Route creation method and route creation device

Abstract

To provide a method which is a method for creating a route for an agent to visit a target, and with which an appropriate route can be efficiently created.SOLUTION: A route creation method creates a visiting route for an agent to visit a plurality of targets by performing the processing below. The plurality of targets are clustered to classify these into a plurality of clusters. Priority information is created that includes the priority of allocating the agent to clusters or the priority of the order of visit to clusters by the agent. A candidate route for the agent is created, which is a candidate for the route of visit, on the basis of the target positions, the agent position, and the priority information. The candidate route is evaluated on the basis of an evaluation index which is an index for evaluating the candidate route. One or more candidate routes are selected as visiting routes on the basis of the evaluation result of the candidate route.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present application primarily relates to a route creation method for creating a visiting route for an agent to visit multiple targets. [Background technology]

[0002] Patent Document 1 discloses a method for assigning multiple delivery destinations to a deliverer and creating a route to follow the delivery destinations. Specifically, the multiple delivery destinations to be assigned are clustered, and each delivery destination is assigned to a delivery cluster. Next, a deliverer is assigned to the delivery cluster. Finally, the delivery destinations that belong to the delivery cluster to which the deliverer is assigned are identified, and a route to follow the identified delivery destinations is created. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-39451 Summary of the Invention [Problem to be solved by the invention]

[0004] The delivery management method of Patent Document 1 uses clustering only to assign deliverers to delivery destinations. Patent Document 1 does not disclose a process for efficiently creating appropriate routes using delivery clusters. Therefore, Patent Document 1 does not fully utilize clustering, and there is room for improvement. Furthermore, this issue is not limited to deliveries. This issue is common to various events in which agents visit targets.

[0005] The present application has been made in light of the above circumstances, and its main purpose is to provide a method for creating a route for an agent to visit a target, which method is capable of efficiently creating an appropriate route. [Means for solving the problem]

[0006] The problem to be solved by the present application is as described above. Next, the means for solving this problem and the effects thereof will be explained.

[0007] According to a first aspect of the present application, there is provided the following route creation method. The route creation method performs the following processes to create a visiting route for an agent to visit multiple targets. The multiple targets are clustered and classified into multiple clusters. Priority information that satisfies the following conditions 1 to 3 is created. Condition 1 is information that assists route search. Condition 2 is information that is created for each cluster. Condition 3 is information that includes the priority of allocation of the agent to the cluster or the priority of the order in which the agent visits the clusters. Based on the position of the target, the position of the agent, and the priority information, a candidate route that is a candidate for the visiting route is created for the agent. The candidate route is evaluated based on an evaluation index that is an index for evaluating the candidate route. Based on the evaluation result of the candidate route, one or more candidate routes are selected as the visiting route.

[0008] According to a second aspect of the present application, there is provided a route creation device having the following configuration. That is, the route creation device includes a communication device and a processing device. The communication device acquires target information and agent information. The processing device creates a visiting route for an agent to visit multiple targets based on the target information and the agent information. The processing device clusters the multiple targets and classifies them into multiple clusters. The processing device creates priority information that satisfies the following conditions 1 to 3. Condition 1 is information that assists route search. Condition 2 is information that is created for each cluster. Condition 3 is information that includes the priority of allocation of the agent to the cluster or the priority of the order in which the agent visits the clusters. The processing device creates candidate routes that are candidates for the visiting route for the agent based on the position of the target, the position of the agent, and the priority information. The processing device evaluates the candidate routes based on an evaluation index that is an index for evaluating the candidate routes. The processing device selects one or more candidate routes as the visiting route based on the evaluation results of the candidate routes. [Effects of the Invention]

[0009] According to the present application, there is provided a method for creating a route for an agent to visit a target, and an appropriate route can be created efficiently. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram of a route generation device according to an embodiment of the present application; [Figure 2] 1 is a flowchart of a route creation method. [Figure 3] FIG. 10 is a schematic diagram of a flow of route creation in a comparative example. [Figure 4] An overview of the route creation process when prioritizing the allocation of agents to clusters. [Figure 5] A table showing the priorities for assigning agents to clusters. [Figure 6] FIG. 2 is a diagram for explaining an equation used in a route search. [Figure 7] An overview of the flow of route creation when priorities are set for the order in which agents visit clusters. [Figure 8] FIG. 8 is a diagram showing information used in the route search in FIG. 7. [Figure 9] FIG. 10 is a diagram showing the distribution of evaluation values ​​of sensors 1 and 2. [Figure 10] FIG. 10 is a diagram showing a method for evaluating candidate routes using sensor evaluation values. DETAILED DESCRIPTION OF THE INVENTION

[0011] Next, an embodiment of the present application will be described with reference to the drawings. A route creation device 1 of this embodiment creates a route for an agent. In detail, the route creation device 1 creates a visiting route, which is a route along which the agent visits multiple targets. Below, a brief description of key terms used in this embodiment will be given.

[0012] An agent is a target that moves along a route to be visited. Therefore, an agent can move autonomously or based on external commands. More specifically, an agent may be a vehicle such as an aircraft, submarine, ship, or vehicle, or an unmanned aerial vehicle such as a UAV, USV, UGV, UUV, or AUV. An agent may also be an industrial robot capable of autonomous movement, or a worker. An agent performs some task on a target. The task may be, for example, measuring information about the target using a sensor, communicating with the target, or exchanging goods with the target. Note that an agent does not necessarily have to perform a task. In other words, the agent may simply approach the target. While this embodiment basically assumes an environment in which multiple agents exist, a route to be visited can also be created even when there is only one agent.

[0013] A target is an object that an agent visits. A target may be an object or a location. If the target is an object, it may be a moving object or an immovable structure. A target is also an object that receives a task from an agent. In some environments, there may be only one target, but in this case, path generation is not required and therefore it is not considered in this specification. Therefore, in the following, it is assumed that there are multiple targets.

[0014] A visit means that an agent reaches a target or a location nearby it. The range indicated by the proximity is interpreted according to the type of agent and the specific content of the task. For example, when an agent is performing a task, the agent is considered to have visited the target if it is close enough to the target to perform the task.

[0015] Next, the route creation device 1 will be described. The route creation device 1 includes a processing device 11, a storage device 12, and a communication device 13. The processing device 11 is a computer such as a PC or a server device. The processing device 11 is, for example, a CPU, and executes various processes by executing programs. For example, the processing device 11 executes a program related to the creation of a visiting route, causing the processing device 11 to execute each process of the route creation method of this embodiment. The storage device 12 is, for example, an HDD, SSD, or flash memory, and stores data. The storage device 12 stores programs, control data, etc. related to the creation of a visiting route. The communication device 13 is, for example, a wired communication module or a wireless communication module, and communicates with external devices.

[0016] The route generation device 1 is installed in, for example, a management room that manages multiple agents. However, the route generation device 1 may be installed in any one of the agents. Alternatively, the route generation device 1 may be installed on the cloud.

[0017] Next, an outline of the process by which the route generation device 1 generates a visiting route will be described with reference to Fig. 2. The flowchart in Fig. 2 is executed by the processing device 11 of the route generation device 1.

[0018] First, the route generation device 1 acquires target information and agent information (S1). The target information and agent information are created by an external device. The route generation device 1 acquires the target information and agent information from the external device via the communication device 13. There are various timings for acquiring the target information and agent information. For example, the route generation device 1 may acquire them periodically, or may acquire them when an instruction to create a visiting route is received. Alternatively, if the situation changes significantly due to a change in the position or orientation of the target, the route generation device 1 may acquire the target information and agent information again and execute the process of creating a visiting route again.

[0019] Target information is information about targets to be used in creating a route. Target information includes at least the position information of each of multiple targets. Target information may further include information about the target's direction, the target's movement speed, the target's movement direction, the type of target, or the task to be performed on the target. Information about the task is, for example, the type of information obtained by the task or the type of sensor to be used on the target. Target information can be created by detecting the surroundings using a detection device such as radar or sonar. Alternatively, if the targets are under management, target information can be created by receiving position information, etc. from each target.

[0020] Agent information is information about an agent to be used for creating a route. Specifically, agent information includes at least the position information of the agent. Agent information may further include information about the agent's direction, the agent's movement speed, the agent's movement direction, the agent's type, or the task the agent is performing. Information about the task is, for example, the type of sensor used by the agent. The method of obtaining agent information is the same as that for target information.

[0021] Next, the path generation device 1 clusters the targets and classifies them into clusters (S2). Clustering refers to classifying multiple targets into one or more clusters. When the number of targets is small, all targets may be classified into one cluster. However, since the present embodiment is premised on the use of multiple clusters, the targets will be treated below as being classified into multiple clusters. Furthermore, the clustering settings may be adjusted so that the number of clusters generated is greater than the number of agents. Clustering is a well-known technique, and therefore will be briefly described below. Clustering in this embodiment is performed mainly based on the position of the target. In other words, targets are classified so that targets located close to each other belong to the same cluster. Specifically, various well-known clustering methods can be employed, and for example, machine learning algorithms such as the k-means method and DBSCAN can be used.

[0022] Next, the route generation device 1 generates priority information (S3). Priority information is information that satisfies the following conditions 1 to 3. Condition 1 corresponds to "information that assists route search." In other words, when generating candidate routes, which will be described later, the priority information is used to efficiently search for appropriate candidate routes. Condition 2 corresponds to "information generated for each cluster." In other words, priority information adds some information to a cluster. Information added to a cluster may be treated as being added to targets belonging to the cluster as well. Furthermore, priority information may be information generated not only for a cluster but also for each agent, in other words, for each combination of a cluster and an agent. Condition 3 corresponds to "information including (1) the priority of assignment of agents to clusters, or (2) the priority of the order in which agents visit clusters." The priority information may include only one of the priorities (1) and (2), or may include both the priorities (1) and (2). Details of priority information will be described later.

[0023] Next, the route generation device 1 generates candidate routes (S4). In detail, the route generation device 1 generates candidate routes based on at least the position of the target, the position of the agent, and the priority information. A candidate route is a candidate for a visiting route.

[0024] In conventional technology, clustering is used only to assign agents to targets. In contrast, in this embodiment, priority information is created for each cluster. The priority information also serves as information to assist in route search. Therefore, by using the priority information during route search, it is possible to create an appropriate route more efficiently than in conventional technology. The creation of candidate routes using priority information will be described in detail later.

[0025] Next, the route generation device 1 evaluates the candidate routes (S5). In detail, the route generation device 1 calculates an evaluation value based on an evaluation index, which is an index for evaluating the candidate routes, and evaluates the candidate routes based on the evaluation value. The evaluation index is, for example, the agent's travel time, the agent's travel distance, the agent's turning angle when traveling along the candidate route, and the quality of the task. The agent's travel time may include the time to perform the task, or may be just the time required for travel. Furthermore, when there are multiple agents, the total value of the travel time, etc., for each agent may be used as the evaluation index. Furthermore, any one of the above-mentioned evaluation indexes may be adopted, or multiple evaluation indexes may be combined. The above-mentioned evaluation indexes are merely examples, and the evaluation value may be calculated using other evaluation indexes.

[0026] Next, the route generation device 1 determines whether it is time to select a route to be visited (S6). The timing to select a route to be visited is determined, for example, using the elapsed time since receiving an instruction to select a route to be visited, the time specified when creating a route to be visited, the number of candidate routes created, or the evaluation value of the candidate route. When using the evaluation value of a candidate route, the selection timing can be determined, for example, as the point when the evaluation value exceeds a threshold, or as the point when improvement in the evaluation value converges. The route generation device 1 repeats the generation of candidate routes (S4) and the evaluation of candidate routes (S5) until it is time to select a route to be visited. Note that the route generation device 1 may repeat the clustering (S2), generation of priority information (S3), generation of candidate routes (S4), and evaluation of candidate routes (S5) until it is time to select a route to be visited.

[0027] When the route generation device 1 determines that it is time to select a route to be visited, it selects a route to be visited based on the evaluation results of the candidate routes (S7). In detail, the route generation device 1 selects one or more candidate routes that have good evaluation results for the route to be visited based on the evaluation index. This completes the generation of the route to be visited.

[0028] The created visiting route is transmitted to the agent or a control device that controls the agent's movement via the communication device 13 or other communication means. If multiple visiting routes are selected by the route creation device 1, the manager or management device determines one visiting route from the multiple visiting routes. The agent visits the target based on the determined visiting route, and performs the task if a task is required. Note that the manager or management device may determine the final visiting route after carefully examining or slightly correcting the visiting route selected by the route creation device 1.

[0029] Next, with reference to FIGS. 3 to 5, a detailed description will be given of the processing when the priority information is information including the above-mentioned "(1) priority of allocation of agents to clusters."

[0030] First, referring to the comparative example of FIG. 3, a problem that occurs when priorities for allocation to clusters are not set will be described. In the comparative example, there are agents A1 and A2. In the comparative example, multiple targets are clustered into clusters C1 and C2. As a result, the targets are classified into clusters C1 and C2. Next, agents are assigned to clusters and routes are created. Taking into account the distance between the agents and the clusters, agent A1 is assigned to cluster C1, and agent A2 is assigned to cluster C2. Therefore, a route is created for agent A1 to visit all targets belonging to cluster C1, and a route is created for agent A2 to visit all targets belonging to cluster C2.

[0031] In the comparative example, the number of targets assigned to agent A1 is significantly greater than the number of targets assigned to agent A2. Furthermore, the travel distance of agent A1's route is significantly longer than the travel distance of agent A2's route. Therefore, agent A1's travel time is significantly longer than that of agent A2. For these reasons, targets are not efficiently allocated in the comparative example. Furthermore, since it is not easy to perform clustering taking into account the circumstances of route creation, this problem cannot be solved simply by changing the clustering settings. Furthermore, omitting clustering would eliminate the cluster constraints, but instead would significantly increase the number of candidate routes, which is not realistic because the amount of calculation required to create an efficient route would be enormous.

[0032] In contrast, in this embodiment, as shown in FIGS. 4 and 5, after clustering, priority information including a cluster assignment priority is created for each combination of an agent and a cluster. The priority is a value between 0 and 1. The higher the priority of a combination, the higher the probability that the agents constituting that combination will be assigned to the target of the cluster constituting that combination. For example, as shown in FIG. 5, the priority of the combination of agent A1 and cluster C1 is 1, so agent A1 is more likely to be assigned to the target belonging to cluster C1. On the other hand, the priority of the combination of agent A1 and cluster C2 is 0, so agent A1 is not assigned to the target belonging to cluster C2. Furthermore, the priority of the combination of agent A1 and cluster C3 is greater than 0 and less than 1, so agent A1 may or may not be assigned to the target belonging to cluster C3. In this embodiment, it can also be considered that a priority is set for each agent in one cluster. The sum of the priorities set for one cluster is 1. The priority can be set to any value between 0 and 1. Furthermore, when multiple agents are assigned a priority greater than 0 for a single cluster, the priorities of the agents may be the same or different. If the priorities of the agents are the same, the probability that each agent will be assigned to a target within the cluster is weighted equally. If the priorities of the agents are different, the probability that each agent will be assigned to a target within the cluster is weighted according to the difference in priorities.

[0033] The priority assigned to a combination of an agent and a cluster is, for example, a value corresponding to the distance between the agent and the cluster. That is, the center of the cluster is first determined. The center of the cluster is, for example, the average coordinate position of the targets belonging to the cluster. The center of the cluster may also be the physical center of gravity, weighted by the visit priority of each target belonging to the cluster. The visit priority of a target is set for each target and indicates the degree to which a task should be prioritized for that target. The closer the distance between the agent and the center of the cluster, the higher the priority is set. Another condition is that, as mentioned above, the priority is set so that the sum of the priorities assigned to a cluster is 1. Note that the absolute distance between the agent and the center of the cluster is not important, but rather the relative distance compared to other agents is important. For example, in the example shown in Figure 4, the distance between agent A1 and the center of cluster C1 is much shorter than the distance between agent A2 and the center of cluster C1. Therefore, the priority of the combination of agent A1 and cluster C1 is 1, and the priority of the combination of agent A2 and cluster C1 is 0. The center of the cluster corresponds to the representative position of the cluster. Note that a position other than the center of a cluster may be treated as the representative position of the cluster.

[0034] The index for setting priority is not limited to distance. For example, if an agent is a ship or the like that takes time to change direction, priority may be set taking into account the orientation of the agent. Alternatively, if there are many targets near agent A1 and few targets near agent A2, the priority of agent A2 may be set higher so that agent A2 is more likely to be assigned to targets.

[0035] In this embodiment, the combination of agent A1 and cluster C3 has a priority of 0.4, and the combination of agent A2 and cluster C3 has a priority of 0.6. Therefore, agent A1 and agent A2 are allowed to share the work of assigning targets belonging to cluster C3. As a result, as shown in the route generation diagram of FIG. 4, agent A1 and agent A2 share the work of assigning targets belonging to cluster C3. As a result, the number of targets assigned between agents can be made closer to uniform, allowing targets to be shared efficiently. Furthermore, even when there are clusters located close to both agent A1 and agent A2, it becomes easier to generate appropriate candidate routes.

[0036] Below, we will explain the specific process of creating candidate routes using priorities. Methods of route search that take weights such as priorities into account are known, and various publicly known technologies such as optimization algorithms and machine learning algorithms can be applied. Below, we will briefly explain the metaheuristic ACO as an example. ACO is an abbreviation for Ant Colony Optimization. Using metaheuristics, or using ACO among metaheuristics, is one example. As described above, other methods can also be used as long as they are capable of route search that takes weights into account.

[0037] In ACO, after creating a candidate route (S4), the candidate route is evaluated based on an evaluation function (S5). The candidate route created here is called the first candidate route. As shown in equation (3) in Figure 6, the evaluation function is the sum of the maximum visit completion time and the average visit completion time (the maximum and average times required for each agent to visit all of its target destinations). The better the evaluation function for the first candidate route, the more likely it is that a route similar to the first candidate route will be created in subsequent candidate route creations. Specifically, when the first candidate route is subdivided into multiple edges (specifically, line segments connecting targets), edges adopted in the first candidate route are more likely to be adopted in subsequent candidate routes. In ACO, this likelihood of adoption is modeled based on a quantity called a pheromone, and edges with pheromones attached are more likely to be adopted in subsequent route searches.

[0038] In ACO, a path search is performed using formulas (1) and (2) shown in FIG. 6. Formula (2) indicates the probability that an agent will visit a target next after visiting a target. The right side of formula (2) includes a term corresponding to the pheromone described above, indicating that an edge corresponding to the pheromone (in other words, a combination of a target and another target) is more likely to be adopted. Furthermore, formula (1) is the value obtained by multiplying the value indicated by formula (2) by a term related to priority. As a result, the higher the priority, the higher the probability that a target belonging to the corresponding cluster will be visited next. In this embodiment, a path search is performed using formula (1) including priority instead of formula (2), thereby enabling a path search using the above-mentioned priority. Note that the process of searching for a specific path based on these formulas is a well-known technique related to ACO, and therefore a description thereof will be omitted.

[0039] Next, with reference to FIGS. 7 and 8, a detailed description will be given of processing when the priority information is information including the above-mentioned "(2) information including the priority of the order of visiting clusters by agents."

[0040] As with the creation of priority information in (1), when creating priority information in (2), after clustering, priority information is created for each agent-cluster combination, including the priority of the order in which the agent will visit the clusters. Priorities are indicated by the letters a, b, and c. "a" has the highest priority, followed by "b" and then "c." For a cluster, a priority is set only for the combination with one selected agent, and no priority is set for combinations with agents that are not selected.

[0041] When creating priority information, first, a route search is performed for each agent, with the center of the cluster as a tentative target. Then, a route that passes through the center of the created cluster is created and evaluated using an evaluation function. The evaluation function is, for example, the sum of the maximum visit completion time and the average visit completion time for the cluster. Note that travel distance may be used instead of time. In other words, multiple routes are created, each route is evaluated using the evaluation function, and the route with the best evaluation result (i.e., the smallest value of the evaluation function) is adopted.

[0042] Next, priorities are set according to the finally determined route. For example, as shown in Figure 8, if a route is created for agent A2 that passes through clusters C5, C3, and C4 in that order, priorities a, b, and c are set for clusters C5, C3, and C4, respectively. Note that priorities are not limited to alphabetical values, and may be numerical values.

[0043] Next, we will explain how to create candidate routes using priority information. First, the agent selects one cluster to which it is assigned, and creates a route that visits all targets belonging to the selected cluster. This route is called an intra-cluster route. As with other routes, an evaluation function is used to search for intra-cluster routes. The evaluation function is the time it takes to visit all targets belonging to the selected cluster. Note that travel distance may be used instead of time. In other words, multiple intra-cluster routes are created, and each intra-cluster route is evaluated using the evaluation function. The intra-cluster route with the best evaluation result (i.e., the smallest value of the evaluation function) is then adopted. This process is performed for all clusters.

[0044] Then, a candidate route is created by combining the cluster visit order indicated in the priority information with the intra-cluster route.The evaluation function of the candidate route is the sum of the visit completion time of all targets and the visit completion time of each agent, as shown in Figure 8, for example.

[0045] As mentioned above, Patent Document 1 does not disclose the efficient generation of appropriate routes using clusters. Therefore, Patent Document 1 does not divide the process into two stages, such as determining the cluster visit order and the intra-cluster route, but rather directly generates a route for one agent to visit assigned targets. In this case, when there are a large number of targets, the amount of calculation becomes enormous, making it difficult to efficiently generate appropriate routes. In contrast, this embodiment uses a two-stage structure in which the cluster visit order is determined first and then the intra-cluster route is generated. This significantly reduces the number of routes to be considered. As a result, the amount of calculation can be reduced compared to the method of Patent Document 1. In addition, the cluster visit order and the intra-cluster route are each rationally generated routes, increasing the likelihood of generating appropriate routes. As described above, the method of this embodiment can efficiently generate appropriate routes.

[0046] Although Figure 7 shows an example in which there are multiple agents, this method can also be used to create a visiting route even if there is only one agent.

[0047] Next, an evaluation function that takes into account the task execution position will be described with reference to Figures 9 and 10. The process described below is applicable to cases where the priority information includes either "(1) the priority of allocation of agents to clusters, or (2) information including the priority of the order in which agents visit clusters."

[0048] In the following, it is assumed that an agent executes a task using a sensor on a target. The sensor may be, for example, a camera, radar, sonar, or acoustic sensor. The task is to measure the target using the sensor. The quality of the task is, for example, whether the amount of information obtained by measuring the target is large or useful.

[0049] Figure 10 shows evaluation function 1 and evaluation function 2. Evaluation function 1 is the same as the evaluation function shown in Figures 6 and 8. Evaluation function 2 is the sum of the evaluation values ​​of information collected by sensors. In other words, the higher the quality of the task, the larger the value of evaluation function 2 and the better the evaluation of the candidate route.

[0050] The path generation device 1 stores the distribution of evaluation values ​​for each sensor, as shown in Fig. 9. The upper diagram in Fig. 9 shows the distribution of evaluation values ​​for sensor 1. The lower diagram in Fig. 9 shows the distribution of evaluation values ​​for sensor 2. The circle in the center indicates the target, and the arrow indicates the direction of the target. The distribution of evaluation values ​​shown in Fig. 9 is based on the direction of the target, and is determined so that the direction of the target is parallel to the x-axis.

[0051] A common tendency shared by Sensor 1 and Sensor 2 is that the closer the sensor is to the target, the better the evaluation value. This is because the closer the sensor is to the target, the better the measurement results obtained. A unique tendency of Sensor 1 is that, assuming the target's direction of travel is in front, the evaluation value deteriorates at positions further back and forth or left and right from the target, while the evaluation value improves at other positions. This is because Sensor 1 is a camera, and capturing an image of the target from an oblique angle provides more information. On the other hand, Sensor 2 does not have a distribution dependent on the target's direction, and its evaluation value is determined solely by the distance from the target. Sensor 2 is, for example, an acoustic sensor. As such, in this embodiment, the distribution of evaluation values ​​varies depending on the sensor's characteristics. This makes it possible to create a visiting route that improves the quality of tasks according to the sensor's characteristics. The distribution of evaluation values ​​shown in FIG. 9 is merely an example, and a different distribution of evaluation values ​​from that shown in FIG. 9 may be used.

[0052] The route generation device 1 determines a task execution position when generating a candidate route. The task execution position is the position where the task is executed (measured by a sensor in this example). When evaluating a candidate route, the route generation device 1 calculates an evaluation function 2 based on the distribution of evaluation values ​​for each sensor. Specifically, as shown in FIG. 10, the route generation device 1 selects one task execution position and calculates an evaluation value based on the task execution position, the orientation of the target, and the distribution of evaluation values ​​for each sensor. Similarly, the route generation device 1 calculates and accumulates evaluation values ​​for all task execution positions. In this way, the value indicated by the evaluation function 2 can be calculated. The candidate route is evaluated based on both this evaluation function 2 and the above-mentioned evaluation function 1. As a result, the evaluation of a candidate route that increases the quality of the task increases, and a visiting route that increases the quality of the task can be generated.

[0053] Furthermore, when generating candidate routes from the second time onward, the evaluation function of previously generated candidate routes is taken into account. Methods for route search by determining task execution locations are known, and various publicly known techniques, such as optimization algorithms and machine learning algorithms, can be applied. Specifically, PSO, a type of metaheuristics, can be used for route search. PSO stands for Particle Swarm Optimization. Using PSO increases the probability of selecting a location near a task execution location where the past evaluation function value was good. Therefore, by repeating the search, an appropriate route can be efficiently generated. Since PSO is a publicly known technique, a detailed explanation is omitted. Note that using metaheuristics, or using PSO among metaheuristics, is just one example. Other methods for route search by determining task execution locations can also be adopted.

[0054] (Feature 1) As described above, the route creation method of this embodiment performs the following processing to create a visiting route for an agent to visit multiple targets. The multiple targets are clustered and classified into multiple clusters. Priority information that satisfies the following conditions 1 to 3 is created. Condition 1 is information that assists route search. Condition 2 is information that is created for each cluster. Condition 3 is information that includes the priority of allocation of agents to clusters or the priority of the order in which the agents visit clusters. Based on the target location, the agent location, and the priority information, candidate routes that are candidates for the visiting route are created for the agent. The candidate routes are evaluated based on an evaluation index that is an index for evaluating the candidate routes. Based on the evaluation results of the candidate routes, one or more candidate routes are selected as visiting routes.

[0055] Since priority information is added to clusters to search for routes, appropriate routes can be created more efficiently than methods that use clusters without adding priority information.

[0056] (Feature 2) In the route creation method of this embodiment, there are multiple agents for whom a visiting route is created. Priority information is created for each combination of agent and cluster. In Features 2 to 5, the priority information includes the priority of allocation of agents to clusters. When creating a candidate route, it is determined whether or not to visit targets belonging to the cluster in question, depending on the priority.

[0057] Since a priority can be set for allocation to a cluster, targets belonging to a cluster with a higher priority can be visited preferentially.

[0058] (Feature 3) In the route creation method of this embodiment, different priorities are set for each agent for the same cluster.

[0059] When multiple agents are assigned to visit multiple targets, the assignment can be set according to priority.

[0060] (Feature 4) In the route creation method of this embodiment, a priority is set for each cluster according to the distance between the cluster and the target agent.

[0061] Priorities can be set according to the distance between the cluster and the agent, allowing for efficient sharing of tasks to visit targets.

[0062] (Feature 5) In the route creation method of this embodiment, when creating candidate routes, different agents are allowed to be assigned to multiple targets belonging to one cluster.

[0063] Multiple agents can be assigned to targets belonging to one cluster (in other words, multiple agents can share one cluster), so targets can be efficiently shared among multiple agents.

[0064] (Feature 6) In the route creation method of this embodiment, in Features 6 to 8, the priority information includes a priority of the order in which the agent visits clusters. When creating candidate routes, candidate routes are created that visit targets belonging to the clusters in an order according to the priority.

[0065] Targets can be visited in an order that takes into account the priority of the visiting order for each cluster.

[0066] (Feature 7) In the route creation method of this embodiment, a route search is performed to visit representative positions of multiple clusters, and priority information including the priority of the cluster visit order is created. A route search is performed to visit all targets belonging to a cluster, and a route within the cluster is created. A candidate route is created by combining the cluster visit order based on the priority and the route within the cluster.

[0067] Since route creation can be divided into two stages, the amount of calculation can be reduced compared to when route search is performed without dividing it into stages.

[0068] (Feature 8) In the route creation method of this embodiment, there are multiple agents for whom visiting routes are to be created, and candidate routes are created for each agent.

[0069] When multiple agents exist, the amount of calculation increases, so the effect of reducing the amount of calculation can be effectively utilized.

[0070] (Feature 9) In the route creation method of this embodiment, the visited route is a route for executing a task to a target. Evaluation values ​​are set around the target according to their positions. Candidate routes are evaluated based on the evaluation values ​​set at the positions where the task is to be executed for the target.

[0071] A visiting route can be created taking into account the quality of the tasks to be performed for the target.

[0072] (Feature 10) In the route creation method of this embodiment, the visited route is a route for performing measurements of a target using a sensor. Evaluation values ​​are distributed around the target according to the characteristics of the sensor. Candidate routes are evaluated based on the evaluation values ​​set at positions where measurements of the target are performed.

[0073] You can create a visiting route taking into account the quality of measurement for the target.

[0074] (Feature 11) In the route generation method of this embodiment, the evaluation value is set so that the closer to the target, the better the evaluation value tends to be.

[0075] You can create a route to get as close as possible to your target.

[0076] The above-described features 1 to 11 can be combined as follows to realize a route generation method. The same applies to the route generation device. [Method 1] A route generation method having feature 1. [Method 2] A route generation method that is Method 1 and further has Feature 2. [Method 3] A route generation method according to Method 2, further comprising Feature 3. [Method 4] A route generation method according to Method 3, further comprising Feature 4. [Method 5] A route generation method that is Method 3 or 4 and further has Feature 5. [Method 6] A route generation method that is any one of Methods 1 to 5 and further has Feature 6. [Method 7] A route generation method according to Method 6, further comprising Feature 7. [Method 8] A route generation method according to Method 6 or 7, further comprising Feature 8. [Method 9] A route generation method that is any one of Methods 1 to 8 and further has Feature 9. [Method 10] A route generation method according to Method 9, further comprising Feature 10. [Method 11] A route generation method that is Method 9 or 10, and further has Feature 11.

[0077] The preferred embodiment of the present application has been described above, but the above configuration can be modified, for example, as follows. Each modification may be made alone, or multiple modifications may be made in any combination.

[0078] In the above embodiment, the process shown in Fig. 2 is performed by one route generation device 1, but it may be performed by multiple pieces of hardware working together. For example, the hardware that performs clustering and the hardware that performs route search may be separate.

[0079] Setting of the evaluation function taking into consideration the task execution position, which has been described with reference to FIGS. 9 and 10, is not an essential process and can be omitted.

[0080] The functions of the elements disclosed herein can be performed using circuits or processing circuitry, including general-purpose processors, special-purpose processors, integrated circuits, ASICs (Application Specific Integrated Circuits), conventional circuits, and / or combinations thereof, configured or programmed to perform the disclosed functions. A processor is considered a processing circuit or circuitry because it includes transistors and other circuitry. In this disclosure, a circuit, unit, or means is hardware that performs the recited functions or hardware that is programmed to perform the recited functions. The hardware may be hardware disclosed herein or other known hardware that is programmed or configured to perform the recited functions. Where the hardware is a processor, which is considered a type of circuit, the circuit, means, or unit is a combination of hardware and software, and the software is used to configure the hardware and / or processor. [Explanation of symbols]

[0081] 1. Route creation device 11 Processing equipment 12 Storage device 13. Communications equipment

Claims

1. A route creation method for creating a visiting route for an agent to visit a plurality of targets, Clustering the plurality of targets into a plurality of clusters; Create priority information that satisfies the following conditions 1 to 3, Condition 1: Information that assists route search Condition 2: Information created for each cluster Condition 3: The priority of the assignment of the agent to the cluster or the priority of the order of visiting the cluster by the agent is included. generating candidate routes for the agent based on the target location, the agent location, and the priority information, the candidate routes being candidates for the visiting route; evaluating the candidate routes based on an evaluation index that is an index for evaluating the candidate routes; A route generation method comprising: selecting one or more of the candidate routes as the visiting route based on an evaluation result of the candidate routes.

2. The route creation method according to claim 1, There are a plurality of agents for whom the visiting route is to be created, the priority information is created for each combination of the agent and the cluster; the priority information includes the priority of assignment of the agent to the cluster; A route generation method, wherein when generating the candidate route, it is determined whether or not to visit the target belonging to the cluster in question, depending on the priority.

3. 3. The route creation method according to claim 2, A path generation method, wherein different priorities are set for each agent for the same cluster.

4. The route creation method according to claim 3, A path generation method, wherein the priority is set for each cluster according to the distance between the cluster and the target agent.

5. The route creation method according to claim 3, A route creation method that allows different agents to be assigned to multiple targets belonging to one cluster when creating the candidate routes.

6. The route creation method according to claim 1, the priority information includes the priority of the order of visiting the clusters by the agent; A route creation method, wherein when creating the candidate route, the candidate route is created to visit the targets belonging to the cluster in an order according to the priority.

7. 7. The route creation method according to claim 6, performing a route search for visiting representative positions of the plurality of clusters, and creating priority information including the priority of the order of visiting the clusters; performing a route search to visit all of the targets belonging to the cluster and creating a route within the cluster; The route generation method generates the candidate route by combining the order of visiting the clusters based on the priority and the route within the cluster.

8. The route creation method according to claim 7, There are a plurality of agents for whom the visiting route is to be created, A route generation method that generates the candidate route for each of the agents.

9. The route creation method according to claim 1, the visiting route is a route for performing a task to the target, Setting evaluation values ​​according to positions around the target; A path generation method for evaluating the candidate paths based on evaluation values ​​set at positions where tasks are to be performed relative to the target.

10. The route creation method according to claim 9, the visiting route is a route for performing measurements of the target using a sensor, Distributing the evaluation values ​​around the target in accordance with the characteristics of the sensor; A route generation method for evaluating the candidate route based on the evaluation value set at a position where measurement is performed for the target.

11. The route creation method according to claim 9, A path generation method, wherein the evaluation value is set so that the closer to the target, the better the evaluation value tends to be.

12. a communication device that acquires target information and agent information; a processing device that creates a visiting route for an agent to visit a plurality of targets based on the target information and the agent information; Equipped with The processing device includes: Cluster multiple targets and classify them into multiple clusters. Create priority information that satisfies the following conditions 1 to 3, Condition 1: Information that assists route search Condition 2: Information created for each cluster Condition 3: The priority of the assignment of the agent to the cluster or the priority of the order of visiting the cluster by the agent is included. generating candidate routes for the agent based on the target location, the agent location, and the priority information, the candidate routes being candidates for the visiting route; evaluating the candidate routes based on an evaluation index that is an index for evaluating the candidate routes; A route generation device that selects one or more of the candidate routes as the visiting route based on the evaluation result of the candidate routes.

Images