JP6908908B2 - Robot arm path generator and path generator - Google Patents

Description

The present invention relates to a route generation device and a route generation program that generate an operation path of a robot when picking a randomly piled work by using a robot.

Conventionally, there has been known a random picking technique in which a pile of works is photographed at random, the photographed image is analyzed to recognize the work, and a movement path of a robot arm is generated based on the recognition result. For example, Patent Document 1 discloses a technique of selecting a candidate for a gripping position of a robot hand using a work shape model and generating an operation path of the robot hand based on the selected candidate.

Japanese Unexamined Patent Publication No. 2008-272886

However, according to the technique of Patent Document 1, it is necessary to register the work in the robot and memorize the form of the work before starting picking, which causes a problem that it takes time and man-hours. In addition, there is a problem that the work cannot be proceeded if there is no engineer who can register the work at the site.

Therefore, an object of the present invention is to provide a route generation device and a route generation program capable of automatically generating an operation path of a robot without the trouble of registering a work.

In order to solve the above problems, the route generator of the present invention has the following features.
(1) A device for generating an operation path of a picking robot, which includes a control means for controlling the picking robot, the picking robot includes a hand and an arm for supporting the hand, and the control means is the position of the hand. The control means generates an operation path based on a pose that satisfies a predetermined condition, including a pose generation means for generating a pose including information and posture information.

(2) The pose includes a candidate for a gripping pose in which the hand grips the work, the pose generation means includes the candidate generation means for generating the candidate, and the candidate generation means determines the pose of the work and the size of the work. Based on this, one candidate is generated, and the one candidate is rotated around a predetermined rotation axis to generate a plurality of other candidates.

(3) The pose includes an approach pose formed by the hand before the grip pose and a detachment pose formed by the hand after the grip pose, and the pose generation means includes the approach pose and the detachment pose based on the candidates. To generate.

(4) The pose generation means calculates an inverse kinematics solution for the generated pose and excludes candidates associated with a pose for which there is no solution.

(5) The pose generation means detects interference between objects based on a predetermined interference condition, and excludes candidates associated with the pose in which the interference is detected.

(6) At this time, the interference conditions include an H region that virtually represents the hand, an A region that virtually represents the arm, a Y region that virtually represents the container for the work, and the work and the work. and HW region representing the hand virtually, H, a, Y, H where each adding a predetermined margin to the HW area, a, Y, and HW extension region, and J regions included from the hand tip at a predetermined distance, for ,
(Condition 1) The H, A, and Y regions do not interfere with each other.
(Condition 2) The H, A, Y, and HW regions do not interfere with each other.
(Condition 3) The H, A, and Y regions do not interfere with each other, or even if they do interfere, the interference position corresponds to the J region.
(Condition 4) The H, A, and Y extended regions do not interfere with each other.
(Condition 5) The H, A, Y, and HW extension regions do not interfere with each other.

(7) The pose generation means is for the generated pose.
(B) Satisfy (Condition 3) at the joint position of the approaching pose,
(B) Satisfy (Condition 3) at the candidate joint position,
(C) Satisfy (Condition 3) and (Condition 5) at the joint position of the withdrawal pose.
If any one of (a) to (c) is not satisfied, the candidates associated with the pose are excluded.

Based on the candidates, the pose generating means generates an approaching motion in which the hand moves from the approaching pose to the grasping pose and a leaving motion in which the hand moves from the grasping pose to the leaving pose, and the pose generating means generates, for each generated motion, the pose generating means.
(2) Subdivide the approaching motion in chronological order and satisfy (Condition 3) at the joint position for each hour.
(E) Subdivide the withdrawal movement in chronological order and satisfy (Condition 3) at the joint position for each hour.
Is inspected, and if either (2) or (e) is not satisfied, the candidates associated with the operation are excluded.

Further, in order to solve the above problems, the route generation program of the present invention includes the control means according to any one of (1) to (8) above.

According to the path generation device and the path generation program of the present invention, the gripping pose is generated based on the pose of the work and the size of the work, and the operation path of the picking robot is automatically generated based on the generated gripping pose. There is no need to register the work in advance, and it has the excellent effect of reducing labor and man-hours.

It is the schematic of the picking system using the path generator which shows one Embodiment of this invention. It is a block diagram of the picking system of FIG. It is a schematic diagram which shows the process of estimating the axis candidate point from the normal vector of a point cloud data. (A) provisional radius r _i: a _{(r min ≦ r i ≦ r} max i = 0~100) models generated from, in schematic view showing a state of rotating the shaft candidate points around z-axis with the origin O be. (B) It is a graph which plotted the average value of the degree of agreement for every _{temporary radius r i.} It is (a) perspective view, (b) side view, (c) front view which shows the gripping position of a work. It is explanatory drawing about the generation of the candidate of a grasping pose. It is explanatory drawing about the generation of approach pose and withdrawal pose. It is a schematic diagram which shows (a) H, A, Y region about condition 1 and (b) H, A, Y extended region about condition 4. It is a schematic diagram which shows (a) H, A, Y, HW region about condition 2, and (b) H, A, Y, HW extended region about condition 5. It is a schematic diagram which shows the enlarged view of (a) H, A, Y, J region and (b) J region concerning condition 3. It is a flowchart which shows the process flow of a route generation apparatus, a route generation program. It is a flowchart which shows the flow of the process of calculating the pose and size of a work.

Hereinafter, an embodiment in which the present invention is embodied in a calculation PC used in a picking system will be described with reference to the drawings. In the following description, a world coordinate system based on the entire picking system, a work coordinate system based on each work W, and a hand coordinate system based on the hand 32 of the picking robot 3 are used. Specifically, in the world coordinate system, the direction from the back to the front of the picking robot 3 is the x-axis positive direction, the direction from the left side to the right side in the front view is the y-axis positive direction, and the vertical direction from the bottom to the top is the z-axis positive direction. The direction is defined as the x-axis positive direction along the central axis passing through the center O of the cylindrical work W, and the direction from the left side to the right side of the columnar paper. The y-axis positive direction and the vertical direction from bottom to top are the z-axis positive directions (see FIGS. 5 and 6), and the hand coordinate system is the z-axis positive direction from the base end to the tip of the hand 32 (see FIGS. 5 and 6). (See FIG. 7). Further, the "pause" includes two elements, a position indicating a certain point in a predetermined xyz space and a posture indicating the amount of rotation with the xyz axis as the rotation axis.

FIG. 1 is a schematic view of the picking system 1. The picking system 1 includes a picking robot 3 that picks randomly stacked work Ws, a 3D scanner 4 that scans the three-dimensional shape of the work W surface, and a picking robot 3 based on measurement intermediate data received from the 3D scanner 4. It is composed of a calculation PC 2 that functions as a route generation device that calculates the route on which the camera operates. At this time, a route generation program may be installed on the arithmetic PC 2 to implement the function. The 3D scanner 4 is also used to scan the pallet 7 as a container for the work W and acquire the 2D data and the 3D data of the pallet 7.

The picking robot 3 is composed of a hand 32 that grips the work W and an arm 31 that supports the hand 32 in a predetermined posture. The hand controller 6 that controls the hand 32 and the robot controller 5 that controls the picking robot 3 are connected to the calculation PC 2 and operate the hand 32 or the picking robot 3 based on the control signal received from the calculation PC 2. At this time, it is also possible to mount the route generation device on the robot controller 5 or the hand controller 6 and install the route generation program on the robot controller 5 or the hand controller 6.

As shown in FIG. 2, the arithmetic PC 2 includes a control unit 11 that controls the picking robot 3, a storage unit 12 that stores data calculated by the control unit 11, a 3D scanner 4, a robot controller 5, a hand controller 6, and the like. It consists of an interface (not shown) for exchanging signals and data with an external device.

The control unit 11 includes a pallet detection unit 21 that detects the pallet 7 based on the 3D data of the pallet 7, and a work recognition unit 13 that recognizes the position, axial direction (posture), and size of the work W to be gripped. It is composed of a route generation device that generates an operation route of the picking robot 3 or a route generation unit 22 that functions as a route generation program. The operation path at this time indicates an operation path in which the picking robot 3 grips the work W.

The pallet detection unit 21 reads the point cloud data extracted from the pallet edge portion from the storage unit 12, collates it with the scanned 3D data, detects the pallet 7, and positions the pallet 7. In addition to the image data, the depth information of the pallet 7 is registered in the storage unit 12 in advance.

The work recognition unit 13 includes a point cloud data generation unit 14 that converts the measurement intermediate data received from the 3D scanner 4 into point cloud data P, and a size estimation unit 15 that estimates the size of the work W based on the point cloud data P. It is composed of a model generation unit 16 that generates a model M using the estimated size as a parameter, and an axial direction estimation unit 17 that estimates the axial direction using the generated model M.

The path generation unit 22 is composed of a grip pose generation unit 24 that generates a pose in which the picking robot 3 grips the work W, and the grip pose generation unit 24 is a candidate generation unit that generates a plurality of poses that are candidates for the grip pose. Includes 25.

Here, the point cloud data P refers to an array (x, y, z) of three-dimensional point information in the world coordinate system. The flow of acquiring the point cloud data P is as follows.
(1) The PC 2 issues a measurement trigger to the 3D scanner 4.
(2) The 3D scanner 4 scans the three-dimensional shape of the surface of the object to be measured.
(3) The 3D scanner 4 transmits the measurement intermediate data to the calculation PC 2.
(4) The arithmetic PC 2 converts the measurement intermediate data into the point cloud data P.
In some cases, the 3D scanner 4 may perform conversion to the point cloud data P. However, since the generation of the point cloud data P requires a large amount of calculation, it is generally converted into the point cloud data P by the arithmetic PC2 using a dedicated library. Further, examples of the three-dimensional shape scanning method include a projector method and a stereo camera method.

The model generation unit 16 inputs the radius r as a parameter, and generates a model of the work W composed of the side surface portions of the semi-cylinder having the radius r and the height r.

Hereinafter, the configuration of the work recognition unit 13 will be described.
As shown in FIG. 3A, the axial direction estimation unit 17 first estimates the normal vector n for each point p included in the point cloud data P. Then, the point q obtained by advancing the point p in the direction opposite to the normal vector n by the radius r is set as the axis candidate point of the cylinder including the curved surface of the normal vector n on the side surface, and the axis candidate point q is registered in the storage unit 12. .. The point q is obtained by the following mathematical formula 1.

Next, the axial direction estimation unit 17 calculates the degree of integration of the axis candidate points q and identifies the points q having a high degree of integration. Then, the origin O of the model M generated in advance is arranged at the point q, and the degree of coincidence D between the model M and the point cloud data P is calculated. The degree of coincidence D is obtained by dividing "the number of points p arranged near the model M" by "the total number of points p included in the semi-cylinder having the radius r and the height r with the model M as the side surface". Be done. The axial direction estimation unit 17 rotates the model M around the z-axis from 0 degrees to 180 degrees, calculates and holds the degree of coincidence D every 5 degrees of the z-angle. Then, the z angle having the highest degree of coincidence D is registered in the storage unit 12 as the axial direction of the model M.

The size estimation unit 15, as shown in FIG. 4 (a), the minimum radius r assumes _min and a maximum radius r _max, r _i (r _min ≦ obtained by Equation 2 below r _i ≦ r _{max: i} = _{Models M i} with radii from 0 to 100) are generated, and the axial direction is estimated for each _{model M i.} Since the generation of the model M _i and the estimation of the axial direction are the same as the processing by the model generation unit 16 and the axial direction estimation unit 17, the description thereof will be omitted.

After that, the size estimation unit 25 calculates the _{degree of coincidence D i between the model M i} and the point cloud data P for _{each radius r i} and the point q with a high degree of integration. Matching degree D _i, like matching degree D, include a "number of p points located near the model M _i", the semi-cylindrical with a radius r _i, the height r _i to the side surface of the "model M _i It is obtained by dividing by "the total number of points p".

Size estimator 25, a model M _i, is rotated from 0 degrees around the z-axis to 180 degrees, z angle 5 degrees each and the highly integrated each axis candidate point q matching degree D _i (r _i, q) Ask for. The size estimator 25, matching degree D _i (r _i, q) maximum maxD _i (r _i, q) of the calculated, further, maxD _i (r _i, q) of the average with respect to q, the average determine the value maxD _i (r _i). Thereafter, size estimator 25, maxD _i (r _i) outputs a radius r _i which maximizes the actual radius r.

In FIG. 4 (b), shows a graph plotting the maxD _i (r _i). Here, r _min = 2.0 mm and r _max = 8.0 _{mm, and max D i} is plotted _{every time the radius r i} is increased by 0.06 mm. In this example, the _{area around 4.6 mm, which is the r i} with the largest _{maxD i,} is estimated to be the actual radius r of the work W.

The storage unit 12 stores the model M in descending order of the degree of agreement. Specifically, since the point cloud data P includes the surface shapes of a plurality of works W, the storage unit 12 calculates a model M facing the same axial direction as each work W, and models the models in descending order of degree of coincidence. Memorize M.

Subsequently, the configuration of the route generation unit 22 will be described.
First, as shown in FIG. 5, the candidate generation unit 25 of the pose generation unit 24 _{grips the model M based on the coordinates (x o} , _yo , z _o ) of the origin O, the vector in the axial direction, and the radius r. Generate one pose candidate. Further, the gripping position is determined by a support method in which the hand 32 supports the work W. For example, in the case of a suction type hand, the gripping position is the position Pa (x _o , _yo , z _o + r), and in the case of the pinching type hand, the gripping position is the position Pb (x _o , _yo , z _o ). In the case of a scoop-up hand, the position is Pc (x _o , _yo , z _o −r). At this time, the gripping positions Pa, Pb, and Pc are not the contact points with the actual hand 32, but are, for example, the midpoints of the left and right claws in the case of a pinch-type hand, and in the case of a "" type scoop-up type hand. Indicates the lower end of "". The candidate generation unit 25 generates a candidate 56 for one gripping pose based on any of these positions Pa, Pb, and Pc.

As shown in FIG. 6, the candidate generation unit 25 generates one candidate 56 based on the pose and size of the work W, rotates the generated one candidate 56 around a predetermined axis, and has a predetermined step angle. Another candidate 56 is generated for each. Specifically, one candidate 56 is generated based on the origin O of the model M for the position, the x-axis of the work coordinate system (the axial direction of the model M) for the posture, and the radius r of the model M for the size (FIG. 6 (a)). Then, one candidate 56 is rotated around the central axis (x-axis of the work coordinate system, + direction) in a range of −90 to +90 degrees in 5 degree increments to generate a plurality of other candidates 56. Next, the one candidate 56 mentioned above is rotated 180 degrees around the Wz of the work coordinate system to further generate a new one candidate 56 (FIG. 6B), and the new one candidate is generated. The 56 is rotated in 5 degree increments in the range of -90 to +90 degrees around the central axis (x-axis of the work coordinate system, -direction) to generate a plurality of other candidates 56, and finally 73 pieces. Candidate 56 is generated. As shown in FIG. 6C, the candidate generation method can be applied to a planar work and a ring-shaped workpiece. For example, a planar workpiece can be adsorbed or a ring can be generated. One candidate for a pose to be inserted into a shaped work can be generated, and 73 candidates 56 can be generated by rotating the range of −180 degrees to +180 degrees in 5 degree increments with the z axis in the work coordinate system as the central axis. The step size can be changed as appropriate.

Based on each of the candidates generated in this way, as shown in FIG. 7, the pose generation unit 24 has an approach pose 52 in which the hand 32 approaches the candidate 56 and a departure pose in which the hand 32 leaves the candidate 56. Generate 54. Specifically, the approach pose 52 is a movement of the pose of the candidate 56 as a base point in the direction from the tip end portion of the hand 32 toward the base end portion (hand coordinate system z-axis, − direction) by a predetermined distance. The detachment pose 54 is a movement of the candidate 56 in the vertically upward direction (world coordinate system z-axis, + direction 58) by a predetermined distance from the pose of the candidate 56 as a base point.

Further, at this time, the pose generation unit 24 sets the linear interpolation operation generated between the candidate 56 and the approaching pose 52 as the approaching operation 57, and the linear interpolation operation generated between the candidate 56 and the leaving pose 54 as the leaving operation. It is set to 58.

Here, as shown in FIGS. 8 to 10, the storage unit 12 stores predetermined area information for detecting interference between objects. The area information includes an H region that virtually represents the hand, an A region that virtually represents the arm, a Y region that virtually represents the container for the work, and the work W and the hand 32 that holds the work W. and HW region to represent, H, a, Y, H obtained by adding a predetermined margin to the HW area, a, Y, and HW expansion region includes J region included from the hand 32 tip to a predetermined distance d.

Next, the pose generation unit 24 describes the following (condition 1) to (condition 5) interference conditions for each of the generated candidates 56, the approach pose 52, and the withdrawal pose 54, (a) to (e). Carry out the inspection. If any of the inspections (a) to (e) fails, the pose generation unit 24 excludes the candidate 56 to be inspected.
(Condition 1) The H, A, and Y regions do not interfere with each other (see FIG. 8A).
(Condition 2) The H, A, Y, and HW regions do not interfere with each other (see FIG. 9A).
(Condition 3) The H, A, and Y regions do not interfere with each other, or even if they interfere with each other, the interference position corresponds to the J region (see FIG. 10).
(Condition 4) The H, A, and Y extended regions do not interfere with each other (see FIG. 8 (b)).
(Condition 5) The H, A, Y, and HW extended regions do not interfere with each other (see FIG. 9B).
(B) There is an inverse kinematics solution for the approach pose 52, and (Condition 3) is satisfied at the joint position of the approach pose 52.
(B) There is an inverse kinematics solution for candidate 56, and (condition 3) is satisfied at the joint position of candidate 56.
(C) There is an inverse kinematics solution for the withdrawal pose 55, and (Condition 3) and (Condition 5) are satisfied at the joint position of the withdrawal pose 55.
(2) The approaching motion is subdivided in chronological order, and (Condition 3) is satisfied at the joint position for each hour.
(E) The withdrawal motion is subdivided in chronological order, and (Condition 3) is satisfied at the joint position for each hour.
At this time, if the inspection of (c) fails, the detachment pose 54 is translated toward the center position of the pallet 7 in the xy direction of the world coordinate system by a predetermined distance, and the (c) It is preferable to re-examine. The work W arranged at the corner of the pallet 7 can be separated from the pallet 7 and adjusted so as to avoid interference between the work W and the pallet 7. Further, in the case of subdividing in time series, it is preferable to inspect each 0.1% of the movement of the picking robot 3, that is, every 2 to 3 mm of moving distance. It is desirable that the margin amount when carrying out the inspections (a) to (e) is 2 mm or less.

Finally, the control unit 11 generates a path in which the picking robot 3 operates based on the candidate 56 generated by the pose generation unit 24 and not excluded by the pose generation unit 24. At this time, the control unit 11 reads out the start pose 51 and the end pose 55 stored in the storage unit 12 in advance, and automatically generates a route from the start pose 51 to the approach pose 52 under (condition 4). Further, a route from the withdrawal pose 54 to the end pose 55 is automatically generated under (condition 5). It is desirable that the margin amount when automatically generating the path from the start pose 51 to the approach pose 52 and the path from the departure pose 54 to the end pose 55 is about 5 mm.

For the automatic generation of the route, for example, the KPIECE algorithm can be used. When there are a plurality of candidates 56 that have not been excluded, routes are automatically generated in parallel for these candidates, and the fastest generated route is adopted as the actual picking route. On the other hand, when all the candidates 56 are excluded or the route is not generated within a predetermined time, the generation of the route is stopped and the model M having the next highest degree of matching is taken out.

Next, the processing flow of the route generation unit 22 having the above configuration will be described with reference to FIG. First, the 3D scanner 4 scans the pallet 7 in which the work W is randomly piled up, and detects the pallet 7 based on the scanned 3D data (S1). Next, the 3D scanner 4 scans the work W, the work recognition unit 13 recognizes the work W based on the scan result, and stores the model M as the recognition result in the storage unit 12 in descending order of the degree of matching ( S2). When the recognition of the work W is completed, the route generation unit 22 takes over the processing.

In the route generation unit 22, the candidate generation unit 25 of the pose generation unit 24 takes out one model M having the highest degree of matching from the storage unit 12 (S3), and generates the gripping pose candidate 56 based on the model M (S3). S4). When the candidate 56 is generated, the pose generation unit 24 inspects all the candidates 56 and excludes the candidate 56 that has failed the inspection (S5). When all the candidates 56 are excluded (S6: No), one model M having the next highest degree of matching is taken out (S3). When the candidates 56 remain, the control unit 11 automatically generates routes in parallel for all the candidates 56. If no route is generated within a predetermined time for all the candidates 56 (S7: No), one model M having the next highest degree of agreement is taken out (S3). On the other hand, when the route is generated within a predetermined time, the route that can generate the route fastest is output. After that, the picking robot 3 is controlled to operate according to the output path, and picking is performed (S8).

Here, the processing flow of the work recognition unit 13 will be described with reference to FIG. The 3D scanner 4 scans the work W and transmits the measurement intermediate data to the arithmetic PC 2. The arithmetic PC2 that has received the measurement intermediate data generates the point cloud data P (S21), and estimates the size assuming that the _{radius is r min to r max (S22).} After that, a model is generated using the real radius r (S23), and the model M is applied to the point cloud data P to estimate the axial direction (S24). Here, the main part of work recognition is the generation of a model using the real radius r and the estimation in the axial direction. This main processing is performed for all the work W included in the point cloud data P, and the storage unit 12 sorts and stores the model M in descending order of the degree of coincidence (S25).

According to the path generation device and the route generation program having the above configuration, the gripping pose 53 is generated based on the position, axial direction (posture), and size of the pose of the work W, and the generated gripping pose 53 is used as the work W. Since the candidate 56 is automatically generated by rotating around the axis, it is not necessary to register the work W, which has an excellent effect of reducing labor and man-hours. Further, since the route is automatically generated based on the H, A, Y, and HW expansion regions based on the generated candidate 56, when the picking robot 3 is actually operated, the arm 31, the hand 32, the work W, and the pallet are generated. It is possible to better prevent contact and collision with 7. Further, since (Condition 3) is provided and the interference in the region J is ignored, the sensitive detection is eliminated, and the tip of the hand 32 reaches the work W arranged at the corner of the pallet 7 for picking. It also has the effect of being able to.

In addition, the present invention is not limited to the above-described embodiment, and the configuration of each part can be arbitrarily changed and implemented without departing from the spirit of the invention. For example, the size of the margin can be set to a different value for each region.

1 Picking system 2 PC for calculation
3 Picking robot 4 3D scanner 5 Robot controller 6 Hand controller 7 Pallet 11 Control unit 12 Storage unit 13 Work recognition unit 14 Point cloud data generation unit 15 Size estimation unit 16 Model generation unit 17 Axial direction estimation unit 21 Pallet detection unit 22 Route generation Part 24 Pose generation part 25 Candidate generation part 31 Arm 32 Hand 51 Start pose 52 Approach pose 53 Grip pose 54 Withdrawal pose 55 End pose 56 Grip pose candidate 57 Approach movement 58 Withdrawal movement S area (Sh: H area, Sh': H extended area, Sa: A area, Sa: A extended area, Sy: Y area, Sy: Y extended area, Shw: HW area, Shw': HW extended area, Sj: J area, Sm: margin)
O Origin P Point cloud data M model W work q-axis candidate point p Each point included in the point cloud data r radius n normal vector

Claims (8)

A device that generates the operation path of a picking robot.
Equipped with control means to control the picking robot
The picking robot includes a hand and an arm that supports the hand.
The control means includes a pose generation means for generating a pose including the position information and the posture information of the hand.
The pose includes a candidate for a gripping pose in which the hand grips the work, an approaching pose formed by the hand before the gripping pose, and a detachment pose formed by the hand after the gripping pose.
The pose generation means detects interference between objects based on predetermined interference conditions, and detects interference between objects.
The interference condition virtually represents the H region that virtually represents the hand, the A region that virtually represents the arm, the Y region that virtually represents the container for the work, and the work and the hand holding the work. Regarding the HW region, the H, A, Y, HW expansion region in which a predetermined margin is added to each of the H, A, Y, and HW regions, and the J region included in a predetermined distance from the tip of the hand.
(Condition 1) The H, A, and Y regions do not interfere with each other.
(Condition 2) The H, A, Y, and HW regions do not interfere with each other.
(Condition 3) The H, A, and Y regions do not interfere with each other, or even if they do interfere, the interference position corresponds to the J region.
(Condition 4) The H, A, and Y extended regions do not interfere with each other.
(Condition 5) H, A, Y, HW extended regions do not interfere with each other.
Including any one or more of
The control means is a path generation device, characterized in that the operation path is generated based on the pose that satisfies a predetermined condition including the interference condition. The pose generation means includes a candidate generation means for generating the candidate.
The candidate generation means generates one candidate based on the position of the work, the axial direction of the work, and the size of the work, and rotates the one candidate around a predetermined rotation axis to rotate the other candidate. The route generation device according to claim 1, wherein the candidate of the above is generated. The route generating device according to claim 1 or 2, wherein the pose generating means generates the approaching pose and the leaving pose based on the candidate. The path generating device according to any one of claims 1 to 3, wherein the pose generating means calculates a solution of inverse kinematics for the generated pose and excludes the candidate associated with a pose for which the solution does not exist. The route generation device according to any one of claims 1 to 4, wherein the pose generation means excludes the candidate associated with the pose in which interference is detected. The pose generation means refers to the generated pose.
(B) Satisfy (Condition 3) at the joint position of the approaching pose,
(B) Satisfy (Condition 3) at the candidate joint position,
(C) Satisfy (Condition 3) and (Condition 5) at the joint position of the withdrawal pose.
The route generation device according to any one of claims 1 to 5 , which excludes candidates associated with the pose when any one of (a) to (c) is not satisfied. Based on the candidate, the pose generation means generates an approaching motion in which the hand moves from the approaching pose to the gripping pose and a leaving motion in which the hand moves from the gripping pose to the leaving pose.
The pose generation means is used for each generated motion.
(2) Subdivide the approaching motion in chronological order and satisfy (Condition 3) at the joint position for each hour.
(E) Subdivide the withdrawal movement in chronological order and satisfy (Condition 3) at the joint position for each hour.
The route generator according to any one of claims 1 to 6 , which excludes candidates associated with the operation when either (2) or (e) is not satisfied. A route generation program that, when executed in a computer, causes the computer to function as the route generator according to any one of claims 1 to 7.

Images