JP7490193B2 - ROBOT, MOVEMENT ROUTE GENERATION DEVICE AND ITS PROGRAM, AND MOVEMENT PREDICTION DEVICE - Google Patents

Description

The present invention relates to a robot that can move autonomously and appropriately even in a crowded environment where there are many people, objects, and other obstacles, a movement path generation device and program for generating a movement path for the robot, and a movement prediction device.

Recently, robots that can move autonomously in an environment where humans coexist have appeared. Some of these robots move autonomously from a specified starting point to a goal point by sensing the status of obstacles, including humans, and avoiding future interference with the obstacles. Such robots do not simply avoid humans unilaterally, but are required to take into account human movements and other obstacles and to act cooperatively with humans to effectively avoid interference. In other words, when moving in such an environment, in addition to passive actions in response to human actions, such as interference avoidance actions and stopping actions, it is necessary for the robot to convey its intentions to humans through these actions and actively encourage the humans to take their own interference avoidance actions. For example, in a case where many people are crowded on the robot's movement path through a gap that is too large for the robot to pass through, it becomes necessary to encourage the people around the gap to move while moving the robot toward the gap in order to secure a passage for the robot.

As a result of their own research, the inventors have proposed various robots that plan routes based on the above-mentioned interactions and move around (see Patent Documents 1 and 2, etc.). In particular, Patent Document 2 discloses a cooperative movement method in which, when interference with humans is predicted, the robot not only moves past humans while keeping a certain amount of space between them, but also moves close to or in contact with humans depending on the situation.

Patent Document 3 discloses an autonomous mobile device that moves to a destination in a shorter distance while reducing the sense of anxiety it gives to humans when passing by them. This autonomous mobile device includes a movement path calculation unit that calculates multiple movement paths according to the surrounding environment and ranks each movement path in terms of operation efficiency corresponding to its length, a movement candidate selection unit that selects movement candidates according to the order of operation efficiency, a repulsion determination unit that determines whether the movement candidates selected by the movement candidate selection unit give a sense of anxiety to people in the vicinity, and a guidance operation instruction unit that sets the movement candidate with the highest operation efficiency ranking that does not give a sense of anxiety to humans as the movement path. The repulsion determination unit calculates a virtual repulsive force that the autonomous mobile device gives to humans, and determines whether or not there is a sense of anxiety based on a preset repulsive force threshold value, assuming that the greater the repulsive force, the more anxiety the human feels. A known SFM (Social Force Model) is used to calculate the repulsive force here. This SFM is a model that introduces virtual forces into an environment where there are multiple mobile agents, such as pedestrians, and predicts the movement of the agents by combining the gravitational forces from the destination to the agents and the repulsive forces from stationary obstacles such as walls and other agents.

JP 2019-84641 A JP 2020-46759 A Japanese Patent No. 5539596

When a robot moves autonomously in a crowded environment where many people are present, it will be in close proximity to many people, so a robot movement plan that takes into account the actions of the person who moves with the robot and the other person who moves with the human is required. In each of the above patent documents, the propagation of movement to multiple surrounding people that will affect the robot when it moves along a predetermined route in the crowded environment is not taken into consideration. When considering the propagation of movement, it is necessary to predict the movement of each person, and there is a method using the above-mentioned SFM for this movement prediction. However, in the conventional SFM, the cognitive state of an agent that can move within the model is not taken into consideration. For example, the repulsive force from other agents approaching the target agent from behind also affects the prediction of the target agent's movement, but in the real space, it is unnatural for a human to recognize and avoid a human behind, and so on, and therefore it is not possible to more accurately predict the movement of the agent. In addition, in the conventional SFM, in a case where a robot encourages the movement of a human while making contact with the human, as disclosed in Patent Document 2, the contact force is not taken into consideration, so it is not possible to simulate the movement of an agent that reflects the influence of the contact force.

The present invention was devised to solve these problems, and its purpose is to provide a robot, a movement path generation device and program thereof, and a movement prediction device that can construct a movement prediction model that includes surrounding people and the like and is suitable for generating a path for a robot that moves autonomously in a crowded environment, and that can generate an optimal movement path while taking into account the movement efficiency of the robot and the burden on surrounding people and the like.

In order to achieve the above object, the present invention is mainly directed to a robot that moves autonomously while avoiding obstacles in the vicinity, the robot comprising a detection device that detects position information and speed information of the obstacles, and a control device that controls the autonomous movement from the detection results of the detection device while taking into account the status of the obstacles, the control device comprising a candidate path search means that generates multiple candidate paths as candidates for a path along which a target agent, which is a target agent consisting of the robot and a movable mobile obstacle among the obstacles, moves, based on the position information and speed information of the obstacles, an optimal path extraction means that extracts an optimal path from each of the candidate paths, and an operation command means that issues an operation command for the autonomous movement based on the optimal path, the optimal path extraction means generating a motion command for each of the agents based on the interaction between the robot and the obstacles, the motion command command for each of the agents ... The system is equipped with a movement prediction simulation unit that simulates the movement state, and a final decision unit that decides the optimal route based on the movement efficiency of the target agent and the impact on the moving obstacle from the simulation results of the movement prediction simulation unit, the movement prediction simulation unit predicts changes in the position and speed of each agent over time using virtual forces that interact between each agent and between each agent and an object that is the obstacle and is always stationary, and the final decision unit calculates a first index value related to the movement efficiency and a second index value related to the load on the moving obstacle for each of the candidate routes using a previously stored formula, and the candidate route with the lowest route cost obtained by combining these index values is selected as the optimal route.

According to the present invention, it is possible to generate an optimal movement route that more accurately considers the interaction between the robot and the surrounding people by comprehensively judging the movement efficiency of the robot and the movement load consisting of the physical and psychological burden on the surrounding people based on a movement prediction model that includes surrounding people, etc., suitable for generating a route for a robot that moves autonomously in a crowded environment. In other words, it is possible to select an optimal route from multiple candidate routes while balancing the movement efficiency of the robot and the movement load of the people who take evasive action due to the movement of the robot. For example, even if the movement load of the surrounding people, etc. is somewhat high, if the movement efficiency of the robot is significantly good, the candidate route at that time will be selected as the optimal route, and conversely, even if the movement efficiency of the robot is somewhat high, if the movement load of the people, etc. is significantly low, the candidate route at that time will be selected as the optimal route.

In addition, the movement prediction model of the present invention takes into account the contact force when a robot moves while in contact with a human, and can change the state of repulsive force taking into account the recognition range of humans and others around the robot, making it possible to more appropriately predict the movements of each agent, such as a robot or a human, when a robot moves in a crowded environment while in contact with surrounding humans.

FIG. 2 is a block diagram that shows a schematic configuration related to the movement of the robot 10 according to the present embodiment. FIG. 4 is a conceptual diagram for explaining generation of candidate routes by a candidate route searching means. 13A to 13F are conceptual diagrams for explaining the movement prediction simulation of each agent in a time series in the movement prediction simulation unit.

The following describes an embodiment of the present invention with reference to the drawings.

Figure 1 shows a block diagram that shows only the configuration related to the movement of a robot 10 according to this embodiment. In this diagram, the robot 10 functions as a moving body that moves autonomously from a preset movement start point (start point) to a specified target point (goal point). When the robot 10 predicts future interference with obstacles such as walls or people during its movement, the robot searches for a movement path that allows efficient movement while reducing the impact on surrounding people and other people, and controls its operation along the movement path.

The obstacles in this case include fixed obstacles that remain stationary in the environment, such as walls that exist around the movement path of the robot 10, and mobile obstacles that can move in the environment, such as humans and moving objects. Note that in the following, multiple humans will be described as moving obstacles, but movement paths can also be generated for other moving objects by using the same process described below.

In addition, although not particularly limited, in this embodiment, the following three types of paths are targeted as movement paths that proceed while avoiding obstacles. That is, the movement paths here include an avoidance path that avoids interference with humans while maintaining a state in which the distance between the robot 10 and the human is equal to or greater than a predetermined value, an approach path that allows the robot 10 and the human to pass each other in an approaching state without touching each other while the distance between them is less than the predetermined value, and a contact path that assumes that when the robot 10 passes a human, it makes contact with the human and takes action to secure a passage for the robot 10.

As shown in FIG. 1, the robot 10 is provided with an operating unit 11 consisting of mechanisms and devices configured to enable various operations, a detection device 12 that detects environmental information around the robot 10, and a control device 13 that performs autonomous movement control of the robot 10 based on the detection results of the detection device 12 while taking into account the status of the obstacles.

The operating unit 11 includes a moving device 14 consisting of a mechanism and its power source for autonomously moving within a specified range, and an arm 15 consisting of a mechanism and its power source for operating within a specified space. The operating unit 11, including the moving device 14 and the arm 15, is composed of all known members, mechanisms, devices, etc., and is not an essential part of the present invention, so detailed illustrations and explanations of each component will be omitted.

The detection device 12 is composed of a surrounding environment detection unit 18 that detects the position and speed information of humans and other obstacles present around the robot 10, and a direction detection unit 19 that detects the direction of the human. The detection results of each detection unit 18, 19 are sequentially transmitted to the control device 13.

The surrounding environment detection unit 18 is not particularly limited, but may be a known distance measurement sensor such as a laser range finder that detects the position of each surface of an obstacle around the robot 10 based on the reflection state of objects, including humans, caused by the irradiation of laser light around the robot 10. In other words, the surrounding environment detection unit 18 measures the two-dimensional position in a planar view of each point in the point cloud that constitutes the surface of various obstacles, using the robot 10 as a reference, and calculates speed information of the moving obstacle based on the measured value.

In addition, other sensors and devices, such as sensors using GPS, can also be used as the surrounding environment detection unit 18, as long as they are capable of detecting the positions of various obstacles present around the robot 10.

The orientation detection unit 19 is not particularly limited, but may be a known depth sensor such as KINECT (registered trademark), and the facial area of the human is identified by skeletal recognition, and the orientation of the facial area of the human relative to the robot 10 is detected using the position information of the human obtained from the detection results of the surrounding environment detection unit 18.

Note that other sensors or devices capable of similar detection may be used as the orientation detection unit 19. Furthermore, the surrounding environment detection unit 18 and the orientation detection unit 19 may be configured as an integrated device or system as long as they are capable of acquiring the various types of information described above.

The control device 13 searches for the optimal movement path so that the robot 10 can move autonomously from a preset start point to a specified goal point while minimizing the impact on obstacles, and issues operation commands to the operation unit 11.

The control device 13 is provided either integrally with the robot 10 or separately, and is composed of a computer including a processor such as a CPU and storage devices such as memory and a hard disk. Programs are installed in the computer to function as the following means.

Next, the specific configuration of the control device 13 will be described.

The control device 13 includes an interference prediction means 21 that uses the detection results from the surrounding environment detection unit 18 to estimate the possibility of humans interfering with the robot 10 in the future, a recognition state estimation means 22 that uses the detection results from the orientation detection unit 19 to estimate the human's recognition state of the robot 10 at the time of detection, a storage means 23 that stores predetermined information according to the estimation results from the interference prediction means 21 and the recognition state estimation means 22, a candidate path search means 24 that generates multiple candidate paths for the robot 10's movement path according to the obstacle situation when the robot 10 moves toward the goal point, an optimal path extraction means 25 that extracts the optimal path from among these candidate paths, and an operation command means 26 that commands the operation unit 11 to operate so that the robot 10 can move autonomously along the optimal path.

As disclosed in JP 2020-46759 A and other publications already proposed by the present inventors, the interference prediction means 21 obtains the current movement status of the human from the detection results of the surrounding environment detection unit 18 and determines whether or not there is a possibility that the robot 10 will interfere with the human in the future. That is, here, personal areas of a predetermined size are set around the robot 10 and the human, and if it is predicted that these personal areas will at least partially overlap in the future, it is determined that there is a possibility of interference, and if not, it is determined that there is no possibility of interference. Specifically, based on the velocity vector of the human from the detection results of the surrounding environment detection unit 18, the distance between the robot 10 and the human's trunk centers is calculated from the position where the robot 10 and the human are side-by-side. Then, when the distance is shorter than the separation distance of each personal area, it is determined that there is a possibility of interference, that is, "degree of interference present," and when not, it is determined that there is no possibility of interference, that is, "no possibility of interference."

The recognition status estimation means 22 estimates the human's recognition status of the robot 10 based on the direction of the human's face detected by the direction detection unit 19 as follows. That is, first, the visual field range in which the surrounding humans and objects can be recognized is specified as the recognizable range. Here, the human's recognizable range is preset as the range of the human's visual field angle (e.g., 200 degrees). The robot 10's recognizable range is set as the range of obstacles detected by the detection device 12. Then, when the relative distance between the robot 10 and the human becomes equal to or less than the preset distance, if the robot 10 is present within the human's recognizable range based on the direction of the human's face, which has eyes that can see the robot 10, it is determined that the human recognizes the robot 10 and that "there is a degree of recognition". On the other hand, if this is not the case, it is determined that the human does not recognize the robot 10 and that "there is no degree of recognition".

When the interference prediction means 21 determines that there is a degree of interference with each human around the robot 10, the storage means 23 stores the position information and speed information of that human and the cognitive state of that human estimated by the cognitive state estimation means 22.

The candidate path search means 24 generates multiple candidate paths for the robot 10 using a method disclosed in JP 2020-46759 A and other publications already proposed by the present inventors. That is, as shown in FIG. 2, the current position of the robot 10 is set as a start point SP, and a straight path (dashed line in the figure) is set that connects the start point SP and the goal point GP with a straight line. Then, if the interference prediction means 21 determines that interference with a human H may occur in the future on the straight path, a way point (Way Point: hereinafter referred to as "WP") of a path that avoids interference with the human H is set on both the left and right sides of the human H. Furthermore, if interference with another human H may occur in the future on another straight path that connects the WP and the goal point GP with a straight line, the next WP is set in the same manner on both the left and right sides of the other human H, and this process is repeated. Finally, a candidate route PT (four routes in the example in Figure 2) is generated that passes through each WP set in this way in order.

The multiple candidate paths generated as described above include an avoidance path in which the robot 10 does not enter the personal area of the human H, an approach path in which the robot 10 enters the personal area of the human H but the torsos of the robot 10 and the human H do not come into contact, and a contact path in which the robot 10 avoids the human while making contact with the human using the arm 15 of the robot 10. In this embodiment, the contact path is generated only when a deadlock state, which will be described later, may occur in the robot 10 based on the simulation results of the movement prediction simulation unit 28, which will be described later, but is not limited to this condition.

The optimal trajectory extraction means 25 includes a movement prediction simulation unit 28 that simulates over time the movements of the robot 10 due to the interactions between the robot 10 and humans present in the vicinity when the robot 10 moves along each of the candidate paths searched for by the candidate path search means 24, and a final decision unit 25 that determines the optimal path from among the candidate paths based on the results of the simulation by the movement prediction simulation unit 28, taking into consideration the movement efficiency of the robot 10 and the impact on the surrounding humans.

The movement prediction simulation unit 28 uses a new movement prediction model that the inventors have expanded on the conventional SFM (Social Force Model), which is a movement prediction model based on an equation of motion, to simulate the movement of the robot 10 and the human over time, taking into account the movement of the robot 10 and the interaction with obstacles around it, until the robot 10 moves to the goal point.

The SFM is a model that predicts the movement of each agent, such as a human or a robot, assuming that each agent moves in the direction of a virtual force vector acting on the agent. In this SFM, a resultant force vector is calculated by combining a virtual gravitational force from the destination of the target agent and a virtual repulsive force from objects in a stationary environment such as fixed obstacles such as walls and other agents, and the target agent is estimated to move according to the resultant force vector. However, when generating an optimal path taking into account the physical and psychological burdens imposed on surrounding people due to the movement of the robot 10, if the movement of people around the robot 10 is predicted simply using the conventional SFM, the optimal movement path cannot necessarily be extracted sufficiently for the following reasons. First, as a first reason, the conventional SFM also takes into account the repulsive force from other agents such as humans approaching the target agent from behind, but in real space, it is unnatural for a human to recognize and avoid a person behind. The second reason is that in conventional SFM, when the robot 10 is moving, a deadlock state occurs in which the repulsive force and the attractive force balance each other, depending on the state of the other agents and the location of the destination. In such a case, a contact path is generated as a candidate path in which the robot 10 uses contact to influence the other agent that receives the most repulsive force, but the contact force at this time is not taken into account in conventional SFM.

Therefore, in the newly developed extended model created by the inventors, the degree of recognition in the recognition state estimation means 22 and the contact force acting on the target human on the contact path determined by the candidate path search means 24 are further taken into consideration, and the movement of each agent, i.e., the robot 10 and the humans present around it, is predicted during the movement process of the robot 10.

The movement prediction simulation here is performed sequentially in time series between the set WPs for each candidate route. In other words, a first simulation is performed in the section from the start point where the first WP is set as the destination of the robot 10, and then a similar simulation is performed in the next section where the destination of the robot 10 is changed from that state to the next WP, and the simulation is performed over time in each section between the WPs up to the goal point of the robot 10.

Specifically, here, when predicting the movement of each agent for each simulation section, a resultant force vector F acting on each target agent to be predicted is calculated. The resultant force vector F is calculated at a predetermined timing (e.g., every 0.5 seconds) during the virtual movement of the robot 10, and each agent is predicted to move according to the corresponding resultant force vector F at each timing. Note that in the simulation, the destination set for each person is a predetermined point that is virtually set along the movement direction calculated based on the detection results by the detection device 12.

The resultant force vector F is the sum of an attractive force vector ^FG from the destination of the target agent, a repulsive force vector F _i ^H from other agents i present around the target agent, a repulsive force vector F _j ^o from objects j that are stationary obstacles such as walls present around the target agent, and a contact force vector F _k ^C that the robot 10 applies to the human on the contact path, and is calculated by the following formula.

In the above formula (2), m is the weight of the target agent, and v ⁰ (t) is the desired velocity vector of the target agent, which are preset to predetermined values depending on the type of agent, such as the robot 10 or a human. Also, v(t) is the current velocity vector of the target agent, which is known in the case of a robot 10, and which is estimated from the detection results by the detection device 12 in the case of a human. Furthermore, τ is a preset time constant equivalent to the timing interval at which the simulation is performed.

ここで、ｋｐ、ｋｓは、斥力の大きさに関する係数であり、事前になされた実験等の結果に基づいて、エージェントの動きが自然になる所定の一定値に設定される。また、ｄ_ｉ、ｄ_ｊは、斥力を発する他のエージェントｉやオブジェクトｊと対象エージェントとの離間距離であり、ｖ_ｉ、ｖ_ｊは、これらの間の単位方向速度ベクトルであり、それぞれ、検出装置１２での検出結果から特定される。更に、ｒは、ロボット１０や人間のすれ違いに必要となる円形の最小領域の半径であり、ｓは、対象エージェントに斥力が及ぼされる範囲に相当するパーソナルエリアの半径である。これらｒ、ｓは、例えば、ｒ＝０．６ｍ、ｓ＝１ｍ等の一定値に予め設定される。更に、θは、従来のＳＦＭにおけるランプ関数である。 In the above equations (3) and (4), fp _i ^H and fp _j ^O are defined as force vectors generated when an agent i collides with another agent j, and fs _i ^H and fs _j ^O are defined as psychological repulsive force vectors that affect the psychological factors of the agent i and the object j, and are calculated using the following equations.

Here, kp and ks are coefficients related to the magnitude of the repulsive force, and are set to predetermined constant values that make the agent's movement natural based on the results of experiments conducted in advance. Furthermore, d _i and d _j are the distances between the target agent and other agents i or objects j that generate repulsive forces, and v _i and v _j are unit direction velocity vectors between them, which are respectively specified from the detection results of the detection device 12. Furthermore, r is the radius of the minimum circular area required for the robot 10 and a human to pass each other, and s is the radius of the personal area corresponding to the range in which the repulsive force is exerted on the target agent. These r and s are set in advance to constant values such as r=0.6 m and s=1 m. Furthermore, θ is a ramp function in conventional SFM.

Furthermore, in calculating the repulsive force vectors F _i ^H and F _j ^O in the above formulas (3) and (4), the degree of recognition by the recognition state estimation means 22 is taken into consideration. That is, as in the above formula (5), in order to prevent the target agent from receiving psychological repulsion from other agents or objects that the target agent is not aware of, when the degree of recognition aw is zero, that is, when there is "no recognition," the psychological repulsive force vectors fs _i ^H and fs _j ^O are set to zero, and the repulsive force vectors F _i H and F _j ^O are calculated. In other words, the psychological repulsive force vectors fs _i ^H and fs _j ^O are taken into consideration ^and act only when there is "recognition."

In the movement prediction simulation unit 28, for example, as shown in FIG. 3, a movement prediction simulation of the robot 10 and surrounding humans A to D is performed. In other words, for each agent of the robot 10 and surrounding humans A to D, a resultant force vector F is calculated at regular intervals according to the respective recognizable ranges R identified by the recognition state estimation means 22. Then, the movement state of each agent is estimated from the resultant force vector F.

First, the WP set in the candidate route is set as the destination P of the robot 10, and then a simulation of the movement prediction of each agent is performed. At the timing of FIG. 3(A), the robot 10 receives a repulsive force (dashed arrow in the figure) from the wall W, and due to this influence, the robot 10 moves slightly inward from the straight line route to the destination P. Then, at the timing of FIG. 3(B) when the robot 10 has further advanced toward the destination P, the movement of the robot 10 is also influenced by a repulsive force (dash-dotted arrow in the figure) from the human B. After that, when the robot 10 advances further as shown in FIG. 3(C), the human A, who is approaching the robot 10, moves to the right in the figure due to the contact force (thick arrow in the figure) from the robot 10, as shown in FIG. 3(D), and the repulsive force (dash-dotted arrow in the figure) is propagated to the humans B and C due to the movement of the human A, and the humans B and C also move as shown in FIG. 3(E). Then, as shown in FIG. 1F, when the robot 10 reaches destination P, the next WP is set as the destination, and a movement prediction simulation is performed in the same manner.

The final decision unit 29 calculates the following two types of index values corresponding to the movement load of the robot 10 and people around it using a pre-stored formula for each candidate route, and selects the most appropriate optimal route from among the candidate routes based on the route cost calculated by combining these index values.

The specific processing performed by the final decision unit 29 is described below.

First, the index values obtained are a first index value relating to the movement efficiency of the robot 10 and a second index value relating to the load imposed on people around the robot 10 based on the movement prediction results from the movement prediction simulation unit 28.

The first index value corresponds to a movement cost E, which is the sum of a first parameter related to the direction of travel of the robot 10 toward the destination in the movement simulation and a second parameter related to the avoidance direction of the robot 10 that is perpendicular to the direction of travel and allows the robot 10 to avoid humans or objects.

In the first parameter, the action of the robot 10 moving straight ahead at a constant speed is set as the lowest cost action, and the acceleration/deceleration of the robot 10 in the direction of travel is set as the cost increase/decrease parameter.

In the second parameter, the greater the avoidance width when the robot 10 avoids interference, the worse the movement efficiency of the robot 10 becomes, so the movement in the avoidance direction perpendicular to the direction of travel of the robot 10 is used as the cost increase/decrease parameter.

上式において、ｖ_ｙｔは、時刻ｔにおけるロボット１０の進行方向の速度であり、ｖ_ｙ０は、同進行方向の基準速度であり、ｖ_ｘｔは、同回避方向の速度であり、ｖ_ｘｔ－１は、１ステップ前の同回避方向の速度である。また、Ａは、前記進行方向の重み係数であり、Ｂは、前記回避方向の重み係数であり、所定値（例えば、Ａ＝１、Ｂ＝１）が予め設定される。 In other words, the movement cost E is calculated by the following formula by synthesizing the changes in the movement state of the robot 10 between each WP simulated by the movement prediction simulation unit 28 from the start point to the goal point.

In the above formula, v _yt is the speed of the robot 10 in the advancing direction at time t, v _y0 is the reference speed in the advancing direction, v _xt is the speed in the avoidance direction, and v _xt-1 is the speed in the avoidance direction one step before. Also, A is a weighting coefficient for the advancing direction, and B is a weighting coefficient for the avoidance direction, which are preset to predetermined values (for example, A=1, B=1).

上式において、ｎはロボット１０が認識した人数に相当する。 The second index value is an influence degree D obtained by combining the sum of the repulsive force vectors F i H and F _j ^o of each agent obtained by the simulation results in the movement prediction simulation unit 28 and the contact force vector F _k ^C when the robot 10 comes into contact with a human, since it is necessary to select a route that will have the least movement and psychological impact on people in the vicinity when the robot 10 moves from the start point to the _goal ^point . This influence degree D is obtained by the following formula.

In the above formula, n corresponds to the number of people recognized by the robot 10.

Next, the route cost C for each candidate route is calculated by multiplying the travel cost E and the influence degree D by weighting factors a and b and adding them up as shown in the following formula. Note that the weighting factors a and b are set to constant values based on the results of prior experiments, e.g., a=1.7, b=0.3.

Then, the candidate route with the smallest route cost C is determined as the optimal route, and a movement route of the robot 10 along this optimal route is generated. Here, the movement of the robot 10 between each WP on the optimal route is performed along a movement trajectory based on the simulation results.

Next, a modified example of this embodiment will be described. In the following description, the same reference numerals will be used for components that are the same as or equivalent to those in the above embodiment, and descriptions will be omitted or simplified except for points that differ from the above embodiment.

In this modified example, the control device 13 treats a human (approaching individual) approaching the robot 10 from the periphery of the robot 10 as a target agent among the moving obstacles, predicts multiple routes that the approaching individual may take in the future as candidate routes, identifies the optimal route among these, and determines the movement route of the robot 10 so as to avoid the approaching individual by matching it with the movement of the approaching individual along the optimal route.

In other words, the candidate path searching means 24 in this modified example predicts the future path of an approaching individual that avoids the robot 10 and other agents such as other humans as candidate paths, using a procedure similar to that performed for the robot 10 in the above embodiment, and derives multiple candidate paths, including the aforementioned contact path in which the approaching individual comes into contact with other agents.

In the optimal path extraction means 25 according to this modified example, first, the movement prediction simulation unit 28 simulates the movement state of each agent over time for each candidate path of the approaching individual, by the movement of the robot 10 accompanying the movement of the approaching individual and the interaction of each obstacle, using the same procedure as in the previous embodiment. Then, the final decision unit 29 calculates the movement cost E, which is a first index value relating to the movement efficiency of the approaching individual, which is the target agent, but with the robot 10 replaced in the previous embodiment, and the influence degree D, which is a second index value relating to the burden on the human, similar to the previous embodiment. Thereafter, the movement cost C of the target approaching individual is calculated in the same manner as in the previous embodiment, and the candidate path with the smallest movement cost C becomes the optimal path for the approaching individual.

Then, in the final decision unit 29, the path of the robot 10 when the target approaching individual moves along the optimal path is identified as the movement path of the robot 10 based on the simulation results by the movement prediction simulation unit 28. In other words, the movement path is the path of the robot 10 simulated under conditions in which the target approaching individual moves along the optimal path. Furthermore, the operation command means 26 issues an operation command to the operation unit 11 so that the robot 10 moves autonomously along the determined movement path.

According to the above modified example, in addition to the operation control mode in the above embodiment in which the robot 10 moves to the target location while avoiding the crowd, or alternatively to the above control mode, a control mode is possible in which the robot 10, which is already in a crowd, operates to avoid an approaching person who is trying to pass by it.

The control device 13 described above includes a cognitive state estimation means 22 that estimates the agent's cognitive state of the surroundings, and a movement prediction simulation unit 28 (movement prediction simulation means) that simulates the agent's movement over time. Therefore, the control device 13 also functions as a movement prediction device that predicts the movement of each agent over time when, during the process of a specific agent's movement, there are other agents that can move around the agent and objects that are always stationary.

The control device 13 also has the candidate route search means 24 and the optimal route extraction means 25, and functions as a movement path generation device that generates a movement path for the autonomously moving robot 10 that moves autonomously while avoiding obstacles present in the surrounding area.

The robot according to the present invention is not limited to the autonomous mobile robot 10 described in the above embodiment, but may be a mobile body capable of autonomously moving within a given space, such as an automobile vehicle, ship, or aircraft, or a manipulator such as a robot arm that moves within a given range of space, and the generation of the optimal path during such movement can be performed using the same configuration and procedure as described above.

In addition, the configuration of each part of the device in this invention is not limited to the illustrated configuration example, and various modifications are possible as long as they produce substantially the same effect.

10 Robot 11 Operation unit 12 Detection device 13 Control device (movement path generation device, movement prediction device)
22 Cognitive state estimation means 24 Candidate route search means 25 Optimal route extraction means 26 Operation command means 28 Movement prediction simulation unit (movement prediction simulation means)
29 Final Decision Section

Claims (7)

In a robot that moves autonomously while avoiding obstacles in the surrounding area,
a detection device that detects position information and speed information of the obstacle; and a control device that controls the autonomous movement based on a detection result of the detection device while taking into consideration a state of the obstacle;
the control device comprises: a candidate route searching means for generating a plurality of candidate routes which are candidates for a route along which a target agent, which is a target agent composed of the robot and a movable obstacle among the obstacles, moves, based on position information and speed information of the obstacle; an optimal route extraction means for extracting an optimal route from among the candidate routes; and an operation command means for issuing an operation command for the autonomous movement based on the optimal route;
the optimum route extraction means comprises: a movement prediction simulation unit which simulates a movement state of each of the agents through an interaction between the robot and the obstacles for each of the candidate routes; and a final determination unit which determines the optimum route based on a movement efficiency of the target agent and an influence on the moving obstacles from a simulation result in the movement prediction simulation unit;
the movement prediction simulation unit predicts changes in position and speed of each of the agents over time by utilizing virtual forces acting between the agents and between each of the agents and the obstacle object that is always stationary;
the final determination unit calculates, for each of the candidate routes, a first index value related to the movement efficiency and a second index value related to the load imposed on the moving obstacle by a pre-stored formula, and selects the candidate route having the smallest route cost obtained by combining these index values as the optimal route ;
said action command means issues the action command with the optimal route as the movement route of the robot when the robot itself is the target agent, while when the moving obstacle approaching the robot is the target agent, said action command is issued by the movement prediction simulation unit to move the robot along the movement route of the robot simulated under conditions where the moving obstacle moves along the optimal route . In a robot that moves autonomously while avoiding obstacles in the surrounding area,
a detection device that detects position information and speed information of the obstacle; and a control device that controls the autonomous movement based on a detection result of the detection device while taking into consideration a state of the obstacle;
the control device comprises: a candidate route searching means for generating a plurality of candidate routes which are candidates for a route along which a target agent, which is a target agent consisting of the robot and a movable obstacle among the obstacles, moves, based on position information and speed information of the obstacle; an optimal route extraction means for extracting an optimal route from among the candidate routes; and an operation command means for issuing an operation command for the autonomous movement based on the optimal route;
the candidate route search means generates the candidate route including a contact route in which the target agent avoids the moving obstacle while contacting the other agents,
the optimum route extraction means comprises: a movement prediction simulation unit which simulates a movement state of each of the agents through an interaction between the robot and the obstacles for each of the candidate routes; and a final determination unit which determines the optimum route based on a movement efficiency of the target agent and an influence on the moving obstacles from a simulation result in the movement prediction simulation unit;
the movement prediction simulation unit predicts changes in position and speed of each of the agents over time using virtual forces acting between each of the agents and between each of the agents and the obstacle object that is always stationary, calculates a resultant force vector that is a combination of a virtual attractive force from the destination of each of the agents, a virtual repulsive force acting between the agent and the object and the agent, and a contact force at the time of contact, and estimates that each of the agents will move in accordance with the resultant force vector;
The robot is characterized in that, in the final decision unit, for each of the candidate routes, a first index value related to the movement efficiency and a second index value related to the load applied to the moving obstacle are calculated using a pre-stored formula, and the candidate route having the lowest route cost obtained by combining these index values is estimated to be selected as the optimal route . The robot according to claim 2, characterized in that in the final decision unit, a movement cost is calculated as the first index value by adding up a first parameter related to the moving direction of the target agent and a second parameter related to an avoidance direction of the target agent for avoiding the obstacle, which is perpendicular to the moving direction, and a degree of influence of the repulsive force and the contact force is calculated as the second index value, and the path cost is calculated for each of the candidate paths by multiplying the movement cost and the degree of influence by a weighting coefficient and adding them up, and the candidate path having the lowest path cost is selected as the optimal path. The detection device is provided so as to be capable of detecting the orientation of the moving obstacle,
the control device further comprises a cognitive state estimation means for estimating a cognitive state of each of the agents based on a detection result by the detection device,
the cognitive state estimation means specifies, for each agent, a visual field range in which the agent can recognize people and objects around the agent as a recognizable range;
In the movement prediction simulation unit, a force vector generated at the time of a collision and a psychological force vector defined as a psychological factor are calculated by a previously stored formula, and then the force vectors are summed up to obtain the repulsive force;
3. The robot according to claim 2 , wherein the psychological force vector is defined to act only when the object or another agent is within the perceptible range. A movement path generation device that generates a movement path for an autonomous moving robot while avoiding obstacles in the vicinity,
a candidate route searching means for generating a plurality of candidate routes which are candidates for a route along which a target agent, which is a target agent composed of said robot and a movable obstacle among said obstacles, can move, based on position information and speed information of said obstacle, and an optimal route extraction means for extracting an optimal route which is optimal from each of said candidate routes,
the optimum route extraction means comprises: a movement prediction simulation unit which simulates a movement state of each of the agents through an interaction between the robot and the obstacles for each of the candidate routes; and a final determination unit which determines the optimum route based on a movement efficiency of the target agent and an influence on the moving obstacles from a simulation result in the movement prediction simulation unit;
the movement prediction simulation unit predicts changes in position and speed of each of the agents over time by utilizing virtual forces acting between the agents and between each of the agents and the obstacle object that is always stationary;
In the final determination unit, for each of the candidate routes, a first index value related to the movement efficiency and a second index value related to the load on the moving obstacle are calculated using a pre-stored formula, and the candidate route having the lowest route cost obtained by combining these index values is selected as the optimal route, and when the robot itself is the target agent, the optimal route is set as the movement route, while when the moving obstacle approaching the robot is the target agent, the movement prediction simulation unit sets as the robot's route simulated under conditions in which the moving obstacle moves along the optimal route as the movement route. A program for causing a computer of a movement path generating device to generate a movement path for an autonomous moving robot while avoiding obstacles in the vicinity,
making the computer function as a candidate route searching means for generating a plurality of candidate routes which are candidates for a route along which a target agent, which is a target agent composed of the robot and a movable obstacle among the obstacles, may move, based on the position information and speed information of the obstacle, and an optimal route extracting means for extracting an optimal route which is optimal from each of the candidate routes;
the optimum route extraction means comprises: a movement prediction simulation unit which simulates a movement state of each of the agents through an interaction between the robot and the obstacles for each of the candidate routes; and a final determination unit which determines the optimum route based on a movement efficiency of the target agent and an influence on the moving obstacles from a simulation result in the movement prediction simulation unit;
the movement prediction simulation unit predicts changes in position and speed of each of the agents over time by utilizing virtual forces acting between the agents and between each of the agents and the obstacle object that is always stationary;
a movement path generation device that determines, for each of the candidate routes, a first index value related to the movement efficiency and a second index value related to the load on the moving obstacle using a pre-stored formula, and selects the candidate route having the lowest route cost obtained by combining these index values as the optimal route; and when the robot itself is the target agent, the optimal route is selected as the movement path, while when the moving obstacle approaching the robot is the target agent, the movement prediction simulation unit selects a route of the robot simulated under conditions in which the moving obstacle moves along the optimal route as the movement path. 1. A movement prediction device for predicting the movement of a predetermined agent over time when other movable agents and objects that are always stationary are present around the agent during the agent's movement, comprising:
a cognitive state estimation means for estimating a cognitive state of each of the agents based on a position and an orientation of each of the agents separately obtained, and a movement prediction simulation means for simulating a movement state of each of the agents over time based on an interaction between each of the agents and the object when a predetermined one of the agents moves,
the cognitive state estimation means specifies, for each agent, a visual field range in which the agent can recognize people and objects around the agent as a recognizable range;
The movement prediction simulation means uses a pre-stored formula and separately obtained position information and speed information of each of the agents to calculate a resultant force vector of a virtual attractive force from the destination of each of the agents, a virtual repulsive force acting between the object and each of the agents, and a contact force when a specific agent comes into contact with another agent, and it is estimated that each of the agents will move according to the resultant force vector.The repulsive force is calculated by separately calculating a force vector generated at the time of a collision and a psychological force vector defined as a psychological factor, and then calculating these force vectors, and the psychological force vector is defined to act only when the object or another agent is within the perceptible range.This movement prediction device is characterized in that

Images