JP7378640B2 - Robot control device and robot control method - Google Patents

Description

The present disclosure relates to a robot control device and a robot control method that control a manipulator.

2. Description of the Related Art On a factory production line, etc., there is a pick-and-place process in which parts and products (hereinafter referred to as "works") are gripped and transported by a manipulator. In pick-and-place, if the operating speed and acceleration of the robot are not appropriate, excessive inertia force and moment are generated on the workpiece and the gripping portion of the manipulator, causing the workpiece to fall. To solve this problem, a technique has been proposed that determines appropriate operating conditions by considering the force and moment generated in the workpiece.

Patent Document 1 discloses that a simulation is performed using a simulation model of a robot including an elastic holding body, and when the load moment generated on the holding body is larger than a threshold value, the simulation is performed so that the load moment is below the threshold value. A simulation device that changes execution conditions is disclosed.

Japanese Patent Application Publication No. 2019-141939

D. Verscheure et al. “Time-Optimal Path Tracking for Robots: A Convex Optimization Approach” IEEE Trans. Automatic Control, 2009 B. Amos et al. “OptNet: Differential Optimization as a Layer in Neural Networks”Proceedings of the 34th International Conference on M achine Learning, 2017

In Patent Document 1, it is necessary to repeatedly change the simulation execution conditions until the load moment becomes equal to or less than a threshold value, so there is a problem that it takes time to determine appropriate execution conditions in consideration of the force and moment generated on the workpiece. There is. In addition, the simulation execution condition that can be automatically adjusted is only the maximum acceleration, and it is difficult to adjust operating conditions with a high degree of freedom such as the robot's operating trajectory and velocity profile. There is a problem in that it is not possible to shorten the operation time until the object is grasped.

The present disclosure has been made to solve the above-mentioned problems, and is to quickly obtain a command trajectory and speed profile of a manipulator that satisfies constraint conditions such as force and moment generated on a workpiece and shortens operation time. The purpose of the present invention is to provide a robot control device and a robot control method that can perform the following tasks.

The robot control device according to the present disclosure calculates a speed profile of the manipulator based on a preset command trajectory to the manipulator, constraint conditions regarding the manipulator, and an evaluation index based on an operation time of the manipulator. a calculation section; a gradient calculation section that calculates a gradient regarding the command trajectory of the operation time based on the speed profile to obtain gradient information; and a gradient calculation section that corrects the command trajectory based on the gradient information to obtain a corrected command trajectory. and a control section that controls the manipulator to follow the corrected command trajectory.

Further, the robot control method according to the present disclosure calculates a velocity profile of the manipulator based on a preset command trajectory to the manipulator, constraint conditions regarding the manipulator, and an evaluation index based on the operation time of the manipulator. a step of calculating a gradient regarding the command trajectory of the operation time based on the speed profile to obtain gradient information; and a step of correcting the command trajectory based on the gradient information to obtain a corrected command trajectory. and a step of controlling the manipulator to follow the correction command trajectory.

According to the present disclosure, a robot control device and a robot control method calculate a speed profile based on an evaluation index based on constraint conditions related to a manipulator and an operation time of the manipulator, and calculate a command trajectory based on a gradient regarding a command trajectory of the operation time. In order to correct this, it is possible to quickly obtain a correction command trajectory and a speed profile of the manipulator that satisfy the constraint conditions and shorten the operation time.

1 is a block diagram showing an example of a robot control device in Embodiment 1. FIG. 2 is a diagram showing an example of a configuration including a robot control device in Embodiments 1 to 4. FIG. FIG. 3 is a block diagram showing an example of a speed calculation unit in Embodiments 1 to 4. FIG. FIG. 3 is a block diagram illustrating an example of a gradient calculation unit in Embodiments 1 to 4. FIG. FIG. 3 is a block diagram showing an example of a control unit in Embodiments 1 to 4. FIG. 5 is a flowchart illustrating an example of the operation of the robot control device in the first embodiment. FIG. 3 is a block diagram showing an example of a robot control device in a second embodiment. FIG. 3 is a block diagram showing an example of a command trajectory correction section in Embodiment 2. FIG. 7 is a flowchart illustrating an example of the operation of the robot control device in Embodiment 2. FIG. FIG. 3 is a block diagram showing an example of a robot control device in Embodiment 3. FIG. 12 is a flowchart illustrating an example of the operation of the robot control device in Embodiment 3. FIG. 7 is a block diagram showing an example of a robot control device in Embodiment 4. FIG. 12 is a flowchart illustrating an example of the operation of the robot control device in Embodiment 4. 3 is a diagram showing the hardware configuration of a robot control device in embodiments 1 to 4. FIG.

Embodiment 1.
FIG. 1 is a block diagram showing an example of a robot control device 100 according to the first embodiment. FIG. 1 is a block diagram composed of a robot control device 100, an actuator 110, a manipulator 1, and a workpiece 112. Further, FIG. 2 is a diagram showing an example of a configuration including the robot control device 100 in the first embodiment. The robot control device 100 controls the actuators 110 installed at each joint of the manipulator 1, which is a vertically articulated robot, so that a gripping section 111 installed at the tip of the manipulator 1 picks and picks up a workpiece 112 to be gripped. Realize the behavior of the place. The surrounding environment 113 is, for example, a camera, and outputs, for example, an image when the manipulator 1 performs a pick-and-place operation to a display device (not shown).

The robot control device 100 includes a constraint storage section 2, a command trajectory storage section 3, a speed calculation section 4, a slope calculation section 5, a command trajectory correction section 6, a command point sequence calculation section 7, and a control section 8. Equipped with.

The constraint storage unit 2 stores parameters of constraints set in advance for the manipulator 1. The constraint condition is at least one of the joint angular velocity, joint angular acceleration, joint torque, hand speed, hand acceleration of the manipulator 1, the force generated in the object (workpiece 112) gripped by the manipulator 1, and the moment generated in the gripped object. There are two conditions. As an example, when the constraint condition is related to hand speed v _h , the constraint condition is "v _min ≦v _h ≦ v _max ". Here, v _min is the lower limit value of the hand speed v _h , and v _max is the upper limit value of the hand speed v _h . The parameters of the constraint conditions in this case are v _min and v _max . The constraints and the parameters of the constraints are not limited to these. For example, the constraint condition may be “|v _h |≦v _th ” and the parameter of the constraint condition may be v _th . Here, |v _h | is the absolute value of the hand speed v _h , and v _th is a threshold value of the hand speed v _h .

Although a case has been described in which the parameters of the constraint conditions are set in advance, the robot control device 100 may include a grip constraint learning section (not shown). The gripping constraint learning unit uses machine learning to learn constraint parameters for at least one of the force and moment generated in the workpiece 112, and stores the obtained constraint parameters in the constraint storage unit 2. It's okay. Specifically, the gripping constraint learning unit acquires the force and moment generated in the workpiece 112 when the manipulator 1 grips the workpiece 112 using a sensor (not shown), and uses the acquired values as parameters of the constraint condition. learn.

The command trajectory storage unit 3 stores a preset command trajectory to the manipulator 1. As an example, when the command trajectory is given as a spline curve, the command trajectory storage unit 3 stores pairs of positions of waypoints on the spline curve and values of parametric variables of the curve at that position (values from 0 to 1). memorize it. Alternatively, the command trajectory storage section 3 may store only the positions of waypoints on the spline curve. In this case, the value of the parameter is calculated from the distance between waypoints on the spline curve. Further, the command trajectory may be given as a B-spline curve, a Bezier curve, or the like other than a spline curve. The command trajectory storage section 3 stores a corrected command trajectory from the command trajectory correction section 6, which will be described later. At this time, the preset command trajectory is discarded, and the corrected command trajectory is newly stored.

The speed calculation unit 4 calculates the speed profile of the manipulator 1 based on a preset command trajectory for the manipulator 1, constraint conditions regarding the manipulator 1, and an evaluation index based on the operation time of the manipulator 1. That is, the speed calculation section 4 issues a command to shorten the operation time of the manipulator 1 within the range of the constraint conditions based on the command trajectory from the command trajectory storage section 3 and the parameter of the constraint condition from the constraint condition storage section 2. Calculate the velocity profile on the trajectory. Alternatively, if the robot control device 100 includes a grip constraint learning section, the speed calculation section 4 calculates the speed profile based on the command trajectory and the parameter of the constraint learned by the grip constraint learning section. Here, the evaluation index based on the operation time of the manipulator 1 is an evaluation function used by the profile calculation unit 44, which will be described later. This evaluation function is a function expressed mathematically using the acceleration and velocity of the manipulator 1 as variables through a parameter. In addition, the velocity profile represents a temporal change in the velocity of each joint when the manipulator 1 moves on a commanded trajectory. The speed calculation unit 4 calculates a speed profile using an optimization problem that minimizes the operation time expressed by an evaluation function.

FIG. 3 is a block diagram showing an example of the speed calculation unit 4 in the first embodiment. The speed calculation unit 4 includes a command interpolation calculation unit 41, a dynamics calculation unit 42, a constraint coefficient calculation unit 43, and a profile calculation unit 44.

The command interpolation calculation unit 41 interpolates the curve of the command trajectory from the command trajectory storage unit 3 at N points using a preset natural number N, and calculates information regarding the position at the interpolation point on the command trajectory and the parameter of the command trajectory. Calculate the first and second derivatives. As an example, if the command trajectory is given as a spline curve, the joint positions (or joint angles) of each axis of the manipulator 1 are expressed as a cubic piecewise polynomial with respect to the parametric variables. , it is possible to calculate not only the position (or angle) of the interpolation point, but also the first and second derivatives with respect to the parametric variables.

The dynamics calculation unit 42 performs kinematics calculation and dynamics calculation of the manipulator 1 using the position, first-order differential, and second-order differential at the interpolation point from the command interpolation calculation unit 41, and calculates the kinematics calculation result and the dynamics. Outputs scientific calculation results. Kinematic calculation is to calculate the velocity, acceleration, angular velocity, and angular acceleration of each joint of the manipulator 1, each link, the grip part 111, etc. from the velocity, acceleration, angular velocity, and angular acceleration of each joint of the manipulator 1. That's true. In addition, dynamic calculation refers to the torque generated in each joint of the manipulator 1 and the torque generated in the grip part 111 from the velocity, acceleration, angular velocity, and angular acceleration of each joint, each link, and the grip part 111 of the manipulator 1. The purpose is to calculate the force and moment.

The constraint coefficient calculation unit 43 calculates coefficients of the constraint based on the kinematic calculation results and the dynamic calculation results from the dynamic calculation unit 42 and the constraint parameters from the constraint storage unit 2. As an example, when a constraint regarding the force generated in the grip portion 111 is expressed by the following formula (1), the coefficients of the constraint are a, b, and c. That is, the coefficients of the constraint conditions are the coefficients included in each term in a relational expression that includes a plurality of variables. Here, the variables include acceleration u and velocity x at the interpolation point in Equation (1).

The profile calculation unit 44 calculates a speed profile based on the constraint coefficients from the constraint coefficient calculation unit 43. That is, the profile calculation unit 44 calculates the velocity profile of the manipulator 1 on the commanded trajectory and the operation time of the manipulator 1 by optimization calculation based on the evaluation function based on the operation time of the manipulator 1 and the coefficients of the constraint conditions. do.

Note that the method by which the constraint coefficient calculation unit 43 calculates the coefficient of the restriction condition and the method by which the profile calculation unit 44 calculates the speed profile are shown in Non-Patent Document 1.

As described above, the speed calculation unit 4 can calculate a speed profile that reduces the operating time within the range of the constraints set in the constraint storage unit 2. Furthermore, by adding the force and moment generated in the work 112 to the constraint conditions, it is possible to suppress problems such as the work 112 falling or excessive force being generated in the work 112 during pick and place of the work 112. can. In addition, the method described in Non-Patent Document 1 is a method called convex optimization in which a global optimal solution can be easily obtained, so that the velocity profile can be calculated at high speed.

Returning to FIG. 1, the gradient calculation section 5 calculates the gradient regarding the command trajectory of the operation time of the manipulator 1 based on the speed profile from the speed calculation section 4, and outputs it as gradient information. Specifically, the gradient calculation unit 5 uses automatic differentiation to calculate the gradient of the operation time with respect to the command trajectory by differentiating the operation time of the manipulator 1 from the profile calculation unit 44 with respect to the command trajectory, and calculates the slope information with respect to the command trajectory. Output as .

FIG. 4 is a block diagram showing an example of the gradient calculation unit 5 in the first embodiment. The gradient calculation section 5 includes a command interpolation gradient calculation section 51, a dynamic gradient calculation section 52, a constraint coefficient gradient calculation section 53, and a profile gradient calculation section 54. The command interpolation gradient calculation section 51, the dynamic gradient calculation section 52, the constraint coefficient gradient calculation section 53, and the profile gradient calculation section 54 are the command interpolation calculation section 41, the dynamics calculation section 42, and the constraint condition coefficient calculation section 43, respectively. , and calculates the slope of the calculation result performed by the profile calculation unit 44.

The profile gradient calculation unit 54 calculates the gradient related to the coefficient of the operation time constraint by calculating the gradient related to the coefficient of the optimization problem based on a method similar to the method shown in Non-Patent Document 2 as an example. calculate.

The constraint coefficient gradient calculation unit 53 inputs the gradient of the coefficient of the constraint from the profile gradient calculation unit 54 and calculates the differential of the calculation procedure of the constraint coefficient calculation unit 43 based on the chain rule, thereby calculating the operating time. Calculate the slope with respect to the kinematic calculation result of and the slope with respect to the dynamic calculation result of operation time.

The dynamic gradient calculation unit 52 inputs the gradient related to the kinematic calculation result of the operating time and the gradient related to the dynamic calculation result of the operating time from the constraint condition coefficient gradient calculation unit 53, and calculates the calculation procedure of the dynamic calculation unit 42. By calculating the differential based on the chain rule, we can calculate the slope with respect to the position at the interpolation point on the command trajectory of the operation time, the slope of the first derivative with respect to the parameter of the command trajectory of the operation time, and the parameter of the command trajectory of the operation time. Compute the slope with respect to the second derivative with respect to .

The command interpolation gradient calculation unit 51 calculates a gradient related to the position at the interpolation point on the command trajectory from the dynamic gradient calculation unit 52, a gradient related to the first-order differential with respect to the parameter of the command trajectory, and a gradient related to the second-order differential with respect to the parameter of the command trajectory. is input, and the gradient of the command trajectory of the operation time is calculated by calculating the differential of the calculation procedure of the command interpolation calculation unit 41 based on the chain rule. The command interpolation gradient calculation unit 51 outputs the gradient regarding the command trajectory of the operation time as gradient information.

Returning to FIG. 1, the command trajectory correction section 6 corrects the command trajectory so as to reduce the operating time of the manipulator 1 based on the gradient information from the gradient calculation section 5, and outputs it as a corrected command trajectory. The command trajectory correction unit 6 corrects the command trajectory by using any one of a gradient descent method, a conjugate gradient method, or a quasi-Newton method based on gradient information. Examples of gradient descent methods include the steepest descent method, the momentum method, and the accelerated gradient method. The correction command trajectory is stored in the command trajectory storage section 3.

Based on the slope information, the command trajectory correction section 6 corrects the command trajectory so that the operation time of the manipulator 1 is shortened, so that the operation time of the manipulator 1 can be shortened. Further, since the command trajectory correction section 6 uses the slope information, the command trajectory can be corrected efficiently and at high speed.

The command point sequence calculation section 7 calculates a command point sequence for each predetermined sampling period based on the corrected command trajectory stored in the command trajectory storage section 3 and the speed profile from the speed calculation section 4. This sampling period is a period when the control unit 8, which will be described later, calculates the current value to the actuator 110. As an example, if the correction command trajectory is given as a spline curve, the command point sequence calculation unit 7 converts between the time and the parameter of the curve of the correction command trajectory based on the speed profile, and converts the parameter By combining with the piecewise polynomial of the spline curve of the correction command trajectory input, the point sequence of the position command to the manipulator 1 is calculated at each sampling.

The control unit 8 controls the manipulator 1 to follow the correction command trajectory. That is, the control unit 8 controls the manipulator 1 to follow the command point sequence from the command point sequence calculation unit 7.

FIG. 5 is a block diagram showing an example of the control unit 8 in the first embodiment. The control section 8 includes a feedforward control section 81, a feedback control section 82, and a current value calculation section 83.

The feedforward control unit 81 performs filter processing such as smoothing on the command point sequence from the command point sequence calculation unit 7, and outputs it as a smoothed command point sequence. The feedforward control unit 81 calculates and outputs the feedforward value of the current input to the actuator 110 by applying the modeled inverse transfer function to the command point sequence from the command point sequence calculation unit 7. . The inverse transfer function is an inverse function to the transfer function of the actuator 110 that is the controlled object. If the disturbance generated in the actuator 110 can be detected by the sensor, the feedforward control unit 81 may apply a modeled inverse transfer function to the disturbance signal. In this case, the feedforward control unit 81 outputs the disturbance signal to which the inverse transfer function is applied, including it in the feedforward value of the current. An example of the disturbance is vibration caused by a worker touching the manipulator 1.

The feedback control unit 82 performs feedback control so that the actuator 110 follows the smoothed command point sequence from the feedforward control unit 81, and calculates and outputs a feedback value of the current input to the actuator 110.

The current value calculation section 83 calculates the current value input to the actuator 110 based on the feedforward value of the current from the feedforward control section 81 and the feedback value of the current from the feedback control section 82 .

FIG. 6 is a flowchart showing an example of the operation of the robot control device 100 in the first embodiment. That is, FIG. 6 is a flowchart showing an example of the robot control method in the first embodiment.

As shown in FIG. 6, when robot control is started by means not shown, the speed calculation unit 4 calculates a command trajectory to the manipulator 1 set in advance, constraint conditions regarding the manipulator 1, and operation time of the manipulator 1. The speed profile of the manipulator 1 is calculated based on the evaluation index (step ST1).

The gradient calculation unit 5 calculates the gradient regarding the command trajectory of the operation time of the manipulator 1 and outputs it as gradient information (step ST2).

The command trajectory correction unit 6 corrects the command trajectory based on the slope information and outputs the corrected command trajectory (step ST3).

The command trajectory storage section 3 stores the corrected command trajectory (step ST4).

The command point sequence calculation unit 7 calculates a command point sequence based on the corrected command trajectory and the speed profile (step ST5).

The control unit 8 controls the actuator 110 so that the manipulator 1 follows the correction command trajectory (step ST6).

By means not shown, it is determined whether or not to continue controlling the robot (step ST7).

If the determination in step ST7 is "Yes", the process returns to step ST6 and control of the robot is continued. If the determination in step ST7 is "No", control of the robot ends. Control of the robot ends, for example, when the manipulator 1 reaches the end point of the correction command trajectory. Alternatively, this is a case where it is determined that the manipulator 1 has operated abnormally. This determination is performed by means not shown.

According to the first embodiment described above, the speed profile is calculated based on the evaluation index based on the constraint conditions regarding the manipulator 1 and the operation time of the manipulator 1, and the command trajectory is calculated based on the gradient of the command trajectory of the operation time. In order to perform the correction, it is possible to quickly obtain a correction command trajectory and a speed profile for the manipulator 1 that satisfy the constraint conditions and shorten the operation time.

Embodiment 2.
In the second embodiment, a command trajectory is generated based on peripheral information around the manipulator 1 acquired in advance in the peripheral environment 113.

FIG. 7 is a block diagram showing an example of the robot control device 100a in the second embodiment. 7 differs from FIG. 1 in that the robot control device 100a includes a surrounding environment information storage unit 9, a command trajectory generation unit 10, and a speed profile storage unit 11 in addition to the components of the robot control device 100. . Further, FIG. 7 differs from FIG. 1 in that a command trajectory correction section 6a is provided instead of the command trajectory correction section 6. The components other than the surrounding environment information storage section 9, the command trajectory generation section 10, the speed profile storage section 11, and the command trajectory correction section 6a are the same as those shown in FIG. 1, so the description thereof will be omitted.

The surrounding environment information storage section 9 stores surrounding information around the manipulator 1. Specifically, the surrounding environment information storage unit 9 stores information such as the position and shape of obstacles around the manipulator 1 acquired from the surrounding environment 113 as surrounding information. The data structure for storing peripheral information may be a basic shape such as a point cloud, a voxel, a polygon mesh, a rectangular parallelepiped, or a combination of a plurality of basic shapes. Furthermore, the data structure may be a bounding volume hierarchy that can speed up distance calculations performed by the distance function calculation unit 61, which will be described later.

The command trajectory generation section 10 generates a command trajectory based on the surrounding information stored in the surrounding environment information storage section 9. Specifically, the command trajectory generation unit 10 generates a command trajectory such that the manipulator 1 does not interfere with surrounding obstacles based on surrounding information, and updates the command trajectory stored in the command trajectory storage unit 3. I do. As an algorithm for generating the command trajectory, for example, RRT (Rapidly-Exploring Random Tree) and PRM (Probabilistic RoadMap) may be used.

The command trajectory generated by the command trajectory generation unit 10 is not a preset initial trajectory, but is a command trajectory that avoids interference with obstacles around the manipulator 1, and is generated autonomously. Therefore, the command trajectory generation unit 10 can automatically generate a command trajectory by simply inputting, for example, the start point and end point of the command trajectory.

The command trajectory correction unit 6a corrects the command trajectory based on the command trajectory stored in the command trajectory storage unit 3, the surrounding information stored in the surrounding environment information storage unit 9, and the slope information from the slope calculation unit 5. Correct it and output it as a corrected command trajectory. The correction command trajectory is stored in the command trajectory storage section 3.

FIG. 8 is a block diagram showing an example of the command trajectory correction section 6a in the second embodiment. The command trajectory correction section 6a includes a distance function calculation section 61, a barrier function calculation section 62, a barrier function gradient calculation section 63, and a command trajectory correction value calculation section 64.

The distance function calculation unit 61 calculates the value of the distance function between the command trajectory and surrounding obstacles based on the command trajectory stored in the command trajectory storage unit 3 and the surrounding information stored in the surrounding environment information storage unit 9. Calculate.

The barrier function calculation unit 62 configures a barrier function such that the value of the function diverges when the value of the distance function becomes less than a certain value, and calculates the value of the barrier function based on the value of the distance function from the distance function calculation unit 61. Calculate.

The barrier function gradient calculation section 63 calculates the gradient of the barrier function value from the barrier function calculation section 62 with respect to the command trajectory. The barrier function gradient calculation unit 63 calculates the gradient using automatic differentiation, for example.

The command trajectory correction value calculation section 64 combines the gradient regarding the command trajectory of the operation time of the manipulator 1 from the gradient calculation section 5 and the slope regarding the command trajectory of the value of the barrier function from the barrier function gradient calculation section 63 to obtain gradient information. Based on this gradient information, a correction value for the command trajectory is calculated so that the sum of the operation time and the value of the barrier function becomes small.

Returning to FIG. 7, the speed profile storage section 11 stores the speed profile from the speed calculation section 4. Since the robot control device 100a includes the speed profile storage section 11, the timing at which the speed calculation section 4 calculates the speed profile can be arbitrarily set. That is, the velocity profile can be calculated not only immediately before the control unit 8 controls the actuator 110, but also while the manipulator 1 is not operating or while the manipulator 1 is performing another operation.

The command point sequence calculation unit 7 calculates a command point sequence for each predetermined sampling period based on the corrected command trajectory stored in the command trajectory storage unit 3 and the speed profile stored in the speed profile storage unit 11. .

FIG. 9 is a flowchart showing an example of the operation of the robot control device 100a in the second embodiment. That is, FIG. 9 is a flowchart showing an example of the robot control method in the second embodiment. Steps ST1 to ST7 in FIG. 9 are the same as steps ST1 to ST7 in FIG. 6, so detailed explanation will be omitted here.

As shown in FIG. 9, when robot control is started by means not shown, the command trajectory generation section 10 generates a command trajectory based on the surrounding information stored in the surrounding environment information storage section 9. (Step ST8).

The speed calculation unit 4 calculates the speed profile of the manipulator 1 based on the command trajectory to the manipulator 1, the constraint conditions regarding the manipulator 1, and the evaluation index based on the operation time of the manipulator 1 (step ST1).

The speed profile storage unit 11 stores the speed profile (step ST9).

The gradient calculation unit 5 calculates the gradient regarding the command trajectory of the operation time of the manipulator 1 and outputs it as gradient information (step ST2).

The command trajectory correction unit 6a calculates the gradient of the value of the barrier function with respect to the command trajectory, and outputs the gradient information combined with the gradient calculated in step ST2 as gradient information (step ST10).

The command trajectory correction section 6a corrects the command trajectory based on the gradient information and outputs the corrected command trajectory (step ST3).

The command trajectory storage section 3 stores the corrected command trajectory (step ST4).

The command point sequence calculation unit 7 calculates a command point sequence based on the corrected command trajectory and the speed profile (step ST5).

The control unit 8 controls the actuator 110 so that the manipulator 1 follows the correction command trajectory (step ST6).

By means not shown, it is determined whether or not to continue controlling the robot (step ST7).

If the determination in step ST7 is "Yes", the process returns to step ST6 and control of the robot is continued. If the determination in step ST7 is "No", control of the robot ends.

According to the second embodiment described above, the command trajectory generation unit 10 generates the command trajectory based on the surrounding information from the surrounding environment 113, so that the operation time can be shortened while avoiding interference with obstacles. The correction command locus and velocity profile of the manipulator 1 can be determined at high speed.

Embodiment 3.
In the third embodiment, the manipulator 1 is controlled using an input/output device 12 such as a tablet.

FIG. 10 is a block diagram showing an example of a robot control device 100b in the third embodiment. FIG. 10 differs from FIG. 1 in that the robot control device 100b is connected to the input/output device 12 by means not shown. Note that the robot control device 100b has the same configuration as the robot control device 100 in Embodiment 1, so a description thereof will be omitted. Note that the robot control device 100b may have the same configuration as the robot control device 100a in the second embodiment.

The input/output device 12 displays operation information of the manipulator 1 and peripheral information acquired from the surrounding environment 113 on a screen. The motion information is, for example, an image of the manipulator 1 in motion. The input/output device 12 outputs motion information input by the worker to the robot control device 100b. For example, the input/output device 12 may allow the operator to input the start and end points of the operation of the manipulator 1 through a touch panel on the screen or an audio interface, or may allow the operator to input the start and end points of the operation of the manipulator 1 by tracing the screen of the tablet with a finger. , a rough command trajectory of the grip portion 111 of the manipulator 1 may be input. This command trajectory is stored in the command trajectory storage section 3. Note that the operator can also use the input/output device 12 to stop the manipulator 1 while it is in operation. In this case, the input/output device 12 displays "stop operation" on the screen and causes the worker to touch the display, thereby transmitting a command to stop the work to the robot control device 100b. Thereby, the robot control device 100b stops the operation of the manipulator 1.

FIG. 11 is a flowchart showing an example of the operation of the robot control device 100b in the third embodiment. That is, FIG. 11 is a flowchart showing an example of the robot control method in the third embodiment. Steps ST1 to ST6 in FIG. 11 are the same as steps ST1 to ST6 in FIG. 6, so detailed explanation will be omitted here.

As shown in FIG. 11, when robot control is started by means not shown, the speed calculation unit 4 calculates a preset command trajectory for the manipulator 1, constraint conditions regarding the manipulator 1, and operation time of the manipulator 1. The speed profile of the manipulator 1 is calculated based on the evaluation index (step ST1).

The gradient calculation unit 5 calculates the gradient regarding the command trajectory of the operation time of the manipulator 1 and outputs it as gradient information (step ST2).

The command trajectory correction unit 6 corrects the command trajectory based on the slope information and outputs the corrected command trajectory (step ST3).

The command trajectory storage section 3 stores the corrected command trajectory (step ST4).

The command point sequence calculation unit 7 calculates a command point sequence based on the corrected command trajectory and the speed profile (step ST5).

The control unit 8 controls the actuator 110 so that the manipulator 1 follows the correction command trajectory (step ST6).

By means not shown, it is determined whether or not to continue controlling the robot (step ST11).

If the determination in step ST11 is "Yes", the process returns to step ST6, and control of the manipulator 1 is continued. If the determination in step ST11 is "No", control of the robot ends. Control of the robot ends, for example, when the manipulator 1 reaches the end point of the correction command trajectory. Alternatively, this is a case where it is determined that the manipulator 1 has operated abnormally. Alternatively, this is a case where a command to "stop operation" is sent from the input/output device 12 to the robot control device 100b. This determination is performed by means not shown.

According to the third embodiment described above, the operation of the manipulator 1 can be visualized by controlling the manipulator 1 using the input/output device 12.

Embodiment 4.
In the fourth embodiment, the manipulator 1 is controlled while the manipulator 1 is installed on the movable frame 13.

FIG. 12 is a block diagram showing an example of a robot control device 100c in the fourth embodiment. FIG. 12 differs from FIG. 1 in that the manipulator 1 is installed on a movable frame 13. FIG. 12 also shows that the robot control device 100c includes a constraint storage unit 2c instead of the constraint storage unit 2, a speed calculation unit 4c instead of the speed calculation unit 4, and a slope calculation unit 5 instead of the slope calculation unit 5. It differs from FIG. 1 in that it includes a gradient calculation section 5c. Components other than the constraint storage section 2c, speed calculation section 4c, and slope calculation section 5c are the same as those shown in FIG. 1, and therefore their description will be omitted. Note that the robot control device 100c is based on the robot control device 100 in the first embodiment, but may be based on the robot control device 100a in the second embodiment or the robot control device 100b in the third embodiment.

The constraint storage unit 2c stores parameters of constraints set in advance for the manipulator 1. The constraint conditions include the joint angular velocity, joint angular acceleration, joint torque, hand speed, hand acceleration of the manipulator 1, the force generated in the object gripped by the manipulator 1, the moment generated in the object, and the condition in which the manipulator 1 is installed. conditions regarding at least one of the reaction force applied to the movable frame 13 and the torque component of the reaction force.

Although a case has been described in which the parameters of the constraint are set in advance, the robot control device 100c may include a reaction force constraint learning section (not shown). The reaction force constraint learning unit uses machine learning to learn parameters of constraint conditions for at least one of the reaction force applied to the movable frame 13 on which the manipulator 1 is installed and the torque component of the reaction force, and learns the parameters of the constraints using machine learning. The parameters of the conditions may be stored in the constraint storage unit 2c. Specifically, the reaction force constraint learning unit acquires the reaction force applied to the movable frame 13 and the torque component of the reaction force using a sensor (not shown) during the operation of the manipulator 1, and determines the constraint conditions based on the acquired values. Learn as a parameter.

The speed calculation unit 4c calculates a command trajectory that shortens the operation time of the manipulator 1 within the range of the constraint conditions based on the command trajectory from the command trajectory storage unit 3 and the constraint parameters from the constraint condition storage unit 2c. Calculate the velocity profile of

The configuration of the speed calculation unit 4c is the same as that of the speed calculation unit 4 shown in FIG. This differs from the speed calculation unit 4 in that it includes the calculation of the torque component of , and that the constraint coefficient calculation unit 43 calculates the coefficients of the constraint including the constraints regarding the reaction force and the torque component of the reaction force. Note that when the robot control device 100c includes a reaction force constraint learning section, the speed calculation section 4c calculates the speed profile based on the command trajectory and the parameters of the constraint learned by the reaction force constraint learning section.

The gradient calculation unit 5c calculates the gradient regarding the command trajectory of the operation time of the manipulator 1, and outputs it as gradient information.

The configuration of the gradient calculation unit 5c is the same as that of the slope calculation unit 5 shown in FIG. This differs from the gradient calculation unit 5 in that it calculates the gradient related to the kinematic calculation result of the operating time and the gradient related to the dynamic calculation result of the operating time, also taking into consideration the components. Further, the profile gradient calculation section 54 differs from the gradient calculation section 5 in that the profile gradient calculation section 54 calculates the gradient by also taking into account the reaction force applied to the movable frame 13 during the operation of the manipulator 1 and the torque component of the reaction force.

As described above, since the robot control device 100c includes the constraint storage section 2c, the speed calculation section 4c, and the slope calculation section 5c, vibrations caused by reaction force on the movable frame 13 can be suppressed, and the manipulator 1 operation time can be shortened.

FIG. 13 is a flowchart showing an example of the operation of the robot control device 100c in the fourth embodiment. That is, FIG. 13 is a flowchart showing an example of the robot control method in the fourth embodiment. Steps ST1 to ST7 in FIG. 13 are the same as steps ST1 to ST7 in FIG. 6, so detailed explanation will be omitted here.

As shown in FIG. 13, when robot control is started by means not shown, the speed calculation unit 4c calculates a preset command trajectory for the manipulator 1, constraint conditions regarding the manipulator 1, and operation time of the manipulator 1. The speed profile of the manipulator 1 is calculated based on the evaluation index (step ST1). The dynamics calculation unit 42 in the speed calculation unit 4c performs kinematic calculation and dynamic calculation including calculation of the reaction force applied to the movable frame 13 and the torque component of the reaction force when the manipulator 1 operates. Further, the constraint coefficient calculation unit 43 in the speed calculation unit 4c calculates coefficients of constraint conditions, including constraints regarding the reaction force applied to the movable frame 13 and the torque component of the reaction force.

The gradient calculation unit 5c calculates the gradient regarding the command trajectory of the operation time of the manipulator 1 and outputs it as gradient information (step ST2). The constraint condition coefficient gradient calculation unit 53 in the slope calculation unit 5c calculates the slope related to the kinematic calculation result of the operation time and the operation, taking into account the reaction force applied to the movable platform 13 during the operation of the manipulator 1 and the torque component of the reaction force. Calculate the gradient regarding the time dynamics calculation results. Further, the profile gradient calculation section 54 in the gradient calculation section 5c calculates the gradient regarding the coefficient of the constraint condition of the operation time, taking into consideration the reaction force applied to the movable frame 13 and the torque component of the reaction force.

The command trajectory correction unit 6 corrects the command trajectory based on the slope information and outputs the corrected command trajectory (step ST3).

The command trajectory storage section 3 stores the corrected command trajectory (step ST4).

The command point sequence calculation unit 7 calculates a command point sequence based on the corrected command trajectory and the speed profile (step ST5).

The control unit 8 controls the actuator 110 so that the manipulator 1 follows the correction command trajectory (step ST6).

By means not shown, it is determined whether or not to continue controlling the robot (step ST11).

If the determination in step ST11 is "Yes", the process returns to step ST6, and control of the manipulator 1 is continued. If the determination in step ST11 is "No", control of the robot ends.

According to the fourth embodiment described above, by considering the reaction force applied to the movable frame 13 during the operation of the manipulator 1 and the torque component of the reaction force, vibration of the movable frame 13 can be suppressed, and the vibration of the movable frame 13 can be suppressed. 1 operation time can be shortened.

Note that in the fourth embodiment, the control of the manipulator 1 performed by the robot control device 100c can also be applied to the movable gantry 13. Furthermore, the present invention is not limited to the case where the manipulator 1 is installed on the movable pedestal 13, but can also be applied when the manipulator 1 is installed on a fixed pedestal (not shown).

The robot control devices 100, 100a, 100b, and 100c and the robot control method in Embodiments 1 to 4 can be applied to other than manipulator 1, which is a vertically articulated robot. For example, it can be applied to a manipulator 1 with any axis configuration, such as a horizontal articulated robot. Moreover, it can be applied to industrial devices other than the manipulator 1.

Here, the hardware configurations of robot control devices 100, 100a, 100b, and 100c in Embodiments 1 to 4 will be described. Each function of robot control devices 100, 100a, 100b, and 100c can be realized by a processing circuit. The processing circuit includes at least one processor and at least one memory.

FIG. 14 is a diagram showing the hardware configuration of robot control devices 100, 100a, 100b, and 100c in Embodiments 1 to 4. Robot control devices 100, 100a, 100b, and 100c can be realized by processor 200 and memory 201 shown in FIG. 14(a). The processor 200 is, for example, a CPU (Central Processing Unit, central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, also referred to as a DSP (Digital Signal Processor)) or a system LSI (Large Scale Integration).

The memory 201 is, for example, RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (registered trademark) (Electrical Non-volatile or These include a volatile semiconductor memory, an HDD (Hard Disk Drive), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a DVD (Digital Versatile Disk).

The functions of each part of the robot control devices 100, 100a, 100b, and 100c are realized by software or the like (software, firmware, or software and firmware). Software and the like are written as programs and stored in the memory 201. The processor 200 reads and executes programs stored in the memory 201 to realize the functions of each section. That is, this program can be said to cause a computer to execute the procedures or methods of the robot control devices 100, 100a, 100b, and 100c.

The program executed by the processor 200 may be an installable or executable file stored in a computer-readable storage medium and provided as a computer program product. Further, the program executed by processor 200 may be provided to robot control devices 100, 100a, 100b, and 100c via a network such as the Internet.

Further, the robot control devices 100, 100a, 100b, and 100c may be realized by a dedicated processing circuit 202 shown in FIG. 14(b). When the processing circuit 202 is dedicated hardware, the processing circuit 202 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate). Array), or a combination of these.

The above describes the configuration in which the functions of each component of the robot control devices 100, 100a, 100b, and 100c are realized by either software or hardware. However, the configuration is not limited to this, and a configuration may be adopted in which some of the components of the robot control devices 100, 100a, 100b, and 100c are realized by software, and other parts are realized by dedicated hardware. .

1 Manipulator, 2, 2c Constraint storage unit, 3 Command trajectory storage unit, 4, 4c Speed calculation unit, 41 Command interpolation calculation unit, 42 Dynamics calculation unit, 43 Constraint coefficient calculation unit, 44 Profile calculation unit, 5. 5c gradient calculation section, 51 command interpolation gradient calculation section, 52 dynamic gradient calculation section, 53 constraint coefficient gradient calculation section, 54 profile gradient calculation section, 6, 6a command trajectory correction section, 61 distance function calculation section, 62 barrier function calculation unit, 63 barrier function gradient calculation unit, 64 command trajectory correction value calculation unit, 7 command point sequence calculation unit, 8 control unit, 81 feedforward control unit, 82 feedback control unit, 83 current value calculation unit, 9 surrounding environment information storage unit, 10 command trajectory generation unit, 110 actuator, 111 gripping unit, 112 workpiece, 113 surrounding environment, 13 moving frame, 200 processor, 201 memory, 202 processing circuit.

Claims (10)

a speed calculation unit that calculates a speed profile of the manipulator based on a preset command trajectory to the manipulator, constraint conditions regarding the manipulator, and an evaluation index based on an operation time of the manipulator;
a slope calculation unit that calculates a slope regarding the command trajectory of the operation time based on the speed profile and uses the slope as slope information;
a command trajectory correction unit that corrects the command trajectory based on the gradient information to obtain a corrected command trajectory;
a control unit that controls the manipulator to follow the correction command trajectory;
A robot control device comprising: The robot control device according to claim 1, further comprising a command trajectory generation unit that generates the command trajectory based on peripheral information around the manipulator. The constraint condition is a condition related to at least one of the manipulator's joint angular velocity, joint angular acceleration, joint torque, hand speed, hand acceleration, a force generated in an object gripped by the manipulator, and a moment generated in the object. A robot control device according to claim 1 or 2. Further comprising a grip constraint learning unit that learns parameters of the constraint condition using machine learning for at least one of the force and the moment,
The robot control device according to claim 3, wherein the speed calculation section calculates the speed profile based on the parameters learned by the gripping constraint learning section. The robot control device according to any one of claims 1 to 4, wherein the constraint conditions include conditions regarding at least one of a reaction force applied to a pedestal on which the manipulator is installed and a torque component of the reaction force. further comprising a reaction force constraint learning unit that learns parameters of the constraint conditions using machine learning for at least one of the reaction force and the torque component of the reaction force,
The robot control device according to claim 5, wherein the speed calculation section calculates the speed profile based on the parameters learned by the reaction force constraint learning section. The speed calculation unit includes a command interpolation calculation unit that calculates a position at an interpolation point on the command trajectory and a first-order differential and a second-order differential with respect to a parameter of the command trajectory;
a dynamics calculation unit that performs kinematic calculation and dynamic calculation of the manipulator using the position, the first-order differential, and the second-order differential, and outputs the kinematic calculation result and the dynamic calculation result;
a constraint condition coefficient calculation unit that calculates a coefficient of the constraint condition based on the kinematic calculation result, the dynamic calculation result, and the constraint condition;
a profile calculation unit that calculates the speed profile based on the coefficients of the constraint;
The robot control device according to any one of claims 1 to 6, comprising: 8. The gradient calculating section calculates a gradient of the operating time with respect to the commanded trajectory by differentiating the operating time with respect to the commanded trajectory using automatic differentiation, and uses the gradient information as the gradient information. The robot control device according to item 1. 9. The command trajectory correction unit corrects the command trajectory by using any one of a gradient descent method, a conjugate gradient method, or a quasi-Newton method based on the gradient information. The robot control device described in . a step of calculating a velocity profile of the manipulator based on a preset command trajectory to the manipulator, constraint conditions regarding the manipulator, and an evaluation index based on an operation time of the manipulator;
Based on the speed profile, calculating a slope regarding the command trajectory of the operation time to obtain slope information;
correcting the command trajectory based on the gradient information to obtain a corrected command trajectory;
controlling the manipulator to follow the correction command trajectory;
A robot control method comprising:

Images