JP2012011495A - Method and device for converting robot angle data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allow the angle data to be converted even to a redundant degree-of-freedom robot, and to compatibly perform the correction of the tool fore end position error and the avoidance of the hinge angle limit or any obstacle.SOLUTION: A method for converting the robot angular data includes: steps (S301-S303) of calculating the tool fore end position vectors of a robot before the change and a robot after the change considering each geometrical error; a step (S304) of calculating the error vectors between these vectors; a step (S307) of calculating the Jacobian matrix to the angle data of the robot after the change considering the geometrical error; a step (S308) of calculating the angle correction vectors based on the error vectors, and the inverse matrix or the pseudo inverse matrix of the Jacobian matrix; a step (S309) of correcting the angle correction vectors by using the performance function considering the avoidance of the hinge angle limit and the avoidance of any obstacle; and a step (S310) of correcting the angle data of the robot after the change by the corrected angle correction vectors. The processing is iterated until the absolute value (S305) of the error vectors is below the threshold (YES in S306).

Description

  The present invention relates to an angle data conversion method in an operation program for accurately reproducing the operation of the robot before the replacement with the robot immediately after the replacement without changing the teaching when the robot operating on the production line is replaced. Relates to the device.

Many industrial robots are operating mainly in automobile production lines, such as arc welding, spot welding, handling, and painting.
First, the current industrial robot will be briefly described.

  FIG. 9 is a configuration diagram of a general industrial robot. In FIG. 9, reference numeral 101 denotes a robot configured as a manipulator having a plurality of joint axes and links. Each joint shaft has a built-in drive motor with an encoder, and each joint can be driven independently. Reference numeral 102 denotes a control device for the robot 101. The control device 102 controls the motion of the robot 101 by performing feedback control (position control system) based on the encoder signals of the joint axis drive motors. The robot 101 and the control device 102 are connected via a connection cable 101a. Reference numeral 103 denotes a portable teaching operation panel, which is an interface for a teacher to manually (JOG) operate the robot and create / edit an operation program. The portable teaching operation panel 103 mainly includes an operation button group 103a and a display screen 103b. 101b is an end effector provided at the wrist of the robot 101. Various tools are attached to the end effector according to the application.

  If a robot operating on the production line fails for some reason, remove the failed robot, install a normal robot of the same type at the same location, and replace the robot. Depending on the extent of the failure, replacing the broken robot on the spot can restore the production line faster by replacing the robot. However, robots have mechanical individual differences (geometrical errors with respect to ideal kinematic models), and even if the same operation program (angle data) is given, the position and orientation of the tool tip deviates greatly. Correction is required, and it takes time to restore the line. Since the line stop time is a great loss for the manufacturer, there is a demand for improving the tool tip position accuracy (robot replacement accuracy) at the time of robot replacement and minimizing teaching correction work.

  Furthermore, even if the robot is not broken, a robot that has become a bottleneck for reasons such as avoiding obstacles and joint angle limits is used to improve the production line. There is also a desire to restore the line in a short time.

In view of the above demand, the following techniques have been proposed in the past.
In Patent Document 1, when an operation program is exchanged between robots when exchanging robots, the angle data of the conversion source robot stored in the control device of the conversion source robot is converted into an ideal program based on the angle data of the ideal robot without setting error. There is an operation program conversion system that can reproduce the same operation by converting and giving it to the control device of the conversion destination robot, and converting the angle data of the ideal program into the angle data of the conversion destination robot by the control device of the conversion destination robot. It is disclosed. In Patent Document 1, when converting the angle data of the conversion source robot into the angle data of the ideal robot, the angle calibration correction value Δθ _i for each corresponding axis is added to the angle data D _ai for each axis of the conversion source robot. By adding, the corrected angle data D _ai ′ of the conversion source robot is calculated, and the corrected angle data D _ai ′ is calculated by calculation based on kinematics using the calibration-corrected robot constant R ′. The position / posture data Dp is converted into the ideal robot angle data D _a ″ by inverse kinematics using the robot constant R. The ideal robot angle data has different calibration correction values. when converting the angular data of the destination robot, the kinematics of the angle data D _a "ideal robot with robot constant R, and converts the position and orientation data Dp, varying the robot constant R '"Computes the, the R" calibrated corrected robot constant R of the destination robots angle data D _a "the position and orientation data Dp by inverse kinematics of the robot by using the' previous calibration correction value ΔR robots conversion, "by subtracting the 'angle calibration correction value Δθ of the destination robots', the angle data D _a of the destination robot" the angle data D _a is calculating ". Here, the elements of the robot constant R are the link length and the tool length, and the elements of the calibration correction value ΔR are the link length, the tool length, and the number of joint angle elements.

In Non-Patent Document 1, in order to accurately position the hand position xm of a robot to the target position xd for a single robot, a Jacobian calculated in consideration of the geometric error (identified geometric parameters) of the robot. There is disclosed a technique for correcting the angle of the hand position of a robot by multiplying a pseudo inverse matrix J ⁺ of a matrix by a tool tip position error (xd−xm).

  Patent Document 2 discloses a position / posture determination method for a redundant manipulator that calculates a correction amount of each joint angle so as to reduce a deviation amount between a current value and a target value of a tool position / posture of the redundant manipulator. . According to the technique described in Patent Document 2, even if the manipulator has redundancy, it is possible to always determine an appropriate position and orientation, and to prevent the joint from exceeding the motion range and the speed range, The amount of change in the position and orientation of the tool corresponding to this amount is obtained by linear approximation with a Jacobian matrix, the degree of utilization representing the degree of utilization of each joint is calculated based on the margin of movement of each joint, and the above The amount of correction of each joint angle is calculated by applying the amount of deviation to a new relationship composed of the linear relationship and the utilization level. That is, in Patent Document 2, when calculating the pseudo inverse matrix of the Jacobian matrix, weighting based on the margin up to the joint angle limit (that is, by preferentially using a joint with a margin), Considering avoidance of angle limit.

  In Patent Document 3, when the redundant manipulator is controlled, an optimal evaluation function is calculated so that the elbow posture becomes the target posture before the hand position / posture value of the redundant manipulator reaches the target value. A method for controlling an elbow posture of a redundant manipulator based on a function is disclosed. A position and orientation determination method for a redundant degree-of-freedom robot in consideration of obstacle avoidance is disclosed using a null space of a pseudo inverse matrix of a Jacobian matrix.

JP-A-9-128026 JP-A-8-52674 Japanese Patent No. 3207728

D. Payannet et al, Identification and Compensation of Mechanical Errors for Industrial Robots, Proc. Of the 15th ISIR, Vol.2, pp.857-864 (fig.4)

  However, since the angle data conversion method of Patent Document 1 only considers the joint angle origin, link length, and tool length as elements of the geometric error of the robot, if the link (joint axis) is twisted For a large robot that cannot be applied and whose load capacity exceeds 100 kg, the positional accuracy of the tool tip after the robot replacement cannot be improved, so that there is a problem that it takes time to correct the teaching. Further, the angle data conversion method of Patent Document 1 has a problem that program conversion cannot be performed for a robot with a redundant degree of freedom.

  As disclosed in Non-Patent Document 1, if a pseudo inverse matrix of a Jacobian matrix is used, high-accuracy error correction considering axial twist is possible, and it is also applicable to a redundant freedom degree robot. There is no specific method how to determine the redundancy degree of freedom after realizing the error correction at the tool tip.

  In Patent Document 2, the redundancy degree of freedom can be determined so as to avoid the joint angle limit as much as possible. However, when the error correction at the tool tip and the joint angle limit avoidance conflict, the convergence of the error correction becomes slow, or the error There is a problem that does not converge.

In Patent Document 3, the redundancy degree of freedom can be determined so as to avoid the obstacle, but there is a problem that the target position cannot be reached when the error correction at the tip of the tool and the obstacle avoidance conflict.
In the end, even if the prior art is utilized to the maximum extent, it is impossible to achieve both correction of the tool tip position / posture error when exchanging the robot and avoiding joint angle limit and obstacle avoidance.

  The present invention has been made in view of such problems, and can convert angle data even for a redundant degree of freedom robot, and further, correction of tool tip position error, joint angle limit avoidance and obstacle avoidance. An object of the present invention is to provide an angle data conversion method capable of achieving both of the above and an apparatus for executing the angle data conversion method.

  In order to solve the above problem, the present invention replaces different robots with known geometric errors with respect to an ideal kinematic model so that the robot behavior before replacement can be accurately reproduced by the robot after replacement. A method for converting angle data of a previous robot, the error vector between a tool tip position vector of a robot before replacement and a robot after replacement each considering a geometric error, and the replacement considering a geometric error. An angle correction vector is calculated based on an inverse matrix or pseudo inverse matrix of a Jacobian matrix with respect to the angle data of the rear robot, and each step of correcting the angle data of the post-exchange robot based on the angle correction vector is provided. It is composed.

  According to the present invention, the angle correction vector is calculated using the pseudo inverse matrix of the Jacobian matrix in consideration of the geometric error data, so that highly accurate error correction can be performed even for the redundant degree of freedom robot.

  Preferably, according to the present invention, after the step of calculating the angle correction vector and before the step of correcting the angle data, according to the distance to joint angle limit avoidance and obstacle avoidance of the robot after replacement. The method further includes the step of correcting the angle correction vector based on an evaluation function calculated as a linear combination of potential functions. More preferably, the step of correcting the angle correction vector calculates a maximum increase gradient vector of the evaluation function, calculates an inner product of the angle correction vector and the maximum increase gradient vector, and the inner product is positive Further, each step of removing a component in the direction of the maximum increasing gradient vector from the angle correction vector is provided. According to the present invention, a linear combination of potential functions related to joint angle limit avoidance and obstacle avoidance is used as an evaluation function, and the component of the maximum increasing gradient vector direction is removed from the angle correction vector, that is, the evaluation function value is not increased. Therefore, the correction of the tool tip position error and the joint angle limit avoidance and obstacle avoidance can both be achieved. Further, in the step of calculating the inner product, when the angle formed by the angle correction vector and the maximum increase gradient vector is less than a predetermined threshold, a plurality of weight coefficient values in the evaluation function are changed to increase the maximum A step of recalculating the gradient vector may be further included. As a result, it is possible to avoid a state (deadlock state) in which it is difficult to perform error correction without increasing the evaluation function.

  Further, the process from the calculation of the angle correction vector to the process of correcting the angle data may be repeated until the absolute value of the angle correction vector becomes smaller than a predetermined threshold. As a result, the error between the tool tip position vectors of the pre-replacement robot and the post-replacement robot can be made sufficiently small, and the operation of the robot before the replacement can be accurately reproduced by the robot after the replacement.

A preferred aspect of the present invention is that the robot before replacement is accurately reproduced by the robot after replacement between different robots with known geometric errors with respect to the ideal kinematic model. A method for converting angle data,
(1) Calculate the position and orientation vector of the tool tip with respect to the angle data by forward kinematics considering the geometric error of the robot before replacement,
(2) After the replacement, initialize the angle data of the robot,
(3) Calculate the position and orientation vector of the tool tip with respect to the post-exchange angle data by forward kinematics taking into account the geometric error of the post-exchange robot,
(4) calculating an error vector between the pre-replacement position and orientation vector and the post-replacement position and orientation vector;
(5) Calculate a Jacobian matrix for the post-exchange angle data in consideration of the geometric error of the post-exchange robot,
(6) calculating an inverse matrix or a pseudo inverse matrix of the Jacobian matrix,
(7) Multiply the error vector by the inverse matrix or the pseudo inverse matrix to calculate an angle correction vector;
(8) Calculate the maximum increasing gradient vector of the evaluation function for joint angle limit avoidance and obstacle avoidance of the robot after replacement,
(9) When an inner product of the angle correction vector and the maximum increase gradient vector is positive, a component in the maximum increase gradient vector direction is removed from the angle correction vector;
(10) Update the post-exchange angle data by adding the angle correction vector to the post-exchange angle data;
(11) Repeat the process of returning to (3) until the absolute value of the error vector is equal to or less than a threshold value.
It is configured as a feature that it is processed by the procedure.

  In the step (9) of a preferred embodiment, when the angle formed by the angle correction vector and the maximum increase gradient vector is less than a predetermined threshold, the process returns to the step (8), and a plurality of weights in the evaluation function The coefficient value may be changed and the maximum increasing gradient vector may be recalculated.

  In the angle data conversion method of the present invention described above, either the pre-exchange robot or the post-exchange robot can be a robot on a simulator (ideal) with zero geometric error. As a result, once the motion program is converted into a nominal motion program that includes the robot angle data on the simulator (ideal) and saved in the external storage device, a large time lag occurs in the replacement of the robot, and the identity of the control device Even if it is not guaranteed before and after the replacement, the operation program for the replaced robot can be obtained without error by converting the nominal operation program to include the angle data of the replaced robot.

  The present invention can also be configured as an angle data conversion device for executing the angle data conversion method. Such an apparatus can also be composed of, for example, a program written to execute the above-described functions, a CPU that executes processing according to the program, and a memory that stores data necessary for the processing. In addition, this apparatus can be configured as an angle data conversion apparatus by itself, but can also be configured as a device incorporated inside a robot control apparatus, for example.

It is a flowchart of the angle data conversion method concerning 1st Example of this invention. It is a flowchart of the angle data conversion method concerning 2nd Example of this invention. It is a flowchart which shows the operation | movement principle of angle data conversion of this invention. It is a flowchart of the correction process of an angle correction vector. It is a block diagram of the system which mounted the angle data conversion method of this invention. It is a figure which shows the specific example of an auxiliary storage device. It is a figure which shows the specific example of geometric error data. It is a figure which shows the operation principle of correction of an angle correction vector. It is a block diagram of an industrial robot.

  Hereinafter, specific examples of the method of the present invention will be described with reference to the drawings.

FIG. 1 shows a flowchart of a method for converting angle data of a robot in a first embodiment according to the present invention.
An outline of the angle data conversion method according to the present embodiment will be described, assuming that the real robot before replacement is Robot A and the real robot after replacement is Robot B.

  First, in the state where the robot A is connected to the control device, the geometric error data of the robot A (data indicating the individual difference between the mechanical design value of the robot and the mechanical dimensional error, Data required to eliminate individual differences in robot motion caused by errors (hereinafter also referred to simply as “error data”) from the auxiliary storage device to the main storage device in the control device, and The ID information added to the error data is collated with the ID information acquired from the actual robot A (S101). If the ID information of the error data matches the ID information of the actual robot A, that is, if it is determined that the error data loaded in the main memory is that of the robot A, the robot A is removed from the production line, The robot B is accurately installed at the same place, connected to the same control device, and the robot is exchanged (S102). After exchanging the robot, the geometric error data of robot B is copied to the auxiliary storage device, further loaded from the auxiliary storage device to the main storage device in the control device, and the ID information appended to the error data The ID information acquired from the actual robot B is collated (S103). If the ID information of the error data matches the ID information of the actual robot B, that is, if it is determined that the error data loaded in the main storage device is that of the robot B, it is stored in the main storage device. For the operation program for robot A, the angle data is converted for robot B, and the operation program is overwritten and saved (S104). Here, the angle data is converted so that the error between the tool tip position of the robot A reflecting the error data of the robot A and the tool tip position of the robot B reflecting the error data of the robot B becomes sufficiently small.

  FIG. 5 shows a configuration diagram of a system in which the angle data conversion method of the present invention is implemented. In addition, the part shown with the dotted line in FIG. 5 is a part which concerns on 2nd Example mentioned later. As shown in FIG. 5, this system includes a robot 101, a control device 102 for controlling the robot 101, a portable teaching operation panel 103 that can be operated and input by a teacher for controlling and teaching the robot 101, It has. As shown in FIG. 9, the robot 101 and the control device 102 are connected via a connection cable 101a.

  The portable teaching operation panel 103 incorporates an auxiliary storage device 103c. A geometric error data storage unit 103c1 for storing at least geometric error data of the robot 101 is allocated to the memory area of the auxiliary storage device 103c. As the auxiliary storage device 103c, as shown in FIG. 6, a device that can be easily attached and detached, such as a CF card or a USB memory, is suitable.

  By making the auxiliary storage device 103c detachable, after loading the geometric error data of the robot A into the main storage device, the auxiliary storage device 103c is temporarily detached from the portable teaching operation panel 103, and the robot B using a personal computer or the like. The geometric error data of the robot B is copied to the auxiliary storage device 103c, the auxiliary storage device 103c is attached to the portable teaching operation panel 103 again, and the geometric error data of the robot B is loaded into the main storage device. it can.

  The control device 102 loads the geometric error data of the robot 101 stored in the auxiliary storage device 103c to the main storage device 102a while the main storage device 102a and the robot 101 are connected to the control device 102. And geometric error data loading / collating means 102b for collating the ID information appended to the geometric error data with the ID information acquired from the robot 101. In the memory area of the main storage device 102a, an error data storage unit 102a1 that stores geometric error data loaded / verified by the load verification unit 102b, and an operation program storage unit 102a2 that stores an operation program for operating the robot 101. And are assigned. A nonvolatile memory is suitable as the main storage device 102a.

  Further, the control device 102 includes a conversion source operation program reading means 102c for reading the operation program stored in the operation program storage unit 102a2 as a conversion source operation program, and a tool reflecting the geometric error data of the two robots. The angle data conversion unit 102d that converts the angle data included in the conversion source operation program, and the operation program in which the angle data is converted by the angle data conversion unit 102d so that the error between the tip positions is sufficiently small. Conversion destination operation program writing means 102e for overwriting the storage unit 102a2.

  The geometric error data load / verification unit 102b acquires the robot ID information added to the geometric error data and the actual robot 101 connected to the control device 102 via the connection cable 101a (FIG. 9). Compare and compare the ID information. If there is a mismatch in the ID information and a collation error occurs, the geometric error data is not copied (loaded) to the main storage device 102a, and the user (operator) is informed that the collation error has occurred. Be notified. As this notification method, for example, an error is displayed on the display screen of the portable teaching operation panel 103. As specific ID information, an encoder serial number (manufacturing number) can be used as shown in FIG. This is because each axis encoder signal cable is always included in the connection cable 101a between the actual robot 101 and the control device 102, and the encoder status and serial number are always communicated in addition to the encoder pulse value. The geometric error data storage unit 102a1 stores a plurality of collated geometric error data. At that time, the stored order is also stored, so that the error data of the robot before conversion (conversion source) and the error data of the robot after conversion (conversion destination) can be reliably identified.

  As described above, since the geometric error data is loaded from the auxiliary storage device while the actual robot is connected to the control device, the error data can be easily and reliably collated, and the data setting error can be confirmed. Can be surely prevented.

Next, details (conversion principle) of the angle data conversion process in S104 of FIG. 1 or 102d of FIG. 5 will be described.
FIG. 3 shows a detailed flowchart of angle data conversion processing between robots.

In FIG. 3, a robot 1 indicates a conversion source (before replacement) robot, and a robot 2 indicates a conversion destination (after replacement) robot.
First, in S300, the angle data θ1 (joint angle vector) of the robot 1 is input. Next, the position and orientation vector Ptcp1 (θ1) of the tool tip is calculated by forward kinematics (forward kinematics) calculation considering the geometric error data of the robot 1 (S301). Next, the angle data θ2 of the robot 2 is initialized (S302). If the robot 1 and the robot 2 are of the same type, θ1 may be copied to θ2. In S303, in the same manner as in S301, the tool tip position and orientation vector Ptcp2 (θ2) is calculated by forward kinematics (forward kinematics) calculation in consideration of the geometric error data of the robot 2. In S304, an error vector ΔPtcp = Ptcp1 (θ1) −Ptcp2 (θ2) of the tool tip positions and orientations of the robot 1 and robot 2 is calculated. In S305, the absolute value of the error vector ΔPtcp is calculated. In S306, it is determined whether the absolute value is less than a preset threshold value. If it is less than the threshold, the angle data θ2 of the robot 2 is output and the process is terminated (S311). If it is greater than or equal to the threshold value, the processing from S307 to S310 described below is executed, and the process returns to S303.

  In S307, the Jacobian matrix J2 (θ2) is calculated in consideration of the geometric error data of the robot 2. The hand position / posture vector P2 of the robot 2 can be generally expressed as P2 = f (θ2, α) using the function of the joint angle vector θ2 of the robot 2 and the geometric error data α. In S307, the Jacobian matrix J2 (θ2) is calculated by partially differentiating f (θ2, α) with θ2 as in the following equation.

J2 (θ2) = ∂f (θ2, α) / ∂θ2
In S308, the angle correction vector Δθ2 = Inv (J2 (θ2)) · ΔPtcp of the robot 2 is calculated. Here, Inv (J2 (θ2)) is an inverse matrix or pseudo inverse matrix of the Jacobian matrix J2 (θ2).

  In S309, when the robot 2 is a redundant degree of freedom robot, the angle correction vector Δθ2 is corrected to Δθ2 ′ in consideration of joint angle limit avoidance and obstacle avoidance according to the method described later. When the robot 2 is not in a redundant degree of freedom, no correction is made if there is a sufficient margin to the joint angle limit and there is no fear of a collision with an obstacle. In S310, the angle correction vector Δθ2 ′ or Δθ2 calculated up to S309 is added to the angle data θ2 of the robot 2. That is, the angle data θ2 is corrected as θ2 = θ2 + Δθ2 ′ or θ2 = θ2 + Δθ2.

  When the angle correction data θ2 is corrected in S310, the process returns to S303 again, and the same processing is executed on the corrected angle data θ2 until the absolute value of the error vector becomes less than the threshold value in S306. By repeating the processing from S303 to S310, the tool tip position / posture before the robot replacement and the tool tip position / posture after the robot replacement can be matched as much as possible. Both forward kinematics and Jacobian computations can be calculated using a 4 × 4 homogeneous transformation matrix describing the relative relationship between the joint coordinate systems fixed to each link. In the homogeneous transformation matrix, all DH parameters (including link torsion), rotation angle offset (encoder origin), and joint axis compliance can be taken into account, and error data from an ideal geometric model is accurate (in advance) Measurement), the tool tip position and orientation of the actual robot also coincide with each other with high accuracy.

FIG. 4 shows a detailed flowchart of the processing of S309 shown in FIG.
In S401, an evaluation function K that considers joint angle limit avoidance and obstacle avoidance is calculated. Usually, this evaluation function is calculated as a linear combination of potential functions according to the joint angle limit or the distance to the obstacle. That is, K = c1 · U1 + c2 · U2 ++... + Cn · Un. U1 to Un are potential functions that have larger values as the distance decreases. Also, c1 to cn are evaluation function weight coefficients.

  In S402, the maximum increasing gradient vector φ = ∇K of the evaluation function K is calculated. Here, ∇ is a vector differential operator, and the gradient vector of the potential function can be calculated. In FIG. 8A, 801 is the contour of the evaluation function (potential function), 802 is the maximum value of the evaluation function, 803 is the maximum increasing gradient vector φ, and 804 is the angle correction vector Δθ2. As shown in FIG. 8A, the direction of the gradient vector φ (803) indicates the steepest increase direction of the potential function in the joint space, and the size of φ gives the maximum gradient. Here, if φn = ∇Un, it can be expressed as φ = c1, φ1 + c2, φ2 + ... cn · φn, where φ is a linear combination of the maximum increasing gradient vector φn of each potential function Un (from c1 to cn is its weighting factor) ).

In S403, the angle α formed by the angle correction vector Δθ2 and the maximum increasing gradient vector φ is calculated based on the inner product.
In S404, it is determined whether the formed angle α is less than 90 degrees. When the angle formed as shown in FIG. 8A is 90 degrees or more, the angle correction vector Δθ2 does not increase the evaluation function K, and thus the process ends without correcting Δθ2 (S408). When the formed angle is less than 90 degrees as shown in FIG. 8B, it is determined whether the formed angle α is less than a preset threshold value (usually a very small value) (S405). If it is equal to or greater than the threshold, in S406, the component in the direction of the maximum increasing gradient vector φ is removed from the angle correction vector Δθ2. That is,
Δθ2 '= Δθ2-(Δθ2 · φ) φ / | φ |
Is corrected based on the above, and the process ends. In FIG. 8B, reference numeral 805 denotes a component of the angle correction vector Δθ2 in the direction of the maximum increasing gradient vector φ, which corresponds to the second term on the right side of the above equation. Reference numeral 806 denotes a corrected angle correction vector Δθ2 ′.

  By correcting Δθ2 in this way, the tool tip position / posture error ΔPtcp can be reduced without increasing the evaluation function K (without reaching the joint angle limit or colliding with an obstacle). .

  However, when the angle α formed as shown in FIG. 8C is less than the threshold value, it is difficult to correct the error without increasing the evaluation function K (deadlock state), and therefore the weighting factor c1 of the evaluation function K in S407. Change cn to return to S401. As shown in FIG. 8D, by changing the value of the weighting factor, the direction of the maximum increasing gradient vector φ changes, so that a deadlock state can be avoided. In FIG. 8D, reference numeral 803 denotes a maximum increasing gradient vector φ before the weighting coefficient is changed, which can be expressed as φ = 0.8φ1 + 0.2φ2 (for convenience, φ = c1, φ1, + c2, φ2). Reference numeral 807 denotes the maximum increasing gradient vector φ ′ after changing the weighting coefficient, which can be expressed as φ = 0.2φ1 + 0.8φ2.

  The coefficient of the evaluation function K is not absolute and can be changed within a certain range. Even if the coefficient of the evaluation function is changed within a certain range, the safety aspect (avoidance of collision with obstacles) is still Fully guaranteed. As a constraint condition at the time of changing the coefficient, for example, ΣCn = const (constant) may be set, or may be changed in a range of a predetermined value <Cn <predetermined value. Alternatively, a plurality of sets of C1 to Cn may be prepared in advance and selected as appropriate.

  As described above, since the angle correction vector is calculated using the pseudo inverse matrix of the Jacobian matrix reflecting the geometric error data, the error correction can be performed with high accuracy even for the redundant freedom degree robot. . In addition, since the linear combination of potential functions related to joint angle limit avoidance and obstacle avoidance is used as an evaluation function, the angle correction vector is corrected in a direction that does not increase the evaluation function value while appropriately changing the weighting coefficient of the evaluation function. It is possible to achieve both correction of the tip position error and joint angle limit avoidance and obstacle avoidance.

  FIG. 2 shows a flowchart of a robot angle data conversion method according to the second embodiment of the present invention. It is the same as in the first embodiment that the robot before replacement is robot A and the robot after replacement is robot B.

  The system for executing the angle data conversion method according to the second embodiment is substantially the same as the system configuration shown in FIG. 5, but differs in the following points. That is, as shown in FIG. 5, in addition to the geometric error data storage unit 103c1, an operation program storage unit 103c2 is further allocated to the memory area of the auxiliary storage device 103c. The conversion source operation program reading means 102c of the control device 102 can not only read the operation program stored in the operation program storage unit 102a2, but also the nominal operation stored in the operation program storage unit 103c2 of the auxiliary storage device. The program can be read as a conversion source program. Further, the conversion destination operation program writing means 102e can not only write the operation program in the operation program storage unit 102a2, but also save it in the operation program storage unit 103c2 of the auxiliary storage device. Since the other points are the same as those of the first embodiment, the second embodiment will be described using the same reference numerals.

  In FIG. 2, S201, S203, and S204 correspond to S101, S102, and S103 of FIG. The difference from the first embodiment is in the processing of S202 and S205. First, in S202, the angle data of the operation program of the robot A stored in the operation program storage unit 102a2 of the main storage device is converted into the angle data of the nominal robot and converted to the operation program storage unit 103c2 of the auxiliary storage device. Save the nominal motion program with the specified angle data. Here, the nominal robot is an ideal robot whose geometric error data is all zero, and is a robot model on a robot simulator.

  In S205, the angle data in the nominal motion program stored in the motion program storage unit 103c2 of the auxiliary storage device is converted from the nominal robot to the robot B based on the geometric error data of the robot B loaded in S204. To the operation program storage unit 102a2 of the main storage device.

As described above, in the second embodiment, the angle data is not directly converted from the robot A to the robot B, but is converted through the nominal robot.
Such conversion is suitable when a large time lag occurs in the exchange from robot A to robot B. For example, since robot C broke down on production line 1, remove it, bring robot A from production line 2 that is not in operation, and install it at the place where robot C was located on production line 1 (robot C was sent for repair). ) Since the production amount increased after a while, the suspended production line 2 was to be reactivated, and the robot B was brought to the surface and installed where the robot A was. In such a case, there is a considerable time lag in exchanging from robot A to robot B on the production line 2, and during that time, the control device to which robot A was connected may be diverted. Identity is not always guaranteed. Therefore, since the geometric error data cannot be correctly verified, all the angle data of the robot A is once converted into a nominal robot and saved in an external auxiliary storage device. To convert the angle data from robot A to nominal robot, the geometric error data of robot 1 in FIG. 3 is converted into the geometric error data of robot A, and the geometric error data of robot 2 is all set to zero. Can be executed in exactly the same procedure. The same procedure can be used to convert angle data from the nominal robot to robot B.

  Thus, in the angle data conversion method of the present invention, the angle conversion can be performed in the same procedure without distinguishing between the actual robot and the nominal robot, so that there is versatility and various conversion modes can be selected.

  Although the above is each Example of this invention, this invention is not limited to the said example, It can change arbitrarily suitably within the scope of the present invention.

  It can also be used when the motion program created by the robot simulator is initially input to the actual robot.

101 Robot 101a Connection cable 101b Tool 102 Control device 102a Main storage device 102a1 Geometric error data storage unit 102a2 Operation program storage unit 102b Geometric error data load / verification unit 102c Conversion source operation program reading unit 102d Angle data conversion unit 102e Conversion destination operation program writing means 103 Portable teaching operation panel 103a Operation button group 103b Display screen 103c Auxiliary storage device 103c1 Geometric error data storage unit 103c2 Operation program storage unit 801 Contour line 802 of evaluation function Maximum point 803 of evaluation function Maximum increasing gradient vector 804 of evaluation function Angle correction vector 805 Maximum increasing gradient vector direction component 806 of angle correction vector Modified angle correction vector 807 Weighting factor is changed The maximum increase in the gradient vector

Claims (9)

By converting the angle data of the robot before replacement so that the robot behavior before replacement can be accurately reproduced by the robot after replacement between different robots with known geometric errors for the ideal kinematic model. There,
An error vector between the tool tip position vectors of the pre-replacement robot and the post-replacement robot considering geometric errors, and an inverse matrix or pseudo-inverse of the Jacobian matrix for the angle data of the post-replacement robot considering geometric errors. Calculate the angle correction vector based on the matrix,
An angle data conversion method comprising the steps of correcting angle data of the post-replacement robot based on the angle correction vector.   After the step of calculating the angle correction vector and before the step of correcting the angle data, it is calculated as a linear combination of potential functions according to the distance to the joint angle limit avoidance and obstacle avoidance of the robot after replacement. The angle data conversion method according to claim 1, further comprising a step of correcting the angle correction vector based on the evaluated function. The step of correcting the angle correction vector includes:
Calculating a maximum increasing gradient vector of the evaluation function;
Calculating the inner product of the angle correction vector and the maximum increasing gradient vector;
3. The angle data conversion method according to claim 2, comprising each step of removing the component in the maximum increasing gradient vector direction from the angle correction vector when the inner product is positive.   In the step of calculating the inner product, when an angle formed by the angle correction vector and the maximum increase gradient vector is less than a predetermined threshold, a plurality of weight coefficient values in the evaluation function are changed to increase the maximum increase gradient vector. The angle data conversion method according to claim 3, further comprising the step of recalculating.   The process from the step of calculating the angle correction vector to the step of correcting the angle data is repeated until the absolute value of the angle correction vector becomes smaller than a predetermined threshold value. The angle data conversion method according to item. By converting the angle data of the robot before replacement so that the robot behavior before replacement can be accurately reproduced by the robot after replacement between different robots with known geometric errors for the ideal kinematic model. There,
(1) Calculate the position and orientation vector of the tool tip with respect to the angle data by forward kinematics considering the geometric error of the robot before replacement,
(2) After the replacement, initialize the angle data of the robot,
(3) Calculate the position and orientation vector of the tool tip with respect to the post-exchange angle data by forward kinematics taking into account the geometric error of the post-exchange robot,
(4) calculating an error vector between the pre-replacement position and orientation vector and the post-replacement position and orientation vector;
(5) Calculate a Jacobian matrix for the post-exchange angle data in consideration of the geometric error of the post-exchange robot,
(6) calculating an inverse matrix or a pseudo inverse matrix of the Jacobian matrix,
(7) Multiply the error vector by the inverse matrix or the pseudo inverse matrix to calculate an angle correction vector;
(8) Calculate the maximum increasing gradient vector of the evaluation function for joint angle limit avoidance and obstacle avoidance of the robot after replacement,
(9) When an inner product of the angle correction vector and the maximum increase gradient vector is positive, a component in the maximum increase gradient vector direction is removed from the angle correction vector;
(10) Update the post-exchange angle data by adding the angle correction vector to the post-exchange angle data;
(11) Repeat the process of returning to (3) until the absolute value of the error vector is equal to or less than a threshold value.
An angle data conversion method characterized in that the processing is performed according to the following procedure.   In the step (9), when the angle formed by the angle correction vector and the maximum increasing gradient vector is less than a predetermined threshold, the process returns to the step (8), and a plurality of weight coefficient values in the evaluation function are obtained. The angle data conversion method according to claim 6, wherein the maximum increase gradient vector is recalculated after changing.   The angle data conversion according to any one of claims 1 to 7, wherein one of the pre-replacement robot and the post-replacement robot is a robot on a simulator (ideal) having no geometric error. Method.   The apparatus for performing the angle data conversion method of any one of Claims 1 thru | or 8.