Learning device, inference device, learning method, inference method, learning program, and inference program

Abstract

To reduce the labor required to set up a simulation of the behavior of a group of moving objects when the group moves while being influenced by factors that affect the group's movement.SOLUTION: A learning data acquisition unit 101 acquires learning data, which is data used for learning, indicating group movement influence factors that are factors affecting the movement of a group of two or more moving objects, and object movement influence factors that are factors affecting the fluid movement of a viscous object. A learning unit 102 regards a formation of the group when the group moves as a shape of the viscous object when the viscous object moves in a fluid manner, and uses the learning data to learn the formation of the group when the group moves while being influenced by the group movement influence factors, based on the shape of the viscous object when the viscous object moves in a fluid manner while being influenced by the object movement influence factors.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a technology for learning the movement patterns of a group of moving objects. [Background technology]

[0002] Patent Document 1 discloses a technique for simulating the behavior of a group of moving bodies, that is, a unit (a group of people, vehicles, etc.). More specifically, the technology of Patent Document 1 simulates the battle situation based on the equipment specifications of the own and enemy forces, the weather and topography of the battle area, etc., and calculates the engagement situation between the own and enemy forces. [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 Patent Document 1, when executing a simulation function, it is necessary to create in advance rules for calculation processing that take into account the equipment specifications of the own and enemy forces, the weather and terrain of the battle area, etc. In Patent Document 1, it is necessary to create rules for this calculation processing for each unit. Therefore, when simulating the movements of hundreds of units, there is a problem in that manually creating such rules is a heavy burden in terms of time and human costs.

[0005] The main objective of the present disclosure is to solve these problems. More specifically, the main purpose of the present disclosure is to reduce the effort required to set up a simulation of the behavior of a group of people as they move while being influenced by factors that affect their movement, such as weather and terrain. [Means for solving the problem]

[0006] The learning device according to the present disclosure includes: a learning data acquisition unit that acquires learning data, which is data used for learning, indicating a group movement influence factor that is a factor that influences the movement of a group of two or more moving objects and an object movement influence factor that is a factor that influences the fluid movement of a viscous object; The learning unit regards the formation of the group when the group moves as the shape of the viscous object when the viscous object moves in a fluid manner, and uses the learning data to learn the formation of the group when the group moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when the viscous object moves in a fluid manner while being influenced by the object movement influencing factor. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to reduce the effort required for setting up a simulation of the behavior of a group of people as they move while being influenced by factors that affect the movement of the group. [Brief explanation of the drawings]

[0008] [Figure 1] 2 is a diagram illustrating an example of a functional configuration of a learning device (formation learning) according to the first embodiment. FIG. [Figure 2] 1 is a diagram illustrating an example of a functional configuration of a learning device (placement learning) according to the first embodiment. [Figure 3] FIG. 3 is a diagram showing a processing flow of the learning device (formation learning) according to the first embodiment. [Figure 4] FIG. 4 is a diagram showing a processing flow of the learning device (placement learning) according to the first embodiment. [Figure 5] 1 is a diagram illustrating an example of a functional configuration of an inference device according to a first embodiment. [Figure 6] FIG. 3 is a diagram showing a processing flow of the inference device according to the first embodiment. [Figure 7] 1 is a diagram illustrating an example of a hardware configuration of a learning device (formation learning) according to the first embodiment. [Figure 8] 1 is a diagram illustrating an example of a hardware configuration of a learning device (placement learning) according to the first embodiment. [Figure 9] 1 is a diagram illustrating an example of a hardware configuration of an inference device according to a first embodiment. [Figure 10] FIG. 2 is a diagram illustrating an example of a supervised learning algorithm according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments will be described with reference to the drawings. In the following description of the embodiments and the drawings, the same reference numerals denote the same or corresponding parts.

[0010] Embodiment 1 In this embodiment, a group of two or more moving bodies will be described as a group of two or more people and / or vehicles. In the following, a group of people and / or vehicles will be referred to as a unit. A unit is, for example, a combat unit. In addition, "people and / or vehicles" will be written as "people / vehicles." In this embodiment, a method is described for determining the formation of a unit during movement and the placement of the people / vehicles that make up the unit within the formation, taking into account environmental information based on the battlefield, when issuing action commands to a unit during a training simulation. In this embodiment, the operation phases include a learning phase and a utilization phase. In the learning phase, machine learning (hereinafter simply referred to as learning) is performed on unit formations and personnel / vehicle placements to generate inference models for unit formations and personnel / vehicle placements. In the exploitation phase, the inference model generated in the learning phase is used to determine unit formations and personnel / vehicle placement. The system according to this embodiment includes learning device 100 and learning device 200 that operate in the learning phase, and inference device 300 that operates in the utilization phase. Below, we will explain the learning phase and the utilization phase in that order.

[0011] <Learning Phase> ***Configuration Description*** FIG. 1 shows an example of the functional configuration of a learning device 100 according to this embodiment. FIG. 7 shows an example of the hardware configuration of the learning device 100 according to this embodiment. The learning device 100 performs machine learning on the unit formation and generates an inference model for the unit formation. The operating procedure of learning device 100 corresponds to a learning method, and the program that realizes the operation of learning device 100 corresponds to a learning program.

[0012] The learning device 100 is a computer. As shown in FIG. 7, the learning device 100 includes, as hardware, a processor 911, a main memory device 912, an auxiliary memory device 913, a communication device 914, and an input / output device 915. 1, the learning device 100 includes, as functional components, a learning data acquisition unit 101 and a learning unit 102. The functions of the learning data acquisition unit 101 and the learning unit 102 are realized by, for example, a program. The auxiliary storage device 913 stores a program that realizes the functions of the learning data acquisition unit 101 and the learning unit 102. These programs are loaded from the auxiliary storage device 913 to the main storage device 912. Then, the processor 911 executes these programs to perform the operations of the learning data acquisition unit 101 and the learning unit 102, which will be described later. FIG. 7 schematically shows a state in which the processor 911 is executing a program that realizes the functions of the learning data acquisition unit 101 and the learning unit 102.

[0013] 1, a learning data acquisition unit 101 acquires learning data 1-1 (110) and learning data 1-2 (120). The learning data acquisition unit 101 outputs the learning data 1-1 (110) and learning data 1-2 (120) to a learning unit . The processing performed by the learning data acquisition unit 101 corresponds to learning data acquisition processing.

[0014] The learning data 1-1 (110) and the learning data 1-2 (120) are data used for learning in the learning unit 102. For example, the learning data 1-1 (110) includes the current location (current position) of the unit, the current location (current position) of the threat, the direction of the threat, and environmental information. A threat is a threat to the movement of the unit, such as an enemy force that is the opponent in combat. Environmental information is information about the surrounding environment when the unit moves. The environmental information indicates, for example, the weather conditions when the unit moves, the topography of the area through which the unit moves, etc. For example, the learning data 1-2 (120) indicates the location of an obstacle, the threat level of the threat, the size of a viscous object, and the viscosity of the viscous object. The viscous object is, for example, a Newtonian fluid. The obstacle is an obstacle that impedes the movement of a unit. The obstacle is, for example, a building on the movement path of a unit. The threat level of the threat is, for example, a numerical value that quantifies the combat capability of the threat. The learning data 1-2 (120) is treated as correct answer data.

[0015] The current location of the unit, the current location of the threat, the direction of the threat, and environmental information shown in the learning data 1-1 (110), and the location of obstacles and the threat level of the threat shown in the learning data 1-2 (120) are factors that affect the movement of the unit (hereinafter referred to as unit movement influence factors). The unit movement influence factors correspond to the group movement influence factors. Furthermore, the size and viscosity of the viscous object shown in the learning data 1-2 (120) are factors that affect the fluid movement of the viscous object (hereinafter referred to as object movement influencing factors).

[0016] The learning unit 102 performs learning using the learning data 1-1 (110) and the learning data 1-2 (120) to generate a first inference model 130, which is an inference model for the formation of units. The learning unit 102 regards the formation of the units when they move as the shape of a viscous object when it moves fluidly, and performs learning using the learning data 1-1 (110) and the learning data 1-2 (120). More specifically, the learning unit 102 learns the formation of the units when they move while being influenced by the unit movement influence factors, based on the shape of the viscous object when it moves fluidly while being influenced by the object movement influence factors. Through learning, the learning unit 102 generates an inference model 130 that can infer the optimal formation corresponding to the first inference data 310 described below in the utilization phase. The learning unit 102 stores the generated first inference model 130 in the first inference model memory device 140. In FIG. 1, the first inference model memory device 140 is an external device, but the first inference model memory device 140 may also be an auxiliary memory device 913 of the learning device 100. The processing performed by the learning unit 102 corresponds to a learning process.

[0017] FIG. 2 shows an example of the functional configuration of learning device 200 according to this embodiment. FIG. 8 shows an example of the hardware configuration of the learning device 200 according to this embodiment. The learning device 200 performs machine learning on the placement of people / vehicles within a unit formation and generates an inference model on the placement of people / vehicles within the formation. The operating procedure of learning device 200 corresponds to a learning method, and the program that realizes the operation of learning device 200 corresponds to a learning program.

[0018] The learning device 200 is a computer. As shown in FIG. 8, the learning device 200 includes, as hardware, a processor 921, a main memory device 922, an auxiliary memory device 923, a communication device 924, and an input / output device 925. 2, the learning device 200 includes, as functional components, a learning data acquisition unit 201 and a learning unit 202. The functions of the learning data acquisition unit 201 and the learning unit 202 are realized by, for example, a program. The auxiliary storage device 923 stores a program that realizes the functions of the learning data acquisition unit 201 and the learning unit 202. These programs are loaded from the auxiliary storage device 923 to the main storage device 922. Then, the processor 921 executes these programs to perform the operations of the learning data acquisition unit 201 and the learning unit 202, which will be described later. FIG. 8 schematically shows a state in which the processor 921 is executing a program that realizes the functions of the learning data acquisition unit 201 and the learning unit 202.

[0019] 2, the learning data acquisition unit 201 acquires learning data 2-1 (210) and learning data 2-2 (220). The learning data acquisition unit 201 outputs the learning data 2-1 (210) and learning data 2-2 (220) to the learning unit 202. The processing performed by the learning data acquisition unit 201 corresponds to learning data acquisition processing.

[0020] The learning data 2-1 (210) and the learning data 2-2 (220) are data used for learning in the learning unit 202. For example, the learning data 2-1 (210) indicates the formation, on-board equipment, positions of people / vehicles, and unit behavioral standards. The formation is the formation of a unit when it moves. On-board equipment is equipment carried by people or equipment mounted on vehicles. The position of people / vehicles is the current position (current coordinates) of people / vehicles within the formation. For example, training data 2-2 (220) shows durability and range. Durability is a numerical value that represents the unit's durability in combat with a threat. Range is the range when projected from a person / vehicle. Training data 2-2 (220) is treated as correct data.

[0021] The formation, onboard equipment, personnel / vehicle positions, and unit behavior criteria shown in learning data 2-1 (210) and the endurance and range shown in learning data 2-2 (220) are factors that affect the placement of personnel / vehicles within the formation (hereinafter referred to as placement influence factors).

[0022] The learning unit 202 performs learning using the learning data 2-1 (210) and the learning data 2-2 (220) to generate a second inference model 230, which is an inference model for the positioning of people / vehicles within a unit. In other words, through learning by the learning unit 202, a second estimation model can be obtained that estimates the position within the formation that a person / vehicle should next take. The learning unit 202 views the arrangement of people / vehicles within a unit as the arrangement of electrons within a molecule, and uses learning data 2-1 (210) and learning data 2-2 (220) to learn the arrangement of people / vehicles within the unit that is influenced by the arrangement influence factors, based on the probability of electrons existing within the molecule. Through learning, the learning unit 202 generates an inference model as the second inference model 230 that can infer the positioning of people / vehicles within a unit corresponding to the formation information and second inference data 320 described below during the utilization phase. The learning unit 202 stores the generated second inference model 230 in the second inference model memory device 240. In Figure 2, the second inference model memory device 240 is an external device, but the second inference model memory device 240 may also be an auxiliary memory device 923 of the learning device 200. The processing performed by the learning unit 202 corresponds to a learning process.

[0023] 1 and 2, learning device 100 and learning device 200 are assumed to be realized by separate hardware. However, learning device 100 and learning device 200 may be realized by the same hardware. Furthermore, learning device 100, learning device 200, and inference device 300 may be realized by the same hardware. Furthermore, the learning device 100, the learning device 200, and the inference device 300 may reside on a cloud server.

[0024] The learning algorithm used by the learning unit 102 and the learning unit 202 can be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. In the following, an example in which a neural network is applied as the learning algorithm will be described. The learning units 102 and 202 learn the unit formation and the placement of people / vehicles within the unit, for example, by so-called supervised learning according to a neural network model. Here, supervised learning refers to a method of providing a learning device with pairs of input and result (label) data to learn features in the learning data and infer a result from the input.

[0025] A neural network is composed 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.

[0026] For example, consider a three-layer neural network as shown in Figure 10. When multiple inputs are input to the input layer (X1-X3), the values ​​are multiplied by weight W1 (w11-w16), and the multiplied value is input to the hidden layer (Y1-Y2). The result is further multiplied by weight W2 (w21-w26), and the multiplied value is output from the output layer (Z1-Z3). This output result varies depending on the values ​​of weights W1 and W2.

[0027] In this embodiment, two neural networks are used.

[0028] The neural network of the learning device 100 performs so-called supervised learning in accordance with the learning data 1-1 (110) and the learning data 1-2 (120).

[0029] That is, the neural network of the learning device 100 learns by inputting learning data 1-1 (110) into the input layer and adjusting the weights W1 and W2 so that the result output from the output layer approaches learning data 1-2 (120) (correct answer).

[0030] The learning unit 102 generates the first inference model 130 by performing the above-described learning.

[0031] The neural network of the learning device 200 performs so-called supervised learning in accordance with the learning data 2-1 (210) and the learning data 2-2 (220).

[0032] That is, the neural network of the learning device 200 learns by inputting learning data 2-1 (210) into the input layer and adjusting the weights W1 and W2 so that the result output from the output layer approaches learning data 2-2 (220) (correct answer).

[0033] The learning unit 202 generates the second inference model 230 by performing the above-described learning.

[0034] ***Explanation of Operation*** 3 shows an example of the operation of the learning device 100. The operation of the learning device 100 will be described below with reference to FIG.

[0035] In step 101, the learning data acquisition unit 101 acquires learning data 1-1 (110) and learning data 1-2 (120) (correct answers). 3 shows an example in which the learning data acquisition unit 101 simultaneously acquires learning data 1-1 (110) and learning data 1-2 (120) (correct answer). As long as the learning data 1-1 (110) and learning data 1-2 (120) (correct answer) are acquired in association with each other, the learning data acquisition unit 101 may acquire the learning data 1-1 (110) and learning data 1-2 (120) (correct answer) at different times.

[0036] In step 102, the learning unit 102 performs so-called supervised learning using the learning data 1-1 (110) and the learning data 1-2 (120) (correct answer). As described above, the learning unit 102 performs learning about the formation of units when they move (formation learning) to generate a first inference model 130.

[0037] In step 103, the learning unit 102 stores the first inference model 130 in the first inference model storage device 140.

[0038] 4 shows an example of the operation of the learning device 200. The operation of the learning device 200 will be described below with reference to FIG.

[0039] In step 201, the learning data acquisition unit 201 acquires learning data 2-1 (210) and learning data 2-2 (220) (correct answer). 4 shows an example in which the learning data acquisition unit 201 simultaneously acquires learning data 2-1 (210) and learning data 2-2 (220) (correct answer). As long as the learning data 2-1 (210) and learning data 2-2 (220) (correct answer) are acquired in association with each other, the learning data acquisition unit 201 may acquire the learning data 2-1 (210) and learning data 2-2 (220) (correct answer) at different times.

[0040] In step 202, the learning unit 202 performs so-called supervised learning using the learning data 2-1 (210) and the learning data 2-2 (220) (correct answer). As described above, the learning unit 202 performs learning (placement learning) about the placement of people / vehicles within the unit when the unit moves, and generates a second inference model 230.

[0041] In step 203, the learning unit 202 stores the second inference model 230 in the second inference model storage device 240.

[0042] <Utilization phase> ***Configuration Description*** FIG. 5 shows an example of the functional configuration of inference device 300 according to this embodiment. FIG. 9 shows an example of the hardware configuration of inference device 300 according to this embodiment. The inference device 300 performs inference using the first inference model 130 and inference using the second inference model 230, and ultimately generates output data 330 indicating the positions of people / vehicles within the unit. The inference device 300 then outputs the output data 330 to the unit movement control processing device 500. The operating procedure of inference device 300 corresponds to an inference method, and the program that realizes the operation of inference device 300 corresponds to an inference program.

[0043] Reasoning device 300 is a computer. As shown in FIG. 9, the inference device 300 includes, as hardware, a processor 931, a main memory device 932, an auxiliary memory device 933, a communication device 934, and an input / output device 935. 5, the inference device 300 has, as its functional configuration, a first inference model management unit 301, a second inference model management unit 302, a first inference data acquisition unit 303, a second inference data acquisition unit 304, a first inference unit 305, and a second inference unit 306. The functions of the first inference model management unit 301, the second inference model management unit 302, the first inference data acquisition unit 303, the second inference data acquisition unit 304, the first inference unit 305, and the second inference unit 306 are realized, for example, by a program. The auxiliary memory device 933 stores programs that realize the functions of the first inference model management unit 301, the second inference model management unit 302, the first inference data acquisition unit 303, the second inference data acquisition unit 304, the first inference unit 305, and the second inference unit 306. These programs are loaded from the auxiliary memory device 933 to the main memory device 932. The processor 931 then executes these programs to perform the operations of the first inference model management unit 301, the second inference model management unit 302, the first inference data acquisition unit 303, the second inference data acquisition unit 304, the first inference unit 305, and the second inference unit 306, which will be described later. Figure 9 schematically shows the state in which processor 931 is executing programs that realize the functions of first inference model management unit 301, second inference model management unit 302, first inference data acquisition unit 303, second inference data acquisition unit 304, first inference unit 305, and second inference unit 306.

[0044] 5, the first inference model management unit 301 acquires the first inference model 130 and manages the first inference model 130. The first inference model management unit 301 also outputs the first inference model 130 to the first inference unit 305. Although not shown in Figure 5, the first inference model management unit 301 obtains the first inference model 130 from the first inference model storage device 140. As described above, the first inference model 130 is an inference model for inferring the formation of a unit, generated by formation learning by the learning device 100. The processing performed by the first inference model management unit 301 corresponds to the first inference model management processing.

[0045] The second inference model management unit 302 acquires the second inference model 230 and manages the second inference model 230. The second inference model management unit 302 also outputs the second inference model 230 to the second inference unit 306. Although not shown in Figure 5, the second inference model management unit 302 obtains the second inference model 230 from the second inference model storage device 240. As described above, the second inference model 230 is an inference model for inferring the placement of people / vehicles within a unit formation, generated by placement learning by the learning device 200. The processing performed by the second inference model management unit 302 corresponds to the second inference model management processing.

[0046] The first inference data acquisition unit 303 acquires the first inference data 310 output from the inference data output device 400. The first inference data acquisition unit 303 outputs the first inference data 310 to the first inference unit 305. The processing performed by the first inference data acquisition unit 303 corresponds to the first inference data acquisition processing.

[0047] The first inference data 310 is inference data used for inference by the first inference unit 305 . For example, the first inference data 310 includes the same information as the learning data 1-1 (110). That is, the first inference data 310 includes the current location (current position) of the unit, the current location (current position) of the threat, the direction of the threat, and environmental information. The current location of the unit, the current location of the threat, the direction of the threat, and environmental information shown in the first inference data 310 are factors that affect the movement of the unit (hereinafter referred to as unit movement influence factors). The unit movement influence factors correspond to mass movement influence factors.

[0048] The second inference data acquisition unit 304 acquires second inference data 320 output from the inference data output device 400. Furthermore, the second inference data acquisition unit 304 acquires information on the unit's formation (hereinafter referred to as formation information) from the first inference unit 305. The second inference data acquisition unit 304 outputs the second inference data 320 and the formation information to the second inference unit 306. The processing performed by the second inference data acquisition unit 304 corresponds to the second inference data acquisition processing.

[0049] The second inference data 320 is inference data used for inference by the second inference unit 306 . For example, the second inference data 320 includes information excluding the "formation" from the learning data 2-1 (210). That is, the second inference data 320 indicates the onboard equipment, the positions of people / vehicles, and the behavioral standards of the unit. The on-board equipment, positions of people / vehicles, and behavioral criteria indicated in the second inference data 320 are factors that affect the placement of people / vehicles within the unit (hereinafter referred to as placement influencing factors).

[0050] The first inference unit 305 uses the first inference data 310 and the first inference model 130 to infer the unit formation when the unit moves while being influenced by the unit movement influencing factors. That is, by inputting the first inference data 310 to the first inference model 130, the first inference unit 305 can infer the optimal unit formation when the unit moves while being influenced by the unit movement influencing factors indicated in the first inference data 310. The first inference unit 305 outputs the inference result, that is, formation information indicating the unit formation, to the second inference data acquisition unit 304. The processing performed by the first inference unit 305 corresponds to the first inference processing.

[0051] The second inference unit 306 uses the second inference data 320, the formation information, and the second inference model 230 to infer the placement of people / vehicles affected by placement influencing factors in the unit formation inferred by the first inference unit 305. That is, the second inference unit 306 can infer the optimal placement of people / vehicles in the formation indicated in the formation information when the unit moves while being influenced by the placement influencing factors indicated in the second inference data 320. The second inference unit 306 outputs output data 330, which is the inference result, to the unit movement control processing device 500. The output data 330 indicates the positions of personnel / vehicles, which is the arrangement of personnel / vehicles within the unit's formation. The position of the person / vehicle shown in the second inference data 320 is the current position (current coordinates) of the person / vehicle within the formation, while the position of the person / vehicle shown in the output data 330 is the next position (next coordinates) that the person / vehicle should take within the formation. The processing performed by the second inference unit 306 corresponds to the second inference processing.

[0052] The unit movement control processing device 500 processes the movement of people / vehicles in the training simulation based on the positions of people / vehicles in the output data 330. The unit movement control processing device 500 applies combinatorial optimization technology to movement costs, but the scheduling method is not limited to this.

[0053] The above describes an example in which the inference device 300 performs inference using the first inference model 130 generated by the learning device 100 and the second inference model storage device 240 generated by the learning device 200. Alternatively, the inference device 300 may use the first inference model 130 generated by a device other than the learning device 100 and the second inference model 230 generated by a device other than the learning device 200. In this case, the first inference model 130 is an inference model generated by comparing the formation of the unit when it moves to the shape of a viscous object when it moves fluidly, and learning the formation of the unit when it moves based on the shape of the viscous object when it moves fluidly.Furthermore, the second inference model 230 is an inference model generated by comparing the arrangement of people / vehicles within the unit to the distribution of electrons within a molecule, and learning the arrangement of people / vehicles within the unit based on the probability of electrons existing within a molecule.

[0054] ***Explanation of Operation*** 6 shows an example of the operation of inference device 300. The operation of inference device 300 will be described below with reference to FIG.

[0055] In step 301, the first inference data acquisition unit 303 acquires first inference data 310 from the inference data output device 400. Then, the first inference data acquisition unit 303 outputs the first inference data 310 to the first inference unit 305.

[0056] In step 302, a first inference unit 305 inputs first inference data 310 into a first inference model 130 to infer the formation of units. That is, the first inference unit 305 acquires the first inference data 310 and instructs the first inference model management unit 301 to acquire the first inference model 130. Then, the first inference model management unit 301 acquires the first inference model 130 from the first inference model storage device 140 and outputs the first inference model 130 to the first inference unit 305. The first inference unit 305 acquires the first inference model 130 from the first inference model management unit 301 and uses the first inference model 130 and the first inference data 310 to infer the optimal formation of the units.

[0057] In step 303, the first inference unit 305 outputs formation information indicating the formation obtained by inference to the second inference data acquisition unit 304.

[0058] In step 304, the second inference data acquisition unit 304 acquires the second inference data 320 from the inference data output device 400. In addition, the first inference model management unit 301 acquires the formation information output in step 303. Then, the second inference data acquisition unit 304 outputs the second inference data 320 and the formation information to the second inference unit 306.

[0059] In step 305, the second inference unit 306 inputs the second inference data 320 and formation information into the second inference model 230 to infer the positioning of personnel / vehicles within the unit's formation. That is, the second inference unit 306 acquires the second inference data 320 and the formation information and instructs the second inference model management unit 302 to acquire the second inference model 230. Then, the second inference model management unit 302 acquires the second inference model 230 from the second inference model storage device 240 and outputs the second inference model 230 to the second inference unit 306. The second inference unit 306 acquires the second inference model 230 from the second inference model management unit 302 and uses the second inference model 230, the second inference data 320, and the formation information to infer the optimal placement of people / vehicles within the formation.

[0060] In step 306, the second inference section 306 outputs output data 330 indicating the position of the person / vehicle obtained by inference to the unit movement control processing device 500.

[0061] The unit movement control processing device 500 performs movement processing of people / vehicles in a simulation based on the output data 330. This makes it possible to automatically create action commands for units.

[0062] ***Explanation of the effect of the embodiment*** According to this embodiment, it is possible to derive the optimal formation of a unit and the optimal placement of people / vehicles within the formation when the unit moves while being influenced by factors that affect the unit's movement. Furthermore, according to this embodiment, there is no need to manually formulate behavior rules in advance, which can save labor.

[0063] In the present embodiment, an example has been described in which supervised learning is applied as the learning algorithm of the learning unit 102 and the learning unit 202. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can be applied in addition to supervised learning.

[0064] The learning unit 102 may also generate the first inference model 130 using a plurality of training data 1-1 (110) and training data 1-2 (120). Similarly, the learning unit 202 may also generate the second inference model 230 using a plurality of training data 2-1 (210) and training data 2-2 (220). Furthermore, the learning unit 102 may use multiple pieces of data collected in the same area or multiple pieces of data collected in different areas as the multiple pieces of learning data 1-1 (110) and learning data 1-2 (120). Similarly, the learning unit 202 may use multiple pieces of data collected in the same area or multiple pieces of data collected in different areas as the multiple pieces of learning data 2-1 (210) and learning data 2-2 (220). Furthermore, the learning data acquisition unit 101 may start acquiring learning data 1-1 (110) and learning data 1-2 (120) for a new region during learning, or may stop acquiring learning data 1-1 (110) and learning data 1-2 (120) for a specific region during learning. Similarly, the learning data acquisition unit 201 may start acquiring learning data 2-1 (210) and learning data 2-2 (220) for a new region during learning, or may stop acquiring learning data 2-1 (210) and learning data 2-2 (220) for a specific region during learning. Furthermore, the learning unit 102 may re-train a first inference model 130 generated using learning data 1-1 (110) and learning data 1-2 (120) for a certain region by applying learning data 1-1 (110) and learning data 1-2 (120) for another region, thereby updating the first inference model 130. Similarly, the learning unit 202 may re-train a second inference model 230 generated using learning data 2-1 (210) and learning data 2-2 (220) for a certain region by applying learning data 2-1 (210) and learning data 2-2 (220) for another region, thereby updating the second inference model 230.

[0065] Deep learning can also be used as the learning algorithm for the learning units 102 and 202. Deep learning is a learning algorithm that learns to extract features themselves. The learning units 102 and 202 may also perform machine learning according to other known methods, such as genetic programming, functional logic programming, and support vector machines.

[0066] The procedure described in this embodiment is an example. Therefore, it is possible to carry out only a part of the procedure described in this embodiment. Furthermore, at least part of the procedure described in this embodiment may be combined with a procedure not described in this embodiment. Furthermore, the configuration and procedures described in this embodiment may be modified as necessary.

[0067] ***Additional hardware configuration information*** Here, a supplementary explanation of the hardware configurations of learning device 100, learning device 200, and inference device 300 will be provided. The processor 911, the processor 921, and the processor 931 are each an IC (Integrated Circuit) that performs processing. The processor 911, the processor 921, and the processor 931 are each a CPU (Central Processing Unit), a DSP (Digital Signal Processor), or the like. The main memory device 912, the main memory device 922, and the main memory device 932 are each a RAM (Random Access Memory). The auxiliary storage device 913, the auxiliary storage device 923, and the auxiliary storage device 933 are each a ROM (Read Only Memory), a flash memory, an HDD (Hard Disk Drive), or the like. The communication device 914, the communication device 924, and the communication device 934 are each electronic circuits that perform data communication processing. The communication device 914, the communication device 924, and the communication device 934 are, for example, communication chips or NICs (Network Interface Cards). The input / output device 915, the input / output device 925, and the input / output device 935 are a mouse, a keyboard, a display, etc., respectively.

[0068] Furthermore, the auxiliary storage device 913, the auxiliary storage device 923, and the auxiliary storage device 933 each store an OS (Operating System). At least a part of the OS is executed by the processor 911, the processor 921, and the processor 931. The processor 911, the processor 921, and the processor 931 each execute a program that realizes a functional configuration while executing at least a part of the OS. The processor 911, the processor 921, and the processor 931 execute the OS, thereby performing task management, memory management, file management, communication control, and the like.

[0069] In addition, at least one of information, data, signal values, and variable values ​​indicating the results of processing by the functional configuration of the learning device 100 (learning data acquisition unit 101 and learning unit 102) is stored in at least one of the main memory device 912, the auxiliary memory device 913, and the register and cache memory within the processor 911. Similarly, at least one of information, data, signal values, and variable values ​​indicating the results of processing by the functional configuration of the learning device 200 (learning data acquisition unit 201 and learning unit 202) is stored in at least one of the main memory device 922, the auxiliary memory device 923, and the register and cache memory within the processor 921. Similarly, at least one of information, data, signal values ​​and variable values ​​indicating the results of processing of the functional configuration of the inference device 300 (first inference model management unit 301, second inference model management unit 302, first inference data acquisition unit 303, second inference data acquisition unit 304, first inference unit 305 and second inference unit 306) is stored in at least one of the main memory device 932, auxiliary memory device 933, registers and cache memory within the processor 931. Furthermore, the programs that realize the functional configurations of learning device 100, learning device 200, and inference device 300 may each be stored on a portable recording medium such as a magnetic disk, flexible disk, optical disk, compact disk, Blu-ray (registered trademark) disk, DVD, etc. Portable recording media on which these programs are stored may then be distributed.

[0070] Furthermore, the "part" of at least one of the functional configurations of learning device 100, learning device 200, and inference device 300 may be read as a "circuit," "step," "procedure," "process," or "circuitry." Furthermore, learning device 100, learning device 200, and inference device 300 may each be realized by a processing circuit, such as a logic integrated circuit (IC), a gate array (GA), an application specific integrated circuit (ASIC), or a field-programmable gate array (FPGA). In this case, the functional configurations of learning device 100, learning device 200, and inference device 300 are each realized as part of a processing circuit. In this specification, the term "processing circuitry" refers to a generic concept that encompasses a processor and a processing circuit. That is, a processor and a processing circuit are each specific examples of "processing circuitry."

[0071] Various aspects of the present disclosure are summarized below as appendices. (Appendix 1) a learning data acquisition unit that acquires learning data, which is data used for learning, indicating a group movement influence factor that is a factor that influences the movement of a group of two or more moving objects and an object movement influence factor that is a factor that influences the fluid movement of a viscous object; a learning unit that compares the formation of the group when the group moves to the shape of the viscous object when it moves in a fluid manner, and that uses the learning data to learn the formation of the group when the group moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor. (Appendix 2) The learning data acquisition unit The learning device described in Appendix 1 acquires learning data in which the group movement influence factors indicate at least one of the position of the group, the surrounding environment when the group moves, the position of obstacles to the group's movement, and the position of threats to the group, and the object movement influence factors indicate at least one of the size of the viscous object and the viscosity of the viscous object. (Appendix 3) a learning data acquisition unit that acquires learning data, which is data used for learning and indicates a placement influence factor that is a factor that influences the placement of the two or more moving objects in a group of the two or more moving objects; a learning unit that views the arrangement of the two or more moving bodies in the group as the arrangement of electrons in a molecule, and uses the learning data to learn the arrangement of the two or more moving bodies in the group that is influenced by the arrangement influencing factor based on the probability of electrons existing in the molecule. (Appendix 4) The learning data acquisition unit The learning device according to claim 3, which acquires learning data indicating, as the placement influence factors, at least one of the formation of the group when the group moves, the equipment mounted on the two or more moving bodies, the behavioral standard of the group, the durability value of the group, and the range when projected from the two or more moving bodies. (Appendix 5) a first inference model management unit that manages a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of a viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management unit that manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as the distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition unit that acquires first inference data, which is data used for inference and indicates a population movement influencing factor that is a factor that influences the population movement; a second inference data acquisition unit that acquires second inference data, which is data used for inference and indicates a placement influence factor that is a factor that influences the placement of the two or more moving objects within the group; a first inference unit that uses the first inference data and the first inference model to infer the formation of the group when the group moves while being influenced by the group movement influencing factor; An inference device having a second inference unit that uses the second inference data, the second inference model, and the group formation inferred by the first inference unit to infer the positioning of the two or more moving bodies in the group formation inferred by the first inference unit, as influenced by the positioning influence factor. (Appendix 6) The first inference data acquisition unit acquire first inference data indicating at least one of the location of the group, the surrounding environment when the group moves, the location of an obstacle to the movement of the group, and the location of a threat to the group as the group movement influencing factor; The second inference data acquisition unit An inference device as described in Appendix 5, which acquires second inference data indicating at least one of equipment mounted on the two or more moving bodies and behavioral standards of the group as the placement influence factor. (Appendix 7) A learning data acquisition process in which the computer acquires learning data, which is data used for learning and indicates a group movement influence factor, which is a factor that influences the movement of a group of two or more moving objects, and an object movement influence factor, which is a factor that influences the fluid movement of a viscous object; a learning process in which the computer compares the formation of the group when it moves to the shape of the viscous object when it moves in a fluid manner, and uses the learning data to learn the formation of the group when it moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor. (Appendix 8) a learning data acquisition process in which the computer acquires learning data, which is data used for learning and indicates placement influence factors that are factors that influence the placement of the two or more moving objects in a group of the two or more moving objects; The computer views the arrangement of the two or more moving bodies in the population as the arrangement of electrons in a molecule, and uses the learning data to learn the arrangement of the two or more moving bodies in the population that is influenced by the arrangement influencing factor based on the probability of electrons existing in the molecule. (Appendix 9) a first inference model management process in which a computer manages a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of a viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management process in which the computer manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as the distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition process in which the computer acquires first inference data, which is data used for inference and indicates a population movement influencing factor, which is a factor that influences the population movement; a second inference data acquisition process in which the computer acquires second inference data, which is data used for inference and indicates a location influence factor, which is a factor that influences the location of the two or more moving objects within the group; a first inference process in which the computer uses the first inference data and the first inference model to infer the formation of the group when the group moves while being influenced by the group movement influencing factor; An inference method comprising a second inference process in which the computer uses the second inference data, the second inference model, and the formation of the group inferred by the first inference process to infer the positioning of the two or more moving bodies in the formation of the group inferred by the first inference process, as influenced by the positioning influence factor. (Appendix 10) a learning data acquisition process for acquiring learning data, which is data used for learning and indicates a group movement influence factor, which is a factor that influences the movement of a group of two or more moving objects, and an object movement influence factor, which is a factor that influences the fluid movement of a viscous object; a learning program that causes a computer to execute a learning process in which the formation of the group when the group moves is likened to the shape of the viscous object when it moves in a fluid manner, and using the learning data, learns the formation of the group when it moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor. (Appendix 11) a learning data acquisition process for acquiring learning data, which is data used for learning and indicates placement influence factors that are factors that influence the placement of the two or more moving objects in a group of the two or more moving objects; a learning process in which the arrangement of the two or more moving bodies in the group is viewed as the arrangement of electrons in a molecule, and the arrangement of the two or more moving bodies in the group that is influenced by the arrangement influencing factor is learned using the learning data based on the probability of electrons existing in the molecule. (Appendix 12) a first inference model management process for managing a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of the viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management process that manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as a distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition process for acquiring first inference data, which is data used for inference and indicates a population movement influencing factor, which is a factor that influences the population movement; a second inference data acquisition process for acquiring second inference data, which is data used for inference and indicates a placement influence factor, which is a factor that influences the placement of the two or more moving objects within the group; a first inference process for inferring the formation of the group when the group moves while being influenced by the group movement influence factor, using the first inference data and the first inference model; An inference program that causes a computer to execute a second inference process that uses the second inference data, the second inference model, and the group formation inferred by the first inference process to infer the positioning of the two or more moving bodies in the group formation inferred by the first inference process, as influenced by the positioning influence factor. [Explanation of symbols]

[0072] 100 Learning device, 101 Learning data acquisition unit, 102 Learning unit, 110 Learning data 1-1, 120 Learning data 1-2, 130 First inference model, 140 First inference model storage device, 200 Learning device, 201 Learning data acquisition unit, 202 Learning unit, 210 Learning data 2-1, 220 Learning data 2-2, 230 Second inference model, 240 Second inference model storage device, 300 Inference device, 301 First inference model management unit, 302 Second inference model management unit, 303 First inference data acquisition unit, 304 Second inference data acquisition unit, 305 First inference unit, 306 Second inference unit, 310 First inference data, 320 Second inference data, 330 Output data, 400 Inference data output device, 500 Unit movement control processing device, 911 Processor, 912 Main memory, 913 auxiliary memory, 914 communication device, 915 input / output device, 921 processor, 922 main memory, 923 auxiliary memory, 924 communication device, 925 input / output device, 931 processor, 932 main memory, 933 auxiliary memory, 934 communication device, 935 input / output device.

Claims

1. a learning data acquisition unit that acquires learning data, which is data used for learning, indicating a group movement influence factor that is a factor that influences the movement of a group of two or more moving objects and an object movement influence factor that is a factor that influences the fluid movement of a viscous object; a learning unit that compares the formation of the group when the group moves to the shape of the viscous object when it moves in a fluid manner, and that uses the learning data to learn the formation of the group when the group moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor.

2. The learning data acquisition unit The learning device of claim 1 acquires learning data in which the group movement influence factors indicate at least one of the position of the group, the surrounding environment when the group moves, the position of obstacles to the group's movement, and the position of threats to the group, and the object movement influence factors indicate at least one of the size of the viscous object and the viscosity of the viscous object.

3. a learning data acquisition unit that acquires learning data, which is data used for learning and indicates a placement influence factor that is a factor that influences the placement of the two or more moving objects in a group of the two or more moving objects; a learning unit that views the arrangement of the two or more moving bodies in the group as the arrangement of electrons in a molecule, and uses the learning data to learn the arrangement of the two or more moving bodies in the group that is influenced by the arrangement influence factor based on the probability of electrons existing in the molecule.

4. The learning data acquisition unit The learning device according to claim 3, wherein learning data is acquired that indicates, as the placement influence factors, at least one of the formation of the group when the group moves, the equipment mounted on the two or more moving bodies, the behavioral standards of the group, the durability value of the group, and the range when projected from the two or more moving bodies.

5. a first inference model management unit that manages a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of the viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management unit that manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as a distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition unit that acquires first inference data, which is data used for inference and indicates a group movement influencing factor that is a factor that influences the movement of the group; a second inference data acquisition unit that acquires second inference data, which is data used for inference and indicates a placement influence factor that is a factor that influences the placement of the two or more moving objects within the group; a first inference unit that uses the first inference data and the first inference model to infer the formation of the group when the group moves while being influenced by the group movement influencing factor; An inference device having a second inference unit that uses the second inference data, the second inference model, and the group formation inferred by the first inference unit to infer the placement of the two or more moving bodies in the group formation inferred by the first inference unit, as influenced by the placement influence factor.

6. The first inference data acquisition unit acquire first inference data indicating at least one of the location of the group, the surrounding environment when the group moves, the location of an obstacle to the movement of the group, and the location of a threat to the group as the group movement influencing factor; The second inference data acquisition unit The inference device according to claim 5 , further comprising: acquiring second inference data indicating at least one of equipment mounted on the two or more mobile bodies and a behavioral standard of the group as the location influencing factor.

7. a learning data acquisition process in which the computer acquires learning data, which is data used for learning and indicates a group movement influence factor, which is a factor that influences the movement of a group of two or more moving objects, and an object movement influence factor, which is a factor that influences the fluid movement of a viscous object; a learning process in which the computer compares the formation of the group when it moves to the shape of the viscous object when it moves in a fluid manner, and uses the learning data to learn the formation of the group when it moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor.

8. a learning data acquisition process in which a computer acquires learning data, which is data used for learning and indicates a placement influence factor, which is a factor that influences the placement of two or more moving objects in a group of two or more moving objects; The computer views the arrangement of the two or more moving bodies in the population as the arrangement of electrons in a molecule, and uses the learning data to learn the arrangement of the two or more moving bodies in the population that is influenced by the arrangement influence factor based on the probability of electrons existing in the molecule.

9. a first inference model management process in which a computer manages a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of the viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management process in which the computer manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as the distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition process in which the computer acquires first inference data, which is data used for inference and indicates a population movement influencing factor, which is a factor that influences the population movement; a second inference data acquisition process in which the computer acquires second inference data, which is data used for inference and indicates a location influence factor, which is a factor that influences the location of the two or more moving objects within the group; a first inference process in which the computer uses the first inference data and the first inference model to infer the formation of the group when the group moves while being influenced by the group movement influencing factor; An inference method comprising a second inference process in which the computer uses the second inference data, the second inference model, and the formation of the group inferred by the first inference process to infer the positioning of the two or more moving bodies in the formation of the group inferred by the first inference process, as influenced by the positioning influence factor.

10. a learning data acquisition process for acquiring learning data, which is data used for learning and indicates a group movement influence factor, which is a factor that influences the movement of a group of two or more moving objects, and an object movement influence factor, which is a factor that influences the fluid movement of a viscous object; a learning program that causes a computer to execute a learning process in which the formation of the group when the group moves is likened to the shape of the viscous object when it moves in a fluid manner, and using the learning data, learns the formation of the group when it moves while being influenced by the group movement influencing factor, based on the shape of the viscous object when it moves in a fluid manner while being influenced by the object movement influencing factor.

11. a learning data acquisition process for acquiring learning data, which is data used for learning and indicates placement influence factors that are factors that influence the placement of two or more moving objects in a group of two or more moving objects; a learning program that causes a computer to execute a learning process in which the arrangement of the two or more moving bodies in the group is viewed as the arrangement of electrons in a molecule, and the learning data is used to learn the arrangement of the two or more moving bodies in the group that is influenced by the arrangement influencing factor based on the probability of electrons existing in the molecule.

12. a first inference model management process for managing a first inference model, which is an inference model generated by learning the formation of a group of two or more moving objects when the group moves, based on the shape of the viscous object when the viscous object moves in a fluid manner, by comparing the formation of the group when the group moves with the shape of the viscous object when the viscous object moves in a fluid manner; a second inference model management process that manages a second inference model, which is an inference model generated by viewing the arrangement of the two or more moving objects in the population as a distribution of electrons in a molecule and learning the arrangement of the two or more moving objects in the population based on the probability of electrons being present in the molecule; a first inference data acquisition process for acquiring first inference data, which is data used for inference and indicates a population movement influencing factor, which is a factor that influences the population movement; a second inference data acquisition process for acquiring second inference data, which is data used for inference and indicates a location influence factor, which is a factor that influences the location of the two or more moving objects within the group; a first inference process for inferring the formation of the group when the group moves while being influenced by the group movement influencing factor, using the first inference data and the first inference model; An inference program that causes a computer to execute a second inference process that uses the second inference data, the second inference model, and the group formation inferred by the first inference process to infer the positioning of the two or more moving bodies in the group formation inferred by the first inference process, as influenced by the positioning influence factor.

Images