5 is a flowchart of the route calculation rule inference model generation process according to this embodiment. The route calculation rule inference model generation process is a process for generating a route calculation rule inference model that infers a route calculation rule used when calculating a route between two nodes in the non-terrestrial network 500. The route calculation rule inference model generation process is a process that is executed subsequent to the learning data generation process (see Fig. 4).

[0051] In step S31, the learning unit 116 trains the route empirical inference model using one piece of learning data stored in the learning data 150. In step S32, the learning unit 116 measures the accuracy of the route empirical inference model using accuracy measurement data prepared in advance. If the accuracy exceeds a predetermined threshold (step S32 → YES), the learning unit 116 proceeds to step S33, and if the accuracy is less than the predetermined threshold (step S33 → NO), the learning unit 116 returns to step S31 to process the next piece of learning data. In step S33, the learning unit 116 stores the route empirical inference model in the route calculation rule inference model database 160.

[0052] 6 is a flowchart of the route setting process according to this embodiment. At the start of the route setting process, the positions and velocities of the satellites 511 to 514 for a predetermined future period are stored in the orbit information 133, and environmental information including meteorological conditions around the ground stations 521 to 524 is stored in the environmental information 134. It is also assumed that a route heuristic inference model has already been generated.

[0053] In step S41, the route setting unit 119 starts the process of repeating steps S42 to S46 for each future time of a predetermined period. In step S42, the route setting unit 119 starts the process of repeating steps S43 to S46 for each pair of two nodes included in the non-terrestrial network 500. Hereinafter, the two nodes that are the targets of this process will be referred to as the processing target nodes.

[0054] In step S43, the link analysis unit 111 calculates link information at a future time based on the trajectory information 133 and the environmental information 134. In step S44, the network model generation unit 112 generates a network model for the future (at a future time) based on the link information calculated in step S43.

[0055] In step S45, the inference unit 118 infers a route calculation rule to be used for calculating a route between the nodes to be processed. In step S46, the route calculation unit 113 calculates a route between the nodes to be processed using the route calculation rule inferred in step S45 in the network model generated in step S44.

[0056] In step S47, the route setting unit 119 generates a route table for each node included in the non-terrestrial network 500 based on the route calculated in step S46. In step S48, the route setting unit 119 transmits the route table created in step S47 to the nodes. Each node sets a route table and forwards communication packets in accordance with the route table.

[0057] <Features of the Route Control Device> The route control device 100 calculates link information between nodes based on the future positions (see trajectory information 133) of the nodes constituting the non-terrestrial network 500 and environmental information 134 such as weather. Next, the route control device 100 calculates a route calculation rule using a route calculation rule inference model based on the link information, and calculates a route between two nodes using the route calculation rule. Next, the route control device generates a route table for each node based on the route and transmits it to the node. Each node forwards communication packets based on the route table.

[0058] Each node forwards communication packets based on the routing table generated by the routing control device 100, enabling communication in the non-terrestrial network 500. By calculating routes by inferring route calculation rules, the routing control device 100 can calculate routes with faster communication than if all route calculation rules were used to calculate and select routes. Furthermore, the routing control device 100 generates a routing table by calculating routes in advance based on the future positions and environments of the nodes. This enables stable communication even in non-terrestrial networks where the topology and environment change drastically.

[0059] By using a route rule-of-thumb inference model that infers a route calculation rule that calculates a route that best satisfies a predetermined evaluation condition, communication that satisfies the evaluation condition becomes possible in the non-terrestrial network 500. For example, by using the packet arrival rate as an evaluation condition, communication that satisfies the predetermined packet arrival rate becomes possible even if there is a large delay.

[0060] <<Modification: Scenario>> In the above embodiment, the scenario 141 (see FIG. 3) is a scenario related to the two nodes indicated by the target node 142, but it may also be a scenario related to all nodes. In this case, the link information 144 is link information related to all links in the non-terrestrial network 500, and the communication determination result 145 and the label 146 include the communication determination result and the label for each pair of two nodes.

[0061] <<Modification: Route heuristic inference model>> In the above-described embodiment, a route heuristic inference model is provided for each pair of two nodes in the non-terrestrial network 500. The pair may be included in the explanatory variables of the route heuristic inference model, forming a single route heuristic inference model.

[0062] <Other Modifications> Although several embodiments and modifications of the present invention have been described above, these embodiments are merely examples and do not limit the technical scope of the present invention. For example, the route control device 100 includes a learning data generation unit 115 and a learning unit 116 and generates a route heuristic inference model, but a separate model generation device may generate the route heuristic inference model. The route control device 100 may calculate routes and create a route table using route calculation rules inferred by the route heuristic inference model generated by the model generation device.

[0063] The present invention can take on various other embodiments, and various modifications such as omissions and substitutions can be made without departing from the spirit of the present invention. These embodiments and modifications are included in the scope and spirit of the invention described in this specification, etc., and are also included in the scope of the invention described in the claims and their equivalents.

[0064] <Hardware Configuration> The routing control device 100 according to the embodiment described above is realized by a computer 900 having a configuration as shown in Fig. 7, for example. Fig. 7 is a hardware configuration diagram showing an example of a computer 900 that realizes the functions of the routing control device 100 according to the embodiment. The computer 900 includes a CPU 901, a ROM 902, a RAM 903, an SSD 904, an input / output interface 905 (referred to as an input / output I / F (Interface) in Fig. 7), a communication interface 906 (referred to as a communication I / F in Fig. 7), and a media interface 907 (referred to as a media I / F in Fig. 7). The computer 900 may include a hard disk drive (HDD) instead of the SSD 904, or may include a HDD in addition to the SSD 904.

[0065] The CPU 901 operates based on programs stored in the ROM 902 or the SSD 904, and performs control by the control unit 110 in Fig. 2. The ROM 902 stores a boot program executed by the CPU 901 when the computer 900 starts up, programs related to the hardware of the computer 900, and the like. The CPU 901 controls an input device 910 such as a mouse or keyboard, and an output device 911 such as a display or printer, via an input / output interface 905. The CPU 901 acquires data from the input device 910 and outputs generated data to the output device 911 via the input / output interface 905.

[0066] The SSD 904 stores programs executed by the CPU 901 and data used by the programs. The communication interface 906 receives data from other devices (not shown) (e.g., satellites 511-514 and ground stations 521-524) via a communication network and outputs the data to the CPU 901. It also transmits data generated by the CPU 901 to other devices via the communication network. The media interface 907 reads programs or data stored on a recording medium 912 and outputs the programs to the CPU 901 via the RAM 903. The CPU 901 loads the programs from the recording medium 912 onto the RAM 903 via the media interface 907 and executes the loaded programs. The recording medium 912 may be an optical recording medium such as a DVD (Digital Versatile Disk), a magneto-optical recording medium such as an MO (Magneto Optical Disk), a magnetic recording medium, a conductive memory tape medium, or a semiconductor memory.

[0067] For example, when the computer 900 functions as the route control device 100 according to the embodiment described above, the CPU 901 of the computer 900 executes a program 138 (see FIG. 2) loaded onto the RAM 903, thereby realizing the functions of the route control device 100. The CPU 901 reads and executes the program from a recording medium 912. Alternatively, the CPU 901 may read the program from another device via a communication network, or may install and execute the program 138 from the recording medium 912 onto the SSD 904.

[0068] <Effects> The effects of the route control device 100 will be described below.

[0069] The route control device 100 according to the above embodiment includes an inference unit 118 that infers a route calculation rule to be used when calculating a route between two nodes included in a dynamic network (non-terrestrial network 500) based on link information between the nodes that make up the dynamic network, using a route calculation rule inference model (see the route calculation rule inference model database 160), which is a machine learning model. The route control device 100 also includes a route calculation unit 113 that calculates a route between the two nodes using the inferred route calculation rule.

[0070] The routing control device 100 can calculate a route that provides faster communication than calculating and selecting a route using all the route calculation rules, and can therefore perform route calculation (route control) in a dynamic network where the environment and topology change drastically.

[0071] The routing control device 100 according to the above-described embodiment further includes a link analysis unit 111 and a route setting unit 119. The link analysis unit 111 calculates link information based on the position of a node at a future time and environmental information about the node. The inference unit 118 infers the route calculation rule based on the link information at the future time. The route calculation unit 113 calculates a route between two nodes at the future time based on the route calculation rule. The route setting unit 119 creates a route table for nodes in the dynamic network based on the calculated route and transmits the route table to the corresponding node.

[0072] Such a routing control device 100 generates a routing table by calculating routes in advance based on the future positions and environments of nodes, thereby enabling stable communication even in dynamic networks where the topology and environment change drastically.

[0073] The explanatory variables of the learning data 150 used to generate the path calculation rule inference model according to the above-described embodiment are link information between nodes. The objective variable is a path calculation rule, among one or more path calculation rules, used to calculate a path between two nodes included in a network model of a dynamic network configured from links indicated by the link information, which is communicable and best satisfies predetermined evaluation conditions.

[0074] Such a route control device 100 enables communication that satisfies the evaluation conditions in a dynamic network. Examples of the predetermined evaluation conditions include packet arrival rate, bandwidth, delay, and combinations of these, all of which are predicted based on link information that constitutes the route.

[0075] The link information according to the above embodiment includes at least one of propagation distance, received signal strength, location of nodes at the end points of the link, propagation loss, and rain attenuation.

[0076] Such a routing control device 100 calculates routes using a network model based on link information suitable for wireless communication, thereby enabling routing control suitable for real dynamic networks.

[0077] 100 Route control device 110 Control unit 111 Link analysis unit 112 Network model generation unit 113 Route calculation unit 114 Communication determination unit 115 Learning data generation unit 116 Learning unit 117 Trajectory and environment information acquisition unit 118 Inference unit 119 Route setting unit 130 Storage unit 131 Propagation model 132 Route calculation rule 133 Trajectory information 134 Environment information 138 Program 140 Scenario database 144 Link information 150 Learning data 160 Route calculation rule inference model database 500 Non-terrestrial network (dynamic network) 511 to 514 Satellites (nodes) 521 to 524 Ground stations (nodes)

Claims

1. A route control device comprising: an inference unit that infers a route calculation rule to be used when calculating a route between two nodes included in a dynamic network based on link information between nodes that constitute the dynamic network, using a route calculation rule inference model, which is a machine learning model; and a route calculation unit that calculates a route between the two nodes using the inferred route calculation rule.

2. A route control device as described in claim 1, further comprising a link analysis unit and a route setting unit, wherein the link analysis unit calculates the link information based on the position of the node at a future time and environmental information of the node, the inference unit infers the route calculation rule based on the link information at the future time, the route calculation unit calculates a route between the two nodes at the future time based on the route calculation rule, and the route setting unit creates a route table for nodes in the dynamic network based on the calculated route and transmits the route table to the corresponding node.

3. A route control device as described in claim 1, wherein the learning data used to generate the route calculation rule inference model has explanatory variables that are link information between nodes, and a target variable that is a route calculation rule, among one or more route calculation rules, used to calculate a route between two nodes included in a network model of the dynamic network composed of links indicated by the link information, which is communicable and best satisfies specified evaluation conditions.

4. The route control device according to claim 1, wherein the link information includes at least one of a propagation distance, a received signal strength, a location of a node that is an end point of the link, a propagation loss, and an amount of rain attenuation.

Images