JPH06143170A - Auto-nomic conveying device - Google Patents

Abstract

PURPOSE:To perform a conveying work of even an object in a non-graspable state by providing a non-graspable state escaping means to plan operation for a robot to escape from a non-graspable state. CONSTITUTION:A non-graspable state detecting means 4 is operated to detect it based on the calculated result of a grasping position posture calculating means 3 that an object in an initial position or a target position is in a non- graspable state. A non-graspable state escape planning means 5 is operated to plan operation of a robot to prevent the occurrence of a state to be incapable of grasping an object when a non-graspable state is detected by the non- graspable state detecting means 3. A conveying work operation producing means 7 is designed to produce operation of a robot to convey an object from an initial state first provided to a target state first provided through synthesis of the operation plan results of a placing operation planning means 6 and a non-graspable state plan escaping means 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for autonomously carrying an object carrying work by a manipulator.

[0002]

2. Description of the Related Art Conventionally, an action plan for grasping, moving, and placing (pick-and-place) an object is determined by giving an initial state and a target state of the moving object, and operates according to this plan. Research is being conducted on autonomous robots (autonomous transportation devices). The robot can perform work even in an environment where a human cannot perform work, and can perform a work that cannot be performed by a human. Therefore, if the autonomous ability of the robot is improved, it is useful in a wide range of fields.

In such an autonomous robot, the work plan is important, and there are various proposals for this. For example, the inventors of the present invention (10)
Volume 2, No. 2, pp. 273-282, issued in 1992), proposed a method for planning the operation of an autonomous carrier. In this motion planning method, when multiple targets cannot be achieved with a single grasping, moving, or placing task due to the shape of the object or hand, the state of obstacles around it, the range of movement of the robot, etc. Depending on the placement work, the grasping position and grasping posture can be changed (holding)
Can achieve the target state.

Furthermore, the inventors of the present invention, at the Robotics Mechatronics Lecture '92 (June 1992),
We proposed a motion planning method that can reduce the number of placing operations by equipping a finger with a hand with a rotating mechanism. These make it possible to efficiently carry out a grasping and transporting plan of an object from an arbitrary initial state to an arbitrary target state, and the robot can autonomously transport an object based on the obtained plan. it can.

[0005]

However, in these conventional methods, there are obstacles near the grasped object,
When a situation in which the object cannot be grasped (state in which the object cannot be grasped) occurs, it becomes difficult to carry out the work of transporting the object. For this reason,
When working with a robot, it was necessary for humans to prepare the environment in advance to prevent such an unrecognizable state.

On the other hand, in a human living environment or natural environment where effective use of robots is expected, it is often difficult to prepare the environment in advance before work. Therefore, there is a problem that the range of use of the robot is limited. Therefore, there is a demand for a robot capable of autonomous work regardless of environmental conditions.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an autonomous carrying device capable of carrying work even when an object is in an unrecognizable state.

[0008]

SUMMARY OF THE INVENTION According to the present invention, a hand for grasping an object and a grasping position and posture of the hand for carrying an object from an arbitrary position and posture to another arbitrary position and posture are planned. In the autonomous transporting device having the transport grasping operation planning means, an unrecognizable state detecting means for detecting that the object cannot be grasped by the hand from the state of the object at the initial position or the target position, When it is detected, the object or the obstacle in the surrounding environment is moved, and the unrecognizable state escape planning means for planning the operation of the autonomous transporting device for enabling the object to be grasped is provided.

[0009] Further, the above-mentioned incomprehensible state escape planning means is
The feature is that planning is performed by using an operation of pushing an object or an obstacle with a hand and sliding it to move.

[0010]

The autonomous transporting apparatus according to the present invention is configured as described above, and automatically plans an operation to escape from the state in which the object cannot be grasped during work. You can Therefore, the object or the obstacle can be moved so that the autonomous carrier can grasp the object. Therefore, autonomous transportation work can be performed regardless of environmental conditions.

[0011]

【Example】

<Description of Overall Configuration> An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of the carrying device according to the present embodiment. The three-dimensional model 1 of the environment includes information on the shape of the grasped target object, the position and orientation of the target object in the initial state and the target state, the shape of the surrounding environment and its position, orientation, the shape of the robot and the hand, the movable range, and the like. Is listed. Transportation grasp operation planning means 2
Plans the grasping position and orientation of the hand and the movement of the hand for transporting an object from an arbitrary position and orientation to another arbitrary position and orientation. The grasping position / orientation calculating unit 3 calculates the grasping position / orientation of the object in the initial state and the target state from the information of the three-dimensional model 1 of the environment. The unrecognizable state detecting means 4 detects that the object at the initial position or the target position is in the unrecognizable state based on the calculation result of the grasping position / orientation calculating means 3. The ungrasping state escape planning means 5 plans the operation of the robot for eliminating the ungrasping state of the object when the ungrasping state detecting means 3 detects the ungrasping state. The placement operation planning means 6 is the grasping position / orientation calculating means 3
Based on the result of 1., the placement operation of the robot for transporting the object in the initial state in which the object can be grasped to the target state in which the object can be grasped is planned. The transporting work motion creating means 7 synthesizes the motion plan results of the placement motion planning means 6 and the incomprehensible state plan escaping means 5 to change the initially given initial state to the initially given target state. Create a robot motion to carry an object.

Then, the manipulator 8 actually carries out the carrying operation using the parallel two-finger grasping mechanism 9 according to the output of the carrying work operation creating means 7.

The operation of the autonomous object carrier according to this embodiment will be described with reference to the overall flow chart of FIG. First, the grasp position / orientation calculation means 3 determines
Based on the information described in the dimensional model 1, all the position / orientation candidates (grasping candidates) of the robot for grasping the object in the initial position / orientation are calculated (S1). Next, based on the calculation result of the grasping position / orientation calculating unit 3, the ungrasping state detecting unit 4 detects whether the object in the initial state is in the ungrasping state. That is, it is determined whether or not there is a grasping candidate (S2).
It should be noted that it may be determined whether or not a graspable space exists, and a grasping candidate for the grasping position may be determined in a later step.
The same processing as in S1 and S2 is performed on the target position / orientation target object to obtain a grasping candidate for the target state object, and from the result, it is determined whether or not the target state is the incomprehensible state.
3, S4).

According to the determination results of S2 and S4, the initial state,
In both of the target states, there are grasping candidates, and if they are not graspable states, the placement movement planning means 6 plans the movement of the robot for placement work based on the grasping candidates obtained in S1 and S3. Yes (S5). In the figure, the initial state,
The AND gate shows the judgment that the target state can be grasped. Further, the processing content of the placement operation planning means 6 is based on the operation planning method of the autonomous transportation device reported in the above-mentioned inventor's journal of the Robotics Society of Japan (Volume 10, No. 2).

On the other hand, in the process of S2, when the unrecognizable state detecting means 4 detects that the object in the initial state is in the unrecognizable state, the unrecognizable state escape planning means 5
Then, the robot operation for transitioning from the initial state of the unrecognizable state to the identifiable state is planned, and the obtained identifiable state is set as a new initial state (S6). Further, if it is determined in S4 that the target object is in the unrecognizable state, the unrecognizable state escape planning means 5 plans the robot operation for the state transition from the graspable state to the unrecognizable target state. The planned graspable state is set as a new target state (S7).

In this way, when both the initial state and the target state can be grasped, the grasping candidate exists in S2 and S4. Therefore, the placement operation planning means 5 plans the work from the initial state in which the object can be grasped to the target state in which the object can be grasped (S).
5). Then, the transportation work operation creating means 7 combines the operation plan results of the placement operation planning means 6 and the incomprehensible state escape planning means 5 to target from the initially given initial state to the initially given target state. A motion plan of a robot for carrying an object is created (S8).

Then, according to the transportation operation plan thus created, the manipulator 8 having the parallel two-finger grasping mechanism 9 actually carries out the transportation operation of the object (S).
9).

In the following, the processing contents of each processing S1, S3, S6, S7, S8 in the flowchart of the embodiment (FIG. 2) will be described in detail.

<Comprehensible Space in Initial State and Target State> First, the processes S1 and S3 in the flowchart of FIG. 2 will be described. Here, in the present embodiment, a parallel two-fingered hand is adopted as the grasping mechanism, and the object handled here is
A polyhedron composed of planes. Also, when grasping the object stably, the elements (face, ridge, vertex) that make up the object
In the figure, two elements that two fingers contact are called a grasp element pair.

The parallel two-fingered hand is constrained by a plane (grasping plane) which can be formed in the center of the grasping element pair. An example of the grasping surface and its restraint is shown in FIG. Thus, if the line connecting the centers of the two fingers of the hand is the y-axis and the longitudinal direction of the two fingers is the z-axis, the origin of the hand is within the grasping plane and the hand y-axis is constrained to be perpendicular to the grasping plane. To be done.

Due to this constraint, the motion space of the hand becomes a three-dimensional space composed of the two-dimensional position in the grasping plane and the rotation around the y-axis of the hand. The graspable space that does not interfere with surrounding obstacles is obtained by subtracting the collision space of the obstacle from the graspable space of the object in the three-dimensional space. At this time, the hand is separated into two fingers and the main body, and the collision spaces are separately obtained. The reason for this will be described later. Also, these spaces are already known to Laugier
Grasping planning method (C.Laugier, J.Pertin: "Automatic Gra
sping: A Case Study in Accessibility Analysis ", Ad
vancedSoftware in Robotics, pp.201-214, 1984) can be applied for calculation.

An example of the calculation is shown in FIG. In this example, both the object and the obstacle are rectangular parallelepipeds and are placed close to each other on the table. Therefore, the finger cannot be inserted into the space between the object and the obstacle.

The determination as to whether or not the object can be grasped in this state will be described. First, this determination is performed by detecting the graspable space and the collision space on the grasping surface of the object. Here, the grasping surface is located at the center of the surface of the object facing toward the front and the corresponding surface facing in the opposite direction, and the position of the origin of the hand with respect to the hand posture in the drawing exists in this surface. The space a is a space in which the object can be grasped in the grasping plane, and if the hand origin exists in this space a, the hand can grasp the object. The space b indicates the space at the origin of the hand where the finger located at the back in the drawing of the hand collides with the obstacle, and the space c indicates the collision space between the hand body and the obstacle.
Spaces d and e respectively indicate collision spaces of the finger and the hand body with respect to the table.

In this example, since the space a is completely included in the space b, in this state, there is no graspable position,
Judgment is impossible.

<Cause of Unrecognizable State and Escape Method> Next, the incomprehensible state will be considered. The incomprehensible state is
All of the grasping candidates fixed in the object coordinate system cannot be grasped for some reason. For example, as in the example of FIG. 4, the graspable space of the object is completely included in the collision space for the obstacle or the like. In FIG. 4, only the case where the hand interferes with surrounding obstacles including the table at the time of grasping is considered, but there are other possible causes of the inability to grasp. That is, the cause is
The following are possible:

A: When the hand interferes with surrounding obstacles at the time of grasping B: When the arm interferes with surrounding obstacles at the time of grasping C: When there is no intermediate operation route due to the obstacle D: There is a grasping candidate In the case where the manipulator is out of the operating range, however, in the present embodiment, only the method of exiting the state A will be described. This is because these causes generally occur at the same time, but it is considered that there are the most grasping candidates that cannot be grasped only by the cause of A in general. In particular, in a work environment in which the manipulator can be operated, since the manipulator is arranged in a place where it is easy to work, most of the grasping candidates will not be impossible to grasp due to D. Also, it is considered that there are few grasping candidates that cannot be grasped due to B and C unless the obstacle is a considerably large obstacle. Therefore, in many cases, it is possible to escape from the incomprehensible state by escaping from the cause of A.

In order to escape from the unrecognizable state, it is necessary to move the obstacle or the object to create a graspable portion. The following are possible strategies for this.

(A) Movement of obstacle by sliding movement operation (b) Movement of obstacle by pick-and-place operation (c) Movement of object by sliding movement operation Here, in the strategy of (c), This also includes the case where the obstacle that is trying to eliminate the influence is a table. Specifically, FIG.
This is a case where the object is grasped by protruding a little from the edge of the table as shown in (b).

Next, the process S6 of the flowchart of FIG.
The unrecognizable state escape planning means 5 for performing S7 will be described. The incomprehensible state escape planning means 5 has the three strategies described above. However, (b) Pick-and-Plac
The movement of the obstacle by the operation is based on the conventional operation planning method of the autonomous carrier (the present inventors, Journal of the Robotics Society, Vol. 10, No. 2). Therefore, it is possible to move the obstacle to a wide area on the table. For the other two strategies, we will describe how much an object or obstacle should be moved in order to escape from the incomprehensible state. Here, for the sake of simplicity, a sliding motion plan for enabling grasping with one hand posture on one grasping surface will be described. In practice, this processing is performed for all postures with a certain quantization width, and a suitable one is selected from them. Further, when there are a plurality of hand directions capable of grasping the object and there are a plurality of grasping surfaces, the same is done for all grasping surfaces. Here, the sliding movement of the object is a parallel movement without rotation.

<Escape Strategy by Moving Obstacles by Sliding and Moving Operation> When the grasping surface and the table upper surface are parallel,
Since the shape of the collision space does not change even if the obstacle is moved, the movement distance can be obtained using the result calculated in the grasping plan. That is, when the obstacle is slid in any direction and moved, the collision space moves in that direction. Therefore, by moving the collision space of the moving obstacle in parallel,
The moving distance is calculated so that a graspable space for the object that does not overlap with all collision spaces appears.

If the grasping surface and the table upper surface are not parallel, first, the line of intersection between the grasping surface and the table upper surface is obtained, and the direction of movement of the obstacle is determined as a direction parallel to the line of intersection and a direction perpendicular to the line of intersection. Think separately.

Even when the obstacle is moved in the direction parallel to the intersection line, the result calculated by the grasping plan can be used as it is.
That is, if the obstacle is moved in that direction, the collision space moves in parallel. Therefore, it is only necessary to move the collision space in parallel and obtain a movement distance such that a graspable space for the object that does not overlap the collision space appears. FIG. 5 shows an example in which the obstacle of FIG. 4 is moved in the direction parallel to the line of intersection between the grasping surface and the table and toward the end of the table. By such movement, a part of the graspable space a of the object (upper left part in the figure) is separated from the collision spaces b, c, d, e. Therefore, this portion does not overlap the collision spaces b, c, d, and e, and the portion can be grasped. Here, for stable grasping, the graspable space is made to escape by the length of the finger width.

When an obstacle is moved in a direction perpendicular to the intersection line, the shape of the collision space changes, so it is not possible to calculate the moving distance using the result calculated in the grasping plan. Therefore, consider moving an obstacle to the outside of the three-dimensional motion space of the hand constrained by the grasping surface. As a result, the collision space between the hand and its obstacle can be eliminated.

At this time, the distance to be moved also differs depending on which part of the hand the obstacle to be moved may collide with. Here, the hand is divided into a hand body and two fingers, and the collision space is obtained separately, thereby classifying the obstacles. The classification and required movement distance d are shown below.

C1: When the obstacle collides with only one finger d = (Sopen-W) / 2 sin θ C2: When the obstacle collides only with the hand body (the obstacle is outside the two fingers) Case) d = (W-Sopen) / 2 sin θ C3: When the obstacle collides with the hand body only (when the obstacle is between two fingers) d = (max (W, Sopen) + Sclose) / 2 sin θ C4: When an obstacle collides with the hand body and one finger d = (max (W, Sopen) + Sclose) / 2 sin θ C5: When an obstacle collides with both fingers d = ∞ Here, as shown in FIG. 6, Sclose is the distance between the two fingers when the hand grasps the object, and Sopen is the distance between the outer sides of the two fingers when the hand moves to the grasping position of the object. , W is the width of the hand body in the y-axis direction. Of the two directions perpendicular to the line of intersection, the direction of movement was a direction away from the object except C3, and both directions were C3. C5
Is an obstacle existing across the object in a direction perpendicular to the grasping surface, and a table or the like often corresponds to this obstacle.
Here, the obstacle C5 is set to ∞ because the moving distance cannot be obtained by classifying it. FIG. 6 shows the calculation of the movement distance d from the unrecognizable state due to the obstacle in the state C4.

From these results, in order to escape from the ungrasping state when the grasping surface and the table upper surface are not parallel, as shown in FIG. 7, three sides of a rectangle composed of da1, da2, and db (see FIG. Move the obstacle origin O above the middle line. Here, da1 and da2 are movement distances obtained in the direction parallel to the line of intersection between the grasping surface and the table top surface, and db is the movement distance obtained in the vertical direction. Note that FIG. 7 is a view of FIG. 4 seen from the vertically upper side.

When the obstacle is moved, the obstacle is moved from the current position of the obstacle to the thick line in FIG. 7 regardless of whether the object in the unrecognizable state is in the initial state or the target state. Plan robot actions for. Here, since the obstacle is a rectangular parallelepiped, it is assumed that the obstacle slides and moves in the direction opposite to the normal vector of the side surface. The movement of the robot for the sliding movement is a known method (Sawada, Inaba, Inoue:
"Study of pushing operation by manipulator using visual feedback", Proceedings of the 5th Academic Lecture Meeting of the Robotics Society of Japan, pp.599-600, 1987). That is, when pushing with one finger of the hand, the center of gravity of the object, the fingertip, and the moving direction are basically in a straight line. When the moving direction is misaligned, the pushing direction is changed and the object is corrected. To do. Here, since the parallel two fingers are used in the present embodiment, if the two fingers are pressed against the object, the rotation of the object can be suppressed and the object can be slid and moved in a desired direction.
Further, the object can be easily rotated by changing the angle of the tips of the two fingers. In this way, the desired sliding movement can be performed by pushing the target object with the fingers of the hand.

<Escape Method by Moving Object by Sliding and Moving Operation> Basically, it can be obtained by the same method as in the previous section. However, when the result calculated in the grasping plan is used as it is, the space in which the object moves can be the graspable space of the object.

When the grasping surface and the table top surface are not parallel and the movement direction of the object is perpendicular to the line of intersection between the grasping surface and the table top surface, there is a slight difference. That is, the moving distance differs depending on which of the classes mentioned in the previous section the obstacle to be excluded belongs. Therefore, an obstacle that can grasp the object if the influence is eliminated is sought, and the shortest moving distance is selected from those obstacle classes. Further, in this case, there is a possibility that an obstacle of the hand newly increases in the moving direction. Therefore, after obtaining the movement distance in the sliding movement direction, a grasp plan is made in the state after the movement, and if the grasp cannot be made in that state, this operation plan fails.

Further, when the object is moved to the end of the table, the object may drop from the table. for that reason,
Considering the center of gravity of the object and the contact surface with the table, it is checked whether the obtained position is stable.

If the object in the unrecognizable state is the initial state, the robot operation for moving from the initial state to the thick line in FIG. 7 is planned. Further, if the object in the unrecognizable state is the target state, the robot operation for moving from a certain position on the thick line in FIG. 7 to the position in the target state is planned. Here, since the object is a rectangular parallelepiped, it is assumed that the object slides and moves in the direction opposite to the normal vector of the side surface. Similar to the above case, the robot movements for sliding movement are already known methods (Sawada, Inaba, Inoue: “Study on push operation by manipulator using visual feedback”, The 5th Academic Lecture Meeting of the Robotics Society of Japan. Proceedings, pp.5
99-600, 1987).

Next, the carrying work operation creating means 6 for carrying out the processing S8 of the flowchart of FIG. 2 will be described.

<Transportation Work Operation Plan> The transportation work operation creation means 7 includes the placement operation planned by the placement operation planning means 6 and the object escape operation in the initial state planned by the unrecognizable state escape planning means 5. The object escape operation in the target state planned by the unrecognizable state escape planning means 5 is synthesized, and a robot operation for transporting the object from the initially given initial state to the initially given target state is created.

First, if the initial state of the object given first is the unrecognizable state, the incomprehensible state escape planning means 5
The object escape operation in the initial state planned by is inserted before the placing operation planned by the placing operation planning means 6.

Next, if the initially given target state of the target object is the incomprehensible state, the incomprehensible state escape planning means 5
The place where the escape action is inserted differs depending on whether the target escape action in the target state planned by is the action based on the escape strategy by moving the obstacle or the action based on the escape strategy by moving the target. . When it is an operation based on the escape strategy by moving an obstacle, the placement operation planning means 6
Insert before the placing operation planned by. In the case of an operation based on the escape strategy by moving the object, the operation is inserted after the placement operation planned by the placement operation planning means 6.

<Operation Example> FIGS. 8 and 9 show an operation example of escape from an unrecognizable state using this method. FIG. 8 shows an operation example 1. The rectangular parallelepiped in front of FIG. 8 (a) escapes from a state in which it cannot be grasped due to an obstacle to a state in which it can be grasped by a sliding movement operation, and the target in FIG. 8 (d). Place in position. FIG. 8B shows a case where a strategy of sliding and moving an object is used, and FIG. 8C shows a case of using a strategy of sliding and moving an obstacle. That is, FIG. 8 (b)
In the above example, the object is slid and moved in a direction parallel to the grasping surface by a predetermined distance to create a graspable space, and the object is grasped. Further, in the example of FIG. 8C, the obstacle is moved in the direction perpendicular to the grasping surface so that the grasp can be performed by the hand.

FIG. 9 shows a second operation example. In this example, the object is grasped by moving to the end of the table from a state in which the object cannot be grasped due to the wide width (FIG. 9A). (FIG. 9 (b)) An example of an operation plan to be placed in the target state (FIG. 9 (c)).

[0048]

As described above, the ungrasping state detecting means for detecting that the object at the initial position or the target position is in the ungrasping state and the operation of the robot to escape from the ungrasping state are planned. Since it has an ungrasping state escape means, it is possible to automatically plan the operation of the robot to escape the object from the ungrasping state, and carry out the transportation work even if the object is in the ungrasping state. Will be possible.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of an autonomous transportation device according to an embodiment.

FIG. 2 is an overall flowchart of the operation of the autonomous object carrier according to the present embodiment.

FIG. 3 is a diagram illustrating a grasping surface and its constraint.

FIG. 4 is a diagram illustrating an example of a grasping plan.

FIG. 5 is a diagram illustrating movement of a collision space.

FIG. 6 is a diagram illustrating a moving distance of an obstacle.

FIG. 7 is a diagram illustrating the positions of obstacles that can be grasped.

FIG. 8 is a diagram for explaining an operation example 1 of the present embodiment, and FIG. 8A is an initial state of an unrecognizable state of the operation example 1;
(B) is a result of the escape plan for moving the object in the operation example 1,
(C) shows the result of the escape plan for moving the obstacle in the operation example 1, and (d) shows the target state in the operation example 1.

FIG. 9 is a diagram illustrating an operation example 2 of the present embodiment,
(A) is the initial state of the unrecognizable state of operation example 2, (b)
Is the result of the escape plan using the table edge of operation example 2,
(C) has shown the target state of the operation example 2.

[Explanation of symbols]

 1 3D model of environment 2 Transportation grasping motion planning means 3 Grasping position / orientation calculating means 4 Ungrasping state detecting means 5 Ungrasping state escape planning means 6 Placement movement planning means 7 Transportation work movement creating means 8 Manipulator 9 Parallel two-finger grasping mechanism

Claims (2)

[Claims] 1. A hand for grasping an object, and a grasping and grasping motion planning means for planning the grasping position and posture of the hand for conveying an object from an arbitrary position and posture to another arbitrary position and posture. In the autonomous carrying device, the unrecognizable state detecting means for detecting that the object cannot be grasped by the hand from the state of the object at the initial position or the target position, and the object or An unidentifiable state escape planning means for planning the operation of the autonomous transportation device for moving an obstacle in the surrounding environment so that an object can be grasped, and the autonomous transportation device. 2. The autonomous transporting device according to claim 1, wherein the ungrasping state escape planning means plans using an operation of pushing an object or an obstacle with a hand and sliding it to move. Autonomous carrier device.