Route planning device, its application equipment, and route planning method

Abstract

To provide a system that creates, for multiple agents, route planning which maximizes a value without creating extra-fine map information in advance and without causing a deadlock.SOLUTION: A route planning device includes: a business management unit that determines moving destinations of multiple agents; a map information management unit; an agent information management unit; a route plan calculation unit that creates, on the basis of information of the map information management unit and information of the agent information management unit, a route plan for moving each agent to the moving destination determined by the business management unit; and an evaluation unit that makes evaluation for each agent. The route plan calculation unit creates the route plan in such a way that, at each time, a total result of evaluation values calculated by a main purpose accomplishment evaluation unit, a main purpose accomplishment auxiliary unit and an intra-agent distance evaluation unit improves compared with a previous evaluation value. Such a device further includes an activity plan transmission unit that transmits, to all agents, the route plan calculated by the route plan calculation unit.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a route planning device, an application facility thereof, and a route planning method for an area in which a wide variety of agents exist. [Background technology]

[0002] To solve the labor shortage in transporting goods in logistics warehouses and between processes in factories, mobile robots (AGVs) are being developed. A utomated G guided V Vehicle, AMR: A utonomous M obile R The introduction of new technologies such as obot is progressing.

[0003] To introduce such mobile robots, it is necessary to set up the paths that the robots can move through within warehouses and factories (creating a graph consisting of nodes and edges, etc.). The more detailed this setting is, the greater the freedom of the paths that the robots can choose, enabling more efficient transport, but on the other hand, the setting up work requires a lot of man-hours. Furthermore, since the setting up of the paths is required every time the layout of the warehouse or factory changes, the work of setting up detailed paths places a heavy burden (engineering costs) on the operators who manage the warehouse or factory.

[0004] To address this issue, Patent Document 1 presents a method for calculating routes for multiple robots (referred to as vehicles in the patent document) to travel without designing detailed paths.

[0005] More specifically, the control inputs (velocity, angular velocity) for each robot are calculated according to the concept of model predictive control so that each robot does not come into contact with obstacles (pillars or walls inside a building, other robots, etc.) and the difference between each robot's current position and its target position is as small as possible.The movement path (position, posture) is then calculated by integrating the calculated control inputs. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Patent Publication No. 2021-77090 Summary of the Invention [Problem to be solved by the invention]

[0007] According to Patent Document 1, it is possible to calculate a path for a robot to reach a target point without having to design detailed passage information in a warehouse or factory in advance.

[0008] However, because Patent Document 1 follows the concept of model predictive control, it inherits the problems that model predictive control itself has. Model predictive control is a method of calculating control inputs so as to minimize a specific evaluation function, but because a numerical optimization method such as the gradient method is used to minimize (or maximize) the evaluation function, there is a high possibility that the calculated optimal solution will not be a global optimum, but a local optimum.

[0009] As described above, the method of Patent Document 1 may result in a local optimum state, which may result in a large number of situations where the robot is unable to operate (stuck, deadlock).

[0010] The present invention has been made to solve the above problems, and aims to provide a route planning device, application equipment, and route planning method that do not require detailed path information to be set in advance and that do not cause deadlocks. [Means for solving the problem]

[0011] In light of the above, the present invention defines a route planning device comprising: an operations management unit that determines destinations for a plurality of agents; a map information management unit that manages map information for the area where the agents are present; an agent information management unit that manages individual information about the agents; a route plan calculation unit that generates, based on information from the map information management unit and the agent information management unit, a route plan for moving each agent to the destination determined by the operations management unit; a main goal achievement evaluation unit that gives each agent a higher rating the closer the position of the agent gets to a target position; a main goal achievement assistance unit that gives each agent a lower rating if the position of the agent continues to be a predetermined distance or more from the respective target position; and an inter-agent distance evaluation unit that gives each agent a lower rating the closer the distance between the agents gets to a predetermined value that is set for each agent by the agent information management unit; wherein the route plan calculation unit generates the route plan at each time such that the overall result of the evaluation value calculated by the main goal achievement evaluation unit, the main goal achievement assistance unit, and the inter-agent distance evaluation unit is improved compared to the previous evaluation value; and further comprising an action plan transmission unit that transmits the route plan calculated by the route plan calculation unit to all agents.

[0012] Furthermore, the present invention defines this as "a path planning method characterized by: generating a route plan for moving each agent to its destination based on the destinations of multiple agents, map information of the area where the agents are located, and individual information about the agents; determining, for each agent, a first rating that gives a higher rating the closer the agent's position is to the target position; a second rating that gives a lower rating for each agent if the agent's position continues to be more than a predetermined distance from its respective target position; and a third rating that gives a lower rating for each agent the closer the distance between agents is to a predetermined value set for each agent; generating a route plan at each time such that the overall result of the first, second, and third ratings is improved over the previous rating value; and communicating this to all agents, all of which are realized by a computer."

[0013] In the present invention, the term "logistics site" refers to a site where a route planning device is applied, agent information is obtained by beacons and / or surveillance cameras, and an action plan is transmitted to the agent.

[0014] In addition, in the present invention, it is defined as "a parking lot to which a route planning device is applied, information on an agent is obtained by a surveillance camera, and an action plan is transmitted to a digital signage." [Effects of the Invention]

[0015] According to the present invention, each agent can generate a movement route that does not cause a deadlock, taking into consideration safety and operational constraints, without preparing detailed map information in advance.

[0016] Therefore, according to the embodiment of the present invention, the engineering costs for agent route planning can be significantly reduced. Furthermore, since movement routes for all agents are generated simultaneously instead of executing movement plans for each individual agent, overall optimization can be achieved. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a functional block diagram showing an example of the overall configuration of a route planning system according to a first embodiment of the present invention. [Figure 2] 1 is a block diagram showing an example of functions of a route planning device 1 according to a first embodiment of the present invention. [Figure 3a] A diagram showing the coordinates of agents. [Figure 3b] A diagram showing the coordinates of agents. [Figure 4] FIG. 10 is a diagram showing an example of an evaluation function related to the distance between agents. [Figure 5a] FIG. 10 is a diagram showing an example of the characteristics of a conventional evaluation function. [Figure 5b] FIG. 10 is a diagram showing an example of the position of an area plane. [Figure 5c] FIG. 10 is a diagram showing a case where the evaluation function has a characteristic that has a local optimum other than a global optimum. [Figure 5d]A diagram showing an example where a local optimum solution occurs on the circumference of a circle. [Figure 6a] A diagram showing the situation in which two robots are moving to a target point q behind the other robot. [Figure 6b] 10A and 10B are diagrams showing the operation of fixing the axes on a circular arc so that the deviation between the axes is constant; [Figure 7a] FIG. 10 is a diagram showing an example of the characteristics of the evaluation function of Equation (12). [Figure 7b] FIG. 10 is a diagram showing an example of how γi is given. [Figure 8] FIG. 10 is a diagram showing an example of the characteristics of the evaluation function of Equation (13). [Figure 9a] A diagram showing the agent's position, running trajectory, and goal point r on a two-dimensional plane. [Figure 9b] A diagram showing the time change in the deviation between the agent's position and the goal point on a two-dimensional plane. [Figure 10] FIG. 10 is a diagram showing the overall processing procedure of a control system according to a second embodiment of the present invention. [Figure 11] FIG. 10 is a diagram showing a detailed flow of the main purpose achievement assistance evaluation unit 42 shown in the processing function FC6. [Figure 12] FIG. 3 is a diagram showing an example of update timing of the route planning device 1. [Figure 13a] FIG. 1 is a diagram showing the stack state as an initial state. [Figure 13b] FIG. 10 is a diagram showing a state when the speed is reduced. [Figure 13c] FIG. 10 is a diagram showing the shape of an overall evaluation. [Figure 13d] Diagram showing movement from a local optimum. [Figure 13e] A diagram showing the search for a global optimum solution. [Figure 14a] FIG. 6B is a diagram showing the evaluation by the main purpose achievement evaluation unit 43 in the case of FIG. 6B. [Figure 14b] FIG. 10 is a diagram showing an evaluation function for a shape that does not have a local solution. [Figure 14c] A diagram showing a situation in which a stack is avoided. [Figure 15a] A diagram showing the state in which an agent approaches deviation eb1, which causes it to become stuck. [Figure 15b] FIG. 10 is a diagram showing the shape of the overall evaluation value when the penalty is increased. [Figure 15c] A diagram showing a situation in which a stack is avoided. [Figure 16] A diagram showing an example of a warehouse with a mix of workers and automated machines. [Figure 17a] A diagram showing an example of an agent's path generated up to prediction step k=5. [Figure 17b] A diagram showing an automated machine being rerouted in a collision-preventing direction. [Figure 17c] A diagram showing an automated machine stopping to prevent a collision. [Figure 18a] A diagram showing a situation where two automated machines are given target positions behind each other. [Figure 18b] A diagram showing a situation where two automated machines stop working. [Figure 18c] A diagram showing a situation in which a path is generated that avoids collisions by utilizing intersecting passages. [Figure 19] Diagram showing a parking lot with a mix of non-automated and automated vehicles. [Figure 20a] FIG. 10 is a diagram showing an example of a route plan that may occur when the conventional technology is applied to a parking lot. [Figure 20b] FIG. 10 is a diagram showing an example of a route plan generated when the present invention is applied. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0019] In the following, Example 1 will explain an example of the overall configuration of a route planning system, Example 2 will explain the processing flow of route planning, Example 3 will explain a mechanism for escaping from a stuck state, Example 4 will explain an example of application to a logistics site, and Example 5 will explain an example of application to a parking lot. [Example]

[0020] Fig. 1 is a functional block diagram showing an example of the overall configuration of a route planning system according to a first embodiment of the present invention. In Fig. 1, a route planning system 10 includes a route planning device 1 that obtains position information S1 and S2 from a monitoring device 2 and an agent 3, and provides action plan information S3 to the agent 3. To simplify the explanation, the example shown here is one in which there is only one agent 3 (including all moving objects such as robots, vehicles, and people), but the basic configuration remains the same even if the number of agents increases.

[0021] Of these, the agent 3 comprises a communication unit 31, a self-location acquisition unit 32, an other person's location acquisition unit 33, and a path following unit 34. All agents are equipped with these functions. However, if the agent 3 is a person, it is sufficient that the location of the person can be grasped by some means (for example, the monitoring device 2) and that an action plan can be communicated to the person.

[0022] In other words, Agent 3 is a mobile robot, and the mobile robot is equipped with LiDAR ( Li ght D etection A nd R It is equipped with sensors that acquire environmental information such as aging, and also uses this sensor information to implement SLAM ( S imultaneous L localization A nd M When calculating the self-position by point cloud mapping, the position information of other agents 3 can also be acquired as point cloud information. In such a case, the self-position acquisition unit 32 and the other agent's position acquisition unit 33 can be executed simultaneously.

[0023] On the other hand, if the agent 3 is a worker, the agent itself cannot have any specific functions. However, if the worker has a tablet with a wireless communication function such as Wi-Fi or Bluetooth, the functions of the communication unit 31 and the self-location acquisition unit 32 can be substituted. Also, if the worker follows the movement instructions displayed on the tablet, it can be said that the function of the path following unit 34 is substituted. However, since the tablet does not have the function to acquire the location information of other agents 3, the worker is treated as an agent 3 without the other-agent-location acquisition unit 33.

[0024] If the agent 3 is a mobile robot, the communication unit 31 receives command values ​​S3 for the action plan (route plan) distributed from the action plan transmission unit 15 in the path planning device 1, and the path following unit 34 controls each actuator (motor, etc.) so as to follow the received path plan. In addition, position information (self-position and other person's position) S2 acquired by LiDAR-SLAM is transmitted from the communication unit 31 to the agent information management unit 13 in the path planning device 1. When LiDAR-SLAM is used, map information of the management area can also be acquired, so if there is a change in the map information, such as a change in the warehouse layout, this information may be transmitted to the map information management unit 12.

[0025] In addition, the position information S1 of the agent 3 within the area may be acquired by an agent position acquisition unit 22 (such as a monitoring camera or beacon) in a monitoring device 2 installed within the area, and passed to the agent information management unit 13 in the route planning device 1 via a communication unit 21 installed in the monitoring device 2.

[0026] 2 is a block diagram showing an example of functions of a path planning device 1 according to a first embodiment of the present invention. In FIG. 2, the path planning device 1 calculates a movement route for each agent 3 in a path planning unit 14 using as input information: agent information S6 obtained from an agent information management unit 13 that manages the positions, IDs, etc. of all agents 3 (including all moving objects such as robots, vehicles, and people) in a management area (for example, a warehouse or a parking lot); task information S4 obtained from a task management unit 11 that manages the task contents (destinations) of the agents in the management area; and map information S5 obtained from a map information management unit 12 that manages map information of the management area. The path plan S3 calculated by the path planning unit 14 is transmitted to the agents 3 in the management area by a behavior plan transmission unit 15.

[0027] According to the information S6, S4, and S5 obtained from the agent information management unit 13, the business management unit 11, and the map information management unit 12, the current location of each agent 3 within the management area, where it has gone, and what it will do are known in chronological order as past and future actions.

[0028] The route planning unit 14 is composed of a main goal achievement evaluation unit 43, an inter-agent distance evaluation unit 41, a main goal achievement assistance unit 42, and a route plan calculation unit 44, and the output of the route plan calculation unit 44 is action plan information S3.

[0029] 2, the main goal achievement evaluation unit 43, the inter-agent distance evaluation unit 41, and the path plan calculation unit 44 are functions that have been conventionally provided, but in the present invention, it can be said that the function of the main goal achievement assistance unit 42 has been added to these. Basically, a path can be planned using the functions of the main goal achievement evaluation unit 43, the inter-agent distance evaluation unit 41, and the path plan calculation unit 44, which are conventional configurations, but if this path plan causes the robot to fall into a state where it cannot operate (a stuck state: stuck, deadlock), the function of the main goal achievement assistance unit 42 allows it to escape from the stuck state.

[0030] First, the functions of the conventional part will be explained. The main goal achievement evaluation unit 43 receives as input task information S4 obtained from the task management unit 11 that manages the task contents (destinations) of agents in the managed area, and map information S5 obtained from the map information management unit 12 that manages map information of the managed area. Using this information, the main goal achievement evaluation unit 43 calculates an evaluation value that is used to accomplish the task that each agent 3 should accomplish, that is, to accomplish movement to a specific position (target point).

[0031] The concept of the processing by the main goal achievement evaluation unit 43 will be explained using Figures 3a and 3b. These figures show the current position of the agent 3 within the management area indicated by XY coordinates.

[0032] When the i-th agent 3 is a mobile robot, it determines its own position (x i ,y i ) and attitude θ i Let p be the position vector composed of i When the target position r i =(rx i ,ry i ,rθ i ) is moved by its own position vector p i and target position r i Deviation e i The goal of path planning is to approach 0 using, for example, equation (1).

[0033]

number

[0034] The equation of motion for the mobile robot in Figure 3a can be given by equation (2), where vi is the moving speed of agent i, ωi is the angular velocity of agent i, and the control input can be given by two inputs (v, ω).

[0035]

number

[0036] On the other hand, as shown in Figure 3b, if Agent 3 is a robot or human worker driven by omni-wheels, it can move freely in the X and Y directions, so the equation of motion follows equation (3). In this case, the control inputs are three: (vx, vy, ω).

[0037]

number

[0038] Moving agent 3 faster improves movement efficiency, but it also increases energy consumption. The energy consumption at each time can be given by equation (4) as a function of the agent's speed and angular velocity.

[0039]

number

[0040] Therefore, in order for agent 3 to reach the target value while suppressing energy consumption, it is desirable to use an evaluation value composed of the deviation e between its own position p and the target point r and the energy consumption E, as shown in equation (5).

[0041]

number

[0042] Here, k means time, k0 is the evaluation start time, and kN is the evaluation end time. α is a weight related to deviation, and β is a weight related to energy consumption. The behavior of agent 3 can be adjusted by adjusting the balance between α and β. For example, if β is made larger than α, energy consumption can be reduced, and energy-saving mode operation can be achieved. On the other hand, if β is set to 0, the fastest route will be calculated without considering energy consumption.

[0043] The function of the main goal achievement evaluation unit 43 is to evaluate equation (6) taking into account equation (5) for all N agents 3 .

[0044]

number

[0045] Next, we will explain the inter-agent distance evaluation unit 41 in Fig. 2, which is also a function of the conventional part. The inter-agent distance evaluation unit 41 receives as input agent information S6 obtained from the agent information management unit 13, which manages the positions, IDs, etc. of all agents 3 within the management area, and map information S5 obtained from the map information management unit 12, which manages map information of the management area. Using this information, the inter-agent distance evaluation unit 41 has the function of correcting the evaluation value of the main objective achievement evaluation unit 43 so that the distance between each agent received from the agent information management unit 13 is equal to or greater than a predetermined value, in order to prevent the agents from coming into contact with each other.

[0046] The distance dij between agent i with position vector Pi and agent j with position vector Pj can be calculated using equation (7). Path planning is performed so that this distance dij is equal to or greater than the margin distance da given to each agent, in other words, so that equation (8) is satisfied.

[0047]

number

[0048]

number

[0049] To realize equation (8), it may be incorporated as a constraint in the optimization calculation of the path planning calculation unit 44, which will be described later. Also, an evaluation function such as equation (9) may be introduced. ε in equation (9) is a small coefficient to avoid division by zero.

[0050]

number

[0051] Equation (9) includes equations (9a) and (9b). Equation (9a) is an evaluation only at a specific time, while equation (9b) indicates all evaluations from the prediction period k = k0 to kN at the evaluation time, in accordance with the concept of model predictive control.

[0052] To simplify the explanation, we will focus on equation (9a), which is an evaluation at a specific time. Figure 4 is a diagram showing the evaluation function of equation (9a), with the distance dij between agents on the horizontal axis and the magnitude of the evaluation function on the vertical axis. Here, since the evaluation function of equation (9a) has the shape shown in Figure 4, the closer the distance dij between agents is to the margin distance da, the larger the value it takes. In other words, by imposing a larger penalty as the distance between agents gets closer, it is possible to prevent contact between the agents 3.

[0053] Next, we will explain the features of the inter-agent distance evaluation unit 41. Here, it is desirable that the margin distance da be changed in accordance with the agent information S6 held by the agent information management unit 13.

[0054] [Table 1]

[0055] Table 1 is an example of a margin distance da designed with a focus on contact safety. It assumes contact for each agent type (worker, slow-speed machine, high-speed machine). Workers move slowly, and even if they come into contact with each other, the likelihood of an accident occurring is low, so the margin distance da for agents is set to 30 cm. On the other hand, if a high-speed machine (such as a forklift) comes into contact with a person, it could result in a serious accident, so the margin distance da is set to 100 cm. Other combinations are designed using similar guidelines.

[0056] [Table 2]

[0057] Table 2 is an example of a design for the margin distance da in addition to Table 1, taking into account the perspective of infection prevention. Here too, contact between each agent type (worker, slow-speed machine, high-speed machine) is assumed. Because the possibility of droplet infection is low between people and machines, the margin distance da is the same as in Table 1. On the other hand, because the possibility of droplet infection increases when people pass each other, it is desirable to ensure social distance by increasing the margin distance da. In both Tables 1 and 2, the margin distance da is determined not only by the characteristics of the agent, but also by factors such as the width of the corridors in the area.

[0058] It is also possible to avoid contact with obstacles by treating obstacles (walls, pillars, etc.) in the map information S5 received from the map information management unit 13 as virtual agents and evaluating the distance di between each agent and the obstacle in a format similar to equation (9). Alternatively, obstacle information may be stored in a cost map C, and equation (10) may be used to evaluate the position pi of each agent in the cost map.

[0059]

number

[0060] The above is the control in a conventional path planning device, and path planning is possible using the functions of the main goal achievement evaluation unit 43, the inter-agent distance evaluation unit 41, and the path plan calculation unit 44. However, if this path planning causes the robot to fall into a state where it cannot operate (a stuck state: stuck, deadlock), it is not possible to escape from the stuck state. The inability to escape from a stuck state will be explained using Figures 5a, 5b, 5c, 5d, 6a, and 6b.

[0061] For example, in Patent Document 1, the current position of the mth robot is pm, the target position is qm, and the deviation is given as em=pm-qm, and an evaluation function such as equation (11) is designed.

[0062]

number

[0063] Here, if there is only one robot (when m = 1) and no obstacles, the evaluation function will have the shape shown in Figure 5a (horizontal axis: deviation, vertical axis: evaluation function), so if we calculate the control input to minimize the evaluation function, we can achieve a state where the deviation e becomes 0 (the robot reaches the target value).

[0064] However, as shown in the area plane of Figure 5b, if there is an obstacle between the robot's current position p1 and the target point q1, the above-mentioned operation cannot be realized. In the case of Figure 5b, to reduce equation (11), the operation of moving in the positive direction on the Y axis is selected. However, when moving in the positive direction on the Y axis, an obstacle is encountered, and it becomes impossible to approach the target point q1 any further.

[0065] To escape from this state, it is necessary to move in the negative direction of the Y axis, but the deviation e1 between the robot's own position p1 and the target point q1 increases, which increases the value of the evaluation function. Therefore, the control input cannot be calculated using the method of Patent Document 1. This situation occurs when the evaluation function has a local optimum other than the global optimum, as shown in Figure 5c.

[0066] A similar situation always occurs when the target point q1 is on the opposite side of the obstacle (passage). For example, if the robot is moving in the positive direction of the X axis in a situation like that shown in Figure 5d, the local optimum will be on the circumference of a circle where the deviation e1 from the target point q1 is a specific distance. To move further in the positive direction of the X axis than this point will increase the evaluation function unless an arc motion is performed that also involves movement in the negative direction of the Y axis. Even if the robot reaches point B by moving in an arc, it will be unable to go any further because there is an obstacle.

[0067] As mentioned above, with the method of Patent Document 1, there is a possibility that the robot will not be able to reach the target point even in an environment where there is only one robot. When there are multiple robots, it becomes even more difficult for the robots to reach the target point.

[0068] For example, consider a situation in which two robots are moving to a target point q behind the other robot, as shown in Figure 6a. If the round robot (m = 1) moves in the negative direction of the Y axis and waits for the square robot (m = 2) to pass, or if it can realize a detour operation (a detour route is generated in the figure), then both robots can reach the target point q.

[0069] However, such a motion cannot be generated unless the deviation e1 of the round robot is increased. The motion realized by reducing the deviation em is as shown in Figure 6b, where the robots move forward and are fixed on an arc where the deviation em is constant, with the robots spaced a distance d apart to avoid collisions.

[0070] In this way, conventionally, when the control input (vi, ωi) is calculated to minimize the evaluation function (equation (11)) with β = 0 in equation (6), and an attempt is made to calculate a route using this result, there is a situation where the agent falls into a local optimum and is unable to move.

[0071] Further examination of the situation in which the agent falls into a local optimum and is unable to move can lead to several patterns. Based on this, the present invention proposes countermeasures corresponding to each pattern.

[0072] The main goal achievement assistance unit 42 of the present invention is introduced to resolve such a situation and assist each agent in reaching the goal point r. First, a situation in which each agent is in a deadlock can be rephrased as the agent's movement speed becoming 0 (zero). This is the first pattern assumed by the present invention.

[0073] In order to avoid the first pattern of the situation, the primary goal achievement assistance unit 42 introduces the evaluation function of equation (12). Note that equation (12) includes equations (12a) and (12b). Equation (12a) is an evaluation only at a specific time, while equation (12b) indicates all evaluations from the prediction period k = k0 to kN at the evaluation time, in accordance with the concept of model predictive control. The following explanation will be given using equation (12a).

[0074]

number

[0075] Equation (12a) has the shape shown in Figure 7a (horizontal axis: v i 2 ,Vertical axis: Magnitude of evaluation function), so the closer each agent's movement speed vi is to 0 (zero), the greater the penalty will be. By adding such a function to the evaluation function, behavior will be generated that will cause the agent to resolve the stopped state.

[0076] More specifically, even if the deviation e between the self-position p and the target point r increases, the total evaluation function will be lower than if the moving speed vi becomes 0 (zero). As a result, the agent can achieve detour behavior in situations like those in Figures 5b, 6a, and 6b. Similarly, the situation in Figure 5d can also be resolved.

[0077] If a penalty is given when the speed vi becomes 0 (zero), the agent will not be able to stop even if it reaches the goal point r. i As shown on the vertical axis (γi), when the deviation ei between the agent's position q and the goal point r becomes equal to or less than a predetermined value eth, γi is changed to 0. If γi = 0, no penalty is given even if the speed vi becomes 0 (zero), so the agent can reach the goal point.

[0078] Note that FIG. 7b is one example of how to give γi, and any method of design may be used as long as γi=0 is satisfied when the deviation ei is 0.

[0079] Furthermore, the situation in which each agent is in deadlock can be rephrased as a state in which the agent remains at a specific position pa. This is the second pattern assumed by the present invention.

[0080] In order to avoid the second pattern of situations, the primary goal achievement assistance unit 42 introduces the evaluation function of equation (13). Note that equation (13) includes equations (13a) and (13b). Equation (13a) is an evaluation only at a specific time, while equation (13b) indicates all evaluations from the prediction period k = k0 to kN at the evaluation time, in accordance with the concept of model predictive control. The following explanation will be given using equation (13a).

[0081] Equation (13a) has the shape shown in Figure 8 (horizontal axis: (Pa-Pi) 2 , vertical axis: magnitude of evaluation function), so the closer each agent's position pi is to pa, the greater the penalty will be. By adding such a function to the evaluation function, behavior that will cause the agent to resolve the stopped state will be generated.

[0082]

number

[0083] It is desirable that the specific position pa be recorded in the map information management unit 12 in advance if the map shape is such that it is possible to predict in advance that a deadlock will occur.

[0084] Furthermore, if the position pi of each agent acquired by the agent information management unit 13 does not change for a certain period of time, the position pi of that agent may be set as the specific position pa described above. If this method is adopted, a large penalty is imposed at the timing when pa = pi is set, so it is expected that the i-th agent will start moving immediately. Once the agent starts moving, it is desirable to cancel the set pa.

[0085] There may also be situations where each agent is not stopped but cannot reach the goal. For example, as shown in Figure 9a, this situation occurs when agent 3 is moving in a circle at a constant speed while remaining at a certain distance from the goal. This is the third pattern assumed by the present invention.

[0086] Figure 9a shows the position Pi of the i-th agent on a two-dimensional plane, as well as its running trajectory and goal point r, and Figure 9b shows the time change in the deviation between the i-th agent's position Pi and the goal point r. At this time, agent 3 is running at a constant speed on a circle (minimum deviation min ei, maximum deviation max ei) while remaining at a certain distance from the goal point ri.

[0087] In the situation shown in Figure 9a, since the agent is moving at a constant speed, it is difficult to impose a penalty using equation (12). Also, since the agent is not staying at a specific position, it is difficult to impose a penalty using equation (13).

[0088] A situation in which each agent is stopped, or as shown in Figure 9a, can be rephrased as a situation in which the minimum value min ei of the deviation ei confirmed within a specified time is not updated. To resolve this situation, a penalty is introduced that evaluates the integral value of the deviation ei of each agent, as shown in equation (14). When traveling on a circular trajectory as shown in Figure 7, the integral value of the deviation increases as time passes, making it more likely that an action that deviates from the circular trajectory will be selected.

[0089]

number

[0090] Equation (14) is similar to the evaluation function in equation (6) where β = 0 in that it evaluates the deviation ei, but it differs in that equation (6) only evaluates the cumulative value of the deviation over a period of "kN - k0". For example, when the time from ta to tb in Figure 9b corresponds to "kN - k0", the evaluation value of equation (6) decreases, and it is determined that appropriate behavior is being generated. On the other hand, equation (14) increases unless the deviation ei is 0, allowing for longer-term evaluation.

[0091] Note that when equation (14) is used, all past deviations ei have an effect, so it is desirable to reset the integral and set the weight ηi to zero each time the minimum value min ei is updated.

[0092] As described above, the main goal achievement assistance unit 42 can realize an escape function by being equipped with a plurality of countermeasure methods corresponding to the above-described patterns. The main goal achievement assistance unit 42 only needs to be equipped with at least one of the above methods. Naturally, it may be equipped with all three of the above-described methods.

[0093] Returning to FIG. 2, the path plan calculation unit 44 integrates the evaluation functions set by the main goal achievement evaluation unit 43, the inter-agent distance evaluation unit 41, and the main goal achievement assistance unit 42, and calculates the control input u so as to minimize the value of the evaluation function.

[0094] Specifically, the control input u is calculated so as to minimize the evaluation function (equation (15)) obtained by integrating the evaluation functions JI set for each element using the weighting coefficient αI. Note that, when a specific method of the main goal achievement assistance unit 42 is to be disabled, the corresponding weighting coefficient αI (I=4, 5, 6) can be set to 0 (zero).

[0095]

number

[0096] The method for calculating the control input u itself follows the well-known concept of model predictive control and does not involve any special ingenuity, so a description thereof will be omitted.

[0097] Once the optimal control input u is determined at each time, time series data of position (x, y) and orientation (θ) can be easily generated using the corresponding equations of motion (equations (2) and (3)). This time series data becomes the path plan for each agent. This path plan itself is the same as that in Patent Document 1 and is a widely known method, so a detailed explanation will be omitted.

[0098] The route plan calculation unit 44 may also determine the pattern and execute a specific countermeasure, or may execute a plurality of countermeasures in sequence and wait for one of them to resolve the problem.

[0099] The calculated path plan is distributed to each agent via the action plan transmission unit 15. [Example]

[0100] An example of the overall configuration of the route planning system has been described in the first embodiment. In the second embodiment, a processing flow in the route planning system will be described.

[0101] The overall processing procedure of the control system according to the second embodiment of the present invention will be described with reference to the flowchart of FIG.

[0102] First, processing functions FC1 to FC3 acquire information necessary for the calculations of the route planning device 1. Processing function FC1 acquires position information S6 of agents within the area. This function corresponds to the agent information management unit 13. Processing function FC2 acquires task content S4 (destination point ri to be moved to) to be given to each agent within the area. This function corresponds to the task management unit 11. Processing function FC3 acquires map information S5 within the area. This function corresponds to the map information management unit 12. Note that map information S5 is updated less frequently than task content and agent position information, so it does not necessarily need to be updated at every step. Note that the processing order of processing functions FC1 to FC3 can be any order, and it is important that all information is gathered before transitioning to processing function FC4 or later.

[0103] The processing function FC4 evaluates whether each agent will move to the target location according to the position S6 and task content S4 of each agent acquired by the processing functions FC1 and FC2. This function corresponds to the main goal achievement evaluation unit 43.

[0104] Processing function FC5 uses the positions S6 of each agent acquired by processing functions FC1 and FC3, and map information S5, to evaluate how each agent will act so as not to come into contact with obstacles or other agents. This function corresponds to inter-agent distance evaluation unit 41.

[0105] In processing function FC6, using the information S4, S5, and S6 acquired by processing functions FC1 to FC3, an evaluation is performed to resolve the situation in which each agent stops while moving to the goal point. This function corresponds to the main goal achievement assistance unit 42.

[0106] The processing order of the processing functions FC4 to FC6 can also be any order, and it is important that all information is gathered before transitioning to processing function FC7 and onwards. As described above, when the idea of ​​model predictive control is used, the processing functions FC4 to FC6 evaluate values ​​from the initial time k0 to the prediction step number (kN-k0) ahead.

[0107] Processing function FC7 calculates the control input for each agent so as to minimize the evaluation function calculated by processing functions FC4 to FC6, and calculates the movement path by integrating the input. This function corresponds to the action plan calculation unit 44.

[0108] Once the path plan is generated in processing function FC7, the process transitions to processing function FC8.

[0109] The processing function FC8 distributes the route plan generated by the processing function FC7 to each agent. This function corresponds to the action plan transmission unit 15.

[0110] A more detailed flowchart of an embodiment of the main purpose achievement assistance evaluation unit 42 shown in the processing function FC6 will be described with reference to FIG.

[0111] 11, the first processing function FC61 of the main purpose achievement assistance evaluation unit 42 determines whether to enable the function of imposing a penalty on deceleration at locations other than the target point, which is the first realization means for pattern 1 in the main purpose achievement assistance evaluation unit 42. If this function is enabled (YES), the process transitions to processing function FC62, and if it is disabled (NO), the process transitions to processing function FC63.

[0112] Whether or not to enable the first implementation method may be switched by the user (area administrator) setting. Furthermore, even if the user decides to enable the first implementation method, this function will be disabled if the agent 3 is near the target point.

[0113] When the process transitions to the processing function FC62, a penalty value for stopping (deceleration) at a point other than the destination point is calculated according to the formula (12).

[0114] When the process transitions to processing function FC63, it is determined whether to enable the function of imposing a penalty on staying at a specific position pa other than the target point, which is the second realization means for pattern 2 in the main purpose achievement assistance evaluation unit 42. If this function is enabled (YES), the process transitions to processing function FC64, and if it is disabled (NO), the process transitions to processing function FC67.

[0115] Whether or not the second implementation method is enabled may be switched by the user's settings. Furthermore, even if the user decides to enable the second implementation method, this function may be disabled if the agent is not stopped.

[0116] In processing function FC64, the position pi where the i-th agent is stopped is given as the specific position pa for penalty calculation. Note that if it is possible to predict in advance where the agent will stop based on the map shape, this position may be set in advance as pa instead.

[0117] When the process transits to the processing function FC65, a penalty value for stopping at a specific position pa other than the destination point is calculated according to the formula (13).

[0118] Processing function FC66 determines whether to enable a function for giving a penalty by using the integral of the deviation between the agent position and the goal point, which is the third realization means for pattern 3 in the main goal achievement assistance evaluation unit 42. If this function is enabled (YES), the process transitions to processing function FC67, and if it is disabled (NO), the process transitions to processing function FC60.

[0119] Whether or not to enable the third implementation means may be switched by the user's settings, but it is desirable to enable it automatically if the user disables the first and second implementation means. Note that, as will be shown in a specific embodiment described later, when the agent is a worker, the worker can resolve a deadlock (stuck) situation by himself / herself, so only in such a case, the first to third implementation means may all be disabled.

[0120] When the process transitions to processing function FC67, the amount of change in the deviation between the agent's position and the goal point at a specific time is analyzed. In processing function FC68, if the minimum value of the deviation acquired in processing function FC67 has changed (YES), the process transitions to processing function FC69b. On the other hand, if the minimum value of the deviation acquired in processing function FC67 has not changed (NO), the process transitions to processing function FC69a.

[0121] In FC69a, a penalty is calculated using the integral value of the deviation ei according to equation (14). In processing function FC69b, an operation is performed to reset the integral value of the deviation ei. If the process of calculating the integral value is performed only when transitioning to processing function FC69a, the process of processing function FC69b does not need to be performed.

[0122] Processing function FC60 integrates the penalties by adding up the penalties calculated by processing functions FC62, FC65, and FC69a. Values ​​for which calculations were not performed by processing functions FC62, FC65, and FC69a are replaced with 0 (zero).

[0123] Next, an example of the update timing of the route planning device 1 will be described with reference to Fig. 12. In this example, the number of agents is N=3 (3A, 3B, 3C), but similar processing can be performed even if the number of agents increases.

[0124] First, it is assumed that all agents are stopped at time t0 when the path planning device 1 starts to operate. When the path planning device 1 receives input such as the position information of each agent 3, it calculates a path plan and distributes the path plan to each agent 3 in sequence.

[0125] When the first route is distributed at time t1, each agent starts moving to the goal point. Since the distance and speed traveled to the goal point r differ for each agent 3, the travel times often do not match.

[0126] Looking at agent 3A, it reaches the target point at time t2, and upon detecting this, the route planning device 1 assigns a new target point to agent 1 and re-plans the route. The re-planning of the route is carried out in the order of input, calculation, and output, and at time t3 the re-planned route can be sent to agent 3A.

[0127] At the time of calculation of this route plan (between t2 and t3), Agent 3B and Agent 3C have already been given routes, so route planning can be done only for Agent 3A. However, planning routes for all agents results in overall optimization, and therefore a better action plan can be realized for the entire area.

[0128] Once the route plan calculation is complete, each agent is given a new route plan at time t3. After that, the same process is performed each time each agent reaches the goal point.

[0129] However, if it becomes necessary to re-execute the route plan, for example, if a load shift occurs in the warehouse and a specific passage becomes impassable, interrupt processing may be performed at time t4 when no agent has reached the target point. [Example]

[0130] In the third embodiment, a mechanism for escaping from a stuck state using the techniques of the above-described embodiments will be described in chronological order.

[0131] First, let us consider the case of Figure 5b where one agent 3 gets stuck. For simplicity's sake, let us assume that there is no inter-agent distance evaluation unit 41 or main goal achievement assistance unit 42, and only a main goal achievement evaluation unit 43. In this case, the shape of the evaluation function will be as shown in Figure 5c. Therefore, in order to improve the overall evaluation (reduce the value of the evaluation function), the path plan calculation unit 44 moves the agent in the positive direction of the y-axis so as to reduce the deviation e.

[0132] Figure 13a shows Figures 5b and 5c side by side, showing the stuck state as the initial state. With these characteristics, the agent is stuck in a local optimum when it should be moving to a global optimum. At this time, as shown in Figure 13a, the value of the evaluation function decreases from f1 to f2 (the overall evaluation improves). After that, the agent cannot move any further because there is a wall between the agent and the target position. If the agent turns back from this state (moves in the negative direction of the y-axis), the deviation e increases, causing the value of the evaluation function to exceed f2, meaning the overall evaluation decreases. Since the path planning calculation unit 44 has the function of calculating a path that improves the overall evaluation, it cannot calculate actions that would result in a decrease in the overall evaluation. Therefore, the agent remains stopped.

[0133] Next, we will explain the operation when the main goal achievement assistance unit 42 is added. Here, we will explain the case where the main goal achievement assistance unit 42 (equation (12), Figure 7a) that focuses on the agent's movement speed is adopted. When the main goal achievement assistance unit 42 is added, as shown in Figure 13a, when the agent begins to decelerate as it approaches the wall, the penalty increases rapidly according to the function shown in Figure 7a. In other words, as the speed decreases, the output of the main goal achievement assistance unit 42 increases, as shown in Figure 13b. As a result, the overall evaluation, which is the sum of the main goal achievement evaluation unit 43 and the main goal achievement assistance unit 42, takes on the shape shown in Figure 13c. In other words, as the agent moves to reduce the deviation e and decelerates as it approaches the wall, the overall evaluation value increases from f1 to f3 (the evaluation decreases).

[0134] When only the main goal achievement evaluation unit 43 is used, the solution falls into a local optimum as shown in Fig. 13a when the deviation e is ea1, but when the main goal achievement assistance unit 42 is added, the solution does not fall into a local optimum as shown in Fig. 13c when the deviation is ea1. This allows the path plan calculation unit 44 to search for a solution that improves the overall evaluation value while reducing the deviation e, as shown in Fig. 13d. When the agent's speed exceeds a certain level, the effect of the penalty by the main goal achievement assistance unit 42 disappears, but when the deviation e falls below ea2 as shown in Fig. 13d, the solution does not fall into a local optimum and the path plan calculation unit 44 can search for a global optimum where the deviation e is 0, as shown in Fig. 13e.

[0135] Next, consider the case of Figure 6b, where two agents 3 become stuck. Figure 14a shows the evaluation of the main goal achievement evaluation unit 43 for each agent in the situation shown in Figure 6b. In the situation in Figure 6b, agent 3 at position p1 makes a detour to move to the destination point, so the evaluation function has a shape with a local solution, as shown in Figure 14a.

[0136] On the other hand, agent 3 at position p2 is moving to the destination point only by moving straight ahead, so the evaluation function for a single agent will have a shape that does not have a local solution, as shown in Figure 14b. However, since the path planning calculation unit 44 does not optimize the behavior of individual agents, but rather the actions of all agents, the overall evaluation value will be optimized by an evaluation function with a shape like that shown in Figure 14c, which is the sum of Figures 14a and 14b. The evaluation function in Figure 14c has a local solution where the total deviation e is eb1. The situation where this local solution occurs corresponds to the situation shown in Figure 6b, where the agents are unable to move away from each other due to the predetermined distance d.

[0137] In contrast, as shown in Figure 15a, when each agent approaches deviation eb1 where stuck occurs, the speed of each agent decreases, and the main goal achievement assistance unit 42 increases the penalty. As the penalty increases, the shape of the overall evaluation value used by the path plan calculation unit 44 changes as shown in Figure 15b, and the value of deviation eb1 is no longer a local solution. Therefore, as shown in Figure 15c, it is possible to avoid a situation where all agents become stuck at the original local solution (eb1). [Example]

[0138] In Example 4, the application of the present invention to a factory or logistics site will be described. Fig. 16 is a simplified diagram of a warehouse 100 in which a worker 3B, who is an agent 3, and an automated machine (forklift) 3A coexist. In the diagram, only one worker 3B and one automated machine 3A are shown for the sake of clarity, but the environment may also have multiple workers and multiple automated machines, or may include a mixture of regular, non-automated machines.

[0139] The worker 3B wears or holds a smart device (not shown). The smart device may be a tablet terminal or a wearable device such as goggles. The smart device can notify and guide the worker 3B of the destination by displaying the route to the destination on the monitor of the tablet terminal or in the goggles.

[0140] The smart device can measure the position of the worker 3B wearing or holding the smart device by wirelessly communicating with the beacon 103 placed in the warehouse area 100. The smart device also has a communication function and can communicate with the control server 104. Furthermore, the situation within the area is monitored by the surveillance camera 101 placed in the warehouse area 100.

[0141] The control server 104 corresponds to the route planning device 1 of the present invention. Note that the functions of the route planning device 1 do not have to be implemented on the same server. For example, only the route planning unit 44, which has a high processing load, may be executed on a specific server. Also, the control server 104 does not necessarily have to be installed inside the warehouse 100.

[0142] The automated machine 3A is equipped with a LiDAR (not shown). By processing the point cloud data collected by the LiDAR, the automated machine creates a warehouse map and estimates its own position at the same time. The SLAM function is implemented. Furthermore, the automated machine 3A is equipped with a controller (not shown) that executes various calculations related to automated driving.

[0143] The controller of the automated machine 3A has a communication function and transmits its own position acquired by SLAM and map information to the control server 104, and receives a route plan from the control server 104. The controller controls the actuators (steering motor, travel motor, etc.) of the automated machine 3A to follow the movement plan. The worker 3B and the automated machine 3A correspond to the agent 3 in this invention.

[0144] Moreover, the monitoring camera 101 placed in the warehouse corresponds to the monitoring device 2 of FIG. 1 in the present invention.

[0145] Hereinafter, assume that the business management unit 11 in Figure 1 has set up a task to have worker 3B retrieve an item from position r1 on shelf 102B, and to have automated machine 3A retrieve an item from position r2 on shelf 102A.

[0146] The path planning device 1 calculates paths that the worker 3B and the automated machine 3A should follow, with the worker 3B being the agent 1 and the automated machine 3A being the agent 2, respectively.

[0147] First, the control server 104 receives the current positions p1 and p2 from the smart device of the worker 3B and the controller of the automated machine 3A, respectively. This function corresponds to the agent information management unit 13 of the present invention. Also, the target points of the worker 3B and the automated machine 3A determined by the work management unit 11 are defined as r1 and r2.

[0148] The route planning unit 14 plans routes to guide the worker 3B from the current position p1 to the target point r1, and the automated machine 3A from the current position p2 to the target point r2. First, when the positions of both agents are sufficiently far apart, the evaluation value of the main goal achievement evaluation unit 43 is dominant, so routes are gradually generated so that each becomes the shortest route.

[0149] Figure 17a shows an example of the paths of each agent generated up to prediction step k = 5. If we draw a path in which automated machine 3A moves in the negative direction of the Y axis from time point (k-5) on the path in Figure 17a, there is a possibility that automated machine 3A and worker 3B will come into contact at time k = 6.

[0150] In this situation, the penalty imposed by the inter-agent distance evaluation unit 41 becomes dominant. Therefore, as shown in Figure 17b, the automated machine 3A generates a path that moves in the negative direction of the X axis to prevent a collision, rather than the shortest path (movement in the negative direction of the Y axis). From this point on, as there is no possibility of contact, paths are sequentially generated so that each agent follows the shortest route.

[0151] If the weight β relating to the energy consumption in equation (6) is set large in the primary goal achievement evaluation unit 43, it is also possible to realize an operation in which the automated machine 3A waits at position p2(4) reached at time k=4 as shown in Figure 17c, without generating a route that detours the automated machine 3A as shown in Figure 17b. Figure 17c shows that the automated machine 3A stops at position P2(4) and does not move between prediction steps k=4 and k=7.

[0152] By utilizing this invention, various route plans can be automatically realized depending on the design of the evaluation function without the need to prepare detailed maps (node ​​and edge information) in advance, thereby significantly reducing engineering costs.

[0153] Note that the operation of Figure 17c cannot be realized due to the effect of the penalty (equation (12)) imposed by the first realization means of the main goal achievement assistance unit 42. Therefore, when prioritizing the reduction of energy consumption, it is desirable to disable the first realization means of the main goal achievement assistance unit 42 and prevent the occurrence of deadlock by using the second and third realization means.

[0154] Next, consider the situation shown in Figure 18a, in which two automated machines 3D (agent 1) and 3C (agent 2) are assigned target positions r1 and r2 behind their respective initial positions p1 and p2, as a situation in which a deadlock could occur in the prior art (Patent Document 1).

[0155] In the prior art, as shown in Figure 18b, which is similar to Figure 6b described above, the automated machines 3C and 3D stop at a position some distance forward (a distance that will not cause a collision). This is a problem specific to situations where only the main goal achievement evaluation unit 43 and the inter-agent distance evaluation unit 41 are effective.

[0156] When the present invention is used, the situation initially becomes as shown in Figure 18b, and the vehicles begin to decelerate to avoid contact. After that, when the penalty by the first realization means of the main goal achievement assistance unit 42 increases compared to the evaluation index (equation (11) with r replaced with q) that attempts to minimize the deviation e in the main goal achievement evaluation unit 43, a route is generated that will move even if it increases the deviation e. In other words, a route that avoids contact by utilizing intersecting passages is generated, as shown in Figure 18c.

[0157] As described above, by using this invention, it is possible to automatically generate a path that will not cause a deadlock, even in situations where a deadlock would occur using conventional technology. This eliminates the need for remote operation by a service technician when a deadlock occurs, which is expected to reduce operational costs.

[0158] Unlike the automated machine 3A, the worker 3B can move around freely at his own will, and so it is expected that he may not follow the route instructions given to the smart device.

[0159] If the worker 3B does not follow the route instructions, the route planning device 1 will not treat the worker 3B as a controllable agent, but as an obstacle that must not be contacted.

[0160] As soon as it is confirmed that the route instruction is not followed, the route planning device 1 executes an interrupt process (time t4 in FIG. 12) to carry out a route plan for the agent excluding the worker 3B.

[0161] If the worker 3B again transmits from the smart device his / her intention to follow the route instruction, the route planning device 1 similarly executes route planning for all agents including the worker 3B by interrupt processing. [Example]

[0162] In the fifth embodiment, a case where the present invention is applied to a route plan in a parking lot will be described.

[0163] 19 is a simplified diagram of a parking lot 200 in which non-automated vehicles 3E and automated vehicles 3F coexist. For ease of understanding, the diagram shows only one non-automated vehicle 3E and one automated vehicle 3F, but it is also possible for there to be multiple non-automated vehicles and multiple automated vehicles, or for there to be no non-automated vehicles at all.

[0164] Non-automated vehicle 3E is equipped with a car navigation system, so GNSS ( G lobal N Navigation S atellite SAlthough some non-automated vehicles have a self-location acquisition function and a communication function using a vehicle navigation system (e.g., a navigation system), this information may not be able to be linked with the route planning system of the present invention. As a result, in a parking lot environment, the non-automated vehicle 3E does not have the functions that the agent 3 should have. However, it is possible to substitute for the functions that the non-automated vehicle 3E should have using the configuration described below.

[0165] The automated vehicle 3F is capable of acquiring its own position using GNSS and LiDAR. The automated vehicle 3F is also equipped with a controller (not shown) that executes various calculations related to automated driving. The controller is also equipped with a communication function, and is capable of communicating with a control server 104 that manages the operation of the parking lot, and traveling within the parking lot according to a route plan transmitted by the control server 104.

[0166] Surveillance cameras 101 are installed at various locations in the parking lot 200 to monitor the positions of vehicles and available spaces within the parking lot. The surveillance cameras 101 correspond to the monitoring device 2 of the present invention. Therefore, the surveillance cameras 101 function as an alternative to the self-position acquisition unit 22 of the non-automated vehicle 3E.

[0167] Digital signage 105 is installed in various locations in the parking lot 200, and can display route guidance to vacant areas. By displaying the route plan distributed by the control server 104 on the digital signage 105, it is possible to guide the non-automated vehicle 3E. Note that, as with the example of the logistics warehouse described above, the control server 104 does not necessarily have to be installed within the parking lot 200.

[0168] In the parking lot 200, the task assigned to each agent is to move to an empty space. Therefore, the task management unit 11 has a function of assigning, from among the empty spaces acquired by the surveillance camera 101, either the empty space closest to the current location of each agent or the empty space closest to the facility entrance as the destination point ri (i=1, ..., N).

[0169] The route planning device 1 calculates a route plan for moving each agent from its current position p1, p2 to the destination point r1, r2 calculated by the operation management unit 11.

[0170] Figure 20a shows an example of a path plan that can occur when applying the conventional technology (Patent Document 1) to a parking lot. Similar to the situations shown in Figures 5b and 5d, each agent will deadlock at the shortest distance (p1(6) and p2(2)) to the goal point ri, which is sandwiched between obstacles (walls or other parked vehicles).

[0171] Figure 20b shows an example of a route plan generated when the present invention is applied to the same situation. When a deadlock occurs in Figure 20a, the penalty value calculated by the primary goal achievement assistance unit 42 increases, so a control input is calculated that moves away from the shortest distance positions (p1(6) and p2(2)). This makes it possible to generate a route that moves away from the shortest distance positions (p1(6) and p2(2)) and then gradually approaches the target points r1 and r2.

[0172] Similar to the example of the logistics warehouse, the non-automated vehicle 3E, unlike the automated vehicle 3F, can move around freely at the driver's discretion, and is therefore expected to not follow the route instructions given to the digital signage 105.

[0173] If the non-automated vehicle 3E does not follow the route instructions, the route planning device 1 will not treat the non-automated vehicle 3E as a controllable agent, but as an obstacle that must not be contacted. As soon as it is confirmed that the non-automated vehicle 3E does not follow the route instructions, the route planning device 1 executes a route plan for the agent excluding the non-automated vehicle 3E by interrupt processing (time t4 in FIG. 12).

[0174] Although the embodiments of the present invention have been described in detail above using a logistics warehouse and a parking lot as examples, it goes without saying that the application of the present invention is not limited to these cases. For example, the present invention can also be used to generate routes for transport vehicles in ports, or routes for robots moving within theme parks. [Explanation of symbols]

[0175] 1: Path planning device 2: Monitoring device 3: Agent 10: Path planning system 11: Business management department 12: Map information management department 13: Agent information management section 14: Route planning section 15: Action Plan Communication Department 21: Communications Department 22: Agent position acquisition unit 31: Communications Department 32: Self-location acquisition part 33:Other position acquisition part 34: Path following unit 41: Agent distance evaluation unit 42:Main objective achievement support section 43: Main Objective Achievement Evaluation Section 44: Path planning calculation unit

Claims

1. an operation management unit that determines the destinations of a plurality of agents; a map information management unit that manages map information of the area where the agent is located; an agent information management unit that manages individual information of the agent; a route plan calculation unit that generates a route plan for moving each agent to the destination determined by the operation management unit based on information from the map information management unit and the agent information management unit; a main goal achievement evaluation unit that evaluates each agent higher as the position of the agent approaches the target position; a main goal achievement assistance unit that gives a low evaluation to each of the agents when the agent continues to be at a predetermined distance or more from its target position, and that gives a low evaluation to each of the agents as the moving speed decreases so that an action is generated to resolve the stopped state when the agent is at a predetermined distance or more from its target position; an inter-agent distance evaluation unit that evaluates each of the agents lower as the distance between the agents approaches a predetermined value set for each agent by the agent information management unit, the path plan calculation unit generates the path plan so that, at each time, a comprehensive result of evaluation values ​​calculated by the main goal achievement evaluation unit, the main goal achievement assistance unit, and the inter-agent distance evaluation unit is improved compared to a previous evaluation value; The path planning device further comprises an action plan transmission unit that transmits the path plan calculated by the path plan calculation unit to all agents.

2. The route planning device according to claim 1 , a route planning device characterized in that the primary goal achievement assistance unit assigns a lower evaluation to each agent when the agent's position is a predetermined distance or more from its target position and the lower the moving speed, and / or the lower the evaluation to each agent when the agent is closer to a predetermined position set in the map information management unit.

3. The route planning device according to claim 1 , a route planning device characterized in that the primary goal achievement assistance unit assigns a lower evaluation to each agent when the agent's position is a predetermined distance or more from its target position and the slower the agent's moving speed is, and / or the closer each agent is to a predetermined position set in the map information management unit, and further assigns a lower evaluation to each agent when the deviation between the agent's position and its target value does not decrease within a predetermined time.

4. an operation management unit that determines the destinations of multiple agents with different shapes; a map information management unit that manages map information of the area where the agent is located; an agent information management unit that manages individual information of the agent; a route plan calculation unit that generates a route plan for moving each agent to the destination determined by the operation management unit based on information from the map information management unit and the agent information management unit; a main goal achievement evaluation unit that evaluates each agent higher as the position of the agent approaches the target position; a main goal achievement assistance unit that gives a low evaluation to each of the agents when the agent continues to be at a predetermined distance or more from its target position, and that gives a low evaluation to each of the agents as the moving speed decreases so that an action is generated to resolve the stopped state when the agent is at a predetermined distance or more from its target position; an inter-agent distance evaluation unit that evaluates each of the agents lower as the distance between the agents approaches a predetermined value set in the agent information management unit according to the form of each agent, The route planning calculation unit, at each time, and generating the path plan so that the overall result of the evaluation value calculated by the inter-agent distance evaluation unit is improved over the previous evaluation value; The path planning device further comprises an action plan transmission unit that transmits the path plan calculated by the path plan calculation unit to all agents.

5. an operations management unit that determines the destinations of multiple agents, including people; a map information management unit that manages map information of the area where the agent is located; an agent information management unit that manages individual information of the agent; a route plan calculation unit that generates a route plan for moving each agent to the destination determined by the operation management unit based on information from the map information management unit and the agent information management unit; a main goal achievement evaluation unit that evaluates each agent higher as the position of the agent approaches the target position; a main goal achievement assistance unit that gives a low evaluation to each of the agents when the agent continues to be at a predetermined distance or more from its target position, and that gives a low evaluation to each of the agents as the moving speed decreases so that an action is generated to resolve the stopped state when the agent is at a predetermined distance or more from its target position; an inter-agent distance evaluation unit that evaluates each of the agents lower as the distance between the agents approaches a predetermined value set in the agent information management unit according to the form of each agent, The route planning calculation unit, at each time, and generating the path plan so that the overall result of the evaluation value calculated by the inter-agent distance evaluation unit is improved over the previous evaluation value; The route planning device further comprises an action plan transmission unit that transmits the route plan calculated by the route plan calculation unit to the autonomous body as a control command value and to a person as a recommended route.

6. generating a route plan for moving each agent to its destination based on destinations of the agents, map information of the area where the agents are located, and individual information of the agents; a first evaluation for each agent that is higher the closer the agent's position is to the target position; a second evaluation for each agent that is lower the agent's position continues to be at a predetermined distance or more from the target position so that an action is generated to resolve the stopped state; and a third evaluation for each agent that is lower the agent's position approaches a predetermined value set for each agent; A path planning method, characterized in that, at each time, the path plan is generated so that the overall result of the first evaluation, the second evaluation, and the third evaluation is improved over the previous evaluation value, and the path plan is transmitted to all agents, the method being realized by a computer.

Images