Inference device, learning device and unit arrangement place proposal method

Abstract

To propose a unit arrangement place adapted to a geography without using manpower.SOLUTION: An inference device 20 according to the present disclosure comprises: a data acquisition unit 21 that acquires an object area in which a unit acts, a current place of the unit, a target point in movement of the unit, and a movement restriction area; and an inference unit 22 that creates a slime mold network which connects the current place and the target point, avoiding the movement restriction area within the object area on the basis of a learned model for inferring the slime mold network which connects the current place and the target point using the object area, the current place, the target point, and the movement restriction area as input, and outputs a unit arrangement place that is a standby place until when the unit starts movement at the current place to the target point on the slime mold network.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device that suggests unit placement locations. [Background technology]

[0002] Patent Document 1 discloses a conventional information processing device for unit command training that uses the battle situation calculated from weather and terrain to make a situation judgment in accordance with situation judgment rules, and then uses the situation judgment results to create orders in accordance with order creation rules. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2005-283030 Summary of the Invention [Problem to be solved by the invention]

[0004] In the above-described conventional information processing apparatus, when creating a command, it is necessary to manually set situation judgment rules, command creation rules, and other action rules in advance.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide an inference device that can propose unit placement locations that are appropriate for the terrain without human intervention. [Means for solving the problem]

[0006] The inference device of the present disclosure comprises: The system is equipped with an inference unit that acquires information on the target area in which the unit will operate, the unit's current location, multiple destination points for the unit's movement, and movement restriction areas, calculates multiple routes from the current location to the multiple destination points based on a trained model for inferring routes from the current location to the multiple destination points, and outputs the intersection of the multiple routes as the unit's placement location, which is a waiting area until the unit begins moving to one of the multiple destination points. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to propose unit placement locations that are appropriate for the terrain without relying on human intervention. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a configuration diagram of a learning device according to a first embodiment. [Figure 2] 4 is a flowchart showing the operation of the learning device according to the first embodiment. [Figure 3] 1 is a configuration diagram of an inference device according to a first embodiment. [Figure 4] 4 is a flowchart showing the operation of the inference device according to the first embodiment. [Figure 5] FIG. 10 is a configuration diagram of a learning device and an inference device according to a second embodiment. [Figure 6] 10 is a flowchart showing the operation of the learning device according to the second embodiment. [Figure 7] 10 is a flowchart showing the operation of the inference device according to the second embodiment. [Figure 8] FIG. 1 is a schematic diagram of a neural network. DETAILED DESCRIPTION OF THE INVENTION

[0009] The inference device disclosed herein takes into account the terrain of the field when conducting a training simulation related to unit command training, for example, and proposes optimal unit placement locations for units to reach each of multiple target points.

[0010] In this disclosure, a unit is a person, a moving object, or a group of these. A unit moves to a unit location at the command of a commander. The unit location is a location where the unit waits until the commander gives instructions, i.e., until the unit begins moving to a target location.

[0011] In addition, in this disclosure, terrain refers to geographical information including plains, mountains, rivers, or buildings in the target area in which the unit operates. The unit mainly moves on plains, and mountains, rivers, and buildings are included in movement restriction areas.

[0012] Conventionally, when proposing unit placement locations using an information processing device, it was necessary to set up in advance for each field mountains, rivers, obstacles and other movement restriction areas, as well as corresponding unit behavior rules.

[0013] Because multiple fields are set up for each training content, setting movement restriction areas and behavior rules for each field requires a significant amount of manpower and time, which has become a further obstacle to the demand for increased training opportunities.

[0014] Furthermore, because movement restriction areas and unit behavior rules are set rule-based, unit placement locations are also determined based on the rules. Therefore, even if a unit placement location is determined, it may not necessarily be optimal in terms of movement time or obstacle avoidance, requiring manual correction. In contrast, the information processing device disclosed herein proposes optimal unit placement locations without manual intervention.

[0015] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following description, like components are denoted by like reference numerals, and their names and functions are the same or similar. Therefore, detailed descriptions thereof may be omitted.

[0016] Embodiment 1 <Learning Phase> ***Configuration Description*** An information processing device 100 according to the first embodiment includes a learning device 10, an inference device 20, a trained model storage unit 30, and a unit placement location input unit 40. The learning device 10 will be described with reference to FIG. 1. FIG. 1 is a configuration diagram of the learning device 10. The learning device 10 includes a data acquisition unit 11 and a model generation unit 12.

[0017] The data acquisition unit 11 acquires the target area, the current location of the unit, the target location, the movement restriction area, and the slime mold network as learning data.

[0018] Slime molds are single-celled organisms, specifically those known as true slime molds. Slime molds are known to calculate rational routes connecting multiple feeding areas while avoiding light sources. Each route is generated so that it is the shortest overall route between feeding areas and has multiple branches. Furthermore, each route has the characteristic that the more optimal routes are displayed in thicker colors.

[0019] In this disclosure, a slime mold network refers to a set of routes generated by a slime mold when a unit's current location and destination are considered to be feeding grounds, and mountains, rivers, obstacles, the areas of operation of other units with different chains of command, and other movement-restricted areas are considered to be light sources. Each route connects the current location and one of the destinations, avoiding the movement-restricted areas. When there is one destination, the route is the shortest route between the current location and the destination.

[0020] In other words, the data acquisition unit 11 acquires, as learning data, information on the slime mold network generated by the slime mold moving through an area considered to be the target region, and information on any of the locations that were feeding grounds at the time the slime mold network was generated as the current location and target location, and information on any location that was a light source at the time the slime mold network was generated as the movement restricted area.

[0021] The training data may be input as an image showing a slime mold network, for example, an image showing a slime mold network generated by slime mold moving through an area where feeding areas and light sources are arranged to mimic real terrain.

[0022] When acquiring learning data from an image, the data acquisition unit 11 acquires an image showing the branching points (nodes) of the slime mold network and the lines (links) connecting the branching points (nodes), as well as information on the locations of feeding sites and light sources in the image. Here, the entire image is considered to be the target area, the feeding sites are considered to be the current location or destination, and the light sources are considered to be movement-restricted areas. Next, the data acquisition unit 11 identifies the branching points (nodes) and lines (links) through image processing and acquires their coordinates as the slime mold network.

[0023] The model generation unit 12 learns the relationship between the current location, destination point, and movement-restricted area of ​​the unit and the slime mold network based on the learning data output from the data acquisition unit 11. Furthermore, the model generation unit 12 receives the target area, the current location, destination point, and movement-restricted area of ​​the unit as input and generates a trained model that infers a slime mold network that connects the current location and destination point within the target area while avoiding the movement-restricted area.

[0024] The learning algorithm used by the model generation unit 12 may be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. As an example, a case where a neural network is applied will be described.

[0025] The model generation unit 12 learns the slime mold network, for example, by supervised learning according to a neural network model. Here, supervised learning refers to a method in which pairs of input and result (label) data are provided to a learning device, which learns the features of the learning data and infers the result from the input.

[0026] A neural network consists of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer, or two or more layers.

[0027] Figure 8 is a schematic diagram of a neural network. For example, in a three-layer neural network like the one shown in Figure 8, when multiple inputs are input to the input layer (X1-X3), the values ​​are multiplied by weight W1 (w11-w16) and input to the intermediate layer (Y1-Y2), and the result is further multiplied by weight W2 (w21-w26) and output from the output layer (Z1-Z3). This output result changes depending on the values ​​of weights W1 and W2.

[0028] In the present disclosure, the neural network learns a slime mold network that connects the unit's current location and target location within the target area while avoiding the restricted movement areas, using so-called supervised learning, in accordance with learning data consisting of a combination of the target area, the unit's current location, the target location, the restricted movement areas, and the slime mold network acquired by the data acquisition unit 11.

[0029] In other words, the neural network learns by inputting the target area, the unit's current location, the destination point, and the movement restriction area into the input layer, and adjusting the weights W1 and W2 so that the results output from the output layer approach the slime mold network.

[0030] The model generation unit 12 generates and outputs a trained model by executing the above-described learning.

[0031] The trained model storage unit 30 stores the trained model output from the model generation unit 12.

[0032] The unit placement location input unit 40 connects the information processing device 100 with external software and inputs information about the unit placement location to the external software. The external software is, for example, software that simulates the behavior of units.

[0033] The hardware configuration of the learning device 10 will now be described. The learning device 10 includes a processor, a main memory device, an auxiliary memory device, a communication device, and an input / output device. The auxiliary memory device stores a program that implements the functions of the data acquisition unit 11 and the model generation unit 12. The main memory device also reads the program from the auxiliary memory device. The processor executes the program.

[0034] The learning device 10 is used to learn the slime mold network in the information processing device 100, but may be a device separate from the information processing device 100, connected to the information processing device 100 including the inference device 20, for example, via a network. The learning device 10 and the inference device 20 may also be built into the information processing device 100. Furthermore, the learning device 10 and the inference device 20 may reside on a cloud server. In this case, the information processing device 100 may be equipped with a communication device and a trained model storage unit 30 for communication, and may perform information communication with the learning device 10 and the inference device 20.

[0035] ***Explanation of Operation*** Next, the learning process of the learning device 10 will be described with reference to Fig. 2. Fig. 2 is a flowchart showing the learning process of the learning device 10.

[0036] In step b1, the data acquisition unit 11 acquires the target area, the unit's current location, the destination point, the movement-restricted area, and the slime mold network. Note that, although the target area, the unit's current location, the destination point, the movement-restricted area, and the slime mold network are acquired simultaneously, it is sufficient if these pieces of information are input in association with each other, and they may be acquired at different times.

[0037] In step b2, the model generation unit 12 uses the learning data acquired by the data acquisition unit 11, which consists of a combination of the target area, the unit's current location, the target location, the movement-restricted area, and the slime mold network, to learn a slime mold network that connects the current location and the target location within the target area, avoiding the movement-restricted area, through so-called supervised learning, using the target area, the current location, the target location, and the movement-restricted area as input, and generates a learned model.

[0038] In step b3, the trained model storage unit 30 stores the trained model generated by the model generation unit 12.

[0039] <Utilization phase> ***Configuration Description***

[0040] 3 is a configuration diagram of the inference device 20 related to the information processing device 100. The inference device 20 includes a data acquisition unit 21 and an inference unit 22.

[0041] The data acquisition unit 21 acquires the target area, the unit's current location, the target location, and the movement-restricted area. Here, the movement-restricted area may be accompanied by terrain attribute information indicating the terrain or use of the area. The terrain attribute information may be, for example, mountains, rivers, obstacles, or the movement areas of other units. The data acquisition unit 21 also sets the movement-restricted area in the target area.

[0042] At this time, the data acquisition unit 21 may acquire information on the target area and the movement restricted area using image data, maps, or GIS data showing the topography of the target area, and set the movement restricted area.

[0043] Furthermore, after setting the movement-restricted area obtained from image data, maps, or GIS data, the data acquisition unit 21 may acquire information on the current location and destination point as input from the user of the inference device 20, and further set the movement-restricted area. The operation in this case will be described later. The image data may be satellite observation images obtained using an observation satellite orbiting the Earth.

[0044] The inference unit 22 infers the location of the unit from the slime mold network obtained using the trained model. The inference unit 22 infers a slime mold network that connects the current location and the target location within the target area while avoiding the restricted movement areas by inputting the target area, the current location of the unit, the target location, and the restricted movement areas acquired by the data acquisition unit 21 into the trained model stored in the trained model storage unit 30. The inference unit 22 also outputs the location of the unit on the slime mold network.

[0045] Here, each point in the inferred slime mold network is not necessarily connected by a single line, but is composed of multiple routes and branches to maintain robustness. As mentioned above, the slime mold network forms an optimal route connecting multiple points, so intersections composed of multiple branches can be considered important places on the way to reaching the goal.

[0046] The hardware configuration of the inference device 20 will now be described. The inference device 20 includes a processor, a main memory device, an auxiliary memory device, a communication device, and an input / output device. The auxiliary memory device stores a program that realizes the functions of the data acquisition unit 21 and the inference unit 22. The main memory device loads the program from the auxiliary memory device. The processor executes the program.

[0047] ***Explanation of Operation*** Next, a process for obtaining a unit placement location using the inference device 20 will be described with reference to Fig. 4. Fig. 4 is a flowchart showing the operation of the inference device 20. The method by which the inference device 20 obtains a unit placement location will be referred to as a unit placement location suggestion method.

[0048] In step c1, the data acquisition unit 21 acquires the target area, the unit's current location, the destination location, and the movement-restricted area. Here, the target area and the movement-restricted area may be acquired from image data, a map, or GIS data showing the topography of the target area. In this case, the data acquisition unit 21 acquires the area shown in the image data, map, or GIS data as the target area. Furthermore, the data acquisition unit 21 acquires the movement-restricted area by image processing or by topographical attribute information set in advance in the map or GIS data. Step c1 is a data acquisition process in the unit placement location suggestion method.

[0049] In step c2, the data acquisition unit 21 associates and sets information about the movement-restricted area on the target area acquired in step c1. After setting the movement-restricted area previously acquired in step c1, the data acquisition unit 21 may further set a movement-restricted area based on input from the user of the inference device 20. The additional movement-restricted area to be set is, for example, the movement area of ​​another unit, or an area where it has become difficult for a unit to move due to changes in the terrain after the image data, map, or GIS data was created. Step c2 is a movement-restricted area setting process in the unit placement location suggestion method.

[0050] In step c3, the inference unit 22 acquires the trained model from the trained model storage unit 30. Next, the inference unit 22 inputs the target area, the unit's current location, the destination point, and the movement restriction area into the trained model to obtain a slime mold network. Step c3 is a slime mold network generation step in the unit placement location suggestion method.

[0051] In step c4, the inference unit 22 selects an intersection of multiple routes constituting the slime mold network obtained in step c3 as a candidate unit placement location. Here, the point selected as the intersection may be a circular area with a certain range. Step c4 is a unit placement location candidate selection process in the unit placement location suggestion method.

[0052] In step c5, the inference unit 22 determines, as a proposed unit placement location, a point with a large number of branches from among the intersections of multiple routes that are candidate unit placement locations selected in step c4. Step c5 is a unit placement location determination step in the unit placement location proposal method.

[0053] Here, the inference unit 22 may determine, as the unit placement location, a point among the intersections of the routes where the angle formed by adjacent routes is a certain value or more, a point where the distance between nearby intersections is a certain value or more, or a point where the constituent routes are thicker.

[0054] In step c6, the information processing device 100 inputs the determined unit placement location to the unit placement location input unit 40. The unit placement location input unit 40 inputs the unit placement location to software external to the information processing device 100. This makes it possible to automatically generate a unit placement location proposal that takes into account obstacles on the way to reaching the target point, without manual correction. Step c6 is a unit placement location input step in the unit placement location proposing method.

[0055] When moving from step c5 to step c6, the user may instruct the execution of input. In this case, after step c5, the inference unit 22 waits for the user's input before performing the operation of step c6.

[0056] ***Other Configurations*** <Variations of the learning phase> In this embodiment, a case where supervised learning is applied to the learning algorithm used by the model generation unit has been described, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to supervised learning.

[0057] Furthermore, the learning algorithm used in the model generation unit 12 can be deep learning, which learns to extract the features themselves, or machine learning can be performed according to other known methods, such as genetic programming, inductive logic programming, or support vector machines.

[0058] The model generation unit 12 may learn unit placement locations in accordance with learning data created for multiple information processing devices 100. Note that the model generation unit 12 may acquire learning data from multiple information processing devices 100 used in the same area, or may learn unit placement locations using learning data collected from multiple information processing devices 100 operating independently in different areas.

[0059] Additionally, a user of the information processing device 100 can add or remove information processing devices 100 that collect training data from the targets during the process. Furthermore, a learning device 10 that has trained a slime mold network for one information processing device 100 can be applied to another information processing device 100, and the slime mold network can be retrained and updated for the other information processing device 100. In other words, when multiple information processing devices 100 are used, the learning devices 10 of each information processing device 100 may train a slime mold network using the same training data.

[0060] <Variations of the Utilization Phase> In embodiment 1, the inference unit 22 is described as outputting the unit placement location using a trained model trained by the model generation unit 12 of the learning device 10, but it may also be configured to obtain a trained model from outside the information processing device 100 and output the unit placement location based on this trained model.

[0061] Furthermore, the inference unit 22 uses a trained model that uses a slime mold network to obtain a route connecting the current location to the destination point, but the trained model used here is not limited to this.

[0062] For example, the inference unit 22 may use a trained model based on another method that infers an optimal route connecting the current location to the multiple destination points while avoiding the movement restricted area, using the target area, the current location of the unit, multiple destination points, and the movement restricted area as input. Even in this case, the inference unit 22 may determine, as the unit placement location, a point with the most branches among the intersections of the routes connecting the current location and the multiple destination points.

[0063] Furthermore, even when proposing a unit placement location based on a slime mold network, the inference unit 22 may determine a route connecting the current location and multiple destination points using a pre-stored mathematical model that uses information on the current location, multiple destination points, and movement-restricted areas, without learning the slime mold network. Even in this case, the inference unit 22 may determine, as the unit placement location, a point with the most branches among the intersections of multiple routes from the current location to the multiple destination points.

[0064] ***Description of Effects of First Embodiment*** In this way, the inference device 20 is equipped with an inference unit 22 that acquires information on the target area, the unit's current location, multiple destination points, and movement-restricted areas, calculates multiple routes from the current location to the multiple destination points based on a trained model for inferring routes connecting the current location and the multiple destination points while avoiding the movement-restricted areas, and outputs the intersections of the multiple routes as proposed unit placement locations that will serve as waiting areas until the unit begins moving from its current location to one of the multiple destination points.This makes it possible to obtain unit placement locations on the optimal routes to the multiple destination points without human intervention, taking into account the terrain.

[0065] Furthermore, when proposing unit placement locations based on a trained model for inferring a slime mold network connecting the current location and the destination using the target area, current location, destination, and movement restriction area as input, the characteristics of the slime mold network make it possible to propose unit placement locations based on rational routes.

[0066] In other words, a route generated based on a trained model connects the current location and the destination point via the shortest overall distance and has multiple branches. This allows the inference device 20 to propose a unit placement location that is the optimal point for traveling from the current location to the destination point by inferring one of the intersections on the route as a unit placement location. As a result, the inference device 20 achieves the effect of reducing the effort required for manual corrections.

[0067] Embodiment 2 ***Configuration Description*** An information processing device 100b according to the second embodiment will now be described. Fig. 5 shows the configuration of the information processing device 100b. In Fig. 5, the same reference numerals as those in Fig. 1 or 3 indicate the same or corresponding components.

[0068] The information processing device 100b includes a learning device 10b, an inference device 20b, a learned model storage unit 30, and a learned model storage unit 30b. The learning device 10b further includes a data acquisition unit 11b and a model generation unit 12b compared to the learning device 10. The inference device 20b further includes an inference unit 22b compared to the inference device 20. The other configurations of the information processing device 100b are the same as those of the information processing device 100. The inference unit 22 and the inference unit 22b are also referred to as a first inference unit and a second inference unit.

[0069] The data acquisition unit 11b acquires the slime mold network and the unit placement locations on the slime mold network as learning data. The unit placement locations may be manually set for the slime mold network.

[0070] The model generation unit 12b learns the relationship between the slime mold network and unit placement locations based on the learning data output from the data acquisition unit 11b. The model generation unit 12b also generates and outputs a trained model that uses the slime mold network as input and infers unit placement locations.

[0071] The trained model storage unit 30b stores the trained model output from the model generation unit 12b.

[0072] The inference unit 22b acquires the slime mold network from the inference unit 22 and acquires the trained model from the trained model storage unit 30b. The inference unit 22 also uses the acquired trained model to infer unit placement locations from the slime mold network. That is, the slime mold network is input into this trained model to infer unit placement locations on the slime mold network.

[0073] <Learning Phase> ***Explanation of Operation*** The learning process of the learning device 10b will be described. Here, the operations of the data acquisition unit 11 and the model generation unit 12 are as described above. Fig. 6 is a flowchart relating to the learning process of the data acquisition unit 11b and the model generation unit 12b. In the learning device 10b, the operations of Fig. 2 and Fig. 6 may be performed in parallel, or, for example, the operation of Fig. 6 may be performed after the operation of Fig. 2 is completed.

[0074] In step b4, the data acquisition unit 11b acquires learning data consisting of a slime mold network and unit placement locations on the slime mold network.

[0075] In step b5, the model generation unit 12b learns the relationship between the slime mold network and the unit placement locations by so-called supervised learning in accordance with the learning data acquired by the data acquisition unit 11b, and generates a trained model.

[0076] In step b6, the trained model storage unit 30b stores the trained model generated by the model generation unit 12b.

[0077] <Utilization phase> ***Explanation of Operation*** Next, a process for obtaining a unit placement location using inference device 20b will be described with reference to Fig. 7. Fig. 7 is a flowchart showing the operation of inference device 20b. The method by which inference device 20b obtains a unit placement location will be called a unit placement location suggestion method.

[0078] In step c7, the data acquisition unit 21 acquires the target area, the current location of the unit, the destination point, and the location of the movement restricted area.

[0079] Here, the target area and movement-restricted area may be acquired from image data, maps, or GIS data showing the topography of the target area. In this case, the data acquisition unit 21 acquires the area shown in the image data, maps, or GIS data as the target area. The data acquisition unit 21 also acquires the movement-restricted area by image processing or by topographical attribute information preset in the maps or GIS data. Step c7 is a data acquisition step in the unit placement location suggestion method.

[0080] In step c8, the data acquisition unit 21 associates information about the movement-restricted area with the topography acquired in step c7 and sets it. After setting the movement-restricted area acquired in advance in step c1, the data acquisition unit 21 may further set other movement-restricted areas including the action areas of other units based on input from the user of the inference device 20b. Step c8 is a movement-restricted area setting step in the unit placement location suggestion method.

[0081] In step c9, the inference unit 22 acquires the trained model from the trained model storage unit 30. Next, the inference unit 22 inputs the unit's target area, current location, destination point, and movement restriction area into the trained model to obtain a slime mold network. The inference unit 22 outputs the slime mold network to the inference unit 22b. Step c9 is a slime mold network generation process in the unit placement location suggestion method.

[0082] In step c10, the inference unit 22b acquires the trained model from the trained model storage unit 30b. Next, the inference unit 22b inputs the slime mold network acquired from the inference unit 22 into the trained model and infers a unit placement location. Step c10 is a unit placement location determination step in the unit placement location proposal method.

[0083] In step c11, the information processing device 100b inputs the inferred unit placement location to the unit placement location input unit 40. The unit placement location input unit 40 inputs the unit placement location to software external to the information processing device 100. This makes it possible to automatically generate a unit placement location proposal that takes the terrain into consideration without manual correction. Step c11 is a unit placement location input step in the unit placement location proposal method.

[0084] When moving from step c10 to step c11, the user may instruct the execution of input. In this case, after step c10, the inference unit 22 waits for the user's input before performing the operation of step c11.

[0085] ***Description of Effects of Embodiment 2*** In the inference device 20b configured as described above, similar to the inference device 20, it is possible to propose unit placement locations that are suited to the terrain without manual intervention.

[0086] Furthermore, the inference device 20b infers unit placement locations using a trained model that has learned the relationship between the slime mold network and unit placement locations, thereby proposing unit placement locations based on multiple perspectives including the number of branches at intersections of routes on the network, the angle between adjacent routes, route width, distance between intersections, and other conditions.

[0087] Various aspects of the present disclosure are summarized below as appendices.

[0088] (Appendix 1) an inference unit that acquires information on a target area in which a unit will act, the current location of the unit, a plurality of destination points in the movement of the unit, and movement-restricted areas, calculates routes from the current location to each of the plurality of destination points based on a trained model for inferring routes connecting the current location and the plurality of destination points within the target area while avoiding the movement-restricted areas, and outputs intersections of the routes as unit placement locations that are waiting locations until the unit starts moving from the current location to one of the plurality of destination points; An inference device comprising: (Appendix 2) a data acquisition unit that acquires a target area in which a unit will act, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; an inference unit that creates a slime mold network connecting the current location and the target location within the target area while avoiding the movement-restricted area based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement-restricted area as input, and outputs a unit placement location on the slime mold network that is a waiting location until the unit starts moving from the current location to the target location; An inference device comprising: (Appendix 3) the inference unit outputs a point with a large number of branches among the intersections of a plurality of routes constituting the slime mold network as the unit placement location. 10. The inference device of claim 2. (Appendix 4) A learning device that generates a trained model for inferring a route from a unit's current location to a destination point using a target area in which the unit operates, the unit's current location, the destination point for the unit's movement, and a movement restriction area. (Appendix 5) a data acquisition unit that acquires a target area in which a unit will act, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; a first inference unit that outputs a slime mold network connecting the current location and the target location, which is acquired by the data acquisition unit based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement-restricted area; a second inference unit that uses the slime mold network output by the first inference unit as an input and infers a unit placement location based on a trained model for inferring a unit placement location that is a waiting location until the unit starts moving from the current location to the target location; An inference device comprising: (Appendix 6) a data acquisition step of acquiring a target area in which a unit will operate, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; a slime mold network generation process for generating a slime mold network connecting the current location and the target location within the target area while avoiding the movement restricted area, based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement restricted area as input; a unit placement location determination step of determining a unit placement location on the slime mold network, the unit being a waiting location until the unit starts moving from the current location to the destination location; A unit placement location suggestion method comprising: [Explanation of symbols]

[0089] (10, 10b) learning device, (11, 11b) data acquisition unit, (12, 12b) model generation unit, (20, 20b) inference device, 21 data acquisition unit, (22, 22b) inference unit, (30, 30b) learned model storage unit, 40 unit placement location input unit, (100, 100b) information processing device

Claims

1. an inference unit that acquires information on a target area in which a unit will act, the current location of the unit, a plurality of destination points in the movement of the unit, and movement-restricted areas, calculates routes from the current location to each of the plurality of destination points based on a trained model for inferring routes connecting the current location and the plurality of destination points within the target area while avoiding the movement-restricted areas, and outputs intersections of the routes as unit placement locations that are waiting locations until the unit starts moving from the current location to one of the plurality of destination points; An inference device comprising:

2. a data acquisition unit that acquires a target area in which a unit will act, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; an inference unit that creates a slime mold network connecting the current location and the target location within the target area while avoiding the movement-restricted area based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement-restricted area as input, and outputs a unit placement location on the slime mold network that is a waiting location until the unit starts moving from the current location to the target location; An inference device comprising:

3. the inference unit outputs a point with a large number of branches among the intersections of a plurality of routes constituting the slime mold network as the unit placement location. The inference device according to claim 2 .

4. A learning device that generates a trained model for inferring a route from a unit's current location to a destination point using a target area in which the unit operates, the unit's current location, the destination point for the unit's movement, and a restricted movement area.

5. a data acquisition unit that acquires a target area in which a unit will act, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; a first inference unit that outputs a slime mold network connecting the current location and the target location, which is acquired by the data acquisition unit based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement-restricted area; a second inference unit that uses the slime mold network output by the first inference unit as an input and infers a unit placement location based on a trained model for inferring a unit placement location that is a waiting location until the unit starts moving from the current location to the target location; An inference device comprising:

6. a data acquisition step of acquiring a target area in which a unit will operate, a current location of the unit, a target point in the movement of the unit, and a movement restriction area; a slime mold network generation process for generating a slime mold network connecting the current location and the target location within the target area while avoiding the movement restricted area, based on a trained model for inferring a slime mold network connecting the current location and the target location using the target area, the current location, the target location, and the movement restricted area as input; a unit placement location determination step of determining a unit placement location on the slime mold network, the unit being a waiting location until the unit starts moving from the current location to the destination location; A unit placement location suggestion method comprising:

Images