JP7178066B2 - Robot, its action planning device and action planning program - Google Patents

Description

The present invention relates to a robot that operates autonomously while avoiding future interference or collision with a moving obstacle, such as a human being, according to the moving situation of the moving obstacle, its action planning device, and a program for action planning. Regarding.

As a robot that autonomously moves from a predetermined starting point to a destination, there is a robot that can move while avoiding interference with the moving obstacle by sensing the current movement status of a moving obstacle such as a human being. When such a robot moves in an environment where there are many people, such as a crowd, it is necessary not only to unilaterally avoid people, but also to consider the movement of people and other obstacles, and effectively reduce interference. A cooperative action with a human is required to avoid it. In other words, when moving in such an environment, in addition to passive actions to human actions, such as interference avoidance actions and stop actions, the robot's action intentions are transmitted to humans through these actions, and human interference occurs. It is necessary to actively encourage avoidance behavior. For example, when the robot moves to its destination, when many people are crowded through a gap that the robot cannot pass through, even if the robot is moved toward the gap, In order to secure a passage for the robot, it is necessary to encourage people around the gap to move.

Therefore, robots have been proposed that are capable of not only avoiding moving obstacles, but also acting to encourage avoidance of moving obstacles (for example, see Patent Documents 1 and 2).

JP 2012-130986 A Japanese Patent No. 5409924

However, the robots of Patent Literatures 1 and 2 are based on the premise that they will move away from humans to some extent and pass each other when interference with humans is expected in the future. For this reason, no consideration has been given to moving the robot in a place where the density of people is high and the movement space of the robot is narrow. In addition, in an emergency where a human suddenly appears in front of the robot at a corner, etc., the robot may stop suddenly. Unable to take action.

The present invention has been devised by paying attention to such problems, and its purpose is to provide a robot that can handle moving obstacles while taking psychological consideration of the moving obstacles according to the relative relationship between the robot and the moving obstacles. To provide a robot capable of avoiding action while approaching or contacting the robot, its action planning device, and a program for action planning.

In order to achieve the above object, the present invention mainly comprises a predetermined operating part, a detecting device for detecting position information of a moving obstacle existing in the surroundings, and the operating part based on the detection result of the detecting device. a control device for controlling motion, wherein the control device includes interference degree determination means for determining a possibility that the robot will interfere with the moving obstacle in the future; action planning means for planning actions of the robot so as to move the robot to a position; When the interference degree determining means determines that the robot may interfere with the moving obstacle in the future, the planning means plans the action of the robot so as to avoid the moving obstacle. The interference planning unit includes an emergency avoidance action selection unit that selects an action of the robot that performs collision avoidance when an emergency occurs due to the situation of the robot with respect to the moving obstacle; With respect to collision avoidance, a normal avoidance action selection unit is provided for estimating an approach risk, which is a risk when the robot approaches the moving obstacle, and selecting an action of the robot.

According to the present invention, in an emergency, in addition to stopping the robot, it is also possible for the robot to avoid a collision while approaching a moving obstacle such as a human being according to the relative relationship between the robot and the moving obstacle. In addition to the normal detour trajectory, approach trajectories and contact trajectories can also be used as trajectory candidates. At this time, the risk of approaching the moving obstacle and the movement energy of the robot are taken into consideration, and the route with the same approach risk as when the robot leaves the moving obstacle and detours and the least movement energy is selected. can. Therefore, it is possible to formulate a plan for efficiently and appropriately moving the robot while taking into consideration the psychological aspects of obstacles such as humans.

2 is a block diagram schematically showing only the overall configuration related to motion control of the robot according to the first embodiment; FIG. FIG. 4 is a conceptual diagram for explaining derivation of movement information of a robot and a human; FIG. 3 is a conceptual diagram showing a state in which the robot and the human are side by side from the state in FIG. 2 ; (A) to (C) are conceptual diagrams for explaining determination of a human's degree of recognition of a robot. It is a conceptual diagram for explaining the parameters when obtaining the recognition level. 3 is a block diagram showing the configuration of an interference planning unit according to the first embodiment; FIG. FIG. 4 is a conceptual diagram for explaining each area specified by the existing area specifying function; FIG. 4 is a conceptual diagram for explaining relative distances for determination; (A) is a conceptual diagram for explaining the interference angle, and (B) is a conceptual diagram for explaining the idea of the degree of compromise according to the difference in the interference angle. FIG. 11 is a flowchart for explaining a processing procedure in an action content identification unit; FIG. (A) to (D) are conceptual diagrams for explaining the processing contents of the constant speed trajectory generation function. FIG. 11 is a conceptual diagram for explaining trajectory generation when WP4 is added; (A) and (B) are conceptual diagrams for explaining the processing contents of the acceleration trajectory generation function. 4 is a flowchart for explaining a processing procedure in a trajectory candidate searching unit; FIG. 4 is a conceptual diagram for explaining each term of a cost formula for obtaining interference avoidance energy; (A) to (C) are conceptual diagrams exemplifying trajectories of patterns excluded from trajectory candidates. (A) to (E) are conceptual diagrams for explaining the procedure of trajectory generation in the iterative action search system. FIG. 10 is a conceptual diagram for explaining behavior of a robot when there is no space to pass through; 4 is a flowchart for explaining the action planning procedure of the robot according to the first embodiment; FIG. 11 is a block diagram showing the configuration of an interference planning unit according to the second embodiment; FIG. 4 is a conceptual diagram for explaining types of trajectories; FIG. 11 is a flow chart for explaining the action planning procedure of the robot according to the second embodiment; FIG.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(First embodiment)

FIG. 1 shows a block diagram schematically showing only the overall configuration related to motion control of the robot 10 according to the first embodiment. In this figure, the robot 10 functions as an autonomous robot that autonomously moves from a preset departure point (start position) to a destination (target position), and interference with obstacles is assumed during movement. In this case, it is possible to move while performing various actions for avoiding the interference.

The obstacles here include not only fixed obstacles (fixed obstacles) such as walls existing around the movement path of the robot 10, but also moving obstacles whose positions change over time. In each of the following embodiments, the moving obstacle is assumed to be a person walking or running.

In addition, as the various actions for interference avoidance, in addition to movement for interference avoidance performed only by the robot 10 without expecting human interference avoidance action, human movement for interference avoidance is encouraged. The robot 10 intends to work on a human being, and includes a working action. Actions of the robot 10 that perform these various actions are appropriately selected according to various situations and the like, as will be described later.

Here, as the above-mentioned action action, there is an insistence action in which a human being is expected to behave to avoid interference to some extent. In addition to the movement that suggests a course to make a person aware of it, there are actions such as calling out a voice to encourage the person to avoid, and intentionally contacting the person.

As shown in FIG. 1, the robot 10 includes an action unit 11 composed of mechanisms and devices capable of performing various actions, a detection device 12 for detecting environmental information around the robot 10, and a detection device 12. A control device 13 is provided for controlling the operation of the robot 10 based on the detection result.

The operation unit 11 includes a moving means 14 including a mechanism for autonomous movement within a predetermined range and a power source thereof, an arm 15 including a mechanism for operating within a predetermined space and a power source thereof, and a speaker 16 that emits a predetermined sound in, for example. These moving means 14, arm 15, and speaker 16 are all made up of known members, mechanisms, devices, etc., and are not essential parts of the present invention.

The detection device 12 includes an ambient environment detection sensor 18 for detecting positional information of obstacles such as humans and objects existing around the robot 10, and a recognition state detection sensor 19 for detecting the recognition state of the human H with respect to the robot 10. , and the detection results of the respective sensors 18 and 19 are sequentially transmitted to the control device 13 .

Although the ambient environment detection sensor 18 is not particularly limited, it detects each surface of obstacles around the robot 10 based on the reflection state of objects including humans due to irradiation of laser light around the robot 10. A range sensor such as a known laser range finder is used to detect the position of the part. In other words, the ambient environment detection sensor 18 measures the two-dimensional position of each point in the point group forming the surface portion of various obstacles in plan view with the robot 10 as a reference.

As the ambient environment detection sensor 18, other sensors and devices such as sensors using GPS or the like may be applied as long as the positions of obstacles such as humans existing around the robot 10 can be detected. It is possible.

As the recognition state detection sensor 19, although not particularly limited, a known depth sensor such as KINECT (registered trademark) is used to identify the human face and to detect the human face with respect to the robot 10. Orientation is detected.

As the recognition state detection sensor 19, other sensors and devices capable of similar detection may be used.

The control device 13 issues various motion commands (to be described later) to the motion unit 11 so that the robot 10 can move autonomously to a destination specified in advance while avoiding interference with humans and other obstacles. is done.

The control device 13 is provided integrally with or separately from the robot 10, and is composed of a computer including an arithmetic processing device such as a CPU, a storage device such as a memory and a hard disk, and the like. A program for functioning as each of the following means is installed in the computer.

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

The control device 13 uses the detection result of the surrounding environment detection sensor 18 to determine the degree of interference determination means 21 that determines the possibility of human interference with the robot 10 in the future, and the detection result of the recognition state detection sensor 19. using the recognition level estimation means 22 for estimating the recognition level of humans for the robot 10 at the present time, and the determination result by the interference level determination means 21 and the estimation result by the recognition level estimation means 22, and the surrounding people Action planning means 23 for planning various actions including the action action so as to move the robot 10 to the destination according to the situation of obstacles, and an action planning means 23 for estimating the degree of achievement of intention transmission to the human by the action action. achievement level estimation means 24; plan adjustment means 25 for changing the action plan of the robot 10 according to the result of the achievement level estimation; and an operation command means 26 for performing the operation.

Therefore, the control device 13 functions as an action planning device that plans various actions (movements) for future interference avoidance with the person in the surroundings according to the detection result of the position information of the person. It will be.

(Interference degree determination means)
As shown in FIG. 2, under the condition that the robot 10 and the human H approach each other while moving, the interference degree determination means 21 determines the current state of the human H based on the detection result of the surrounding environment detection sensor 18 as follows. is obtained, and the presence or absence of the possibility that the robot 10 will interfere with the human H in the future is determined.

First, as a premise, movement information consisting of position information and speed information (velocity and direction) of the robot 10 is specified as trajectory information in which a route, which is position information to a destination, is associated with time. Here, the current position of the robot 10 is not particularly limited, but is defined as the position of the central portion of the robot 10 in plan view (hereinafter referred to as "the trunk center of the robot 10"), and the position of the robot 10 in the absolute coordinate system in plan view. It is specified as two-dimensional coordinates (x _R , y _R ). Also, the velocity information of the robot 10 in the absolute coordinate system is specified as a velocity vector V _R consisting of velocity components (v _Rx , v _Ry ) in the two-dimensional coordinate direction. The absolute coordinate system used in this description has a predetermined position in the path P of the robot 10 as the origin, the vertical direction along the path P in FIG. 2 as the x axis, and the horizontal direction as the y axis. do.

On the other hand, the movement information of the human H is obtained based on the positional information of the human H detected by the ambient environment detection sensor 18 .

That is, in the ambient environment detection sensor 18, in a relative coordinate system with the robot 10 as a reference, the position of each point of the point group that constitutes the surface portion of each obstacle such as a human H or an object that exists around the robot 10 is detected. Then, in this interference degree determination means 21, the continuity is determined from the pitch size of each point in the point group representing the surface of the obstacle, and the presence and type of each obstacle is determined according to the continuity. is identified. Here, an obstacle having a width corresponding to the width of the shoulders of human H (hereinafter referred to as "trunk width of human H") is identified as human H, and the point group of the surface of human H is approximated to an ellipse. As a result, the position of the central portion in plan view (hereinafter referred to as "the center of the trunk of the human H") is set as the current position of the human H. As shown in FIG. The current position of the human H in the relative coordinate system thus detected is converted from the coordinates (x _R , y _R ) in the absolute coordinate system at the current position of the robot 10 to the coordinates (x _H , y R ) in the absolute coordinate system. y _H ). In addition, since the surface portion of the same shape that constitutes the human H is displaced due to measurement by the ambient environment detection sensor 18 at predetermined time intervals, the displacement of the coordinates (x _H , y _H ) of the human H over time is Thus, the velocity vector VH of the human _H in the absolute coordinate system consisting of velocity components ( _vHx , _vHy ) is calculated.

Next, based on the current positions (x _R , y _R ) and (x _H , y _H ) of the robot 10 and the human H, the x-axis of each trunk center of the robot 10 and the human H at the present time is calculated every predetermined time. A longitudinal separation l _HR in the direction is calculated.

Then, from the speed information of the robot 10 and the human H, the time St until the robot 10 and the human _H will be aligned side by side in the future is calculated by the following equation.

Next, from time St, as shown in FIG. 3, the horizontal ( _y -axis direction) separation distance _DHR between the trunk centers of the robot 10 and the human H when the robot 10 and the human H are lined up side by side is calculated by the following equation. Calculated by

Finally, based on the separation distance _DHR , the degree of interference, which indicates the possibility of future interference between the robot 10 and the human H when the robot 10 and the human H do not perform an avoidance action at present, is determined.

ここで、前記定数Ｋは、人間Ｈがある物体に接近したときに回避行動を取れるパーソナルスペースの横幅（マージン）に相当する倍数（例えば、１．３倍）とされる。そして、前記離間距離Ｄ_ＨＲが、閾値Ｄ_Ｌ以上の場合は、干渉度が「無」とされ、逆に、当該離間距離Ｄ_ＨＲが、閾値Ｄ_Ｌよりも小さい場合は、干渉度が「有」とされる。つまり、本実施形態においては、前記各パーソナルスペースＰＳ_Ｒ，ＰＳ_Ｈが重なり合う状態が、「干渉」と称される状態となる。 Specifically, as shown in FIG. 3, a _{constant K} (K> 1) Circular personal spaces PS _R and PS _H (broken line portions in the figure) having doubled diameters are virtually set. Then, the degree of interference is determined using the distance D _L obtained by adding the radii R _LR and R _LH of the personal spaces PS _R and PS _H as a threshold value for interference determination.
That is, _DL is represented by the following equation.

Here, the constant K is a multiple (for example, 1.3 times) corresponding to the horizontal width (margin) of the personal space in which the person H can take avoidance action when approaching an object. When the separation distance _DHR is equal to or greater than the threshold value D _L , the degree of interference is determined to be "absent". Conversely, when the separation distance _DHR is smaller than the threshold value D _L , the degree of interference is determined to be ”. That is, in the present embodiment, the state in which the personal spaces PS _R and PS _H overlap is called "interference".

In the interference degree determination means 21, instead of the lateral separation distance _DHR , based on the current movement information of the robot 10 and the human H, the shortest straight line at the time when they will be closest in the future is calculated. The relative distance, which is the distance, may be used to determine the presence or absence of the degree of interference by setting a threshold in the same manner as described above.

(Recognition level estimation means)
When the human H approaches the robot 10, the recognition level estimating means 22 determines the human H's visual recognition level of the robot 10 based on the orientation of the human H's face detected by the recognition state detection sensor 19. "", "Small", and "None".

Specifically, first, from the current position (x _R , y _R ) of the robot 10 and the current position (x _H , y _H ) of the human H detected by the surrounding environment detection sensor 18, the current robot 10 and the human A relative distance, which is a straight line distance between each trunk center of H, is calculated. The current position (x _H , y _H ) of the human H can also be detected using the distance measuring function of the recognition state detection sensor 19 .

Then, when the relative distance D _SHR becomes equal to or less than a preset distance (eg, 5 m), the degree of recognition is estimated as follows.

That is, as shown in FIG. 4A, the robot 10 moves the first angle range θ _R1 (for example, the center line C extending along the face direction of the human H) with respect to the face direction of the human H. ), it is determined that the human H sees the robot 10 in the central visual field, and the degree of recognition is set to "high". Also, as shown in FIG. 2B, the robot 10 exceeds the first angular range θ _R1 with respect to the direction of the face of the human H and falls within the second angular range θ _R2 (for example, the center line C). ), it is determined that the human H sees the robot 10 in the peripheral vision, and the degree of recognition is set to "low". Furthermore, as shown in FIG. 4C, when the robot 10 exists beyond the second angular range θ _R2 with respect to the direction of the face of the human H, the robot 10 is in the peripheral visual field of the human H. It is judged to exist outside, and the degree of recognition is "None".

Specifically, the degree of recognition is determined by the following calculation.

As a premise, the head angle θ _HEAD corresponding to the orientation of the face of the human H is detected by the recognition state detection sensor 19 . As shown in FIG. 5, the head angle θ _HEAD is an inclination angle formed by the center line C extending in the direction of the face of the human H with respect to the y-axis direction. Also, from the current position (x _R , y _R ) of the robot 10 and the current position (x _H , y _H ) of the human acquired by the surrounding environment detection sensor 18, the relative angle θ _HR between the human H and the robot 10 is It is obtained by the following formula (4). Then, when the following equation (5) holds, that is, when the magnitude of the absolute value of the difference between the relative angle θ _HR and the head angle θ _HEAD is 1/2 or less of the first angle range θ _R1 , Awareness is considered to be "high". Further, when the following formula (6) holds, that is, when the magnitude of the absolute value exceeds 1/2 of the first angle range θ _R1 and is equal to or less than 1/2 of the second angle range θ _R2 is considered to have a "small" degree of recognition. Furthermore, when the following equation (7) holds, that is, when the magnitude of the absolute value exceeds 1/2 of the second angle range θ _R2 , the degree of recognition is set to "no".

The degree of recognition in the present embodiment is determined as an instantaneous degree of recognition representing the possibility that the human H visually recognizes the robot 10 at the moment of determination, but the present invention is limited to this. It is also possible to judge whether the degree of recognition is "high", "low" or "no" depending on the state of recognition over time.

This degree of recognition over time is determined, for example, as follows. That is, when the above-described momentary state of "high" recognition level exists for a predetermined number of times or more within a predetermined time up to the present time, the recognition level is set to "high". In addition, if the momentary recognition level "high" state exists more than a predetermined number of times within a predetermined time period in the past, the recognition level is set to "low". Furthermore, in cases other than these, the degree of recognition is set to "none". The various recognition levels described above are appropriately selected according to the application and properties of the robot 10 .

(action plan means)
As shown in FIG. 1, the action planning means 23 plans the route of the robot 10 from the current position to the destination when the interference degree determination means 21 determines that the degree of interference is "no". From the planning unit 28 and the interference time planning unit 29 that plans the route and motion of the robot 10 from the current position to the destination while avoiding interference with humans when it is determined that the degree of interference is "present" Become.

In the non-interference time planning unit 28, since no interference with a human, which is a moving obstacle, is assumed at this time, collision with a fixed obstacle such as a wall detected by the detection device 12 or grasped in advance is avoided. A route with the shortest distance is planned. Then, a trajectory corresponding to the position information and time is generated for the route, and the trajectory is specified as an execution trajectory for moving the robot 10 .

As shown in FIGS. 1 and 6, the interfering time planning unit 29 is a pass-through space determination unit that determines whether or not there is a pass-through space having a pass-through width that is the width required for the robot 10 to pass beside a human. 31, and an interference avoidance action determination unit 32 that determines an interference avoidance action, which is an action of the robot 10 for avoiding interference, according to the presence or absence of a pass-through space.

The pass-through space determining unit 31 determines whether or not there is a pass-through space as follows.

In this determination, similar to the above-described calculation based on the detection result of the surrounding environment detection sensor 18, when the human H will be next to the robot 10 in the future, the horizontal direction (y-axis direction) shown in FIG. ) is _computed . In other words, this width _ls is the width between the person H and a fixed obstacle such as the wall W of the passage P located in the lateral direction of the person H. As shown in FIG. Furthermore, this width l _s is the robot passage width required for the robot 10 to pass. Here, the robot passage width l _s is the diameter of the personal space PS _R of the robot 10 (2R _LR ) is determined. If the robot passage width l _s is wider than the width of the personal space PS _R , it is determined that there is a "passing space", and if not, it is determined that "there is no passing space".

In addition, in the passing space determination unit 31, based on the position information of the identified fixed obstacles and the current movement information of the robot 10 and the human H, the robot 10 will pass next to the human H in the future. Other calculation methods can be adopted as long as it can be estimated whether or not it is physically possible.

As shown in FIG. 6, the interference avoidance action determination section 32 determines the interference avoidance action determination section 34 for a large space that specifies the interference avoidance action when there is a pass-through space, and the interference avoidance action when there is no pass-through space. and a narrow space action determination unit 35 to be specified.

1. Configuration and Processing Contents of Action Determining Unit for Large Space The configuration and processing content of the action determining unit 34 for vast space will be described below.

As shown in FIG. 6, the large space action determination unit 34 includes a behavior content identification unit 37 that identifies the behavior content of interference avoidance with respect to the human based on the relative relationship between the robot 10 and the human at the current time; An execution trajectory specifying unit 38 is provided for specifying an execution trajectory related to the movement route of the robot 10 from the current position to the destination so as to correspond to the action content of the robot 10 specified by the specifying unit 37 .

1-A Description of action content identification unit 37

The action content identification unit 37 includes an existing region identification function 40 that identifies where the robot 10 exists among a plurality of regions set according to the relative position and relative speed with respect to the human H at the current time, and a robot 10 Based on the interference angle calculation function 41 that calculates the respective interference angles of the robot 10 and the human at the time of interference with the human, and each interference angle calculated by the previously obtained degree of recognition and the interference angle calculation function 41, with respect to the human and a concession degree determination function 42 for specifying the action of the robot 10 for interference avoidance by determining the degree of concession, which is the degree to which the robot 10 yields the course, for each of the regions.

First, the existing area specifying function 40 will be described.

In the existing area _specifying function 40, first, the determination relative distance _LRH between the robot 10 and the human at the current time is calculated as will be described later. It is specified whether it exists in any of the three regions. These areas are divided into first, second, and third areas A1, A2, and A3 in order of shortest time required for interference, that is, in order from the human H side to the far side. Here, the first area A1 is a synchronous action area in which the robot 10 alone avoids interference in accordance with the movements of the human H without expecting the human H's interference avoidance action. In the second area A2, by suggesting the traveling direction of the robot 10 to the human H, the robot 10 is expected to take action to avoid interference, and the robot 10 is made to perform assertive actions including actions for avoiding interference. It is considered as an assertive action area. Furthermore, in the third area A3, even if there is a distance between the robot 10 and the human H at present, and the robot 10 performs an action to encourage the human H to avoid interference, This is an area where it is considered difficult to convey the intention. For this reason, the third area A3 is defined as an area outside the interference avoidance action in which the robot 10 is not yet conscious of the interference avoidance action including the action of acting on the robot 10 .

The relative distance for determination _LRH is obtained by converting the interference estimated time from the current time required for interference between the robot 10 and the human H into a distance in the relative direction between the robot 10 and the human H, and is obtained as follows. be done.

First, based on the velocity vectors V _R and V _H of the robot 10 and the human H specified by the detection result of the detection device 12 , the estimated interference time T _if is calculated. Then, from the respective velocities v _Rr and v _Hr of the robot 10 and the human H in the above-mentioned relative direction, which is the extending direction of the straight line L (see FIG. 8) connecting the centers of the trunks of the robot 10 and the human H, the relative A distance obtained by multiplying the speed by the estimated interference time _Tif is obtained as the determination relative distance _LRH . That is, the determination relative distance _LR is calculated by the following equations (8) and (9). Here, φ _R represents the angle formed between the direction of the velocity vector _VR of the robot 10 and the straight line L connecting the robot 10 and the human _H , as shown in FIG. It represents the angle formed between the direction of the velocity vector _VH and the straight line L.

The tuning action area A1 is an area in which the determination relative distance _LRH is within a predetermined boundary distance with respect to the human H. FIG. As shown in FIG. 7, this boundary distance is a critical avoidance distance L obtained by adding a preset action preparation time _Tla to the shortest avoidance distance _LPSE , which is longer than the personal space critical distance _LPS , as will be described later. _IL . This critical avoidance distance L _IL is obtained by the following equations (10) to (12).

The personal space critical distance L _PS means a psychological distance of the human H that the robot 10 does not want the robot 10 to approach any more, and is determined corresponding to the radii R _LR and R _LH of the personal spaces of the robot 10 and the human. is a constant value, for example, 1 m.

The shortest avoidance distance L _PSE is the collision avoidance preparation required for the robot 10 to take some action to avoid an unexpected collision when the robot 10 has intruded into the personal space of the human H. Considering the time _Tpp , it means the distance obtained by multiplying the collision avoidance preparation time _Tpp by the relative speed and adding the personal space critical distance _LPS . Here, the collision avoidance preparation time _Tpp takes a preset constant value.

Furthermore, the critical avoidance distance L _IL is necessary for causing the robot 10 to perform some kind of behavior in anticipation of interference avoidance behavior with respect to the human H before entering the area from the human H to the shortest avoidance distance L _PSE . It means the minimum distance from a human being H. That is, the critical avoidance distance L _IL is a distance obtained by _multiplying the preparation time for action _Tla , which is the minimum time required for the action action, by the relative speed. , means the distance obtained by adding the shortest avoidance distance L _PSE .

The asserted action area A2 has a relative distance for judgment _LRH longer than the critical avoidance distance _LIL , which is the distance from the human H to the boundary of the synchronized action area A1, and is the maximum distance at which the action action of the robot 10 is effective. The area is defined as a range shorter than the farthest working distance, which is the distance from the distant human H. The farthest working distance is set in advance as a constant value (for example, 7 m).

The interference avoidance action outside area A3 is an area farther from the human H than the assertion action area A2.

Next, the interference angle calculation function 41 will be described.

The interference angle calculation function 41 calculates the interference angle for each of the robot 10 and the human at the time when the personal spaces of the robot 10 and the human will contact each other in the future. As shown in FIG. 9(A), the interference angle is determined by the directions of movement of the robot 10 and the human H (directions of arrows in FIG. 9), and the respective personal spaces from their trunk centers C _R and C _H. This is the angle formed between _PSR and _PSH and the direction toward the interference point P, which is the point of contact. Hereinafter, the interference angle of the robot 10 is defined as an angle θ _R , and the interference angle of the human H is defined as an angle θ _H.

These interference angles θ _R and θ _H are values specified based on the detection results of the detection device 12, that is, the velocity vectors V _R and V _H of the robot 10 and the human H, and the interference point P , and the position vectors S _R and S _H in the directions between the trunk centers C _R and C _H of the robot 10 and the human H, respectively, are calculated by the following equations. In the following equation, the coordinates of the point of interference P are (p, q), and the coordinates of the trunk centers C _R and C _H at the time of interference are (x _R , y _R ) and (x _H , y _H ). and

Next, the concession degree determination function 42 will be described.

In this concession degree determination function 42, the difference between the existing area of the robot 10 specified by the existing area specifying function 40 and the interference angles θ _R and θ _H calculated by the interference angle calculating function 41 is determined as follows. , and the degree of human recognition estimated by the recognition degree estimating means 22, the degree of concession _PR is determined, and the action content of the robot 10 corresponding to the degree of _concession PR is specified.

As illustrated on the left side of FIG. 9B, when the difference between the interference angles θ _R and θ _H is large, the robot 10 should avoid to the left and the human H should avoid to the right. Each avoidance direction is easy to understand, such as being judged. On the other hand, when the difference between the interference angles θ _R and θ _H is small, as exemplified on the right side of FIG. Therefore, a threshold _{Δθ t} for the difference between the interference angles θ _R and _{θ H is} set.

In this embodiment, the degree of _{concession PR} of the robot 10 and the action corresponding to the degree of concession PR are determined according to the following various situations, but the numerical values used as the conditions are not particularly limited. Various modifications are possible as long as the degree of concession P _R of the robot 10 is set higher as the difference between the interference angles θ _R and θ _H is smaller.

Further, if the robot 10 is present within the asserted action area A2 at the present time, the recognition level estimated by the recognition level estimation means 22 is also considered to determine the _concession level PR of the robot 10 . Here, the degree of concession P _R of the robot 10 is set to be low because it is believed that the greater the degree of recognition, the more likely the human H will be expected to avoid interference.

In this case, when the degree of recognition is "high" and the difference between the interference angles θ _R and θ _H is equal to or greater than the threshold value Δθ _t (hereinafter referred to as "there is a difference in the interference angles"), the robot 10 is equivalent to the human H. , the concession degree _PR of the robot 10 is set to 50%. On the other hand, when the degree of recognition is “high” and the difference between the interference angles θ _R and θ _H is less than the threshold Δθ _t (hereinafter referred to as “there is no difference in the interference angles”), the robot 10 can move a little more. In addition, the concession degree P _R of the robot 10 is set at 75%. Therefore, in these cases, that is, when the robot 10 is present in the asserted action area A2 at the present time and the degree of recognition is "high", the human H is physically and psychologically affected by the robot 10's course suggestion. An assertive action of the robot 10, which is an appealing action that prompts a change, is planned.

Furthermore, when the degree of recognition is "small" or "no" and there is a difference in the interference angle, the concession degree _PR of the robot 10 is set at 100%. On the other hand, when the degree of recognition is "small" or "no" and there is no difference in the interference angle, the degree of concession _PR of the robot 10 is set to 125% in further consideration of safety. Therefore, in these cases, that is, when the robot 10 exists within the asserted action area A2 and the degree of recognition is "low" or "no", the avoidance behavior of the human H is not expected much, and the human Synchronous action is planned in which only the robot 10 avoids interference according to the movement of H. However, in this case, as the first appealing action, an assertive action of calling out to the human H's mind is also planned in order to increase the robot 10's intention transmission rate.

On the other hand, when the robot 10 is present in the synchronous action area A1 at the present time, only the synchronous action is planned according to the difference in the interference angle regardless of the degree of recognition. That is, in this case, only the synchronous behavior that is the same as when the robot 10 is present in the claimed behavior area A2 and the degree of recognition is "low" or "no" is planned. That is, when there is a difference in the interference angle, the concession degree P _R of the robot 10 is set to 100%, while when there is no difference in the interference angle, the concession degree P _R of the robot 10 is set to 125%.

When the robot 10 exists within the interference avoidance action outside area A3, the degree of concession _PR is not yet determined, and the action of the robot 10 for interference avoidance is not yet planned.

The processing procedure of the action content identification unit 37 having the above configuration will be described below using the flowchart of FIG. 10 .

First, the existing area specifying function 40 specifies in which area the robot 10 currently exists among the synchronization action area A1, the insistence action area A2, and the interference avoidance action outside area A3 (step S101).

When the robot 10 is present in the synchronization action area A1, the interference angle calculation function 41 calculates the interference angles θ _R and θ _H between the robot 10 and the human H (step S102). Then, the concession degree determination function 42 determines the difference between the interference angles θ _R and θ _H (step _S103 ). While planned (step S104), when there is no difference in the interference angle, a synchronous action is planned in which the degree of concession PR of the robot 10 is 125% (step _S105 ).

Also when the robot 10 exists in the asserted action area A2, first, the interference angles θ _R and θ _H are calculated by the interference angle calculation function 41 (step S102). Then, it is confirmed whether the current recognition level of the person H, which has already been estimated by the recognition level estimation means 22, is "high", or whether it is "low" or "none" (step S106).

Next, the concession degree determination function 42 determines whether or not there is a difference in the interference angle between when the degree of recognition is "large" and when the degree of recognition is "small" and "none" (steps S107 and S108). ).

Here, when the degree of recognition is "high" and there is a difference in the interference angle, an assertive action is planned in which the degree of concession PR of the robot 10 is 50% (step _S109 ). On the other hand, when the degree of recognition is "high" and there is no difference in the interference angles, an assertive action is planned in which the degree of concession PR of the robot 10 is 75% (step _S110 ).

When the degree of recognition is "small" or "no" and there is a difference in the interference angle, a synchronous action is planned in which the degree of concession P _R of the robot 10 is 100% (step S111). On the other hand, when the degree of recognition is "small" or "no" and there is no difference in the interference angle, a synchronous action is planned in which the degree of concession PR of the robot 10 is 125% (step _S112 ). Further, it is confirmed whether the assertive behavior of calling out has already been performed (step S113), and only when the assertive behavior of calling out has not been performed yet, the assertive behavior of calling out is planned (step S114). In other words, when the robot 10 is present in the assertive action area A2 at the present time and the degree of recognition is "low" or "no", one call-out, which is one of the assertive actions, is planned along with the synchronizing action. .

In the assertive action area A2, it is also possible to sense the state of human interference avoidance behavior and select assertive action and synchronous action according to the result. For example, if it is determined that the human has insisted on the interference avoidance action, the robot 10 performs the synchronous action, and if it is determined that the human does not change the action, the robot 10 is set to take the assertive action. Also good.

At this point, when the robot 10 is in the interference avoidance behavior outside area A3, the interference avoidance behavior of the robot 10 with respect to humans is not yet planned.

1-B Description of execution trajectory identification unit 38

The execution trajectory identifying section 38 derives an execution trajectory for the movement path of the robot 10 from the current position to the destination in accordance with the degree of _concession PR identified by the action content identifying section 37 . The execution trajectory specifying unit 38 includes a trajectory candidate searching unit 44 that searches for a plurality of patterns of trajectory candidates that can move the robot 10 toward the destination without interfering with humans, and An optimum trajectory extracting unit 45 extracts an optimum trajectory from the robot 10 and uses it as an execution trajectory for actually moving the robot 10 .

First, the trajectory candidate searching section 44 will be described below.

The trajectory candidate search unit 44 includes an interference time calculation function 47 that calculates an interference time, which is the time at which the robot 10 and the human will interfere with each other in the future. A constant-speed trajectory generation function 48 for generating a constant-speed trajectory in the case where the It has a generation function 49 and a trajectory candidate determination function 50 that excludes some trajectories from these constant speed trajectories and acceleration/deceleration trajectories based on predetermined requirements and finally determines trajectory candidates. The constant-velocity trajectory generation function 48 and the acceleration/deceleration trajectory generation function 49 constitute a trajectory generation function that generates various trajectories related to routes that avoid interference at the interference time.

Based on the detection result of the detection device 12, the interference time calculation function 47 determines the future time of interference between the robot 10 and the personal spaces PS _R and PS _H of the human H (see the dashed line portion in FIG. 11A). An interference time, which is an expected time, is calculated. That is, here, from the respective velocity vectors V _R and V _H of the robot 10 and the human H and the position vectors P _R(t0) and P _H(t0) of the robot 10 and the human H at the current time t0, , the time at which the relative distance between the robot 10 and the human H is R _LR +R _LH , which is the sum of the radii of their respective personal spaces PS _R and PS _H , is obtained, and this time is taken as the interference time t _i .

Next, the constant speed trajectory generation function 48 will be described.

The constant-velocity trajectory generation function 48 identifies a route that avoids interference by moving the robot 10 at a constant speed so that the personal spaces of the robot 10 and the human do not overlap each other, as follows. A fast trajectory is required. That is, first, three passage points (Way Points) of the trajectory for avoiding interference with humans are identified. Then, by connecting the current position of the robot 10, each passing point, and the destination with straight lines in the order of movement time, the constant-velocity trajectory is specified. In the following description, the passing points will be referred to as "WP1", "WP2", and "WP3" in order of movement time of the robot 10. As shown in FIG.

As shown in FIG. 6, the constant-speed trajectory generation function 48 includes a temporary trajectory generation function 52 that obtains a temporary trajectory that is a provisional constant-speed trajectory, and a temporary trajectory generation function 52 that adjusts the temporary trajectory to avoid interference as desired. and a trajectory adjustment function 53 that specifies the final constant-velocity trajectory of .

The provisional trajectory generation function 52 derives a provisional trajectory by the following procedure.

That is, first, the position of WP2 (see FIG. 11(B)), which is the point of closest approach where the robot 10, which is closest to the human H in the trajectory, is positioned side by side, is tentatively determined. Here, the temporarily determined position vector _PR(wp2) of WP2 is obtained as follows. That is, the current position SP (coordinates: (x _Rs , y _Rs )) of the robot 10 is obtained from the position of the human H at the interference time t _i obtained by the interference time calculation function 47 (broken line in FIG. 11B). It is set at a position shifted by a predetermined distance away from the human H in the normal direction N with respect to the target traveling direction T, which is linearly directed toward the ground GP (coordinates: (x _Rg , y _Rg )). 15) to (17). In the following equation, _ΔPSG is the relative position vector between the current position SP and the destination GP.
Note that the predetermined distance here is defined as the sum of the radii of the personal spaces PS _R and PS _H of the robot 10 and the human H, R _LR +R _LH , which is the sum of the radii, as _{a reference distance L b .} It is a value obtained by multiplying the obtained concession degree P _R (%).
Also, in the following equation, P _H(ti) represents the position vector of the human H at the interference time t _i .

Thereafter, the positions of WP1 and WP3 are tentatively determined from the tentatively determined position of WP2. Here, WP1, WP2, and WP3 are arranged on the same straight line parallel to the target traveling direction T, as shown in FIG. 11(B). Also, the linear distance between WP1 and WP2 and the linear distance between WP2 and WP3 are each set to have a constant value. For example, the linear distance between WP1 and WP2 is approximately 1 m, and the linear distance between WP2 and _WP3 is the same length as the radius RLR of the robot 10's personal space _PSR . As described above, when WP2 is tentatively determined, each position of WP1 and WP3 before and after WP2 is specified one by one. Then, a route connecting the current positions SP, WP1, WP2, WP3, and the destination GP of the robot 10 with straight lines in this order is planned, and a provisional temporary trajectory TR _t is specified from the route. That is, when the robot 10 passes through the nearest vicinity of the human H, the trajectory is generated so as to form a straight line along the target traveling direction T. FIG. Therefore, among the passing points, WP1 is a near front point through which the robot 10 passes before WP2, which is the point of closest approach, and WP3 is a rear near point through which the robot 10 passes after WP2.

In the present embodiment, three passage points of the temporary trajectory TR _t are provided at WP1, WP2, and WP3, but one additional point (WP4) may be set beyond WP3. As illustrated in FIG. 12, WP4 is placed on a straight line TL that connects the current position SP of the robot 10 and the destination GP, and is located at a certain distance D It is placed at a point away from the destination GP side for _TL . According to this, after the robot 10 completes passing the human H, the robot 10 returns to the original trajectory along the target traveling direction T, and the swelling of the trajectory at the time of passing each other can be suppressed to a minimum. It is useful in the movement of the robot 10 such as.

The trajectory adjustment function 53 performs trajectory adjustment to avoid interference due to the time difference between the movement of the human H and the movement of the robot 10 on the previous temporary trajectory. to explain.

When the robot 10 is moved along the temporary trajectory TR _t at a predetermined steady speed, the time at which the tentatively determined position of WP2 is reached may differ from the previously obtained interference time t _i . Interference with the human H may occur at this time. That is, for example, as in FIG. 11(C), since the provisional trajectory TR _t expands laterally with respect to the initial trajectory along the target traveling direction T, the interference time t when moving along this initial trajectory There is a possibility that WP2 is set at a position that requires a longer movement than the position of the robot 10 of _i (position of (A) in the figure). In this case, the time at which the robot 10 reaches WP2 is later than the interference time t _i . Therefore, as indicated by the solid line in FIG. 11(C), the human H advances further from the position (broken line in the figure) at the time of interference t _i , and interference with the robot 10 may occur. Become.

Therefore, in the trajectory adjustment function 53, if the time _tw2 at which the robot 10 reaches _WP2 is different from the time tw2 when the robot 10 reaches WP2, the trajectory adjustment function 53 performs the following adjustment until these times match completely or substantially. done.

Assuming that the robot 10 moves along the tentative trajectory TR _t specified in the previous stage, a new interference time t _i is obtained by the above-described interference time calculation function 47 . Then, as shown in FIG. 11(D), a new temporary trajectory TR _t is obtained based on the new interference time t _i by the above-described procedure in the temporary trajectory generation function 52, and the robot at the steady speed 10, the time _{tw2 at which WP2 is reached on the new temporary trajectory TRt is} obtained. The above process is repeated, and when the time _tw2 completely or almost coincides with the new interference time _ti , the provisional trajectory is identified as the official constant velocity trajectory.

11 shows an example in which the robot 10 generates a constant-velocity trajectory that avoids the right side and crosses the front of the human H from the right side. , a left-avoiding constant-velocity trajectory in which the robot 10 crosses behind the human H from the left side is also generated. Therefore, the constant-velocity trajectory generating function 48 generates two types of constant-velocity trajectories, ie, right-avoiding and left-avoiding, one each.

Next, the acceleration trajectory generation function 49 will be explained.

The acceleration/deceleration trajectory generation function 49 generates an acceleration/deceleration trajectory that avoids interference between the robot 10 and the human H by accelerating or decelerating the robot 10 moving at the predetermined steady speed in a partial section. be. As the acceleration/deceleration trajectory, first, a plurality of right-avoidance and left-avoidance lateral avoidance shift trajectories for moving the robot 10 to avoid the human H from the right side or the left side while temporarily accelerating or decelerating are generated. .

First, as shown in FIG. 13A, the position of human H at time _tw2 when robot 10 reaches WP2 in the constant-speed trajectory specified by constant _- speed trajectory generation function 48 is defined as reference position PHB. , a plurality of shift positions PH- _s , which are the positions of the human H shifted forward and backward along the moving direction of the human H from the reference position PH- _B at predetermined intervals, are set. Note that the shift from the reference position PH _B here is performed within a preset range. Then, by the procedure similar to the provisional trajectory derivation procedure in the provisional trajectory generation function 52 described above, provisional trajectories are generated when the robot 10 interferes with the human H who has reached each shift position _PHs . That is, in deriving the temporary trajectory, the position of WP2 of the robot 10 is obtained from the above equation (17), and the positions of WP1 and WP3 are specified from the positions of each WP2, as illustrated in FIG. Thus, a route passing through these waypoints is determined. In this embodiment, the interval between each shift position PHs is set to about 0.1 _m , but it is not particularly limited. In this way, it is possible to reduce the load of arithmetic processing for determining trajectory candidates.

Next, for each path, a lateral avoidance speed change trajectory that partially changes the speed of the robot 10 is determined according to the following conditions. When the robot 10 moves near the human H and passes the human H, it is desirable that the robot 10 moves at a steady speed. For this reason, although not particularly limited, in this embodiment, the robot 10 is moved at the preset steady speed after WP1, and the speed of the robot 10 is changed from the current position SP to WP1. to enable. Specifically, for each route, the interference time t _i at which the human H reaches the shift position PH _s is obtained from the velocity vector of the human H, and the interference time t _i is calculated on the route passing through the shift position PH _s . The speed of the robot 10 from the current position SP to WP1 is obtained so that the robot 10 reaches WP2 of . That is, since the straight line distance in the section from WP1 to WP2 in each route and the speed of the robot 10 in that section are both constant values set in advance, the passage time from WP1 to WP2 is constant in each track. Become. Therefore, the passage time is subtracted from the time from the current position SP of the robot 10 to the interference time t _i at which WP2 is reached. The speed from 10 current positions SP to WP1 is calculated, and the lateral avoidance shift trajectory corresponding to each shift position _PHs is specified.

Of these lateral avoidance shift trajectories, the trajectory (broken line in FIG. 13B) approaching the current position of the human H, relative to the constant speed trajectory represented by the solid line in FIG. , the trajectory accelerates with respect to the steady speed. Conversely, with respect to the constant speed trajectory, the trajectory away from the current position of the human H (one-dot chain line in the figure) is a trajectory that decelerates from the steady speed from the current position SP of the robot 10 to WP1. .

As described above, a plurality of patterns of the lateral avoidance shift trajectory are generated for each of the right avoidance and the left avoidance.

Furthermore, the acceleration/deceleration trajectory generation function 49 accelerates or decelerates the robot 10 only in a predetermined section while keeping the originally planned straight line path from the current position SP of the robot 10 to the destination GP as the acceleration/deceleration trajectory of the robot 10. As a result, a straight gear shift trajectory that avoids interference with the human H is generated as follows. That is, when the human H traverses the straight path, the WP2 is set on both front and rear sides of the robot 10 in the traveling direction so that the robot 10 is positioned at a predetermined distance from the personal spaces of the human H so that the personal spaces of the robot 10 and the human H do not overlap. be done. WP1 is set in front of WP2 on the straight path under the same conditions as when the lateral avoidance speed change trajectory was generated, and the speed from the current position SP of the robot 10 to WP1 is obtained, and the straight speed change trajectory is obtained. identified. The straight speed change trajectory includes a trajectory for accelerating movement in a speed change section from the current position SP to WP1 of the robot 10 at a speed higher than the steady speed, and a trajectory for decelerating the speed change section at a speed lower than the steady speed. , and each of these trajectories is generated one by one. Interference avoidance in which the robot 10 advances ahead of the human H so that the robot 10 passes through the point before the human H traverses the straight path by acceleration in the speed change section in the acceleration straight shift trajectory. action is taken. On the other hand, in the deceleration straight shift trajectory, the robot 10 causes the robot 10 to pass the human H first so that the robot 10 passes through the point after the human H traverses the straight path due to deceleration in the speed change section. evasive action is taken.

As described above, the constant-velocity trajectory generation function 48 generates one kind of constant-velocity trajectory each for right avoidance and left avoidance. Further, the acceleration/deceleration trajectory generation function 49 generates, as acceleration trajectories, a lateral avoidance shift trajectory with multiple patterns of acceleration and deceleration for right avoidance and left avoidance, and a straight shift trajectory for each type of acceleration and deceleration. be.

The trajectory candidate identification function 50 excludes predetermined trajectories from the trajectories generated by the constant speed trajectory generation function 48 and the acceleration/deceleration trajectory generation function 49, and the remaining trajectories are used as trajectory candidates.

A trajectory excluded here includes a trajectory that is physically infeasible for the robot 10 . For example, a trajectory on which the robot 10 cannot physically move due to the presence of a fixed obstacle such as a wall, or a speed specified in the acceleration/deceleration trajectory generated by the acceleration/deceleration trajectory generation function 49 may cause the robot 10 to Due to the specifications of , the speed is outside the preset range, and there are trajectories that are practically unworkable.

According to this trajectory candidate specifying function 50, trajectories that are practically unexecutable are excluded from candidates before calculation by the optimum trajectory extraction unit 45, which will be described later. The load can be reduced as much as possible.

A processing procedure in the trajectory candidate search section 44 having the above configuration will be described with reference to the flowchart of FIG. 14 .

First, the interference time calculation function 47 obtains an interference time t _i at which the robot 10 and the human being H will interfere in the future (step S201).

After that, the temporary trajectory generating function 52 in the constant speed trajectory generating function 48 generates a temporary trajectory along which the robot 10 travels at a preset steady speed from the obtained interference time t _i (step S202). Furthermore, in the trajectory adjustment function 53 included in the constant-velocity trajectory generation function 48, whether the time _tw2 when the robot 10 reaches _WP2 on the previously obtained temporary trajectory matches the previously identified interference time ti. It is determined whether or not (step S203). If these times do not match, the processes of steps S201 and S202 are repeated until they match. As a result, when the time _tw2 at which the robot 10 reaches _WP2 coincides with the previously specified interference time ti, the provisional trajectory is determined as a constant velocity trajectory (step S204).

Next, in the acceleration/deceleration trajectory generation function 49, an acceleration/deceleration trajectory is generated in which the robot 10 accelerates or decelerates in a portion of the progress of the robot 10 at the steady speed (step S205). Here, the acceleration/deceleration is changed by changing the interference time at an arbitrary interval around the interference time t _i on the constant-velocity trajectory, and shifting the position of the human H at each interference time along the traveling direction. A lateral avoidance shift trajectory is generated. Furthermore, here, a straight speed change trajectory is also generated in which the robot 10 accelerates and decelerates in a partial section on the straight trajectory from the current position to the destination.

Finally, in the trajectory candidate identification function 50, trajectories that do not meet predetermined requirements are excluded from the trajectories generated by the constant speed trajectory generation function 48 and the acceleration/deceleration trajectory generation function 49, and constant speed trajectories and acceleration/deceleration trajectories are selected. A plurality of trajectory candidates consisting of trajectories are determined (step S206).

Next, the optimal trajectory extraction unit 45 will be described below.

The optimum trajectory extraction unit 45 calculates, as a trajectory cost, an interference avoidance energy _Er corresponding to the kinetic energy for avoiding interference for each of the plurality of trajectory candidates searched by the trajectory candidate search unit 44, and calculates the trajectory cost. The fewest trajectory candidates are extracted as execution trajectories.

That is, the interference avoidance energy E _r is divided into 1st, 2nd, . . . , n−1, n, n+1, . It is calculated by summing first and second energy elements _EG and _ED at each point and integrating the sum from the current position to the destination. In other words, the interference avoidance energy E _r is a value obtained by integrating changes in the kinetic energy of the robot at each point on the trajectory.

Specifically, the interference avoidance energy E _r is obtained by the following cost expression (18) using the first and second energy elements E _G and E _D as parameters.

Here, as shown in FIG. 15, the direction of the straight path connecting the current position SP of the robot 10 and the destination GP is defined as the target traveling direction (G direction), and the normal direction perpendicular to the straight path is the traveling normal. At each point on the trajectory, the velocity vector of the robot 10 at that time is resolved into the G direction and the D direction, and the velocity vectors in these directions are used. Therefore, VGN in the above equation (18) represents the velocity vector in the G direction at point n on the trajectory, and _VG0 represents the velocity vector in the _G direction in the robot 10 at the current position. Further, V _DN represents the D-direction velocity vector at the n ground, and V _Dn-1 represents the D-direction velocity vector at the point n-1, which is one before the n point. Furthermore, A and B are weighting constants for each energy element that are set to predetermined values in advance, but they may be the same number, such as setting each value to 1, for example.

In the above equation (18), the first energy element EG is the _G direction of the robot 10 at the current position SP, based on the fact that the robot 10 continues to move at a constant speed is the behavior with the lowest trajectory cost. is an element corresponding to the speed change due to the acceleration/deceleration motion of the robot 10 in the same direction with respect to the initial speed of .

The second energy element ED is an element that corresponds to the speed change in the direction _D from the previous point, based on the fact that the larger the avoidance width when the robot 10 avoids interference, the worse the movement efficiency becomes. ing.

According to the optimum trajectory extracting unit 45 described above, the best execution trajectory that is directed toward the destination GP in a constant speed state from the steady state before the avoidance action and that does not deviate significantly laterally from the destination GP. can be quantitatively extracted from each trajectory candidate.

In the trajectory candidate specifying function 50 of the trajectory candidate search unit 44, the trajectory candidates are further narrowed down as follows, thereby further reducing the computational load during cost calculation in the optimal trajectory extraction unit 45. can be done.

As for the narrowing down, the following trajectories are excluded from the trajectory candidates to be calculated by the optimum trajectory extraction unit 45 . For example, as shown by the dashed line in FIG. 16A, a trajectory in which the robot 10 avoids the moving human H while decelerating in front of it increases the avoidance width, which increases the trajectory cost. is clear and is excluded from the orbital candidates. Also, as indicated by the dashed line in the figure, a trajectory in which the robot 10 avoids the moving human H while accelerating behind it is also excluded from the trajectory candidates because the avoidance width is similarly large. .

In addition, as illustrated in FIGS. 16B and 16C, the trajectory in which the robot 10 moves so as to stick to the human H increases the trajectory cost, so it is excluded from the trajectory candidates. . That is, the positions of WP1 to WP3 are aligned with the avoidance direction D of the robot 10 with respect to the human H, with the straight line TL connecting the current position SP of the robot 10 and the destination GP in the target traveling direction (see the dashed-dotted line in each figure) as a boundary. A trajectory existing in the area on the opposite side is excluded from the trajectory candidates. In other words, the trajectory of FIG. 16B is a trajectory that avoids the human H from the left side, while the positions of WP1 to WP3 are located in the area on the right side of the straight line TL, and is excluded from the trajectory candidates. . The trajectory shown in FIG. 4C avoids the human H from the right side, while the positions of WP1 to WP3 are located in the area on the left side of the straight line TL, and is excluded from the trajectory candidates.

1-C Repetitive Action Search System in Execution Trajectory Identification Unit 38

The execution trajectory identification unit 38 may be provided with a repetitive action search system that repeats the above-described procedure to generate trajectories when there is a possibility that the robot 10 will interfere with a plurality of humans H in the future. . In this iterative action search system, the above-described trajectory generation procedure in the trajectory candidate search unit 44 is used to generate an avoidance trajectory for a person who is predicted to interfere with the robot 10 first (hereinafter referred to as "shortest interfering candidate"). is searched for, then avoidance trajectories for other humans whose interference is expected next are searched. Then, the trajectory search is repeatedly performed in the order of interference time for all humans expected to interfere with the robot 10 at the current position in the future, and trajectory candidates capable of avoiding all these humans are obtained.

Specifically, first, the situation around the robot 10 at the current position is detected by the detection device 12 (see FIG. 1), and future interference is predicted by the interference degree determination means 21 (see FIG. 1). Multiple people are identified. Among them, the shortest interfering prospective person whose relative distance to the robot 10 is the shortest and who is predicted to interfere with the robot 10 at the fastest is specified.

After that, in the trajectory candidate search unit 44, trajectory candidates are generated by the procedure shown in FIG.

That is, first, as shown in FIG. 2A, the passing point is determined for the shortest interfering prospective person H1 from the current position SP of the robot ₁₀ by the same procedure as described above, and the passing point is determined. Various trajectories (constant velocity trajectory and acceleration trajectory) Kn up to WP3 (H ₁ ), which is the rear neighboring point, are generated. In addition, in each figure of FIG. 17, display of all trajectory candidates is omitted in order to avoid complication of the drawings.

Next, assuming that the robot 10 reaches WP3 (H ₁ ) on various trajectories up to WP3 (H ₁ ), and after that, assume that WP3 (H ₁ ) is the initial position of the robot 10, and similarly: Interference determination up to the destination GP is performed. As a result, when _another human H2 with the possibility of interference is detected, as shown in _FIG . Identification of the shortest interfering prospective person _H2 to be performed is made. On top of that, as shown in FIG. 1(C), with the previous WP3 (H ₁ ) as the initial position, similarly, various trajectories up to WP3 (H ₂ ) as trajectories that can avoid the shortest interfering candidate H ₂ A trajectory Kn is generated.

Here, as shown in FIG. 3C, if the avoidance trajectory for the human H ₂ at the back is planned in the previous procedure, when the robot 10 reaches WP3 (H ₁ ), the human H ₂ at the back will also reach WP3 (H 1 ). H1), and interference avoidance may become impossible as shown in (D) of the figure. Therefore, in order to avoid such a situation, a trajectory correction is performed as exemplified by the one-dot chain line portion in FIG. This trajectory correction is performed when the direction to avoid is the same for the _first person H1 and the _second person H2, whose interference is expected at a later time (in the same figure, the trajectory portion where each person avoids to the right) ). That is, here, from WP1 (H ₁ ), which is the front vicinity point among the passing points of the trajectory for avoiding the first person H ₁ , to the front vicinity of the trajectory for avoiding the second person H ₂ , A partially corrected trajectory is generated that includes a portion KA _n that moves linearly to the point WP1 (H ₂ ).

After that, when there is another human Hn who may interfere with the destination GP, the same process is repeated to connect _WP3 ( _{Hn )} of the last human Hn and the destination GP. , a route from the current position SP of the robot 10 to the destination GP is determined, and a trajectory candidate for the route is generated. The optimal trajectory extracting unit 45 selects the trajectory candidate with the lowest trajectory cost from among the plurality of patterns of trajectory candidates generated in this way, and makes it the final execution trajectory.

According to such a repetitive action search system, when the current robot 10 avoids multiple humans who may interfere with each other in the future, smooth trajectory planning can be performed without causing chattering or getting stuck.

In the process of the iterative action search system described above, a trajectory candidate is generated by the above-described procedure so that a plurality of people who are close to each other are clustered and the cluster is collectively avoided as one person. can also By doing so, it is possible to reduce the load of arithmetic processing at the time of trajectory search.

That is, here, the relative distance between the shortest interfering prospective person whose interference is predicted at a predetermined time and the other person detected by the detecting device 12 is within a predetermined range, that is, the relative distance is equal to or less than a predetermined threshold. A trajectory that avoids the cluster as a single person is generated by estimating these multiple people as a cluster. As the threshold here, a space width (for example, its lowest value) that allows the robot 10 to pass between adjacent humans without interference is applied.

2. Configuration and Processing Details of Narrow Space Action Determining Section The configuration and processing content of the narrow space action determining section 35 will be described below mainly with reference to FIGS. 1 and 6. FIG.

In the small space action determination unit 35, it is predicted that the robot 10 cannot pass by the human at present, and as will be described later, the action of the robot 10, including various actions, is determined according to various conditions. selected.

Here, the approach action includes an action of suggesting a route to the human by changing the moving direction of the robot 10 to avoid interference with the human, and a voice prompting the human to avoid from the speaker 16. and an action of bringing the arm 15 into contact with a human being, and any one or any combination of these action actions is selected.

In addition, as the approach action here, when it is determined that there is a degree of interference by the interference degree determination means 21, according to the recognition level estimated by the recognition level estimation means 22, first, the action action for avoiding interference Re-encouragement actions to be performed as necessary based on the achievement level of the action actions estimated by the achievement level estimating means 24 described later from the first action action performed and the human behavior after the first action action and It is divided into The details of the above appeal behavior will be described later.

This narrow space behavior determination unit 35 includes an arm contact determination unit 55 that determines whether there is a possibility of contact of the arm 15 with a human being, and a remote response unit that determines the action to be taken when the possibility of contact is "no". 56, and a proximity corresponding unit 57 that determines the action to be taken when the possibility of contact is "present".

The arm contact determination unit 55 determines whether or not there is a possibility that the arm 15 will come into contact with a human depending on whether or not a human exists within the movement range of the arm 15 . Whether or not there is a possibility of contact is determined based on the relative positional information between the robot 10 and the human based on the detection result of the detection device 12 and the preset movement range of the arm 15 .

The long-distance support unit 56 determines a remote action group consisting of the following three types of action actions according to the degree of recognition.

In this long-distance action group, when the degree of recognition is "high", the action action of suggesting a course to the human by changing the moving direction of the robot 10 is selected as follows. Specifically, as shown in FIG. 18, the width of the passage space S for the robot 10 to pass through is not sufficiently secured, so the robot 10 is moved toward the central point P of the passage space S. As a result, the route is visually suggested to the human H so that the passing space S for the robot 10 to pass through can be secured.

In the distant working group, when the degree of recognition is "low", the same behavior of suggesting a career path as in the case of the degree of recognition of "high" and the above-mentioned low level of intervention voice "small" are performed. A working action is selected.

Furthermore, in the distant working group, if the degree of recognition is "no", the same action that suggests a career path as in the case of the degree of recognition of "high" and the above-mentioned voice with a strong degree of intervention "high" A working action is selected.

In the proximity handling unit 57, various actions are performed in the following procedure. First, in advance, in order to notify the human H of the approach of the robot 10 from the auditory aspect, the action of calling out a predetermined sound from the speaker 16 is selected.

Then, from among the proximity action groups composed of the following three types of action actions, actions including action actions corresponding to the degree of recognition are determined.

In this proximity action group, when the degree of recognition is "high", the moving direction of the robot 10 is changed as follows. In this case, it is considered that the human H is sufficiently aware of the robot 10 because the voice has been made in advance, and the human H secures the width of the space S through which the robot 10 passes. It is presumed that it is impossible to move for the purpose. Therefore, in this case, for example, as indicated by the dashed-dotted line arrow in FIG. A turning action is selected to change direction. At this time, the robot 10 searches for a new movement route to the destination.

In addition, in the proximity working group, when the degree of recognition is "low", by bringing the arm 15 of the robot 10 into contact with the human H, contact is made to encourage the movement of the human H so as to secure the passing space S. A "small" appealing action is selected.

Furthermore, in the proximity working group, when the degree of recognition is "none", the working action of "large" contact with a stronger degree of intervention than when the degree of recognition is "low" is selected.

These contact-based actions include, for example, the action of pushing the shoulders and back of the human H with the arm 15. With the action of "small" contact, the human H is lightly touched, and with the action of "large" contact, the human H , repeatedly contacting the human H, or contacting the human H so as to receive an unexpected collision when approaching.

(Achievement level estimation means)
The achievement degree estimating means 24 estimates the degree of achievement of conveying the intention of the action to the human by verifying the action state of the human by the action action of the robot 10 from the detection result of the detection device 12 .

Specifically, the achievement level estimation means 24 estimates whether or not the action of the robot 10 based on the action action that has already been performed has been transmitted to the human as an intention to avoid interference between the robot 10 and the human in the future. That is, here, the movement information of the human after the robot 10's working motion is acquired based on the detection result of the detection device 12, and the degree of achievement is specified from the movement information.

The achievement level estimation means 24 includes a first achievement level determination section 59 that determines the level of achievement by the action of the robot 10 planned by the large space action determination section 34, and the narrow space action determination section 35. and a second achievement degree determination unit 60 that determines the achievement degree by the action of the robot 10 that has been determined.

In the first achievement degree determining unit 59, the relative distance, which is the straight line distance between the mutual trunk centers at the time when the human will be closest to the robot 10 in the future, is calculated, and based on the relative distance, , as follows, the degree of achievement is identified by three types of "yes", "no", and "very much". Here, the first evaluation, the degree of achievement "yes", is a case where the intention is presumed to be effectively communicated to a human, and the second evaluation, the degree of achievement "no", is when the intention is This is the case where it is estimated that the intention was not communicated to the human very much, and the third evaluation, the degree of achievement "excessive", is the case where it is estimated that the intention was communicated to the human more strongly than the degree of achievement "existent". is.

That is, here, when the relative distance at the time of closest approach is smaller than the sum of the personal spaces obtained by adding the radii of the personal spaces of the robot 10 and the human, the state of possibility of future interference is eliminated. Therefore, it is determined that there is a need to reconstruct the motion plan of the robot 10, and the degree of achievement is set to "none". On the other hand, when the relative distance is equal to or greater than the predetermined value exceeding the sum of the personal spaces, it is presumed that the person is avoiding the object more than expected, and the degree of achievement is set to "extraordinary". Here, the predetermined value is set in advance, and can be exemplified by 1.5 times the sum of the personal spaces. Furthermore, when the relative distance is equal to or greater than the sum of the personal spaces and is less than the predetermined value, it is estimated that the human is avoiding the robot as expected, and the achievement level is determined to be "yes", and the operation plan of the robot 10 is changed. It is determined that there is no need to reconfigure.

The second achievement degree determination unit 60 determines the achievement degree as follows based on the presence or absence of the space S through which the robot 10 passes. The achievement levels “yes” and “no” here are the same as the concept of the achievement levels “yes” and “no” in the first achievement level determination unit 59 .

That is, here, the existence or non-existence of the passage space S for the robot 10 is determined by the processing in the passage space determination unit 31 based on the movement information of the person after the action action acquired by the detection device 12 . Then, when the passage space S for the robot 10 is ensured, the degree of achievement is determined to be "existing", and it is determined that there is no need for further work or action. On the other hand, if the passage space S for the robot 10 has not yet been secured, the degree of achievement is set to "none" and it is determined that there is a need for another action.

(Plan adjustment means)
The plan adjustment means 25 includes a first adjustment section 62 that determines the action of the robot 10 according to the achievement level determined by the first achievement level determination section 59, and the second achievement level determination section 60. and a second adjustment unit 63 that determines the action of the robot 10 according to the determined degree of achievement.

The first adjustment unit 62 determines the behavior of the robot 10 after the achievement level is specified as follows.

In other words, when the degree of achievement is "no", it is necessary to reconstruct the action plan of the robot 10. Therefore, in the same manner as described above, the action plan means 23 again determines the pass through based on the degree of recognition. Various action plans are made according to the presence or absence of space.

In addition, in the case of the achievement level "extraordinary", it is estimated that humans are avoiding more than expected, so a new trajectory that is straighter toward the destination is generated considering the trajectory cost. be.

Furthermore, if the degree of achievement is "yes", it is estimated that the human being is avoiding as expected, so the movement on the previously specified execution trajectory is left as it is without changing the action plan that has already been decided. Continued.

The second adjustment unit 63 determines the behavior of the robot 10 after the achievement level is specified as follows.

Here, if the degree of achievement is "none", the next appeal action is selected in accordance with the previous action group.

When the remote working group performs the working action in the previous stage, the arm contact determination unit 55 determines whether there is a possibility of the arm 15 coming into contact with the human being, and thereafter, the narrow space action determining unit 35 performs the work. action is taken.

In addition, when the approach action was performed in the proximity action group in the previous stage, it was determined that the passage space for the robot 10 to pass through could not be secured, and a trajectory for moving the robot 10 in a completely different direction was generated. , a reorientation movement along the trajectory is made. In addition, when the direction change movement has already been performed in the proximity action group in the previous stage, the movement is continued as it is.

On the other hand, if the degree of achievement is "yes", the passage space for the robot 10 has been secured, so the movement along the already determined trajectory is continued.

(Action plan procedure for avoiding interference)
The action planning procedure of the robot 10 by the control device 13 will be described below mainly using the flow charts of FIGS. 1, 6 and 19. FIG.

First, after the robot 10 starts moving from a predetermined starting point, the degree-of-interference determining means 21 determines whether or not there is a possibility of future interference between the robot 10 and humans moving around it. is determined (step S301). Then, the action planning means 23 makes an action plan for the robot 10 according to the presence or absence of the degree of interference.

Here, when the degree of interference is "no", the non-interference planning unit 28 moves the robot to a preset destination while avoiding fixed moving objects such as walls without considering the avoidance of interference with humans. A route plan is made to move 10 in the shortest distance, and an execution trajectory corresponding to the plan is generated (step S302). Then, the processes after the determination of the degree of interference are performed at fixed time intervals until the destination is reached (step S303).

On the other hand, if the degree of interference is "yes", an action plan for avoiding future interference between the robot 10 and a human is made in the interference planning section 29 according to various situations.

When performing an action plan for interference avoidance, first, the recognition level estimation means 22 determines the recognition level, which indicates the degree of human recognition of the robot 10, as "high", "low", or "none" ( step S304).

After that, the passage space determining unit 31 determines whether or not there is a passage space when the robot 10 passes by the human (step S305). Then, different action plans are made depending on the presence or absence of the passing space.

A. A case judged to have a passing space

In this case, the following various types of processing are performed in the large space action determination unit 34 .

First, the degree of concession is obtained by the action content identification unit 37 (step S306).

Next, the trajectory candidate search unit 44 generates a plurality of patterns of trajectory candidates for avoiding interference with humans according to the degree of concession obtained previously (step S307).

Then, the optimum trajectory extraction unit 45 determines, as the execution trajectory, a trajectory that enables efficient movement with the lowest trajectory cost from among the plurality of patterns of trajectory candidates generated by the trajectory candidate search unit 44 (step S308).

After that, in consideration of the operation time for generating the execution trajectory, the response time from the start of movement of the robot 10 on the execution trajectory to the action of the human, etc., a predetermined waiting time is provided. The next process is not performed until the time has elapsed (step S309).

After the waiting time has elapsed, the achievement level is estimated by the achievement level estimating means 24 (step S310). Depending on the results, the plan adjustment means 25 determines whether or not the action details need to be adjusted. That is, if the achievement level is "none", the process returns to step S304 to determine the degree of recognition, and the processes after step S305 are performed again.

In addition, in the case of the degree of achievement "extraordinary", the execution trajectory is corrected so that it becomes straighter toward the destination (step S311).

Further, if the degree of achievement is "yes", the already determined execution trajectory is left as it is, and the processes after the determination of the degree of interference are performed at regular time intervals until the destination is reached (step S303).

B. Cases determined as "no passing space"

In this case, the following various processes are performed by the narrow space action determination unit 35 .
First, the arm contact determination unit 55 determines whether or not there is a possibility that the arm 15 will contact a human at the present time (step S312).

B-1. In the case of contact possibility "no"

If the possibility of contact with the arm 15 is "no", the action is selected from the remote action group according to the degree of recognition (step S313).

That is, in the remote working group, when the degree of recognition is "high", a trajectory for changing the moving direction of the robot 10 is generated, and the human is instructed to secure a space for the robot 10 to pass through. Engaging behaviors that provide visual course suggestions are selected.

In addition, in the remote working group, when the degree of recognition is "low", the same action that suggests a course is selected as in the case of the degree of recognition of "high". ” is selected.

Furthermore, in the distant working group, when the degree of recognition is "no", the same behavior that suggests a career path is selected as in the case of the degree of recognition of "high". ” is selected.

Thereafter, the achievement level is determined by the achievement level estimation means 24 (step S314). Here, based on the position information of the person after the action action acquired by the detection device 12, the presence or absence of the space through which the robot 10 passes is determined by the same processing as in step S305 described above. Therefore, when a space for the robot 10 to pass through is ensured, the degree of achievement is set to "yes", and it is determined that there is no need for further interference avoidance action. On the other hand, if the space through which the robot 10 can pass is still not secured, the degree of achievement is set to "none", and it is determined that there is a need for further interference avoidance action.

Then, the plan adjusting means 25 takes the following measures. That is, when the degree of achievement is "yes", the already determined execution trajectory is left as it is, and the processing after the determination of the degree of interference is performed at regular time intervals until the destination is reached (step S303). On the other hand, if the achievement level is "none", the process returns to step S312, and the subsequent processes are repeated.

B-2. If contact possibility is "Yes"

In step S312, if it is determined that there is a possibility of contact with the arm 15, first, a predetermined sound is played from the speaker to notify the human of the approach of the robot 10 from the auditory perspective. 16 is performed (step S315). Then, various actions are selected from the proximity action group according to the degree of recognition (step S316).

In the proximity working group, when the degree of recognition is "high", an action is selected in which the robot 10 retreats from the human and turns the robot 10 in a completely different direction. A new trajectory aiming at the destination is generated while changing.

Further, when the degree of recognition is "low" in the proximity action group, the arm 15 of the robot 10 is brought into contact with the human to encourage the movement of the human so as to secure the passing space. is selected.

Furthermore, in the proximity working group, when the degree of recognition is "none", the working action of "large" contact with a stronger degree of intervention than when the degree of recognition is "low" is selected.

Thereafter, the achievement level is determined by the achievement level estimating means 24 (step S317). Here, based on the position information of the person acquired by the detection device 12, it is determined whether or not there is a space for the robot 10 to pass through by the same processing as in step S305 described above. Therefore, when a space for the robot 10 to pass through is secured, the degree of achievement is set to "yes", and it is determined that there is no need for further action. On the other hand, if the passage space for the robot 10 has not yet been secured, the degree of achievement is set to "no", and it is determined that there is a need for further action.

Then, the plan adjusting means 25 takes the following measures. That is, when the degree of achievement is "yes", the already determined execution trajectory is left as it is, and the processing after the determination of the degree of interference is performed at regular time intervals until the destination is reached (step S303). On the other hand, if the degree of achievement is "None", it is determined that the passage space for the robot 10 to pass through cannot be secured, and a direction change trajectory is generated to move the robot 10 in a completely different direction (step S318). Thereafter, the process returns to the determination of the degree of interference by the degree-of-interference determining means 21 (step S301), and the action plan of the robot 10 is made through the same procedure as described above.

As described above, according to the present embodiment, it is possible for the robot 10 to move in a manner similar to that of a human being, such as when humans pass each other. Moreover, the robot 10 can be efficiently moved toward the destination.

It should be noted that the control device 13 collects sound information such as sound pressure in the environment around the robot 10 with a microphone (not shown) in the above-described behavior of calling out, and considers the surrounding environment according to the sound information. It is also possible to perform sound control in which the loudspeaker 16 emits a sound such as the volume and type of the sound produced.

In addition, in the control device 13, when the arm 15 is in contact with a human being, a sensor that acquires environmental information around the robot 10, such as soft contact when the surroundings of the human being are crowded, is detected. Depending on the results, motion control may be performed to vary the degree of contact with humans.

In addition, the types of actions described above are examples, and various other action actions can be set by the actions of the robot 10 .

Further, in the above-described embodiment, the "degree of interference" is set in two stages, the "level of recognition" is set in three stages, and the "level of achievement" is set in two stages. It is also possible to evaluate with , and set more detailed actions for each stage.

Next, another embodiment of the present invention 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 of the first embodiment, and the description will be omitted or simplified.
(Second embodiment)

The robot 100 according to the second embodiment assumes that the robot 100 can pass through the area inside the personal space of a person, which is a movement obstacle, and that an unexpected collision between the robot 100 and a person can occur while taking into account the alleviation of the discomfort given to the person. It is characterized by being able to perform avoidable trajectory planning.

That is, as shown in FIG. 20, the robot 100 of the present embodiment includes an interference planning section 103 having a configuration different from that of the interference planning section 29 of the first embodiment. It is substantially the same as one embodiment.

When the interference degree determining means 21 determines that the future degree of interference is "presence", the interference planning section 103 plans the behavior of the robot 100 to avoid the human expected to interfere. .

As shown in FIG. 20, this collision time planning unit 103 includes an approaching state determination unit 105 that determines whether the robot 100 at the current position is in the human's personal space, a contact possibility determination unit 106 that determines whether or not there is a possibility that the robot 100 will make physical contact with the human trunk; An emergency avoidance action selection unit 108 that selects an emergency avoidance action that is an action of the robot 100 that performs avoidance, and a normal avoidance action selection unit that selects a normal avoidance action that is an action of the robot 100 that performs collision avoidance in a non-emergency situation. 109.

(Configuration of approaching state determination unit)
In the approaching state determination unit 105, based on the detection result of the detection device 12, the range of the personal space of the human being is specified at the present time in the same manner as in the first embodiment. It is determined whether or not the part is encroaching. That is, if the personal space of the robot 100 is currently within the range of the personal space of a person, it is in the approaching state, and if not, it is in the non-approaching state.

(Configuration of Contact Possibility Determining Unit)
In the contact possibility determination unit 106, based on the detection result of the detection device 12, the relative distance when the robot 100 comes closest to the human is obtained in the same manner as in the first embodiment. , it is determined whether or not there is a possibility of trunk contact in which the trunk portion of the robot 100 contacts a human. That is, here, when the relative distance is shorter than the sum of the trunk widths of the robot 100 and the human trunk specified in advance, the possibility of trunk contact is determined to be "present"; The possibility of the trunk contact is set to "no".

(Configuration of preparation time determination unit)
The preparation time determining unit 107 determines whether or not there is preparation time for working as follows. Here, as in the first embodiment, the relative distance between the robot 100 and the human and their velocity vectors are specified from the detection result of the detection device 12 at the present time, so that they interfere with each other. time is calculated. Then, the time is compared with the action preparation time set in advance for various actions (arm 15, call action, etc.) required for the action action, etc., and if the time is shorter than the action preparation time, the preparation time If the preparation time is longer than the action preparation time, the preparation time is set as "presence".

(Configuration of emergency avoidance action selection unit)
The emergency avoidance action selection unit 108 determines whether the robot 100 is currently in an approaching state in which the personal space of the robot 100 exists within a human's personal space, or in a non-approaching state, and the preparation time for the action action is "none". , a predetermined action is selected from various emergency avoidance actions described later.

The emergency avoidance action selection unit 108 includes an approach trajectory generation function 111 that generates an approach trajectory, which is a trajectory in an approach route that allows the robot 100 to enter a human's personal space and advance to the destination without contacting the human. , a reverse trajectory generation function 112 for generating a reverse trajectory, which is a trajectory in a reverse movement route for advancing the robot 100 in the opposite direction, and a temporary stop function 113 for temporarily stopping the robot 100 .

In the approach trajectory generation function 111, the approach trajectory is generated in each of the following cases based on the determination results of the approach state determination unit 105, the contact possibility determination unit 106, and the preparation time determination unit 107. That is, as the first case, when the robot 100 is currently in a close state in a human's personal space, the possibility of contact between the trunks is "no" and the robot 100 has been in some way. For example, the case of taking action to work. In the second case, when the robot 100 is in the non-approaching state, there is no preparation time for the action and the robot 100 has already taken any action.

The approach trajectory generation function 111 generates an approach trajectory in which the robot 100 moves at a constant speed by the same arithmetic processing as the constant speed trajectory generation function 48 (see FIG. 6) in the first embodiment. Specifically, here, the concession degree P _R in the above equation (17) in the constant speed trajectory generation function 48 is set to 100%. In addition, when the reference distance L _b substituted into the above equation (17) is set to R _LR +R _LH , which is the sum of the radii of the personal spaces of the robot 10 and the human, it is shorter than R _LR +R _LH and A predetermined distance longer than the width of the trunk. As a result, an approach trajectory is generated that allows the robot 100 to pass the human without contacting the human while the robot 100 is intruding into the human's personal space. As exemplified in FIG. 21, this approach trajectory is selected from the two types of approach trajectories PT ₁ so as to allow the human H to pass each other on both the left and right sides, and the trajectory with the smaller movement distance is selected. be. If the robot 100 cannot pass through _{one of the approach routes PT-1} due to the presence of a fixed obstacle or the like, the approach trajectory of the other approach route PT- ₁ is selected. If the robot 100 cannot pass through any of the approach trajectories, either movement on the reverse trajectory by the reverse trajectory generation function 112 or temporary stop by the temporary stop function 113 is selected. .

The reverse trajectory generation function 112 directs the robot in the opposite direction away from the human in each of the following cases based on the determination results of the approach state determination unit 105, the contact possibility determination unit 106, and the preparation time determination unit 107. One type of backward trajectory for conversion movement is generated. That is, as the first case here, when the robot 100 is currently in a close state in which the robot 100 exists in a human's personal space, the possibility of contact between the trunks thereof is "no" and A case where no action is taken from 100 is given. In addition, as a second case, there is a case where the possibility of contact between the trunks is "present" in the approach state. Furthermore, as a third case, there is a case where the preparation time for the action action is "none" in the non-approaching state and the robot 100 has not made any action action so far.

In the temporary stop function 113, the reverse trajectory generated by the reverse trajectory generation function 112 is used when the robot 100 cannot move due to the presence of fixed obstacles on the path, etc. If it is impossible to avoid contact with the trunk of the robot 100, an action to temporarily stop the robot 100 is selected. After that, while the robot 100 is being moved, the various processes in the non-interference time passage unit 28 (see FIG. 1) or the interference time passage unit 103 are performed again.

(Configuration of normal avoidance action selection unit)
In the normal avoidance action selection unit 109, when the personal space of the robot 100 is currently outside the personal space of the human being and the preparation time for the action action is "yes", various normal actions to be described later are performed. A predetermined action is selected from the avoidance actions. This normal avoidance action includes an action of moving the robot 100 along any one of an approach trajectory, a detour trajectory, and a contact trajectory, accompanied by an action action by the robot 100 .

Here, as shown in FIG. 21, as described above, the approach trajectory based on the approach path PT1 is intended for the robot ₁₀₀ to enter the human's personal space PSH, and the robot 100 is intended to enter the human's _H space. This is the trajectory for advancing the robot 100 to the destination GP while passing by the human H at the minimum distance that does not contact the . When the robot 10 moves on this approach trajectory, the arm 15 is set up in preparation for an unexpected collision with the human H, and the robot 100 also makes an action of calling out to the human H that the robot 100 is approaching.

Also, the detour trajectory is a trajectory based on the detour route PT2 in _FIG . In other words, this detour route PT ₂ is such that the personal space PS R of the robot 100 and the personal space PS _R of the human _H are not overlapped by the personal space PS _R of the robot 100 and the personal space PS H of the human H. This is the path that the robot 100 circumnavigates outside of the human _{H's personal space PS H ,} intended not to intrude on H. At the time of movement on this detour trajectory, an appealing action based on _assertive action with a concession degree PR of 50% as described in the first embodiment is performed.

Further, the contact trajectory is a trajectory based _on the contact path PT3 of FIG. This contact path PT3 moves the human H by pushing the human H away from the robot 100 while bringing the arm 15 into contact with the human H _, and the robot 100 passes by the human H after the movement, while the robot 100 moves toward the target. This is the path for advancing the robot 100 to the ground. When the robot 100 moves along the contact trajectory, the robot 100 makes an approach action to the human H and calls out to inform the human H of the planned contact, etc., and the arm 15 is used to press the human H in a predetermined manner. An action is performed to urge the person H to move by applying pressure. Note that the contact trajectory here is intended to move the human H by a predetermined amount in the direction opposite to the robot 100, but when the reaction force of the arm 15 measured by a sensor (not shown) exceeds a predetermined value, may determine that the human H is immobile, and take actions such as temporarily stopping the movement of the robot 100 .

Specifically, the normal avoidance action selection unit 109 includes a detour decision unit 115 that determines whether the detour route PT2 can be set based _on the relative positional relationship between the robot 100 and the human H. A pass-through space determination unit 116 that determines whether or not there is a pass-through width, and a trajectory candidate search that searches for trajectory candidates for advancing the robot 100 toward the destination from detour trajectories, approach trajectories, and/or contact trajectories. an optimal trajectory extraction unit 118 that selects the optimal trajectory from among the plurality of generated trajectory candidates and sets it as an execution trajectory for actually moving the robot 10; and an action content identification unit 119 that identifies the action.

The detour availability determination unit 115 determines whether or not there is a relative distance at which the detour route PT2 can be physically planned based _on the current relative positional relationship between the robot 100 and the human. That is, first, each passing point (WP1 to WP3) is obtained by the same arithmetic processing as that of the trajectory candidate searching section 44 in the first embodiment. Then, regardless of the existence or non-existence of fixed obstacles, is it physically possible in terms of the performance of the robot 100, which is specified in advance, to change the direction and adjust the speed of the robot to reach each passing point? No is determined. As a result, when the same is possible, the detour is “permitted”, and when the same is not possible, the detour is “impossible”.

The pass-through space determination unit 116 determines whether or not there is a pass-through width, which is the width for the robot 100 to actually pass through the side of the human, in the same manner as the pass-through space determination unit 31 of the first embodiment. . That is, when the robot 100 comes closest to the human, if the space next to the human has a width equal to or greater than the pass-through width, the pass-through space is "present", and otherwise, the pass-through space is "absent". It is said that Here, as the pass-through width, a constant value obtained by adding the spare allowance to the trunk width of the robot 100 is set in advance.

The trajectory candidate searching unit 117 includes a detour trajectory generation function 121 that generates the detour trajectory, an approach trajectory generation function 122 that generates the approach trajectory, a contact trajectory generation function 123 that generates the contact trajectory, and these functions. For each trajectory generated in steps 121 to 123, a proximity energy calculation function 124 that obtains a proximity energy as an indicator of approach risk, which is the risk when the robot 100 approaches a human, and a trajectory candidate is determined based on approach risk determination by proximity energy. and a trajectory candidate specifying function 125 to specify.

In the detour trajectory generation function 121, the degree of concession P _R described in the first embodiment is set to 50%, and determined in the same manner as the processing in the trajectory candidate search unit 44 and the optimum trajectory extraction unit 45 in the first embodiment. One trajectory is set as a trajectory candidate for the detour trajectory. Further, here, a detour trajectory for comparison, which is a reference detour trajectory for use in approach risk determination by the trajectory candidate specifying function 125, which will be described later, is also obtained. This detour trajectory for comparison is specified by the same procedure, assuming that there is no fixed obstacle such as a wall obstructing the passage of the robot 100 _on the detour route PT2. Depending on the environment around the robot 100, the detour trajectory for comparison may be the same as the detour trajectory generated in consideration of the existence of fixed obstacles.

The approach trajectory generation function 122 generates an approach trajectory in the same manner as the approach trajectory generation function 111 of the emergency avoidance action selection unit 108. Here, a plurality of patterns of approach trajectories with different moving speeds of the robot 100 are generated. generated.

The contact trajectory generation function 123 generates a contact trajectory along which the robot 100 moves at a steady speed by the same arithmetic processing as the constant speed trajectory generation function 48 in the first embodiment. Specifically, here, the _concession degree P _R in the above equation (17) in the constant speed trajectory generation function 48 is set to 100%, and the robot A minimum required distance is set so that the robot does not come into contact with the human at a position where the human is assumed to have moved by a certain target amount of movement when the human is moved by 100 contacts. Also, here, a plurality of patterns of contact trajectories with different moving speeds of the robot 100 are generated.

The proximity energy calculation function 124 obtains the proximity energy E _l for each trajectory obtained by each of the trajectory generation functions 121 to 123 by the following equation. Here, it is estimated that the approach risk is higher as the proximity energy _El is larger.

In the above equation, A _c is a variable whose parameter is the degree of recognition specified according to the degree of recognition obtained by the recognition degree estimation means 22 in the same manner as in the first embodiment. and when the degree of recognition is "low" and "none". These values are set in advance, and when the degree of recognition is "small" or "no", it is estimated that the approach risk is higher than when the degree of recognition is "large", and a large value is set. .

Also, P _n in the above equation is a variable that takes as a parameter the difficulty of approaching a person who is expected to interfere in the future. In other words, a predetermined value is assigned according to attributes such as the age of the person concerned and situations such as the person holding something. This variable _Pn is assigned a predetermined value according to a predetermined rule based on detection by a sensor or the like (not shown), or may be specified by inputting an arbitrary value from a remote location.

Furthermore, _Dr in the above equation is a variable whose parameter is the orientation of the human relative to the robot 100 when the robot 100 is closest to the human. Here, the orientation of the human with respect to the robot 100 at the time of closest approach is specified from the orientation of the human face detected by the detection device 12, and the variable _Dr is calculated. This variable _Dr is set in stages according to the relative angle representing the orientation of the human face with respect to the robot 100, from the state in which the robot 100 exists in the front position in front of the human to the state in which the robot 100 exists in the rear position behind the human. An increasing value is identified. That is, the variable _Dr is set to gradually increase as the robot 100 moves from the state in which the robot 100 exists directly in front of the human face to the state in which the robot 100 exists from the side to the rear of the human face. Therefore, as the variable _Dr increases, the approach risk increases.

Also, d is the relative distance between the robot 100 and the human when they are closest to each other, which is specified based on each trajectory of the robot 100 . Further, _vR is the component velocity of the robot 100 toward the human when the robot 100 and the human are closest to each other, and _vH is the component velocity of the human toward the robot 100 when the robot 100 and the human are closest to each other. Note that _k1 and k2 _are constant coefficients set in advance.

The proximity energy calculation function 124 described above _obtains the proximity energy El for risk determination for each of the detour trajectory for comparison, which is a virtual detour trajectory, and the approach trajectory and contact trajectory of each pattern.

The trajectory candidate specifying function 125 extracts an approaching trajectory and a contacting trajectory having the same or equivalent value as the _proximity energy El in the reference detour _trajectory for comparison with respect to the approaching trajectory and the contacting trajectory obtained respectively. and is used as a trajectory candidate for the approach trajectory and the contact trajectory. That is, here, the approach trajectory and the contact trajectory having the same or the same approach risk as the detour trajectory for comparison are specified as trajectory candidates. As for the detour trajectory, the detour trajectory generation function 121 determines one trajectory as a trajectory candidate in consideration of the presence of fixed obstacles in the same manner as in the first embodiment.

The trajectory candidate searching unit 117 configured as described above generates trajectory candidates of different types according to the following conditions. That is, when the detour decision unit 115 determines that the detour is “possible” and the pass-through space determination unit 116 determines that the pass-through space is “present,” the trajectory candidates for each of the detour trajectory, the approach trajectory, and the contact trajectory is generated. If the detour decision unit 115 determines that the detour is “impossible” and the passing space determination unit 116 determines that the pass-through space is “present,” trajectory candidates for the approach trajectory and the contact trajectory are generated. Furthermore, regardless of whether detouring is permitted by detour determination section 115, if passage space determination section 116 determines that there is no passage space, only a trajectory candidate for the contact trajectory is generated.

In the optimum trajectory extraction unit 118, for each of the plurality of trajectory candidates, in the same manner as in the optimum trajectory extraction unit 45 (see FIG. 6) of the first embodiment, the movement energy corresponding to the interference avoidance energy _Er of the first embodiment is calculated. E _r is calculated, and the trajectory candidate with the lowest transfer energy E _r is extracted as the optimal execution trajectory with the lowest trajectory cost.

That is, here, the transfer energy _Er for the detour trajectory and the approach trajectory is obtained by the same cost formula as the above formula (18) used in the first embodiment.

なお、上式において、ｒは、ロボット１００の接触により、人間をロボット１００の反対側に移動させる所望量であり、予め一定値に設定される。Ｃは、当該エネルギー要素における重み付け定数であり、予め一定値に設定され、例えば、上式（１８）における他のエネルギー要素の重み付定数Ａ，Ｂよりも大きい値が設定される。 On the other hand, the movement energy E _r for the contact trajectory is as shown in the following equation (20), with the cost C r, which is a parameter corresponding to the amount of movement of the human due to the contact of the robot 100, as the third energy element, the above equation ( It is obtained by further adding to the first and second energy elements _EG and _ED of 18).

In the above formula, r is a desired amount of movement of the human to the opposite side of the robot 100 by contact with the robot 100, and is set to a constant value in advance. C is a weighting constant for the energy element and is set to a constant value in advance, for example, a value larger than the weighting constants A and B of the other energy elements in the above equation (18).

(Action planning procedure in interference planning unit 103)
Next, the action planning procedure of the robot 100 in the interference planning unit 103 will be described below with reference to the flowchart of FIG. 22 as well.

First, the approaching state determination unit 105 determines whether the personal space of the robot 100 is currently located in the personal space of a person, or in the non-approaching state (step S401).

As a result, in the case of the approaching state, the contact possibility determination unit 106 determines whether there is a possibility that the trunk portion of the robot 100 will come into contact with a human in the future (step S402). Here, if the possibility of trunk contact is "yes", the reverse trajectory generation function 112 of the emergency avoidance action selection unit 108 generates a reverse trajectory for moving the robot 100 in the opposite direction away from the human. An action is planned to move the robot along the reverse trajectory generated (step S403). Here, when the movement is impossible, a temporary stop function 113 is used to plan an action to temporarily stop the robot 100 .

On the other hand, when the possibility of body trunk contact is "no" in the approaching state, a different plan is made depending on the presence or absence of some kind of stimulating action so far (step S404). First, since it is assumed that the human H is not ready for the approach of the robot 100 when no action action has been performed so far, the same action as in the case of the possibility of contact with the trunk is "presence". is planned (step S403). That is, at this time, a reverse trajectory is generated, and an action is planned to temporarily stop the robot 100 when the robot 100 cannot move in the reverse direction. Conversely, it is assumed that human beings are prepared to some extent when the action has already been taken. Therefore, in the approach trajectory generation function 111 of the emergency avoidance action selection unit 108, an approach trajectory that avoids the trunk contact between the robot 100 and the human in a state in which the human's personal space is entered without expecting the human's avoidance action is generated. An action is planned to move the robot 100 along the generated approach trajectory (step S405).

On the other hand, when the robot 100 is currently in a non-approaching state outside the human personal space, the action plan of the robot 100 is determined by the following procedure.

First, the preparation time determination unit 107 determines whether or not there is a preparation time required for the robot 100 to act on the human (step S406).

Then, when the preparation time is "no", the emergency avoidance action selection unit 108 selects the emergency avoidance action selection unit 108 according to the presence or absence of the previous action, as in the case where the possibility of body trunk contact in the approaching state is "no". Then, the action of the robot 100 is planned (steps S403 to S405).

On the other hand, when the preparation time is “yes”, the normal avoidance action selection unit 109 performs the following procedure.

First, the detour availability determination unit 115 determines whether there is a relative distance at which a detour route can be physically planned based on the current relative positional relationship between the robot 100 and the human (step S407). In addition, the passing-through space determining unit 116 determines whether or not there is a passing-through space (steps S408 and S409).

Therefore, if the detour is “permissible”, the trajectory candidate searching unit 117 generates trajectory candidates of combinations of different types according to the presence or absence of the passing space. That is, when there is a passing space, a trajectory candidate consisting of a detour trajectory, an approaching trajectory, and a contact trajectory is generated (step S410). On the other hand, when there is no passing space due to the presence of a fixed obstacle such as a wall, a trajectory candidate consisting of contact trajectories is generated (step S411).

In addition, even if the detour is “impossible”, the trajectory candidate searching unit 117 generates trajectory candidates of combinations of different types according to the presence or absence of a passing space. That is, when there is a passing space, a trajectory candidate consisting of an approach trajectory and a contact trajectory is generated (step S412). On the other hand, when there is no passing space, a trajectory candidate consisting of contact trajectories is generated (step S411).

Then, for each of the obtained trajectory candidates, the trajectory candidate with the lowest trajectory cost is determined as the execution trajectory by cost calculation in the optimum trajectory extraction unit 118, and the action content identification unit 119 makes an action according to the type of the execution trajectory. A behavior is identified (step S413).

The action planning process described above is repeated at regular time intervals until the robot 100 reaches the destination (step S414).

According to the second embodiment described above, due to the presence of blind spots in the course of the robot 100, even if an unexpected situation such as the robot 100 suddenly intruding into a person's personal space occurs, the mutual body movement is prevented. It is possible to efficiently move the robot 100 to the destination while avoiding contact with the trunk and considering human psychology. In addition, the robot 100 can select a route to pass the human in a state in which the robot 100 invades the human's personal space while minimizing discomfort to the human, and the robot 100 is more efficient for mutual collision avoidance. Can formulate an action plan.

In addition, the robot according to the present invention is not limited to the autonomous mobile robots 10 and 100 described in each of the above embodiments, and can be autonomously moving within a predetermined space, such as an automobile, a ship, or an air vehicle. In addition to the movable body, it may be a manipulator such as a robot arm that operates within a predetermined range of space, and their motion planning can be performed as described above.

In addition to the humans described in the above embodiments, the moving obstacles of the present invention can also be applied to control for avoiding interference with moving bodies such as bicycles, automobiles, and other autonomous mobile robots, and animals. In this case, when estimating the degree of recognition of a moving obstacle, if the object to be estimated is a moving object controlled by a human, for example, it is possible to estimate the degree of recognition based on the recognition state of the operator in the same manner as described above. . In addition, when the estimation target is not a human like other robots, a sensor, camera, etc. for the other robot to recognize the surrounding environment are provided as the detection device 12 including the recognition state detection sensor 19. Detecting the relative direction of the part as the recognition state, and estimating the degree of recognition based on the detection result in the same procedure as described above can be exemplified.

In addition, the configuration of each part of the device in the present invention is not limited to the illustrated configuration example, and various modifications are possible as long as substantially the same action is exhibited.

REFERENCE SIGNS LIST 10 robot 11 operation unit 12 detection device 13 control device 21 interference level determination means 22 recognition level estimation means 23 action plan means 24 achievement level estimation means 25 plan adjustment means 26 motion command means 32 interference avoidance action determination section 37 action content identification section 38 Execution trajectory identification unit 40 Existing region identification function 41 Interference angle calculation function 42 Concession degree determination function 44 Trajectory candidate search unit 45 Optimal trajectory extraction unit 47 Interference time calculation function 48 Constant speed trajectory generation function (trajectory generation function)
49 Acceleration/deceleration trajectory generation function (trajectory generation function)
100 robot 103 interference planning unit 108 emergency avoidance action selection unit 109 normal avoidance action selection unit 111 approach trajectory generation function 112 reverse direction trajectory generation function 113 temporary stop function 117 trajectory candidate search unit 118 optimum trajectory extraction unit 121 bypass trajectory generation function 122 Approach trajectory generation function 123 Contact trajectory generation function 124 Proximity energy calculation function 125 Trajectory candidate identification function A1 Synchronous action area (first area)
A2 assertive behavior area (second area)
A3 Area outside interference avoidance behavior (third area)
H human (moving obstacle)

Claims (7)

A robot comprising a predetermined action part, a detection device for detecting position information of a moving obstacle existing in the surroundings, and a control device for controlling the action of the action part based on the detection result of the detection device,
The control device includes interference degree determination means for determining the possibility that the robot will interfere with the moving obstacle in the future, and action of the robot so as to move the robot to a target position according to surrounding conditions. An action plan means for planning, and an action command means for giving an action command to the action unit so as to execute the action planned by the action plan means,
The action planning means plans the action of the robot so as to avoid the moving obstacle when the interference degree determining means determines that the robot may interfere with the moving obstacle in the future. including an interference planning unit,
The interference planning unit includes an emergency avoidance action selection unit that selects an action of the robot that performs collision avoidance when the situation of the robot with respect to the moving obstacle is determined to be an emergency, and a collision avoidance action if not, a normal avoidance action selection unit for estimating an approach risk, which is a risk when the robot approaches the moving obstacle, and selecting an action of the robot ;
The normal avoidance action selection unit includes a trajectory candidate search unit that searches for a trajectory candidate for advancing the robot toward the target position, and a trajectory that is optimal from among the plurality of trajectory candidates searched by the trajectory candidate search unit. an optimum trajectory extraction unit that selects and sets an execution trajectory for actually moving the robot,
The trajectory candidate searching unit has a detour trajectory generating function for generating a detour trajectory for advancing the robot to the target position without entering a personal space within a preset range that is considered to be in an approach state to the moving obstacle. and an approach trajectory generation function that generates an approach trajectory for intruding into the personal space and advancing the robot to the target position; a contact trajectory generating function for generating a contact trajectory for advancing the robot toward the target position while moving the contact trajectory; a trajectory candidate identification function that identifies the trajectory candidate based on approach risk determination by energy,
In the optimum trajectory extraction unit, for each trajectory candidate, a moving energy corresponding to a trajectory cost based on kinetic energy is calculated by a preset cost formula, and the trajectory candidate with the lowest moving energy is extracted as the execution trajectory. A robot characterized by : A robot comprising a predetermined action part, a detection device for detecting position information of a moving obstacle existing in the surroundings, and a control device for controlling the action of the action part based on the detection result of the detection device,
The control device includes interference degree determination means for determining the possibility that the robot will interfere with the moving obstacle in the future, and action of the robot so as to move the robot to a target position according to surrounding conditions. An action plan means for planning, and an action command means for giving an action command to the action unit so as to execute the action planned by the action plan means,
The action planning means plans the action of the robot so as to avoid the moving obstacle when the interference degree determining means determines that the robot may interfere with the moving obstacle in the future. including an interference planning unit,
The interference planning unit includes an emergency avoidance action selection unit that selects an action of the robot to perform collision avoidance when the situation of the robot with respect to the moving obstacle is determined to be an emergency, and a collision avoidance action in other cases, a normal avoidance action selection unit for estimating an approach risk, which is a risk when the robot approaches the moving obstacle, and selecting an action of the robot;
The emergency avoidance action selection unit selects the moving obstacleA preset state that is in proximity toAn approach trajectory generation function that generates an approach trajectory that moves the robot to the target position by entering the personal space within the range, and a reverse direction trajectory generation function that generates a reverse trajectory that moves the robot in the opposite direction. function and a temporary stop function that temporarily stops the robot, depending on the situationany of the trajectories is generated bycharacterized in that the movement of the robot along the trajectory or a temporary stop is selectedRurobot. 3. The trajectory candidate identification function extracts a trajectory having the same level of proximity energy as the detour trajectory used as a reference when identifying the approach trajectory and the contact trajectory as the trajectory candidates. 1. The robot according to claim 1. The control device further comprises recognition degree estimation means for estimating a recognition degree representing the degree of recognition of the moving obstacle by the robot, using the detection result of the detection device,
The proximity energy calculation function is characterized in that the proximity energy is obtained using the relative position and relative speed between the robot and the moving obstacle, the degree of recognition, and the attribute of the moving obstacle as parameters. Item 1. The robot according to item 1. The cost formula applied to the contact trajectory is obtained by adding a parameter corresponding to the amount of movement caused by the robot contacting the moving obstacle to the cost formula applied to the detour trajectory and the approach trajectory. 2. The robot according to claim 1 , characterized in that: In an action planning device that operates a robot so as to avoid future interference with a moving obstacle in accordance with the detection result of position information of a moving obstacle that exists in the surroundings,
Interference determination means for determining the possibility that the robot will interfere with the moving obstacle in the future, and action planning means for planning the action of the robot so as to move the robot to a target position according to surrounding conditions. and
The action planning means plans the action of the robot so as to avoid the moving obstacle when the interference degree determining means determines that the robot may interfere with the moving obstacle in the future. including an interference planning unit,
The interference planning unit includes an emergency avoidance action selection unit that selects an action of the robot that performs collision avoidance when the situation of the robot with respect to the moving obstacle is determined to be an emergency, and a collision avoidance action if not, a normal avoidance action selection unit for estimating an approach risk, which is a risk when the robot approaches the moving obstacle, and selecting an action of the robot ;
The normal avoidance action selection unit includes a trajectory candidate search unit that searches for a trajectory candidate for advancing the robot toward the target position, and a trajectory that is optimal from among the plurality of trajectory candidates searched by the trajectory candidate search unit. an optimum trajectory extraction unit that selects and sets an execution trajectory for actually moving the robot,
The trajectory candidate searching unit has a detour trajectory generating function for generating a detour trajectory for advancing the robot to the target position without entering a personal space within a preset range that is considered to be in an approach state to the moving obstacle. and an approach trajectory generation function that generates an approach trajectory for intruding into the personal space and advancing the robot to the target position; a contact trajectory generating function for generating a contact trajectory for advancing the robot toward the target position while moving the contact trajectory; a trajectory candidate identification function for identifying the trajectory candidate based on approach risk determination by energy,
In the optimum trajectory extraction unit, for each trajectory candidate, a moving energy corresponding to a trajectory cost based on kinetic energy is calculated by a preset cost formula, and the trajectory candidate with the lowest moving energy is extracted as the execution trajectory. A robot action planning device characterized by : In a program for performing an action plan for operating a robot so as to avoid future interference with a moving obstacle in the future, according to the detection result of the position information of the moving obstacle existing in the surroundings,
Interference determination means for determining the possibility that the robot will interfere with the moving obstacle in the future, and action planning means for planning the action of the robot so as to move the robot to a target position according to surrounding conditions. to make the computer work as
The action planning means plans the action of the robot so as to avoid the moving obstacle when the interference degree determining means determines that the robot may interfere with the moving obstacle in the future. including an interference planning unit,
The interference planning unit includes an emergency avoidance action selection unit that selects an action of the robot that performs collision avoidance when the situation of the robot with respect to the moving obstacle is determined to be an emergency, and a collision avoidance action if not, a normal avoidance action selection unit for estimating an approach risk, which is a risk when the robot approaches the moving obstacle, and selecting an action of the robot ;
The normal avoidance action selection unit includes a trajectory candidate search unit that searches for a trajectory candidate for advancing the robot toward the target position, and a trajectory that is optimal from among the plurality of trajectory candidates searched by the trajectory candidate search unit. an optimum trajectory extraction unit that selects and sets an execution trajectory for actually moving the robot,
The trajectory candidate searching unit has a detour trajectory generating function for generating a detour trajectory for advancing the robot to the target position without entering a personal space within a preset range that is considered to be in an approach state to the moving obstacle. and an approach trajectory generation function that generates an approach trajectory for intruding into the personal space and advancing the robot to the target position; a contact trajectory generating function for generating a contact trajectory for advancing the robot toward the target position while moving the contact trajectory; a trajectory candidate identification function that identifies the trajectory candidate based on approach risk determination by energy,
In the optimum trajectory extraction unit, for each trajectory candidate, a moving energy corresponding to a trajectory cost based on kinetic energy is calculated by a preset cost formula, and the trajectory candidate with the lowest moving energy is extracted as the execution trajectory. A robot action planning program characterized by :

Images