JP2003280710A - Generation and control method of working track of robot hand - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method used for guiding a robot hand from a designated position to a target position in a work environment wherein obstacles having different sizes and shapes from one another are present, capable of remarkably improving a calculation processing time because of not including calculation of a complicated three-dimensional potential field as compared with a conventional method, and of automatically generating a safe working track of the hand without interfering with an object at all. <P>SOLUTION: This creation and control method of a working track of a universal robot hand is used for creating a world map (environmental map) of the actual world by a model for avoiding the obstacles and a virtual potential in order to determine the spatial working track of the hand (end effector) attached to the tip of a robot arm in a three-dimensional work environment including the obstacles, and for searching for a relay point of a passage in the optimum condition while consulting a potential (movement number) registered in the map. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating and controlling a work trajectory of a robot hand in a three-dimensional work environment including obstacles.

In order to perform some work using the robot manipulator, first, an operation locus for guiding a finger mechanism of a robot hand called an end effector attached to the tip of an arm from a work start position to a work end position (hereinafter, A work route, a work trajectory, or simply a route or a trajectory) is defined, and then two techniques for determining the overall posture (joint movement) of the arm from position / posture information on the route are basic.

These are called trajectory planning problems and inverse kinematics problems, respectively, and a link mechanism that constitutes a hand or an arm contacts or interferes with various objects (hereinafter synonymous with obstacles) in the environment while moving in space. It is important to try to solve the problem with the intention of not causing In particular, the latter has been researched and developed in various ways as a subject to directly control the relative motion of the robot arm, in which the degree of freedom of joints and various motion modes are related. The present invention relates to an autonomous trajectory generation method for a finger mechanism of a robot hand in the planning for all steps of work using such a manipulator. That is, regarding a simple calculation algorithm for quickly generating a work trajectory in which a hand moves while avoiding obstacles or a specific area in place of human intervention or instructions in an environment including various objects. Is.

[0004]

2. Description of the Related Art When a robot arm is controlled to execute a work, it is common to specify details of the work operation by a program or teaching. Although various methods for interactively planning robot movements have been developed today using a simulator having a graphic display function and a collision check function, it still requires a lot of time and effort. Technology and experience for so-called autonomy, which gradually releases human judgments and operations from the stage that requires them, have been steadily accumulated, but unlike mere machine automation, one function of manipulation is taken. However, the barrier to achieving a complete work plan is the ability to adapt to the environment.

Regarding a motion trajectory for moving a robot hand corresponding to a human hand from one position to another position, an expected trajectory is divided into several sections, and each trajectory section is algebraically expressed as a continuous time function. Although it is extremely easy to handle, there are few approaches that detect obstacles and avoid them, and it has been recognized that obstacle detection and avoidance problems are difficult.

[0006] In particular, while admitting that the complexity of the calculation has been clarified in the field of computational geometry, which may cause a computational explosion in the worst case, the theoretical evaluation is extremely effective as a motion of a real robot. It is a rare case, and rather, as a result of research and development from the standpoint that a practical motion plan can be sufficiently realized, for example, repulsion according to the distance from the object or target position in the environment to the robot It is well known that the potential method that incorporates force and attraction has made a great contribution to autonomous robot guidance without experiencing interference with obstacles or near misses.

An approach that develops this idea and improves the potential weaknesses of the artificial potential method itself, namely,
A global method that collects scalar state quantities handled by heat conduction theory and circuit theory as environmental information of the robot world and associates it with the movement of the robot has become a useful tool for behavior planning in the environment with obstacles. However, since the method originally requires the calculation of the process of determining the temperature and potential distribution of the field from the multi-dimensional simultaneous equations in which the heat conduction elliptic equation is differentiated and the nodal equations of the network,
The burden of calculation for three-dimensional mapping in which those state quantities are registered as addresses is extremely large, and it was considered as a development issue.

[0008]

As described above, the potential method has obtained satisfactory results by associating the behavior of the temperature field and the potential field with the movement of the moving body on the plane. The application to the three-dimensional trajectory planning problem operating below exposes an unexpected disadvantage that causes a rapid increase in calculation time and amount of calculation. Efforts that spend most of the computational effort in the preliminary process of such route determination are not always rational in the case of 3D, and it is necessary to review and simplify the environmental information that is effective for moving in 3D space. .

[0009]

The present invention is characterized by the following method when generating and controlling the work trajectory of the robot hand.

(1) In a three-dimensional work environment including obstacles, an obstacle avoidance model and a virtual obstacle avoidance model are used in order to determine the work trajectory in space of a hand (end effector) attached to the tip of a robot arm. This is a method of creating a real world world map (environment map) based on a potential, and searching for a relay point of the path of the robot arm under optimum conditions with reference to the potential (movement number) registered in the map.

(2) In the above method, when there is a specific control target area in which an obstacle is placed in the work environment, the standby potential is designated to temporarily suspend the registration of the movement number in that area. This is a method of creating an environment map that incorporates the measures to avoid the control target area when searching for the optimum route of the robot hand.

(3) A method for constructing the shortest trajectory of a robot hand whose path is less likely to change by thinning (jumping) a relay point to give priority to movement on a straight line with respect to the work trajectory of the guided robot hand. Is.

(4) When the posture of the robot arm corresponding to each point on the guided work trajectory becomes physically unrealizable, a feasible trajectory candidate buried in the optimum search algorithm is used. Is the way to do it.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION 1. The present invention relates to a spatial map that simply models movement in a three-dimensional environment-that is, an obstacle avoidance model and a real-world world map based on a virtual potential ( Hereinafter, it is also referred to as an environment map or a potential map) -and the main pillars of the search algorithm that guides the hand of the manipulator to the target point using the map (map).

The method of creating a map is basically to define a virtual potential (hereinafter, simply referred to as a potential), which is not directly related to a general physical phenomenon, in the environment under consideration according to the following principle. Become.

That is, in this map, in addition to the description of the target position G and the starting position S that the end effector is to reach, the state quantity (potential) of the target point G is set to the minimum, and the farther the target position G is from the four directions. Suppose that a virtual world with increasing potential is condensed. If the potential of the starting position S is determined in this three-dimensional world, the target point G is naturally reached by tracing from the starting point S at a high level to the direction showing the lower potential. The route can be found. This is the basic idea of the search algorithm, and since the spatial distance can be indirectly expressed by the potential amount, it is easy to generate a route having a short moving distance by taking a route in the direction of a lower potential value.

Further, it is an advantage that a collision between the robot and the obstacle can be surely avoided during the movement by creating a map in which the potential is not set with respect to the obstacle and the prohibited area. The problem is how to give such a potential distribution, but here we explain the generation of the most intuitive and practical environment map through a concrete example.

Hereinafter, the consideration environment is defined as a discretized space consisting of a finite number of grid points suitable for handling a computer, and the path of the end effector is determined by using an environment map in which a virtual state quantity is set at each grid point. Explore. Although the above-mentioned search algorithm is generally applied to a three-dimensional space, here, in order to make the description easy to understand, the two-dimensional model shown in (1) of FIG. The cells in the figure mean lattice points, and the state quantities are written therein. The space to be handled is roughly divided into a boundary wall (shaded area in the figure), a free space where the finger mechanism of the hand can move without collision (blank mesh area in the figure), and an obstacle (shaded area in the figure). Each grid point belongs to one of them.

(1) Creation of Potential Map First, the procedure for creating a potential map will be described prior to the creation of a work route.

A) Setting of entry prohibition area As shown in (1) and (2) of FIG. 2, in the present invention, a sphere (in consideration of diversity of movement directions of the end effector in the three-dimensional space) It is represented by a sphere having a radius r circumscribing the end effector. Further, as a calculation model for avoiding a collision with an object (obstacle), an area where the end effector is prohibited to enter (r from the surface of the object) around the object (obstacle).
A band region of + δ: δ is a clearance) and is applied to the boundary wall and obstacle of calculation. In (2) of FIG. 1, this r +
δ is a band-shaped area with unit grid width (mesh area: inaccessible area)
It is defined for convenience. (Details of the collision avoidance model are shown in 2).

B) Setting a virtual potential for the adjacent grid point of the target point G (step 1) After initializing the potential of each grid point to 0, a potential map in free space is created. First, as shown in (3) of FIG. 1, the potential of the target point G (hereinafter, also referred to as a movement number) is set to 1. Next, the grid points adjacent to the target point (in the case of a two-dimensional model,
Orientation: In the case of a three-dimensional model, the front, back, left and right are added to the front and
Check the attributes of grid points located in the direction. If the attribute of the adjacent grid point is a movable free space constituent element, 2 is set as the potential of the grid point as shown in (3) of FIG.

C) Expansion of Potential Setting Range (Step 2) Next, a grid point having a potential of 2 is searched, and the attribute and potential of the grid point adjacent to the grid point are searched. If the potential can be set at the adjacent lattice points, the potential 3 is set as shown in (4) of FIG. In this way, the grid point for which the potential can be set generally has the attribute of free space that can be moved, and the grid point for which the potential has not been set by the previous step (that is, the potential value remains 0 (Registered grid points). Hereinafter, the setting range of the potential is expanded as Step 3, Step 4, ....

D) Completion of potential map creation work (step 10) When the potential setting area is expanded by the above-mentioned sequential operation, the starting point S will eventually be reached as shown in (5) of FIG.
To reach. Looking at the figure, although the number of steps here reaches 10, it is noticed that there are still unregistered grid points. In fact, even if the potential setting range is further expanded, the potentials set thereafter will handle values larger than the potential 11 found at the starting point, and they can be selected as the constituent points of the work route. There is no nature. Therefore, in the present invention, at the same time when the position of the starting point is recognized, the map creating operation is terminated and the operation route of the end effector is searched.

(2) Search Planning of Work Route Here, the procedure for generating a work trajectory using the potential map shown in (5) of FIG. 1 will be described.

First, the starting position S is selected as the first point of the route from the movement numbers (potentials) registered as ll. Next, the search for the second point is selected from the lattice points adjacent to the first point. It is possible that the neighboring grid points with respect to the starting point may be the entry prohibition points of obstacles for which no potential is set, and at most 8 points correspond to selection candidates (usually 26 points for 3D model). . Considering that the optimum grid for moving to this adjacent grid point is the grid point having the smallest potential value, the grid point having the potential value 9 is obviously the adjacent grid point selected as the second point of the work route in this case. It becomes a point.

Next, the third point is searched from the adjacent lattice points for the selected potential 9 by the same operation. Hereinafter, by repeating this optimum search operation, a route finally reaching the target point G having a potential value of 1 is found, which is a set of lattice points having inverted movement numbers in (6) of FIG. (11,9,8,7,6,4,3,2,1)
Is.

2. Collision Avoidance Model with Obstacles In the present invention, the end effector is represented by a sphere having a radius r circumscribing it, and the trajectory along which the center of this sphere moves is defined as the work trajectory of the end effector. In order for the end effector not to contact the obstacle, its center needs to be separated from the obstacle surface by r + δ as shown in (1) of FIG. Here, δ means a clearance for avoiding a collision.

However, if the distance from the obstacle is calculated every time the end effector moves during the search of the work route, the calculation efficiency is very poor. Alternatively, if a strip-shaped portion of r + δ that is conservatively expanded in the same direction from the obstacle surface as shown in (2) of FIG. 2 is set as the entry prohibition area, the work trajectory of the hand that does not collide with the target is this. It is derived from the route search of the space to be excluded.

3. Control of Work Trajectory by Using Standby Potential In the generation of the work trajectory of the end effector, it is an essential condition that avoidance of collision with obstacles existing in the environment is completely achieved. For a spot with a risk of radioactivity, dust, etc. or a local narrow space where it is difficult to move with sufficient margin within the plant,
It is important to avoid them as much as possible from the viewpoint of robot arm protection, and this issue is addressed. "As much as possible" means that planning is given to move the area depending on the environment map creation situation, and it is handled differently from the model that completely prohibits movement such as obstacles. Become a feature.

According to experience, in such a case, by directly adjusting the thermal conductivity or the electrical conductivity in the equation expressing the temperature field or the electric potential field, the assumed potential distribution of the field is successful in controlling the path. Well connected Although there is no effective control parameter such as thermal conductivity in the present invention that handles the movement of the hand with the virtual potential field as the environment, the effect of suppressing the spread of the potential is taken into account when the environment map is created and designated. Try to find a route that avoids the area.

Therefore, it is assumed that, with respect to the work of creating a map in a designated area to be controlled, the movement from a certain grid point of interest to its adjacent grid point temporarily waits until a certain condition is satisfied. In other words, the movement number setting of adjacent grid points is suspended for a while, but in the free space other than the designated control area during that period, the movement numbers continue to be registered at the adjacent grid points one after another from the start of calculation as usual. move on. Here, when the movement to the destination is interrupted, the state amount equal to the number of times of stepping is designated as the standby potential, and is designated by calculation input, and this input designated value is set as the potential of the grid point of interest at the time when stepping is started. It is assumed that when the state quantity (potential) to which (standby potential) is added matches the number of steps for creating the environment map, this state quantity is assigned to the adjacent grid point in the control area.

For example, in the area where the standby potential is set to 5, five steps are required from the registration of the potential at the current position to the registration of the potential of the adjacent grid point. During this stepping, the setting of the potential is expanded five times in the normal free space (standby potential 1), and the movement path of the hand may be preferentially constructed outside the specific control area by this control operation. Occurs.

Below, the control effect of the work route when the standby potential is designated in a specific area is the same as in FIG.
In a three-dimensional environment. FIG. 3 shows the creation of a special environment map by the standby potential and the optimum detour trajectory determined using the map when a specific control target area exists in the environment.

The starting point S and the target point G are displayed in (1) of FIG. The shaded area near the center of the calculation model is a control target area in which the standby potential is set to 5, and the other areas are free spaces in which the normal value 1 is set.

Next, the potential map is created by the method shown in FIG. 1 from the target point G as a starting point to the controlled area. (2) of FIG. 3 shows the state of step 3 in which the potential around the lattice point having the potential of 3 is set. Usually, if the potential of the point of interest is 3,
The potential of the adjacent grid point (the grid point marked with a circle) is 4, but at the adjacent grid point (the grid point marked with a circle in the shaded area) that belongs to the control target area where the standby potential is designated as 5. , 8 is added because this numerical value is added.

Then, in step 4 of (3) of FIG.
The potential of the adjacent grid point with respect to the grid point of interest having the potential of 4 is set to 5. On the other hand, when the adjacent grid point to the target grid point belongs to the control target area, as shown in the figure, the potential 4 of the target grid point in step 4 is used.
It is noted that the standby potential 5 is added to. Normal,
Since the potential of the grid point where the potential is newly set matches the next step number, it becomes the grid point of interest in the next step.

However, since a value larger than usual (8 in this case) is set in the control target area, the grid marked with 8 does not become the grid point of interest in the next step.
In this case, the potentials of the adjacent lattices are not set and wait until the total number of steps matches the potential value of the lattice of interest in the control region. However, this is a principle, and the potential of the adjacent grid point that is aimed at is not always fixed after the grid point of interest has stepped a specified number of times. It may be set.

For example, the lattice point marked 8 in step 4 seems to set the potential 13 to the adjacent lattice point immediately below when it reaches step 8, but in reality, from among the lattice points in work step 6, Since the setting of the potential to this adjacent point precedes, the result is set to 11 instead of 13.

When the steps of creating a map are advanced in the same manner, step 9 in (4) of FIG. 3 is performed, and the grid point of interest at this point is a grid point having a potential of 9. The grid point having this movement number exists in the free space and the control area respectively, and the former sets 10 for the adjacent grid point, but the latter waits only when the value at the adjacent grid point is not set. 14 with the added potential is set at the two grid points marked by the circles in the shaded area.

By repeating the above operation, the potential at the starting point S is determined as shown in (5) of FIG.
The work of creating the potential map with the target point ends at this point (step 11). Finally, the movement path of the hand searched along the maximum gradient from the starting point is the sequence (12,10,8,7,6,5) having the reversed movement number.
3, 2, 1) is shown in (6) of FIG. Apparently, the generation of a route that automatically bypasses a specific control area is confirmed.

As described above, the standby potential is understood as an effective means for avoiding the control area. By the way, in the case of a calculation scenario in which the target point cannot be reached unless the work route passes through the controlled area, the simulation result that correctly recognizes the situation, that is, the optimum route that passes through the controlled area is generated. It is reconfirmed in the process of constructing the work route of the control target area that it has a nature that is essentially different from the obstacle.

4. Reconstructing the work trajectory by thinning out the relay points Since the work path of the hand obtained by the trajectory generation algorithm is composed of multiple grid points (relay points),
If there is a spatial allowance, the point sequence configuration that deletes (jumps) some of them reduces the number of movement steps,
Often a simple trajectory is obtained. In particular, the current method, which is based on the principle of movement on the grid points of free space, is directly affected by the arrangement of various obstacles arranged in the environment, and thus has the drawback that the course of the generation route is frequently changed. is there. In order to minimize the drive error generated in the joint part of the robot arm, the path along which the hand moves is preferably a smooth trajectory with little fluctuation and simple linear movement. Therefore, the generated route is modified and processed according to the following procedure to reconstruct a satisfactory route.

That is, since the work trajectory of the hand generated in the present invention is composed of a plurality of lattice points (relay points),
By thinning out some of the relay points, it is expected that the desired work route will be reconstructed. To explain this in the figure, FIG. 7 shows an example of a simple trajectory generation in which the target point cannot be reached unless the course is changed in the middle of the movement even though the obstacle does not exist. By reconstructing the route generated using the point decimation operation algorithm, the route along which the hand moves can be made the most ideal straight line.

Therefore, in order to explain the basic idea of the present invention, consider a simple work route generated on the two-dimensional plane of FIG. 8 (1) showing the movement route generated by the virtual potential method. . In FIG. 8 (1), the radius for the end effector (hand attached to the tip of the robot hand) and the clearance for avoiding collision are set to 0 for convenience.

First, FIG. 8 showing an interference discrimination route with an obstacle when it is assumed that the vehicle directly moves between the starting point and the target point.
As shown in (2), the state of collision with an obstacle is examined assuming a route that directly connects the starting point S and the target point G. Figure 8
In (2), it is clear that interference with the obstacle has occurred.

In this case, as shown in FIG. 8 (3), which shows the search route of the point closest to the target point from the point of departure that can move without interfering with obstacles, the grid on the original work route is shown. Focusing on the grid point P ₁ which is one before the target point from among the points, the interference with the obstacle is examined assuming a straight line path connecting the starting point and this point. When the relay point P ₁ also interferes with the obstacle, the grid point P ₂ is further focused and the same processing is repeated.
By the above processing, the relay point P where interference with the obstacle does not occur
By finding ₃ , one linear trajectory SP ₃ is obtained.

Next, using the relay point P ₃ as a temporary starting point, a search for a relay point that can be thinned out is performed between P ₃ and G by the same method. Therefore, the process ends when a jumpable relay point as shown in FIG. 8 (4) showing the discovery of a relay point that can move without interfering with an obstacle and reconstruction of the route is the target point. To do.

Regarding the interference check described above, an arbitrary point (x ₁ , y ₁ , z ₁ ) and another point (x ₂ , y ₂ ,
z ₂₎ and the straight line equation connecting the _{(x-x 1) / (} x 2 -x 1) =
(Y−y ₁ ) / (y ₂ −y ₁ ) = (z−z ₁ ) / (z ₂ −
z ₁ ) is used to determine the collision with the designated obstacle (prohibited area) D. This can be done by directly inspecting with a table look whether or not each point on the straight line captured at a certain interval belongs to the entry prohibited area D.

In this way, an operation (jump operation) of thinning out some of the grid points of the path initially generated
In the case of, since the movement on a straight line with a few course changes is prioritized, it is expected that the movement distance will be shortened.

5. Special Note (Chain mechanism consisting of link and joint of robot arm) So far, we have mentioned the establishment of the work trajectory of the hand in robot planning, but this is the autonomous operation of the work that is coupled with the problem of determining the arm posture of the manipulator. Has meaning to Therefore, in order to successfully execute the operation of the arm, it is important to consider multiple provision of information in the design process of the trajectory planning of the hand so that the entire planning can be smoothly deployed. In fact, when deciding the movement of the robot arm, the arm posture that satisfies the position and posture of the hand cannot be physically realized due to the severe movement conditions imposed on the joints and the complexity of the work environment. It is fully anticipated that once they occur, joint solutions cannot be determined. When such a situation is encountered, the initial calculation work plan is not abandoned, that is, attention is paid to other information regarding the trajectory of the hand generated without changing the work start position and the end position.

As described above, the role of the hand was to create an optimal trajectory for creating a virtual potential distribution state from the target point G to the starting point and then returning to the original with the maximum gradient. Instead of uniquely determining the adjacent point following this starting point with the minimum potential, here, among the 26 neighboring points,
The candidates excluding the 9 points far from the target point G are taken out, and the point sequence heading toward the target position with the maximum gradient is focused on. That is, the locus from the starting point to the target point (up to 17
We propose to discover feasible trajectories buried in the optimal search method by registering from the group) in ascending order of reach. Needless to say, if they are used as a means of resolving the insolvency of the inverse problem, it is considered to be useful for the autonomy of intelligent robots aiming at improvement of work ability and flexible adaptation to the environment.

The attitude (orientation) of the arm tip required to handle the inverse kinematics problem can be freely specified in a fixed direction between the work start point and the work end point or by sequential input. , Re-examination of posture data by general-purpose handling using general rotation transformation and interpolation technique will also be useful for solving the above inverse problem. In any case, these can be selected depending on the way of working with the manipulator, and since they are all technically established, they are not mentioned here.

[0053]

[Example] In a three-dimensional environment in which various obstacles that hinder the performance of work are present, it is assumed that the robot arm moves from a designated position to another target position while human intervention and sensor support are performed. In order to autonomously perform the work instead of doing it, it is necessary that the work locus in which the tip (hand) of the arm moves from the work start position to the end position is completely determined. As an embodiment of the present invention that satisfies this requirement, a work trajectory in which the robot hand can safely move without contacting an object (obstacle) at all is reproduced by computer simulation. Furthermore, with the comparison of the calculation time with the potential induction method by the temperature field,
The effectiveness of the present invention having the features described in the claims such as the control of the work trajectory by the standby potential and the reconstruction of the work trajectory by the thinning operation will be verified.

(Embodiment 1) A cubic environment having a side of 2 m as shown in (1) and (2) of FIG. 4 is divided into a 41 × 41 × 41 square lattice, and further, (3) and (3) of FIG. As shown in 4), a work start position S (350, 1) of a robot hand for performing maintenance of a simple experimental device placed therein.
050,700) (unit mm) and work end position G (15
00, 800, 900). In addition, (2) in FIG.
Is a calculation model in which a control target area (depicted by a dotted line) is specified in this environment, and (3) and (4) of FIG. 4 show the specifications of the system including the control target area.

Now, as shown in (3) and (4) of FIG. 4, an end effector simulated by a sphere having a radius of 245 (mm) assumes a clearance of 50 (mm) and has two points SG.
Reproduce the situation of moving between. When the environment map of the action world of the robot hand based on the obstacle avoidance model and the virtual potential described above is created, and the required trajectory for guiding to the target position G through the optimum search is set in the environment with a special control area. Compare with.

That is, (1) of FIG. 4 is a calculation model in which the environment including a simple experimental device and the work start position and end position of a spherical hand are designated, and (2) of FIG. 3 is a calculation model in which the control target area is specified in the environment of FIG.
(3) and (4) are calculation models in which the arrangement of the experimental equipment in the environment and the specifications of the control area are described.

The work trajectories shown by the dotted lines in (1) and (2) of FIG. 5 are the optimum work trajectories of the hand when the control target region is not taken into consideration, while the corresponding solid lines indicate the standby potential in the control region. This is a calculation example designated as 10. Tables 1 and 2 are comparisons of the coordinates and the travel distances of the two travel routes depicted in FIG. That is, Table 1 shows the work locus of movement of the end effector (robot hand), and Table 2 shows the locus of movement of the end effector (robot hand) when the control target area is designated.

[0058]

[Table 1]

[0059]

[Table 2]

The trajectory indicated by the dotted line in FIG. 5 shows a situation in which the work area is a narrow work area where pumps, valves, etc. are present, but the object can be successfully moved without interfering with surrounding objects in calculation. what if,
Considering the local area near the pump as the controlled area and redesigning the trajectory plan with sufficient margin to avoid it,
Although the moving distance to reach the target position G becomes long, the result (a solid line in the figure or a table) in which a safe trajectory of the hand is generated can be confirmed.

That is, (1) of FIG. 5 shows an overall view of the comparison of work trajectories of the end effector (robot hand) depending on the presence or absence of the control area, and (2) of FIG. 5 depends on the presence or absence of the control area. The comparison (top view) of the work trajectory of the end effector (robot hand) is shown. The dotted line is the calculation without setting the standby potential (calculation processing time: 0.0256
[Second], moving distance: 1366 [mm]), and the solid line is a calculation in which the standby potential is specified in the control area (calculation processing time: 0.0363 [second], moving distance: 1866 [m].
m]).

Next, another calculation example of the process of performing the decimation operation of the relay points on the generated work trajectory will be shown. Under the same working environment, the start position S and the end position G are (4
00,600,450) and (1650,1450,90)
The movement trajectory of the hand, which is derived by defining 0), is given by the dotted lines of (1) and (2) in FIG. On the other hand, the solid line in the figure shows the work route that has been processed into a trajectory with few course changes through the thinning algorithm. Most of this relay point was deleted,
It can be seen from Table 3 that the result of prioritizing the linear movement is that the movement distance is shortened by about 10%. Table 3 shows a comparison of work trajectories of the end effector (robot hand) with and without thinning operation. Thus, the reconstruction of the work route by the thinning operation is an effective method for processing the original model-based optimum search route into a shorter trajectory.

[0063]

[Table 3]

That is, (1) of FIG. 6 is an overall view of comparison between the work trajectory of the end effector (robot hand) guided by the optimal search method and the work trajectory reconstructed by the thinning operation of its relay point. 6B shows a top view of a comparison between the work trajectory of the end effector (robot hand) guided by the optimal search method and the work trajectory reconstructed by the decimation operation of the relay points. The dotted line is the trajectory calculation (calculation processing time: 0.039
6 [seconds], moving distance: 1804 [mm]), and the solid line is the trajectory calculation with the thinning operation (calculation processing time: 0.04).
29 [seconds], moving distance: 1615 [mm]).

Finally, the calculation processing time until the work path of the hand is calculated through the conventional calculation of the temperature field has a tendency to vary greatly depending on the number of grid points dividing the space. The present invention does not show a significant change in the calculation time, and the calculation result for the 41 × 41 × 41 lattice division model described above is approximately several thousandths of the time required for orbit determination from the calculation of the three-dimensional temperature field. Was found to be processed in. It is evaluated that the solution to this problem was sufficiently achieved by such an overwhelming speedup.

[0066]

As described above, the algorithm of the present invention for guiding the robot hand from the designated position to the target position in a work environment in which there are obstacles of different sizes and shapes is more effective than the conventional method. Since the complicated calculation of the three-dimensional potential field is not included, the calculation processing time is extremely improved, and a safe hand work trajectory that does not interfere with the object at all is automatically generated. The results are applicable to all robot manipulator tasks.

[Brief description of drawings]

FIG. 1 shows an outline of creating a virtual potential map stepwise in a two-dimensional environment and an optimum trajectory determined by using the environment map.

FIG. 2A shows a collision avoidance condition for an end effector (robot hand) and an obstacle, and FIG. 2A shows an obstacle collision avoidance condition for an obstacle.

FIG. 3 shows a case where a specific control target area exists in the environment,
The outline of the special environment map created by the standby potential and the optimal detour trajectory determined using the map are shown.

FIG. 4 shows (1) a calculation model in which an environment including a simple experimental device and a work start position and an end position of a spherical hand are designated, and (2) designates a control target area in the environment of (1). The calculation model is shown, and (3) and (4) show the calculation model in which the arrangement of the experimental equipment in the environment and the specifications of the control area are described.

FIG. 5 (1) shows an overall view of the comparison of work trajectories of the end effector (robot hand) with and without the control region, and (2) shows comparison of work trajectories of the end effector (robot hand) with and without the control region. FIG.

FIG. 6 (1) shows an overall view of comparison between a work trajectory of an end effector (robot hand) guided by the optimal search method and a work trajectory reconstructed by thinning operation of its relay point, and (2) Shows a top view of a comparison between the work trajectory of the end effector (robot hand) guided by the optimal search method and the work trajectory reconstructed by the decimation operation of its relay points.

FIG. 7 is a diagram showing a work route accompanied by a course change.

FIG. 8 is a diagram showing reconstruction of a work route by a thinning operation.

Claims (4)

[Claims] 1. A model of obstacle avoidance and a virtual potential in order to determine a work locus in space of a hand (end effector) attached to a tip of a robot arm in a three-dimensional work environment including an obstacle. Creating a real world world map (environment map) by using, and generating a general-purpose robot hand work trajectory that searches the relay point of the route under optimum conditions while referring to the potential (movement number) registered in the map. And control method. 2. When a specific control target area different from the obstacle exists in the work environment to which claim 1 is applied, the standby potential is designated to temporarily suspend registration of the movement number in that area. With the function to avoid the control area from the creation of the environment map and the route search that incorporate the measures to
Generation and control method of work trajectory of robot hand. 3. With respect to the work trajectory of the robot hand guided in claim 1 or 2, the shortest trajectory with few path changes is re-established from the decimation (jump) operation of the relay point for giving priority to the movement on a straight line. How to create and control the work trajectory of the robot hand. 4. The posture of the robot arm (that is, the operation of the chain mechanism composed of the link and the joint) corresponding to each point on the work trajectory guided in claim 1, 2 or 3 is not physically realized. A method for generating and controlling a work trajectory of a robot hand, characterized by robustness that utilizes feasible trajectory candidates buried in an optimal search algorithm when possible.