JP2013045265A - Autonomous mobile body - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an autonomous mobile body which searches for a route for reaching a destination while moving along flow that an obstacle moves without approaching the obstacle too much when the autonomous mobile body moves while avoiding the mobile obstacle.SOLUTION: Virtual viscous force and virtual repulsive force are considered to act between an autonomous mobile body and a mobile obstacle, and the virtual viscous force and the virtual repulsive force are included in movement cost which is used in route search technology. In addition, when the movement cost is calculated by tentatively selecting a branch point, a position and speed of the mobile obstacle at a point of time for passing through the branch point are predicted to calculate the virtual viscous force and the virtual repulsive force. Since the movement cost increases when the autonomous body approaches the mobile obstacle too much, and the movement cost moves decreases when the autonomous mobile body moves along flow of the mobile obstacle, a route on which the autonomous mobile body does not approach the mobile obstacle too much, and moves along the flow of the mobile obstacle is searched for.

Description

  In this specification, the technique regarding an autonomous mobile body is disclosed.

Autonomous mobile bodies have been developed that move by searching for a route to a target point when a person designates the target point.
Non-Patent Document 1 discloses a technique for searching for a route along which the autonomous mobile body moves along the crowd flow when the autonomous mobile body moves in the crowd from the current position to the target position. Opportunities to avoid crowded areas and to move in the opposite direction to the crowds because the movement cost is set to be high where the crowd density is high and where the relative speed between the autonomous mobile body and the crowd is large Routes with less are searched.

  In the technique disclosed in Patent Document 1, a viscous resistance coefficient distribution that takes a maximum value at the position of an obstacle and decreases as the distance from the obstacle is virtually set. When a route search is performed with the viscous resistance coefficient distribution set, a route that avoids an obstacle is searched.

Peter Henry and three others, “Learning to Navigate Through Crowded Environments”, 2010 IEEE International Conference on Robotics and Automation, US, 2010, p. 981-986

JP-A-5-297937

  In the technology of Non-Patent Document 1, it is set so that the movement cost is high where the crowd density is high and where the relative speed between the autonomous mobile body and the crowd is high, so avoid crowded places. A route with few opportunities to go in the opposite direction to the crowd is searched. However, it does not take into account that the position of the crowd changes with the passage of time. In the route search technology, branch points are temporarily selected from selectable branch points one after another, and a movement cost is calculated when a route that passes through the temporarily selected branch point is taken. When calculating the movement cost, the movement cost should be calculated based on the environment at the time when the moving object passes through the temporarily selected branch point. The environment at the time when the route search is actually performed is different from the environment when the moving object passes through the temporarily selected branch point. The technology of Non-Patent Document 1 does not consider that the environment changes with the passage of time, and the movement cost in the case of a route that passes through a selected branch point depends on the environment at the time of the route search. I'm calculating. For this reason, while moving along the searched route, it is influenced by an environmental change that is not assumed at the time of searching, and an actually inappropriate route is indicated. In the technique of Non-Patent Document 1, a situation in which the searched route becomes inappropriate while moving along the searched route frequently occurs, and it is often necessary to re-search the route.

Patent Document 1 mainly deals with an obstacle avoidance technique in an environment where a stationary obstacle exists. In the sixth embodiment of Patent Document 1, reference is also made to a moving obstacle. However, although the viscous drag coefficient distribution considered in Patent Document 1 is displayed as viscosity, actually only the distance from the obstacle is displayed. It is a value that depends on, and does not correspond to the relative speed of the autonomous mobile body and the obstacle. Therefore, although the autonomous mobile body searches for a route that is as short as possible to the destination and is separated from the moving obstacle, the autonomous moving body highly evaluates the route that moves along the moving direction of the moving obstacle, and determines the route. The element to search cannot be expected.
Patent Document 1 also mentions a method of setting a velocity distribution space instead of a viscous drag coefficient distribution. However, the velocity distribution space is a distribution space that depends only on the relative velocity of the autonomous moving body and the moving obstacle. It does not depend on the distance between the autonomous mobile body and the moving obstacle. Therefore, even at a position away from the moving obstacle, the route moving along the moving obstacle is highly evaluated. At a position away from the moving obstacle, the route moving along the moving obstacle is highly evaluated although the route distance should be highly evaluated.

  In this specification, when it is necessary to move along a moving obstacle, a route search technique is provided in which a route moving along the moving obstacle is highly evaluated. In addition, in order to perform a route search taking into account that the position of the moving obstacle changes over time, there is a situation in which the searched route becomes inappropriate while moving along the searched route. Disclosed is a technique for searching for a route that is unlikely to occur.

In this specification, the autonomous mobile body which searches the path | route to a target point and moves is disclosed. In particular, in an environment where a person is on a bicycle, a wheelchair, a senior car, etc., or where a person is walking, the autonomous mobile body avoids moving obstacles and travels from the current position to the target position. The present invention relates to a moving object that searches and moves.
As shown schematically in FIG. 1, the autonomous mobile body disclosed in this specification includes a map storage device 4, a target point designation device 6, an autonomous mobile body information acquisition device 8, and a moving obstacle information acquisition device 10. The route search device 12 is provided. The route search device 12 includes a required time calculation device 14, a prediction device 16, a virtual viscous force calculation device 18, a movement cost calculation device 22, and a selection device 24.

The map storage device 4 stores map description data describing the branch point coordinates and whether or not the branch points are connected by a movable route. The map information is stored, for example, in the form of a topological map.
The target point specifying device 6 is operated by the user of the autonomous mobile body, and allows the user to select and specify an arbitrary branch point from among the branch points included in the map description data. .
The autonomous mobile body information acquisition device 8 acquires the position and speed of the autonomous mobile body.
The moving obstacle information acquisition device 10 acquires the position and speed of the moving obstacle.
The route search device 12 performs autonomous route information acquisition by performing route search processing on map description data, such as A star search (which is often described as A ^* search, but is described in katakana in this specification). A route from the current position acquired by the device 8 to the target point specified by the target point specifying device 6 is searched.
The required time calculation device 14 calculates the time required for the autonomous mobile body to move to the branch point temporarily selected by the route search device 12.
The prediction device 16 is a point at which the autonomous mobile body moves to a temporarily selected branch point from the position and speed of the moving obstacle acquired by the moving obstacle information acquisition device 10 and the required time calculated by the required time calculation device 14. Predict the position and speed of moving obstacles.
The virtual viscous force calculation device 18 calculates the virtual viscous force acting on the autonomous moving body from the position and speed of the moving obstacle predicted by the prediction device 16 and the speed when the autonomous moving body moves at the branch point temporarily selected. calculate.
The movement cost calculation device 22 calculates a movement cost including a path distance and a virtual viscous force when the path passes through a temporarily selected branch point.
The selection device 24 selects a branch point having a minimum travel cost calculated by the travel cost calculation device 22 from the temporarily selected branch point group.

  The virtual viscous force is a viscous force that acts between an autonomous mobile body and a moving obstacle when they are placed in a viscous fluid. It is a function that changes depending on two variables, the relative speed of the obstacle. The virtual viscous force increases as the distance between the autonomous moving body and the moving obstacle decreases, and increases as the relative speed between the autonomous moving body and the moving obstacle increases. That is, the virtual viscous force increases as the autonomous moving body and the moving obstacle approach, and the virtual viscous force increases as the relative velocity between the two increases. In addition, the moving obstacle moves while the autonomous moving body is moving toward the destination. The autonomous mobile object of the present invention calculates the time required for the autonomous mobile object to move to the branch point when the route search device 12 temporarily selects one of the branch points and calculates the movement cost. Calculate the position and velocity of moving obstacles in time. Then, the virtual viscous force acting on the autonomous moving body is calculated. Since the selection device 24 selects a route that minimizes the movement cost including the viscous force described above, when the autonomous moving body approaches the moving obstacle, the selection device 24 moves relative to the moving obstacle. Search for a route that does not get too close to the moving obstacle as the speed decreases. Thereby, the autonomous mobile body moves along the flow of the moving obstacle in the vicinity of the moving obstacle without colliding with the moving obstacle.

  If a person is on a moving obstacle, or if the moving obstacle is a pedestrian, the person has a personal space that feels anxious when another person or another object enters, and therefore enters the personal space. It is preferable to search for routes that do not exist.

  In order to search for a route that does not enter the personal space, a virtual repulsive force calculation device that calculates a virtual repulsive force acting on an autonomous moving body from the position of the moving obstacle predicted by the prediction device and the position of the branch point temporarily selected is provided. It is preferable to add. Moreover, it is preferable that the movement cost calculation device calculates a movement cost including a path distance, a virtual viscous force, and a virtual repulsive force when a route passes through a temporarily selected branch point.

As described above, the virtual viscous force increases as the distance between the autonomous moving body and the moving obstacle approaches. If a route with a small virtual viscous force is selected, a route that collides with a moving obstacle is not searched. However, in the vicinity of the moving obstacle, the relationship between the virtual viscous force and the distance for guiding the path that moves along the flow of the moving obstacle does not necessarily coincide with the relation for guiding the path that does not enter the personal space. If a route search is performed by calculating a virtual repulsive force acting between an autonomous mobile body and a moving obstacle, a route that does not enter the personal space can be derived.
For example, when calculating the movement cost using both a virtual repulsive force that rapidly increases as the distance between the autonomous mobile body and the moving obstacle approaches, and a virtual viscous force that gradually increases as the distance approaches, It is possible to search for a route that does not enter the personal space. When an autonomous mobile body moves in a crowd, it becomes possible to move without making the surrounding people feel uneasy.

It is preferable that the virtual repulsive force becomes zero and the repulsive force does not act when the distance exceeds a certain distance. In this case, it is preferable that the distance from the moving obstacle at which the virtual repulsive force becomes zero depends on the moving direction of the moving obstacle. That is, the relationship is preferably long in the traveling direction and short in the backward direction.
A person's personal space is long in the traveling direction and short in the backward direction. By setting up a virtual repulsion field that incorporates this relationship, it is possible to search for a route that does not enter a personal space that a person actually has.

The distance from the moving obstacle where the virtual repulsive force becomes zero is preferably changed depending on the relative speed between the autonomous moving body and the moving obstacle. That is, the relationship is preferably long at high speed and short at low speed.

In this case, when the autonomous moving body and the moving obstacle move relative to each other at high speed, the route search is executed with a higher evaluation as the distance to the moving obstacle is longer from the stage where the distance to the moving obstacle is longer. . When the autonomous mobile body and the moving obstacle relatively move at a low speed, the route search is executed by evaluating a route having a short route distance until it is very close to the moving obstacle. As for the route searched for in this way, a short route is searched within a range that is familiar to human senses and does not cause anxiety.

According to the present invention, when approaching a moving obstacle, a route that moves along the flow of the moving obstacle is searched. In addition, when calculating the travel cost at each branch point, the travel cost is calculated based on the environment when passing through the branch point, so it becomes inappropriate while actually traveling. It is possible to search for a route that is less likely to be searched for and that can continue to actually travel. Appearance of an event that requires a re-search because the searched route becomes inappropriate can be suppressed.
Further, if the virtual viscous force and the virtual repulsive force are used in combination, an autonomous mobile body that does not enter a personal space that a person has can be realized.

The structure of the autonomous mobile body of this invention is represented. It represents re (the distance at which repulsive force becomes zero) that changes depending on the movement direction of the person. It represents re (the distance at which repulsive force becomes zero) that changes depending on the speed of the person. The block diagram of the autonomous mobile body 2 of an Example is represented. The meaning of topological map data is schematically shown. The flowchart of A star search processing procedure is represented. The route determination process by A star algorithm in case a moving body exists in the branch point P5 is shown typically. The route determination process by A star algorithm in case a mobile body exists in the branch point P8 is shown typically.

The features of the embodiment described below will be summarized.
Feature 1: The autonomous mobile body of the embodiment calculates the position and speed of a moving obstacle at an arbitrary time based on the position and speed of the moving obstacle at the current time. Thereby, an optimum route in an environment where the position of the moving obstacle changes is searched in advance.
Feature 2: The moving cost is calculated by setting a repulsive force field that depends on the distance between the moving obstacle and the autonomous moving body, and a viscous force field that depends on the distance between them and the relative speed between them.
Feature 3: In addition to the distance cost, the cost calculated from the virtual field described in Feature 2 is also included in the calculation formula for the movement cost used in the route search process. As a result, the autonomous mobile body does not get too close to a moving obstacle such as a person and moves along the flow of the person.
Feature 4: The distance from the moving obstacle at which the virtual viscous force becomes zero varies depending on the moving direction of the moving obstacle, and is long in the traveling direction and short in the backward direction.
Feature 5: The distance from the moving obstacle where the virtual viscous force becomes zero varies depending on the relative speed between the autonomous moving body and the moving obstacle, and is long at high speed and short at low speed.

FIG. 4 shows a block diagram of the autonomous mobile body 2 of the embodiment. The map storage device 4 stores map data describing the coordinates of the branch points and whether or not the branch points are connected by a travelable route. For example, in the case illustrated in FIG. 5, the coordinates of the branch points P1... And the data indicating that the branch points P1 and P2 are connected are stored. Since data indicating that the branch points P1 and P3 are connected is not stored, this indicates that there is no route directly connecting the branch points P1 and P3. The map storage device 4 stores map information in the above-described topological format, and the amount of data is compressed.
When moving in the plaza Q, it is necessary to search and determine a route moving in the plaza Q. Therefore, branch points P5, etc. are also set in the plaza. The branch points P5... In the square are uniformly distributed in the square Q. In the square Q, it is possible to move directly between any branch points. However, for route search, data indicating that the branch points that can be moved from the branch point are eight branch points arranged adjacent to each other are stored. Since map data according to the conditions is stored, when a route moving in the plaza Q is searched, a route moving in the plaza Q is searched while following an adjacent branch point. In the embodiment described below, since the route search process is performed on the map data when the autonomous mobile body is at the branch point P1, the route that follows the branch points P1, P2, P5 is searched, and the branch point P5 in the square Q is searched. A case will be described in which the route search process is performed again on the map data when moving to, and the route from the starting point P5 in the square Q to the target point G is searched by moving in the square Q.

  The target point designating device 6 is a device operated by the user of the autonomous mobile body 2 and allows an operation of designating a branch point selected by the user from among the branch point groups stored in the map storage device 4. . The target point selected by the user is input to the autonomous mobile body 2. There can be various forms of processes until the user designates the target point. For example, a database that associates facility names with branch points may be prepared, and the user may specify the facility name. The topological map stored in the map storage device 4 may be displayed on the screen and designated by the user. A method such as a link may be used. The autonomous mobile body 2 may be one on which the user boardes, or may move unattended toward a target point designated by the user.

  The encoder 32 calculates the movement amount of the autonomous mobile body 2. The gyro 34 calculates the moving direction of the autonomous mobile body 2. Information on the amount of movement and the direction of movement is sent to the autonomous mobile body information acquisition device 8. The autonomous mobile body information acquisition device 8 identifies the current position and current direction of the autonomous mobile body 2 by integrating the data of the movement amount and the movement direction. Moreover, the speed of the autonomous mobile body 2 is specified from the information of the movement amount and the elapsed time. The autonomous mobile body information acquisition device 8 may specify the position and speed of the autonomous mobile body 2 using, for example, a GPS function.

The LRF 30 detects a moving obstacle existing around the autonomous mobile body 2. Most of the moving obstacles detected here are pedestrians or people who travel on bicycles, wheelchairs or senior cars. The LRF 30 emits laser light, measures the distance until the object is reflected and bounces back, measures the distance to the object, and measures the direction in which the reflected light bounces back, thereby measuring the direction of the object relative to the LRF 30 To decide. When the autonomous mobile body 2 enters the square, it is possible to detect the presence of a moving obstacle over a wide range. When the LRF 30 detects the distance from the autonomous mobile body 2 to the moving obstacle and the direction of the moving obstacle with respect to the autonomous mobile body 2, the current position coordinates of the autonomous mobile body 2 and the current orientation of the autonomous mobile body 2 are taken into account. Thus, the coordinates of the moving obstacle can be specified. The coordinates of the identified moving obstacle are stored in the moving obstacle information acquisition apparatus 10. By specifying the coordinates of the moving obstacle every predetermined time, the moving speed and moving direction of the moving obstacle can be specified.
The moving obstacle information acquisition apparatus 10 acquires the position, speed, and moving direction of the moving obstacle by this method. In addition to the method using the LRF 30, the position, speed, and direction of the moving obstacle may be obtained from an infrastructure such as a security camera grounded in the plaza Q, or a camera mounted on the autonomous mobile body 2 may be used. It is also possible to use information to be obtained.

  The speed of the autonomous mobile body 2 of the present embodiment is kept constant and continues to move at an unreasonable speed. Instead of this, an algorithm for controlling the speed using information such as whether or not the route is a straight line and the distance to obstacles existing in the surroundings may be introduced.

The current position and current speed of the autonomous mobile body 2 acquired by the autonomous mobile body information acquisition device 8 and the position, speed, and direction of the moving obstacle acquired by the moving obstacle information acquisition device 10 are sent to the route search device 12. .
The route search device 12 includes a device 13 that selects a branch point. For example, the case where the route to the branch point P8 shown in FIG. In this case, there are eight branch points that can move from the branch point P8. The device 13 for temporarily selecting branch points sequentially selects movable branch points (in this case, eight branch points). Hereinafter, a case where the branch point P7 is selected will be described. In this embodiment, when the autonomous mobile body 2 is at the departure point P5 in the square, a route search process is performed to search for a route moving to the target point G. For example, the timing for temporarily selecting the branch point P7 is not the timing at which the autonomous mobile body 2 arrives at the branch point P8. When the autonomous mobile body 2 is at the starting point P5 in the plaza, the route is searched by calculating the following.

The route search device 12 includes a required time elapse device 14. The required time lapse device 14 should arrive at the bifurcation point selected by the autonomous mobile body 2 when the route search processing is performed (in this example, when the autonomous mobile body 2 is at the departure point P5 in the plaza). Calculate the time of.
The route search device 12 includes a prediction device 16. The prediction device 16 predicts the position and speed of the moving obstacle at the time when the autonomous mobile body 2 should arrive at the branch point selected temporarily. In the illustrated case, the position and speed of the moving obstacle are specified and stored in the moving obstacle information acquisition device 10 when the autonomous mobile body 2 is at the starting point P5 in the square. The time required for the autonomous mobile body 2 to move to the temporarily selected branch point from that time is specified by the required time elapse device 14. Assuming that the moving obstacle maintains the moving direction and speed, the predicting device 16 can predict the position and speed of the moving obstacle at the time when the autonomous mobile body 2 should arrive at the branch point temporarily selected. .

  Information on the position and speed of the moving obstacle predicted by the prediction device 12 is sent to the virtual viscous force calculation device 18 and the virtual repulsion force calculation device 20. The virtual viscous force calculation device 18 calculates the virtual viscous force using the SPH particle method, and the virtual repulsive force calculation device 20 calculates the virtual repulsive force using the SPH particle method.

The SPH particle method is described in detail in “Joi Sakai,“ Basics and Applications of SPH Particle Method ””. In the SPH method, fluid motion is simulated by approximating a fluid to a large amount of particles. The physical quantity φ (r) at an arbitrary position r is represented by the sum of the physical quantities φ _j of the particles j existing around.

In this example, a single person (pedestrian, passenger in a bicycle / wheelchair / senior car, etc.) was compared with particles in order to model a moving obstacle. The physical quantity is an object at the position r _j on position r phi was set to be a function of the distance r-r _j between positions r _j of the position r and the object. That is, modeling was performed using Equation 1 below.

The virtual viscous force F _vis and virtual repulsive force F _rep derived from Equation 1 are as follows.

∇W _vis (r−r _j ) and ∇W _rep (r−r _j ) are functions of the distance between the autonomous mobile body 2 and the moving obstacle.
The function ∇W _vis (r−r _j ) involved in the virtual viscous force F _vis is zero at a distance of re = r−r _j and increases as the distance becomes shorter.
The function ∇W _rep (r−r _j ) that determines the virtual repulsive force F _rep is also zero at the distance of re = r−r _j and increases as the distance becomes shorter. The function ∇ W _rep (r−r _j ) involved in the virtual repulsive force F _rep increases more sensitively to the decrease in distance than the function ∇ W _vis (r−r _j ) involved in the virtual viscous force F _vis. .
From Equation 2, the virtual viscous force F _vis is expressed by the function ∇ W _vis (r−r _j ) determined by the distance r _e = r−r _j from the autonomous moving body 2 to the moving obstacle, and the autonomous moving body 2 and the movement obstacle. It can be seen that this is the product of the relative velocity v−v _j with the object.
The virtual viscous force F _vis and the virtual repulsive force F _rep have the following relationship with respect to the distance between the autonomous mobile body 2 and the moving obstacle. However, rf <re. The relative velocity v-v _j of a moving obstacle and an autonomous moving body 2 is not intended excessive.
1) When distance> re: The virtual viscous force F _vis and the virtual repulsive force F _rep are zero.
2) When re>distance> rf: virtual viscous force F _vis is dominant.
3) When rf> distance: Virtual repulsive force F _rep is dominant.
If the distance between the autonomous mobile body 2 and the moving obstacle becomes shorter, the virtual viscous force F _vis increases. In the route search technique, a route having a small value is searched. Therefore, when the route search is performed using the virtual viscous force F _vis , it is difficult to search for a route in which the distance between the autonomous mobile body 2 and the moving obstacle becomes short. In the present embodiment, in addition to that, a route approaching the moving obstacle is not searched in order to use the virtual repulsive force F _rep that rapidly increases as the distance to the moving obstacle becomes shorter. When the moving obstacle is a person, the route for entering the personal space that the person has is not searched.
The value of re on the distance r-r _j between the autonomous moving body 2 and the human may be a value that varies with the direction of movement of a human, long in front of the moving direction of the person, to shorten behind direction be able to. As shown in FIG. 2, when re in the person's traveling direction is reF and re in the direction opposite to the person's traveling direction is reB, reF> reB. In this case, when the autonomous mobile body 2 moves in front of the movement direction of the person, a route having a long distance from the person is preferentially searched. When the autonomous mobile body 2 moves behind the movement direction of the person, a route with a short travel distance is preferentially searched. It is possible to secure a personal space that closely matches a personal space that a person has. When re in the human side direction is reL, reF>reL> reB. Alternatively, reL = reB.
Further, the value of re is preferably increased or decreased according to the relative speed between the autonomous mobile body 2 and the moving obstacle. As shown in FIG. 3, when re is 1 when the relative speed is v1 and re is 2 when the relative speed is v2, it is preferable that re1> re2 if v1> v2. It is preferable that the personal space that the person feels when moving relatively at high speed is long, and the personal space that the person feels when moving at a low speed is short.
In the present embodiment, the virtual viscous force F _vis and the virtual repulsive force F _rep are set to zero when the distance between the autonomous mobile body 2 and the moving obstacle is equal to or greater than re. Thereby, the calculation amount of the virtual viscous force F _vis and the virtual repulsive force F _rep is compressed. Instead, a function in which the distance at which the virtual viscous force F _vis becomes zero and the distance at which the virtual repulsive force F _rep becomes zero may be used. Alternatively, the virtual viscous force F _vis and the virtual repulsive force F _rep may be calculated using a function that decreases as the distance increases but does not become zero.

  In the present embodiment, the route in the plaza Q is searched when the autonomous mobile body 2 reaches the starting point P5 in the plaza Q in FIG. FIG. 7 illustrates a point in time when the autonomous mobile body 2 reaches the starting point P5. 70 in the figure is the position of the moving obstacle detected at that time, and an arrow 72 indicates the speed and direction. These are stored in the moving obstacle information storage device 10. There may be multiple moving obstacles. FIG. 7 shows that there are several moving obstacles. The numbers 70 and 72 are given only one of them.

The A star algorithm is used for route search. In the A star algorithm, the movement cost is calculated using the movement cost calculation device 22. For the calculation of the movement cost, the map description data stored in the map storage device 4, the target point information specified by the target point specifying device 6, and the autonomous mobile body 2 acquired by the autonomous mobile body information acquisition device 8 are used. The position and speed, the virtual viscous force acquired by the virtual viscous force calculator 18 and the virtual repulsive force acquired by the virtual repulsive force calculator 20 are used. The virtual viscous force calculation device 18 calculates the virtual viscous force based on the position and speed of the autonomous mobile body 2, the position and speed of the moving obstacle 70, and Equation 2. The virtual repulsive force calculation device 20 calculates a virtual repulsive force based on the position of the autonomous mobile body 2, the position of the moving obstacle 70, and Equation 3. The processing procedure of the A star algorithm used in this embodiment is shown in FIG. Specifically, it is as follows.
1. The current branch point P5 is set as the starting branch point.
2. All branch points Sn that can be directly moved from the current branch point are acquired (step S2). In the case of FIG. 7, n = 6-10.
3. One of the branch points Sk is temporarily selected (step S4). k is selected from 6 to 10.
4). The estimated value f * (Sk) of the movement cost when passing through the branch point Sk is obtained by the formula f * (Sk) = g (Sk) + h * (Sk) (step S6). g (Sk) is a cost required to move from the starting point P5 to the temporarily selected branch point Sk, and h * (Sk) is a moving cost required to move from the temporarily selected branch point Sk to the target point G. Is an estimated value. For h * (Sk), a value proportional to the distance when there is a straight path from the selected branch point Sk to the target point G can be adopted. g (Sk) is a value proportional to the distance from the starting point P5 to the tentatively selected branch point Sk, and the position (in P5) and speed of the autonomous moving body 2 when the autonomous moving body 2 is at the starting point P5. And the position and speed of the moving obstacle 70, a value proportional to the virtual viscous force calculated based on Equation 2, and the position of the autonomous mobile body 2 when the autonomous mobile body 2 is at the starting point P5 (in P5) And the position of the moving obstacle 70 and a value proportional to the virtual repulsion calculated based on Equation 3 are added.
5. It is determined whether or not all branch points Sn (n = 6 to 10) that can be directly moved from the current branch point P5 are selected and the movement cost is calculated (step S8). If there is a branch point for which the movement cost has not been calculated (No in step S8), the branch point for which the movement cost has not been calculated is temporarily selected (step S10), and the process returns to step S6 to calculate the movement cost. .
6). If it is determined in step S8 that the travel cost has been calculated through all selectable branch points Sn (n = 6 to 10) (Yes in step S8), the calculated value of the travel cost is minimized. Is selected (step S12). In the case of FIG. 7, the case where it is calculated that it is the minimum movement cost that passes through the branch point of P8 is illustrated.
7). Include virtual viscous forces in moving cost calculations. The virtual viscous force has a large value for the path along which the autonomous mobile body 2 moves in the vicinity of the moving obstacle in a state where the speed difference from the moving obstacle is large, and thus operates so that such a path is not selected. . That is, when passing through the vicinity of the moving obstacle, it operates so that the route on which the autonomous mobile body 2 moves is preferentially selected in a state where the speed difference from the moving obstacle is small. It operates so that the route along which the autonomous mobile body 2 moves along the moving obstacle is preferentially selected.
8). Include virtual repulsion in movement cost calculations. Since the virtual repulsive force has a large value for a route approaching a moving obstacle, the virtual repulsive force operates so that such a route is not selected. That is, it operates so that a route passing through a position away from the moving obstacle is preferentially selected. When a moving obstacle is a person or when a person is on the autonomous mobile body 2, a route that does not enter a personal space in which a person feels anxiety when another person or another object invades is preferentially selected. Operates as follows.
9. In step S14, it is determined whether or not the branch point selected in step S12 is the target point G. If the selected branch point is the target point G (Yes in step S14), the route to the target point G has been searched and the process is terminated.
10. If the selected branch point is not the target point G (No in step S14), the current branch point in step S2 is replaced with the branch point determined in step S12, and the processes in and after step S2 are repeated.

  FIG. 8 illustrates the contents of processing executed when the current position is updated to the branch point P8. The timing for updating the current position to the branch point P8 and continuing the route search is performed when the mobile body 2 is actually at the branch point P5. When the mobile body 2 is actually at the branch point P5, the route search process to the target point G is completed. Since the calculation is performed when the mobile body 2 does not actually reach the branch point P8, the position and speed of the moving obstacle 70 when the mobile body 2 reaches the branch point P8 cannot be directly detected.

In this embodiment, the time required for the autonomous mobile body 2 to reach the branch point P8 is calculated by the required time calculation device 14. Then, the position and speed of the moving obstacle 70 at that time are estimated. Since the position and speed of the moving obstacle 70 at the time when the autonomous mobile body 2 is at the branch point P5 are detected, assuming that the moving obstacle 70 continues to move at a constant speed, the autonomous mobile body 2 The position and speed of the moving obstacle 70 when reaching P8 can be estimated.
When step S6 is executed by replacing the current branch point in step S2 with the branch point determined in the calculation (P8 in this case) from the actual branch point P5, the estimation is performed as described above. The virtual viscous force and the virtual repulsive force are calculated using the position and speed of the moving obstacle 70, and the moving cost is calculated.

In this embodiment, g (Sk) used for calculating the movement cost is the following equation.
g (Sk) = L [S0, Sk] + W _vis F _vis + W _rep F _rep
Here, L [S0, Sk] is the distance from the branch point S0 determined immediately before to the temporarily selected branch point Sk, F _vis is the virtual viscous force according to the above-described equation 2, F _rep is the virtual repulsive force of the equation 3, and W _vis is a weight parameter of F _vis , and W _rep is a weight parameter of F _rep . W _vis and W _rep are parameters identified by a learning algorithm such as Maximum Entropy Inverse Reinforcement Learning. Since details are described in Non-Patent Document 1, description in this specification is omitted.

  The route search device 12 searches for a route from the departure point S0 to the target point G while at the departure point S0 (for example, the departure point P5 in the plaza), and moves along the route. In the route search, not only the distance of the route but also the virtual viscous force is taken into account, so that the route along the flow of the moving obstacle is preferentially searched. In addition, since virtual repulsion is also taken into account, a route that does not enter a person's personal space is preferentially searched. In addition, when calculating the virtual viscous force and the virtual repulsive force, it takes time for the autonomous mobile body 2 to travel the route, and the position of the moving obstacle changes during that time. The virtual viscous force and virtual repulsive force are calculated according to the position and speed of the moving obstacle. The calculation is based on actual calculation, and the probability that an event in which the calculated path is actually an inappropriate path appears is low. There is no need to re-search for routes frequently. The autonomous mobile body 2 moves in the direction along the flow of the person without entering the person's personal space. Furthermore, when changing the traveling direction, the traveling direction is changed not behind the person but behind the person. The distance with the person can be appropriately maintained according to the movement speed of the person.

  As shown in FIG. 4, the autonomous mobile body 2 includes a local avoidance route search device 26. If the autonomous mobile body 2 detects the surroundings while moving along the searched route and detects an obstacle that needs to be avoided, the autonomous mobile body 2 generates and avoids the avoidance route by the local avoidance route search device 26. As a result of the avoidance, when it becomes impossible to return to the searched route, a re-search process is executed.

Specific examples of the present invention have been described in detail above, but these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.
The technical elements described in this specification or the drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology illustrated in the present specification or the drawings achieves a plurality of objects at the same time, and has technical utility by achieving one of the objects.

2: Autonomous mobile body 4: Map storage device 6: Target point designation device 8: Autonomous mobile body information acquisition device 10: Moving obstacle information acquisition device 12: Route search device 13: Temporary selection device 14: Necessary time Calculation device 16: Prediction device 18: Virtual viscous force calculation device 20: Virtual repulsive force calculation device 22: Movement cost calculation device 24: Selection device 26: Local avoidance route search device 28: Route 30: LRF
32: Encoder 34: Gyro 70: Moving obstacle 72: Vectors P1, P2 indicating speed and direction of the moving obstacle: Branch point
reF: re in the direction of human movement
reL: re in the direction opposite to the direction of human movement
reB: re in the human side
re1: re when the relative speed is v1
re2: re when the relative speed is v2

Claims (4)

It is an autonomous mobile that moves by searching for the route to the target point.
A map storage device storing map description data describing branch point coordinates and whether or not the branch point and the branch point are connected by a movable route;
A target point specifying device that allows a person to select and specify an arbitrary branch point from among branch points included in the map description data;
An autonomous mobile body information acquisition device for acquiring the position and speed of the autonomous mobile body;
A moving obstacle information acquisition device for acquiring the position and speed of the moving obstacle;
A route search device that performs route search processing on map description data and searches for a route from the current position acquired by the autonomous mobile body information acquisition device to the target point specified by the target point specification device,
The route search device has the following devices:
A required time calculation device that calculates the time required for the autonomous mobile body to move to the temporarily selected branch point;
The position and speed of the moving obstacle at the time when the autonomous mobile body moves to the branch point temporarily selected from the position and speed of the moving obstacle acquired by the moving obstacle information acquisition apparatus and the required time calculated by the required time calculation apparatus. A prediction device for predicting
A virtual viscous force calculation device that calculates a virtual viscous force acting on the autonomous mobile body from the position and speed of the moving obstacle predicted by the prediction device and the speed when the autonomous mobile body moves at the branch point temporarily selected;
A movement cost calculation device for calculating a movement cost including a path distance and a virtual viscous force when a path is routed through a selected branch point;
An autonomous mobile body comprising a selection device that selects a branch point that minimizes the travel cost calculated by the travel cost calculation device from a temporarily selected branch point group. A virtual repulsive force calculation device for calculating a virtual repulsive force acting on the autonomous moving body is added from the position of the moving obstacle predicted by the prediction device and the position of the branch point temporarily selected.
The autonomous moving body according to claim 1, wherein the movement cost calculation device calculates a movement cost including a path distance, a virtual viscous force, and a virtual repulsive force when a route passes through a temporarily selected branch point.   The autonomous mobile body according to claim 2, wherein the distance from the moving obstacle at which the virtual repulsive force becomes zero varies depending on the moving direction of the moving obstacle, and is long in the traveling direction and short in the backward direction. .   4. The distance from the moving obstacle at which the virtual repulsive force becomes zero varies depending on the relative speed between the autonomous moving body and the moving obstacle, and is long at high speed and short at low speed. Autonomous mobile body.

Images