JPWO2019098044A1 - Robot motion adjustment device, motion control system, and robot system - Google Patents

Abstract

The robot operation is adjusted so that an excessive load does not act on the work target, and the adjustment is facilitated. The robot controller (111) sends an operation command value to the robot (120), causing the robot (120) to which the end effector (130) is attached to try the work on the work target (200). The external sensor (142) detects a force acting on the end effector (130) due to the trial. The motion adjustment device (112) performs learning using the detection result of the external sensor (142) to adjust and update the motion command value acquired from the robot control device (111).

Description

  The present invention relates to industrial robots, service robots for non-manufacturing industries, and the like. In particular, the present invention includes an operation adjustment device and an operation control system for adjusting the operation of the robot for causing the end effector mounted on the robot to reach a target position and orientation, and the operation adjustment device and the operation control system. It relates to robot systems.

  In conventional industrial robot systems, there are many system configurations in which the relationship between the robot and the work target is precisely positioned, and the robot repeats the work with high speed and high accuracy in the positioned environment. In contrast, in recent years, robot systems that use a plurality of external sensors such as force sensors or vision sensors are increasing. Such a robot system is used in an environment where the robot and the work target are not precisely positioned, and controls the robot operation according to the detection result of the external sensor.

  For example, such a robot system is used in a situation where the position and orientation of the object to be worked on or the surrounding environment is unknown. As another example, such a robot system is used in a situation where the position and orientation of an object to be worked or the surrounding environment changes. Specific examples include a bin picking operation, an insertion operation accompanied by a surface copying operation, and a fitting operation of components such as a connector. In the field of service robots for non-manufacturing industries, it is premised on work under various changing environments, and similarly, the operation of the robot is controlled using a plurality of sensors.

  In a robot control system using these sensors, adjustment of a plurality of control parameters is necessary to adjust the operation of the robot. By appropriately adjusting the control parameters, the operation of the robot becomes appropriate and the performance of the robot system is ensured. However, adjustment of control parameters is not easy and often requires specialized knowledge. Therefore, in order to facilitate the adjustment of the control parameter, some automatic adjustment means have been proposed. For example, Patent Document 1 discloses a robot system that speeds up the operation of a robot by learning.

JP 2017-94438 A

  In the conventional robot system, the magnitude of the load acting on the work target due to the movement of the robot is not considered in learning. Therefore, in the operation of the robot obtained by learning, the load acting on the work target may not be an appropriate magnitude, and an excessive load may act on the work target. An object of the present invention is to obtain an operation adjustment device, an operation control system, and a robot system that can adjust the operation of a robot so that an excessive load does not act on a work target, and can easily adjust the operation of the robot. To do.

  The robot motion adjustment device of the present invention includes a robot equipped with an end effector and a robot control device that controls the motion of the robot. The robot motion adjustment device is used in a robot system in which a robot performs work on a work target. A command value learning unit that performs learning using the detected force acting on the end effector as an input and adjusts an operation command value transmitted from the robot controller to the robot in order to control the operation of the robot is provided.

  The motion control system of the present invention is used in a robot system in which a robot equipped with an end effector performs work on a work target, and transmits a motion command value to the robot to control the robot motion. And a command value learning unit that adjusts the operation command value by performing learning using the force acting on the end effector detected by the sensor as an input.

  In addition, the robot system of the present invention includes a robot on which an end effector is mounted, a robot control device that transmits an operation command value to the robot to control the operation of the robot, and a force that acts on the end effector detected by the sensor. The robot includes a command value learning unit that performs learning as input and adjusts the motion command value, and the robot performs work on the work target.

  According to the motion adjusting device, the motion control system, and the robot system of the present invention, the motion of the robot can be adjusted so that an excessive load does not act on the work target, and the adjustment of the motion of the robot can be facilitated.

It is a block diagram which shows an example of the system configuration | structure of the robot system provided with the motion adjustment apparatus by Embodiment 1 of this invention. It is a figure which shows an example of the concrete hardware constitutions for implement | achieving the robot control apparatus and operation | movement adjustment apparatus by Embodiment 1 of this invention. It is a block diagram which shows the structural example of the operation adjustment apparatus by Embodiment 1 of this invention, and a surrounding block. It is a figure for demonstrating operation | movement of the operation adjustment apparatus by Embodiment 1 of this invention. It is a figure which shows an example of the speed pattern before the update in the robot system by Embodiment 1 of this invention. It is a flowchart which shows an example of the flow of a process of the operation control system by Embodiment 1 of this invention. It is a figure for demonstrating operation | movement of the operation adjustment apparatus by Embodiment 2 of this invention. It is a figure which shows an example of the initial value of the speed pattern in the robot system by Embodiment 2 of this invention. It is a figure which shows an example of the detected value of the force sensor in the robot system by Embodiment 2 of this invention. It is a figure which shows an example of the speed pattern after the update in the robot system by Embodiment 2 of this invention. It is a figure which shows another example of the speed pattern after the update in the robot system by Embodiment 2 of this invention. It is a block diagram which shows the structural example of the operation adjustment apparatus by Embodiment 3 of this invention, and a surrounding block. It is a block diagram which shows the structural example of the command value learning part by Embodiment 3 of this invention, and a surrounding block. It is a figure which shows an example of the operation | work which the robot system by Embodiment 3 of this invention implements. It is a flowchart which shows an example of the flow of a process of the learning process part by Embodiment 3 of this invention. It is a flowchart which shows an example of the flow of the pre-processing performed by the learning process part by Embodiment 3 of this invention. It is a flowchart which shows an example of the flow of the learning process performed in the learning process part by Embodiment 3 of this invention. It is a figure which shows an example of the speed pattern at the time of the trial in the robot system by Embodiment 3 of this invention. It is a figure which shows an example of the force information acquired at the time of the trial in the robot system by Embodiment 3 of this invention. It is a block diagram which shows the structural example of the operation adjustment apparatus by Embodiment 4 of this invention, and a surrounding block. It is a flowchart which shows an example of the flow of the pre-processing performed with the operation | movement adjustment apparatus by Embodiment 4 of this invention. It is a flowchart which shows an example of the flow of the learning process performed with the operation | movement adjustment apparatus by Embodiment 4 of this invention. It is a block diagram which shows the structural example of the operation adjustment apparatus by Embodiment 5 of this invention, and a surrounding block. It is a block diagram which shows the structural example of the action learning part by Embodiment 5 of this invention. It is a block diagram which shows another structural example and peripheral block of the operation | movement adjustment apparatus by Embodiment 5 of this invention.

Embodiment 1.
FIG. 1 is a block diagram showing an example of a system configuration of a robot system 100 provided with an operation adjustment apparatus according to Embodiment 1 of the present invention. As shown in FIG. 1, the robot system 100 includes an operation control system 110, a robot 120, an end effector 130, an internal sensor 141, and an external sensor 142. Further, the motion control system 110 includes a robot control device 111 and a motion adjustment device 112. The robot control device is also called a robot controller.

  The robot control device 111 transmits an operation command value for controlling the operation of the robot 120 to the robot 120 based on the detection results of the internal sensor 141 and the external sensor 142, and controls the operation of the robot 120. An end effector 130 such as a robot hand is attached to the robot 120. The end effector 130 works directly on the work target 200. An appropriate type of end effector 130 is selected according to each operation performed by the robot system 100. A surrounding environment 300 exists around the work target 200.

  The peripheral environment 300 includes, for example, a part to which the work target 200 is assembled, a jig for positioning the work target 200, a tool for processing the work target 200 (such as an electric screwdriver), a parts feeder that supplies the work target 200, and the robot 120. A safety cover that surrounds the belt, a belt conveyor that conveys the work object 200, and the like. In some cases, the external sensor 142 such as a camera that captures a work target is also handled as a part of the surrounding environment. This is because the robot 120 or the end effector 130 may come into contact with the external sensor 142 when the external sensor 142 is fixed at a predetermined position around the robot 120.

  The operation command value output from the robot control device 111 is, for example, information indicating the target position and target posture of the end effector 130 attached to the robot 120 at each time, that is, a position command value. When the motion command value represents the target position of the end effector 130 at each time, the movement speed of the end effector 130 between the times is also represented by the motion command value. Therefore, the position command value can be considered as a speed command value representing the target operation speed of the robot.

  Further, the operation command value output from the robot control device 111 may be a speed command value representing the target operation speed of the robot 120 or the target movement speed of the end effector 130. The target operation speed or the target movement speed is given by a speed between each time point of the operation of the robot 120 or a speed between each point on the route. Further, the motion command value may be an acceleration command value that represents a target acceleration for the motion of the robot 120 or a target acceleration for the movement of the end effector 130. The motion command value may be in various forms as long as it directly controls the motion of the robot 120.

  The motion adjustment device 112 adjusts and updates the motion command value generated by the robot control device 111 according to the detection result of the external sensor 142 and the constraint condition given from the outside. That is, the motion adjustment device 112 adjusts the motion of the robot. In other words, the motion adjustment device 112 adjusts the correspondence relationship between the detection results of the inner world sensor 141 and the outer world sensor 142 and the motion command value output from the robot control device 111, and reflects the adjustment result to change the correspondence relationship. Will be updated. The adjustment of the operation command value can also be referred to as the correction of the operation command value or the correction of the operation command value.

  When the updated operation command value exists, the robot control device 111 outputs the updated operation command value to the robot 120. The motion adjustment device 112 may update the motion command value with reference to not only the detection result of the external sensor 142 but also the detection result of the internal sensor 141. Note that the constraint condition may be stored in advance in the motion adjustment device 112 or the robot control device 111.

  The robot system 100 according to the present embodiment performs two processes: an adjustment process for adjusting and updating an operation command value, and a work process for performing work on the work target 200 using the updated operation command value. In other words, the operation of the robot system 100 includes an adjustment phase and a work phase, and the adjustment process is a process of the robot system 100 in the adjustment phase. The work process is a process of the robot system 100 in the work phase. The motion adjustment device 112 adjusts the motion command value so as to obtain an optimal motion command value in the adjustment process. However, the adjustment process and the work process need not be completely separated. For example, the robot system 100 may be configured so that the motion adjustment device 112 calculates an optimal motion command value as needed while a work on the work target 200 is being performed. In this configuration, the robot system 100 updates the operation command value at a predetermined timing as necessary, for example, when an operation command value more appropriate than the currently used operation command value is calculated. This is the same in the following embodiments.

  FIG. 2 is a diagram illustrating an example of a specific hardware configuration for realizing the robot control device 111 and the motion adjustment device 112. The robot control device 111 and the motion adjustment device 112 are realized by the processor 401 executing a program stored in the memory 402. The processor 401 and the memory 402 are connected by a data bus 403. The memory 402 includes a volatile memory and a non-volatile memory, and temporary information is stored in the volatile memory. Note that the robot control device 111 and the motion adjustment device 112 may be configured integrally or may be configured separately. For example, the robot control device 111 and the motion adjustment device 112 may be connected via a network or the like. Also in the following embodiments, the robot control device 111 and the motion adjustment device 112 can be realized by the same hardware configuration.

  The robot system 100 constitutes a control system in which the operation control system 110 outputs an operation command value based on the data acquired by the internal sensor 141 and the external sensor 142, and the robot 120 operates following the operation command value. ing. Examples of the internal sensor 141 include a sensor that acquires the position of the joint of the robot, a sensor that acquires the operation speed of the joint, and a sensor that acquires the current value of the motor for operating the joint. The robot system 100 constitutes a position control system that positions the end effector 130 by the robot control device 111, the robot 120, and the internal sensor 141. As a sensor for acquiring the position of the joint of the robot, for example, an encoder, a resolver, a potentiometer, or the like that detects the rotation amount of the motor can be considered. In addition, a tachometer or the like can be considered as a sensor for acquiring the operation speed of the joint. In addition, as the internal sensor, a gyro sensor, an inertia sensor, or the like may be used as information on the robot 120 itself.

  By feedback control based on the inner world sensor 141, the robot system 100 constitutes a position control robot system that performs material handling work and the like. Here, the material handling operation is an operation of transferring or transporting materials or parts. This position control robot system is referred to as a feedback control system based on the internal sensor 141. In feedback control based on the internal sensor 141, control parameters include position control gain, speed control gain, current control gain, filter design parameters used for feedback control, and the like. As a filter used for feedback control, a moving average filter, a low-pass filter, a band-pass filter, a high-pass filter, and the like can be considered. Note that the feedback control based on the inner world sensor 141 is control for the robot 120 to operate according to the operation command value. In other words, the feedback control based on the internal sensor 141 is control performed to realize the operation command value.

  On the other hand, the external sensor 142 includes a force sensor, a vision sensor such as a camera, a tactile sensor, and a touch sensor. The external sensor 142 measures the contact state and positional relationship between the robot 120 and the work target 200 or the surrounding environment 300. In the robot system 100, the robot control device 111, the motion adjustment device 112, the robot 120, and the external sensor 142 constitute a sensor feedback control system based on the external sensor 142. Further, the robot system 100 may not use the sensor feedback control based on the sensor signal output from the external sensor 142 but may use the sensor signal from the external sensor 142 as a trigger signal. In this case, the robot system 100 switches control parameters for feedback control by the internal sensor 141 with the trigger signal as a starting point. The sensor feedback control system based on the external sensor 142 is constructed as an outer loop of the position control robot system.

  The sensor feedback control system based on the external sensor 142 is based on acceleration, speed, position and orientation, distance, force, moment, and the like, based on the positional relationship and contact between the robot 120, the robot arm or the end effector 130, and the work target 200 or the surrounding environment 300. Sensing behavior. Furthermore, the sensor feedback control system based on the external sensor 142 controls the operation of the robot 120 so as to obtain a desired positional relationship or force response based on the sensing result. In other words, the sensor feedback control system based on the external sensor 142 corrects the operation command value so as to obtain a desired positional relationship or force response. In the sensor feedback control system based on the external sensor 142, the control parameters include force control gain related to haptic control, impedance parameter, gain related to visual servo control, visual impedance parameter, filter setting parameter used for feedback control, and the like.

  In the case where control is performed based on the inner world sensor 141 and the outer world sensor 142, control parameters that require adjustment may be simply referred to as parameters hereinafter. Here, specific examples of the sensor used as the internal sensor 141 or the external sensor 142 include a current value sensor, a joint position sensor, a joint speed sensor, a temperature distance sensor, a camera, an RGB-D sensor, and a proximity sensor. A tactile sensor, a force sensor, etc. are conceivable. In addition, the measurement target of the internal sensor 141 or the external sensor 142 may be the position / posture of the robot 120, the position / posture of the end effector 130, the position / posture of the work to be the work target 200, the position / posture of the operator, and the like.

  FIG. 3 is a block diagram illustrating a configuration example of the motion adjustment device 112 according to the first embodiment of the present invention and peripheral blocks. FIG. 3 shows a part of the configuration of the robot system 100 extracted. The motion adjustment device 112 includes a command value learning unit 113. In FIG. 3, the sensor 140 is a combination of the inner world sensor 141 and the outer world sensor 142. As described above, various sensors 140 are conceivable. However, in the robot system 100 of the present embodiment, the sensor 140 includes at least a force sensor that detects an external force acting on the end effector 130 due to the operation of the robot 120. This force sensor becomes the external sensor 142. Note that the sensor 140 includes at least a force sensor in the following embodiments.

  The force sensor is used for measuring an external force acting on the end effector 130 and performing force control or impedance control. Controlling the force that the end effector 130 gives to the work target 200 or the surrounding environment 300 is called force control. Controlling the operation of the robot 120 according to the detection result of the force sensor is called force sense control. In the force control, a target work force is set, and the magnitude of the force applied to the work target 200 or the surrounding environment 300 is controlled.

  On the other hand, in impedance control, impedance characteristics (spring, damper, inertia) relating to contact force generated when the end effector 130 and the work target 200 come into contact are defined and used for control. As a case where the contact force is generated, a case where the end effector 130 and the surrounding environment 300 come into contact with each other, a case where the work target 200 held by the end effector 130 and the surrounding environment 300 come into contact, or the like can be considered. The impedance characteristic is represented by an impedance parameter.

  In force control, it is necessary to determine a target value for force control. In impedance control, it is necessary to determine control characteristics using impedance parameters. Furthermore, in both force control and impedance control, it is necessary to determine a gain that contributes to control responsiveness, and there are many adjustment items. In conventional robot systems, many parameter adjustments have been made for the purpose of performing work stably. In this case, the system characteristics including the response of the operation of the robot 120, the mechanical rigidity, and the like are identified, and one parameter set that stably responds regardless of conditions or states is found. However, in the operation of the robot 120 with contact with the work target 200, the contact state between the work target 200 and the end effector 130 changes as the operation progresses. Therefore, the adjustment of the parameter set needs to be performed in consideration of the transition of the contact state. This adjustment has been done on a trial and error basis and has not been easy.

  In the robot system 100 of the present embodiment, the motion adjustment device 112 controls the motion of the robot 120 to be appropriate by updating the motion command value. A constraint condition is input to the motion adjustment device 112. The constraint condition includes an upper limit value or a lower limit value of force information detected by the force sensor. In the following description, it is assumed that the operation command value output from the operation control system 110 is a speed command value. The speed command value is a target moving speed of the end effector 130 for each point on the moving path of the end effector 130. At this time, the time-series speed command value is a speed pattern for each point. The speed command value may be a target operation speed of the robot 120 for each time point during work.

  In the speed pattern, a target speed Vi (i = 1, 2, 3,...) And a target speed switching position Pi (i = 1, 2, 3,...) Are defined. Note that the switching position may be set by a switching time or a parameter for switching. As the parameter for switching, the progress rate of the operation command value based on the position and time is exemplified. The target speed switching position Pi may be a target speed switching start point or a target speed switching completion point. Further, the target speed switching position Pi may be a point where it is ensured that the operation speed detected by the inner sensor 141 falls within a predetermined error range from the target speed.

  FIG. 4 is a diagram for explaining the operation of the operation adjustment device 112 according to the first embodiment of the present invention. Consider the case where the end effector 130 attached to the robot 120 moves from position P0 to position P3 as shown in FIG. A force sensor 143 is attached to the robot 120 as the external sensor 142. The force sensor 143 measures an external force acting on the end effector 130.

  FIG. 5 is a diagram showing an example of a speed pattern before update in the robot system 100 according to the first embodiment of the present invention. In FIG. 5, the horizontal axis represents the position P of the end effector 130, and the vertical axis represents the target moving speed V of the end effector 130. In the speed pattern of FIG. 5, the target speed changes while the end effector 130 moves from P0 to P3. The motion adjustment device 112 updates the speed pattern based on the detection result of the force sensor 143.

  FIG. 6 is a flowchart showing an example of the processing flow of the operation control system 110 according to the first embodiment of the present invention. Here, the constraint conditions include an upper limit value and a lower limit value of force information detected by the force sensor 143 and an upper limit value of work time. First, in step S10, the robot control device 111 determines an initial value of the speed pattern. Next, in step S <b> 11, the robot control device 111 tries the operation by controlling the operation of the robot 120. Note that, as described above, a part of normal work in the robot system 100 may be treated as a trial, such as when adjustment processing and work processing are not completely separated.

  Next, in step S12, the motion adjustment device 112 determines whether the constraint condition is satisfied. That is, in step S12, the motion adjustment device 112 determines whether the detection value of the force sensor 143 is between the upper limit value and the lower limit value defined by the constraint conditions and whether the work time constraint is satisfied. To do. When determining the detection value of the force sensor 143, for example, the maximum value of the detection value is compared with the upper limit value of the constraint condition, and the minimum value of the detection value is compared with the lower limit value of the constraint condition. In step S12, the motion adjustment device 112 may use an evaluation value obtained by calculation from the detection value instead of the detection value itself of the force sensor 143. As an example of the evaluation value, an evaluation value calculated by an evaluation function having the detection value of the force sensor 143 and the tact time as inputs can be considered. In step S12, the motion adjustment device 112 may determine whether or not the evaluation value is within the limit range.

  If it is determined in step S12 that the constraint condition is satisfied, the process of the motion control system 110 is temporarily terminated, and thereafter, the work with the updated speed pattern is performed. On the other hand, when it is determined in step S12 that the constraint condition is not satisfied, the process of the operation control system 110 proceeds to step S13. In step S13, the operation adjustment device 112 adjusts the speed pattern and updates the speed pattern. In step S13, the motion adjustment device 112 adjusts the speed pattern by, for example, calculating a correction coefficient for correction and multiplying the speed pattern when the trial is performed. When the process of step S13 ends, the process of the operation control system 110 returns to step S11.

  The operation control system 110 according to the first embodiment of the present invention performs the processing as described above. As described above, the motion control system 110 according to the first embodiment of the present invention adjusts the speed pattern in a learning manner based on data obtained by a plurality of trials. In other words, the motion control system 110 according to the first embodiment of the present invention adjusts the speed pattern, which is the motion command value, using machine learning or an optimization method.

  In the above description, it is assumed that the upper limit value of the work time is included in the constraint condition, but it may be other conditions instead of the essential condition. Further, instead of being given the upper limit value of the work time as a constraint condition, the constraint condition may be that the work time becomes the shortest after satisfying other conditions. Further, in the above description, the case where the motion control system 110 updates the motion command value so as to satisfy the given constraint condition has been described. However, the motion control system 110 is configured to adjust and update the control parameter. Is also possible. Further, FIG. 1 illustrates a configuration example in which the robot control device 111 and the motion adjustment device 112 are separately provided, but the robot control device 111 may be configured to incorporate the motion adjustment device 112.

  The motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment are configured as described above. According to the motion adjustment device 112, the motion control system 110, and the robot system 100 of the present embodiment, the motion of the robot 120 is adjusted so that the detection value of the force sensor 143 is within a predetermined range. Here, the detection value of the force sensor 143 represents the magnitude of the external force acting on the end effector 130. In other words, the detection value of the force sensor 143 is information indicating the magnitude of the force applied to the work target 200 or the surrounding environment 300 due to the operation of the robot 120. Therefore, according to the motion adjustment device 112, the motion control system 110, and the robot system 100 of the present embodiment, the force applied to the work target 200 or the surrounding environment 300 becomes an appropriate magnitude, that is, the work target 200 or the surrounding area. The operation of the robot 120 can be adjusted so that an excessive load does not act on the environment 300, and the adjustment of the operation of the robot 120 can be facilitated.

  As described above, by using the force sensor 143 to adjust the motion command value so that the force response falls within a desired range, a high-quality robot work that does not damage the work item is realized. can do. Furthermore, high-speed work can be realized by adding work time to the constraint condition.

  Further, although the motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment use the magnitude of the force detected by the force sensor 143 as a constraint condition, the moment, torque, current value, and the like are used. It is also possible to detect and use either the upper limit or the lower limit as a constraint condition. As a result, a limit value can be set for the contact state between the robot 120 or the end effector 130 and the outside world, and an operation command value within a desired range can be searched. As a result, an operation that does not damage the operation target 200 can be realized.

  Furthermore, as a constraint condition, a relative position and orientation with respect to the surrounding environment 300 and a position and orientation of the robot 120 can be added. By adding either the upper limit or the lower limit to the constraint condition, it is possible to realize a robot operation that suppresses interference with the surrounding environment 300 while realizing a high-quality operation. As a result, it is possible to obtain a special effect such as increasing the operating rate of the system. The effects described above can be obtained in other embodiments as well.

Embodiment 2.
The configuration of the motion adjustment device, the motion control system, and the robot system of the present embodiment is the same as that shown in FIG. The motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment divide the motion command given to the robot 120 for a series of work into a plurality of sections and adjust the motion command value for each section. Is. In the following description, it is assumed that the operation command value output from the operation control system is a speed command value.

  FIG. 7 is a diagram for explaining the operation of the operation adjustment device 112 according to the second embodiment of the present invention. As shown in FIG. 7, an operation of moving the end effector 130 attached to the robot 120 from the position P0 to the position P3 is considered. The initial position P0 is the work start point, and the position P3 is the work end point. The end effector 130 passes through the position P1 and the position P2 while moving from the position P0 to the position P3.

  In the robot system 100 according to the present embodiment, the path from the work start point to the work end point is divided into a plurality of sections. In other words, in the robot system 100 of the present embodiment, the operation of the robot 120 from the start of one work to the end of the work is divided into a plurality of sections. Here, the position P0 to the position P1 is set as a section S1, the position P1 to the position P2 is set as a section S2, and the position P2 to the position P3 is set as a section S3. Further, the target moving speed of the section S1 is V1, the target moving speed of the section S2 is V2, and the target moving speed of the section S3 is V3. The robot system 100 according to the present embodiment adjusts and updates the operation command value for each divided section. Specifically, the robot system 100 adjusts the target moving speed of the section S1, the target moving speed of the section S2, and the target moving speed of the section S3.

  In the robot system 100 according to the present embodiment, the positions P1 and P2 that are division points for division into segments are determined in advance according to the work content. The positions P1 and P2 are positions at which the sections are switched, and are sometimes called switching positions. In addition, although the number of sections is illustrated as three here, the number is not limited to three. Furthermore, although the section is spatially defined according to the position here, the time from the start of the work to the end of the work may be divided in time.

  It is assumed that the upper limit value Flim of the detection result of the force sensor 143 is given to the motion control system 110 of the present embodiment as a constraint condition. The processing flow of the operation control system 110 of the present embodiment is basically the same as the flowchart shown in FIG. However, the speed pattern is adjusted for each section. First, in step S10 of FIG. 6, the robot controller 111 determines an initial value of the speed pattern. FIG. 8 is a diagram showing an example of initial values of speed patterns in the robot system 100 according to the second embodiment of the present invention. In FIG. 8, the horizontal axis is the position P of the end effector 130, and the vertical axis is the target moving speed V of the end effector 130. In FIG. 8, the initial value of the speed pattern is V1 = V2 = V3 = Vini.

  Next, in step S <b> 11, the robot control device 111 tries the operation by controlling the operation of the robot 120. FIG. 9 is a diagram illustrating an example of detection values of the force sensor 143 in the robot system 100 according to the second embodiment of the present invention. In FIG. 9, the horizontal axis represents the position P of the end effector 130, and the vertical axis represents the detection value F of the force sensor 143. FIG. 9 shows values detected by the force sensor 143 when the robot 120 is operated with the initial value of the speed pattern shown in FIG.

  Next, in step S12, the motion adjustment device 112 determines whether the constraint condition is satisfied. That is, in step S12, the motion adjustment device 112 determines whether the detection value of the force sensor 143 in each section is equal to or less than the upper limit value Flim defined by the constraint condition. As the detection value of the force sensor 143 used for the determination, for example, the maximum value among the detection values of the force sensor 143 in each section is used. In step S12, when the detection value of the force sensor 143 is equal to or smaller than Flim in all sections, the motion adjustment device 112 determines that the constraint condition is satisfied. On the other hand, in step S12, when there is at least one section in which the detection value of the force sensor 143 exceeds the upper limit value Flim, the motion adjustment device 112 determines that the constraint condition is not satisfied.

  If it is determined in step S12 that the constraint condition is satisfied, the process of the motion control system 110 is temporarily terminated, and thereafter, the work with the updated speed pattern is performed. On the other hand, when it is determined in step S12 that the constraint condition is not satisfied, the process of the operation control system 110 proceeds to step S13. In step S13, the motion adjustment device 112 adjusts the speed pattern so that the target speed of the section in which the detection value of the force sensor 143 exceeds the upper limit value Flim is small, and updates the speed pattern.

  In the example shown in FIG. 9, the detection value Fmax2 of the force sensor 143 exceeds the upper limit value Flim in the section S2. On the other hand, the detection value Fmax1 of the force sensor 143 in the section S1 and the detection value Fmax3 of the force sensor 143 in the section S3 do not exceed the upper limit value Flim. Accordingly, in step S12, the motion adjustment device 112 determines that the constraint condition is not satisfied. In step S13, the motion adjustment device 112 adjusts the speed pattern so that the target speed V2 in the section S2 becomes small. The operation control system 110 according to the second embodiment of the present invention performs the above processing. FIG. 10 is a diagram showing an example of the updated speed pattern in the robot system 100 according to the second embodiment of the present invention. 10, the horizontal axis represents the position P of the end effector 130, and the vertical axis represents the target moving speed V of the end effector 130.

  In the above description, the upper limit value Flim of the detection result of the force sensor 143 is given as a constraint condition. However, it may be added as a constraint condition that the work time is further shortest. In this case, since Fmax1 and Fmax3 do not exceed the upper limit value Flim in FIG. 9, in step S13, the motion adjustment device 112 increases the target speed V1 in the section S1 and the target speed V3 in the section S3. Adjust the speed pattern. By adjusting the speed pattern in this way, the work time can be further shortened. FIG. 11 is a diagram showing another example of the updated speed pattern in the robot system 100 according to the second embodiment of the present invention. In FIG. 11, the horizontal axis represents the position P of the end effector 130, and the vertical axis represents the target moving speed V of the end effector 130.

  When the operation command value is a speed command value, the dividing points P1 and P2 are positions where the target speed is switched as shown in FIGS. The division points P1 and P2 may be start points for switching the target speed, or may be completion points for switching the target speed. Further, the division points P1 and P2 may be points at which the operating speed detected by the inner sensor 141 is guaranteed to be within a predetermined error range from the target speed.

  The motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment are configured as described above. According to the motion adjustment device 112, the motion control system 110, and the robot system 100 of the present embodiment, the motion of the robot 120 is adjusted for each section. Since the operation is adjusted so that the operation is delayed only in the section in which the detection value of the force sensor 143 is larger than the predetermined value, the operation of the entire work is not unnecessarily delayed, and the work object 200 or the surrounding environment 300 is excessively large. Therefore, the operation of the robot 120 can be adjusted so that a large load is not applied, and the adjustment of the operation of the robot 120 can be facilitated. Furthermore, if it is configured so that the operation of the section in which the detection value of the force sensor 143 is smaller than a predetermined value is adjusted so that the operation becomes faster, the operation of the entire work can be made faster.

  As described above, according to the motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment, it can be realized by conventional adjustment by learning and updating the optimum motion command value for each section. It was possible to design detailed operation command values that were not present, and as a result, high-speed and high-quality robot work could be realized.

Embodiment 3 FIG.
FIG. 12 is a block diagram showing a configuration example of the motion adjustment device 112b according to Embodiment 3 of the present invention and peripheral blocks. FIG. 12 shows a part of the configuration of the robot system 100 extracted. The motion adjustment device 112b includes a command value learning unit 113b. The configuration of the motion adjustment device, the motion control system, and the robot system of the present embodiment is the same as that shown in FIG. 1 except that the motion adjustment device 112 is replaced with the motion adjustment device 112b. The operation adjustment device 112b in the present embodiment is different from the operation adjustment device 112 in the second embodiment in that segment information is input. The division information includes information on the initial value of the division position and the initial value of the operation command value in each division. The section position is the position of the dividing point Pi at both ends of each section, for example, a position where the target value of the operation speed is switched. When the inner world sensor 141 or the outer world sensor 142 detects that the end effector 130 has reached a predetermined position, the target value of the operation speed is switched.

  The motion adjustment device, motion control system, and robot system according to the present embodiment adjust the motion command value for each section in the same manner as in the second embodiment. Learning to adjust the command value learning unit 113b to adjust the command value learning unit 113b for each category so that the category in which a collision or the like occurs is a low-speed operation, and to adjust the other categories to a high-speed operation. Can be a container. According to this learning device, it is possible to automatically learn an operation command value that realizes high-speed work. The command value learning unit 113b automatically learns the operation command value corresponding to each section. For the sake of simplicity, description will be made assuming that the motion adjustment device 112b adjusts only the motion command value without adjusting the control parameter.

  In the motion adjustment device, the motion control system, and the robot system according to the present embodiment, the command value learning unit 113b receives the segment information, the constraint condition, the detection value of the sensor 140, and the motion command value before update as the motion command value. Update each. The category information is defined to divide the operation command value into N categories. Each division point for dividing into sections is defined as Pi (i = 0, 1, 2,..., N + 1). Here, N is a natural number. Here, the start point and end point of the operation are also included in the division points, and the start point is P0. A section between the division point immediately before the division point Pi and the division point Pi is called a section Si (i = 0, 1, 2,..., N).

  In the motion adjustment device, the motion control system, and the robot system according to the present embodiment, it is assumed that the classification is defined every time the work state changes. For example, considering a fitting operation using a force sensor, the dividing point Pi is defined before and after the contact phenomenon between the components to be fitted occurs and before and after the contact state changes. The dividing point Pi is defined according to the expected change in the contact state, and the speed of the entire work can be increased by changing to the target values of the operation commands such as the position, speed, and acceleration suitable for each. At this time, it is a feature of the motion adjusting device, the motion control system, and the robot system of the present embodiment that an appropriate position of the dividing point Pi and a command value pattern of each section Si are defined from past trial information.

  The constraint condition input to the command value learning unit 113b is a condition that defines the boundary between the work success and the work failure with respect to the speeded up work. In a high-speed work operation, there is a risk that the end effector 130 may strongly collide with the work target 200 due to an error in position control of the end effector 130 or the like. When a strong collision occurs, the end effector 130 or the work target 200 may be damaged, resulting in work failure. In consideration of such past work failures, generation of operation command values for high-speed and low-impact work by the user defining constraints at the time of design or by defining constraints based on past trial data Is feasible.

  Constraint conditions include position limit range, posture limit range, upper limit value of motion speed, lower limit value of motion speed, upper limit value of force, lower limit value of force, upper limit value of moment, lower limit value of moment, etc. . In particular, when the position / orientation of the robot 120 or the work target 200 can be acquired, the position / orientation of the robot 120 is defined as a constraint condition, and either the upper or lower limit of the relative position / orientation of the robot 120 and the surrounding environment 300 is defined. Limit value can be entered.

  Data acquired by the inner sensor 141 or the outer sensor 142 is referred to as sensor information. For sensor information, preprocessing such as processing for removing noise by filter processing and processing for extracting only values exceeding a threshold value is performed as necessary.

  The operation command value refers to a control command value that can be input to the position control system of the robot system 100. The operation command value is sometimes simply referred to as a command value. The operation of the robot 120 is controlled by the operation of the motor of each axis. The operation command value includes, for example, a position command value, a speed command value, a current command value and the like for controlling the operation of the motor. Further, the motion adjustment device 112b can equivalently generate time series data of the position command value from the speed pattern generated from the profile representing the relationship between the time and the speed, and can input it to the robot control device 111. The operation command value can also be generated inside the robot controller 111.

  The motion adjustment device 112b according to the present embodiment takes out the command value inside the robot control device 111, adjusts the motion command value according to the sensor information obtained as a response when the robot 120 performs the work, and updates it. To do. This is the same in other embodiments. In the following description, it is assumed that the operation command value output from the operation control system is a speed command value. As another configuration, a configuration in which the motion adjustment device 112b passes not the motion command value itself but a parameter necessary for generating the motion command value to the robot control device 111 is also conceivable. For example, the motion adjustment device 112b can input only the segment position and the target value of the motion speed in each segment to the robot control device 111. In this case, the robot control device generates an operation command value based on the input segment position and operation speed target value.

  The motion adjustment device 112b includes a command value learning unit 113b. The command value learning unit 113b adjusts and updates the operation command value. The command value learning unit 113b obtains a new motion command value based on the classification information, the constraint condition, the motion command value before update, and the detection value of the sensor 140. The command value learning unit 113b is designed so that when a new motion command value is obtained, the work speed and work quality are evaluated by an evaluation function, and the work object 200 is not easily broken and searches for a fast motion. In addition, the motion adjustment device 112b may be configured to adjust and update the control parameters used in the robot control device 111. Adjustment and update of the control parameters are also performed by the command value learning unit 113b.

  FIG. 13 is a block diagram illustrating a configuration example of the command value learning unit 113b according to Embodiment 3 of the present invention and peripheral blocks. FIG. 13 shows a part of the configuration of the robot system 100 extracted. The command value learning unit 113b includes a storage unit 114 and a learning processing unit 115. An example of a search method in the command value learning unit 113b will be described with reference to FIG. Here, it is assumed that the number of sections is defined as N = 4 in advance. Further, it is assumed that a speed target value, which is a target speed value defined in each section, is used as an operation command value. In addition, the command value learning unit 113b realizes high-speed work by adjusting the speed target value in each section.

  FIG. 14 is a diagram illustrating an example of work performed by the robot system 100 according to the third embodiment of the present invention. As shown in FIG. 14, the robot system 100 performs an operation of inserting the first component 210 into the second component 310. FIG. 14 illustrates changes in the relative positions of the first component 210 and the second component 310 as the work progresses, in the order of (a), (b), (c), and (d). It shows how work is progressing. The first part 210 corresponds to the work target 200, and the second part 310 corresponds to the surrounding environment 300.

  A hole 211 is provided in the first component 210. On the other hand, the second component 310 is provided with a protrusion 311. When the first component 210 is inserted into the second component 310, the protrusion 311 is inserted into the hole 211. The first component 210 is made of a first material. On the other hand, the second component 310 includes a portion 312 made of the first material and a portion 313 made of the second material. When the first component 210 is inserted into the second component 310, the contact state between the first component 210 and the second component 310 changes.

  In the example illustrated in FIG. 14, the part and the contact state that are in contact with each other change as the work progresses from (b) to (d). Examples of the contact state include the material of each part of the contact portion, the width of the contact portion, and the like. As the contact state changes, the frictional force generated at the contact portion changes. In FIG. 14B, a frictional force between the outer shapes of the first component 210 and the second component 310 is generated. In FIG. 14C, since the contact between the hole 211 and the protrusion 311 is further added, the frictional force is increased. The detection result of the force sensor 143 also changes due to a change in the frictional force generated between the parts. That is, in the part fitting work and the connector inserting work, the reaction force between the parts changes according to the progress of the work. The force sensor detects the reaction force between the components.

  As illustrated in FIG. 13, the command value learning unit 113 b stores the force information detected by the sensor 140 and the speed pattern acquired from the robot control device 111 in the storage unit 114. It is assumed that the robot system 100 can operate the robot 120 by designating a speed pattern when attempting a work for adjusting the operation command value. Based on the force information, speed pattern, classification information, and constraint conditions stored in the storage unit 114, the learning processing unit 115 updates the speed pattern and outputs it to the robot control device 111 as offline processing.

  Here, one speed pattern is stored in the robot control device 111. However, when adjusting the operation command value, the motion adjustment device 112b also applies a plurality of types to one reference speed pattern. Encourage them to try the task using the speed pattern. As a result, when adjusting the operation command value, the robot system 100 tries under various conditions. The motion adjustment device 112b adjusts the motion command value based on data obtained through trials under various conditions. For example, the robot system 100 performs Na trials with different motion command values including motion command values different from the motion command values stored in the robot controller 111. The motion adjustment device 112b inputs data obtained as a result of Na trials, learns once, and updates the motion command value. When N trials are performed with Na trials as one set, the operation command value converges in many cases, and no further improvement occurs. Here, Na and Nb are integers of 1 or more.

  As described above, in the robot system 100 according to the present embodiment, trial is performed using one or more set operation command values, and an evaluation value is generated based on the obtained force sensor data. . The motion adjustment device 112b updates the motion command value based on each evaluation value. In the update of the operation command value, the operation adjustment device 112b generates one or more plural operation command values, and performs the trial again. When there is only one operation command value, if the evaluation value has converged in the graph in which the evaluation value is plotted, the operation adjustment device 112b ends the update of the operation command value. When there are a plurality of operation command values, the motion adjustment device 112b updates the operation command value if the evaluation value converges in a graph in which only the result that minimizes the evaluation value corresponding to the operation command value is plotted. finish. In this case, when a plurality of operation command values have been updated, the operation adjustment device 112b updates the operation command value with the smallest evaluation value.

  FIG. 15 is a flowchart showing an example of a processing flow of the learning processing unit 115 according to the third embodiment of the present invention. As shown in FIG. 15, first in step S100, the learning processing unit 115 performs preprocessing as a preparation stage. Next, in step S200, the learning processing unit 115 performs learning processing.

  FIG. 16 is a flowchart illustrating an example of the flow of preprocessing performed by the learning processing unit 115 according to Embodiment 3 of the present invention. For the purpose of explaining the operation, FIG. 16 also shows an operation performed by blocks other than the learning processing unit 115. First, in step S101, the robot control device 111 sets a control parameter for performing force sense control. Next, in step S <b> 102, the robot control device 111 tries the operation by operating the robot 120. Next, in step S103, the command value learning unit 113b acquires data obtained by the trial. The data obtained by each trial is called trial data. The trial data includes the force information detected in each trial and the velocity pattern used in each trial. The force information is time-series data acquired by the force sensor 143 at predetermined time intervals in each trial, and is also called a force waveform. Next, in step S104, the storage unit 114 stores the data acquired in step S103.

  Next, in step S105, the learning processing unit 115 determines whether or not K or more pieces of trial data have been acquired. Here, K is a natural number and is set in advance. If K or more pieces of trial data have not yet been acquired, the process returns to step S102. On the other hand, if K or more pieces of trial data have been acquired, the process proceeds to step S106. Therefore, when the process proceeds to step S106, K pieces of trial data D1j (j = 1, 2, 3,... K) are acquired and stored in the storage unit 114.

  Next, in step S <b> 106, the learning processing unit 115 defines segment positions based on the K trial data stored in the storage unit 114. The section position is the position of the dividing point at both ends of each section. The position of the dividing point corresponds to the position of the end effector 130, for example. The position of the division point becomes the speed target value switching position. The position of the dividing point may be a start point of target speed switching or a completion point of target speed switching. Further, the position of the dividing point may be a point at which the operation speed detected by the inner sensor 141 is guaranteed to be within a predetermined error range from the target speed.

  The learning processing unit 115 defines the division position based on, for example, an average or variance of K pieces of trial data. The learning processing unit 115 pays attention to the rate of change of the force waveform, and can automatically determine the position of the dividing point by setting the dividing point before and after the force waveform changes greatly. Alternatively, the user can manually determine a point where a state change occurs according to the work content as a division point.

  Next, in step S107, the learning processing unit 115 determines whether or not the division position is defined. If the segment position is not defined, the process returns to step S106. If the segment position is defined, the process proceeds to step S108. Next, in step S108, the learning processing unit 115 defines a speed target value for each section. The speed target value is calculated based on the upper limit value of the force specified by the user and the target tact time.

  Specifically, the learning processing unit 115 sets the standard work speed for completing the work by the target tact time as the overall speed target value Vdn. Next, the learning processing unit 115 defines a speed upper limit value Vmax based on the upper limit value of the force. The relationship between the speed when the end effector 130 collides with the work target 200 or the surrounding environment 300 and the external force applied to the end effector 130 at that time can be obtained in advance based on the rigidity information of the work target. The learning processing unit 115 can obtain the speed upper limit value Vmax with reference to a table or the like that stores this relationship.

  The learning processing unit 115 determines the target speed Vd using the entire speed target value Vdn and the speed upper limit value Vmax. The speed target value Vd is larger than 0 and smaller than the speed upper limit value Vmax. The target speed Vd is set so as to gradually approach Vdn. For example, the learning processing unit 115 defines a plurality of speed target values Vd using random numbers so that the speed parameter varies to some extent under the condition of 0 <Vd <Vdn <Vmax. As described above, in step S108, the learning processing unit 115 determines the speed target value Vd so as to be a value that varies within the determined range. Next, in step S109, the learning processing unit 115 determines whether or not a speed target value is defined. If the speed target value is not defined, the process returns to step S108. If the speed target value is defined, the preprocessing ends. The initial value for performing the learning process is determined by the preprocessing.

  FIG. 17 is a flowchart showing an example of the flow of learning processing performed in the learning processing unit 115 according to Embodiment 3 of the present invention. For the sake of explanation of the operation, FIG. 17 also shows an operation performed by blocks other than the learning processing unit 115. First, in step S <b> 201, the robot control device 111 tries the operation by operating the robot 120. Next, in step S202, the command value learning unit 113b acquires trial data obtained by the trial. Next, in step S203, the storage unit 114 stores the trial data acquired in step S202.

  Next, in step S204, the learning processing unit 115 determines whether or not M pieces of trial data have been acquired. Here, M is a natural number and is set in advance. If M or more pieces of trial data have not yet been acquired, the process returns to step S201. On the other hand, if M or more pieces of trial data have been acquired, the process proceeds to step S205. Accordingly, when the process proceeds to step S205, M pieces of trial data D2j (j = 1, 2, 3,... M) are acquired and stored in the storage unit 114. If M groups of segment positions and speed target values are defined, the trial in step S201 is executed using different groups of segment positions and speed target values. Therefore, when the process proceeds to step S205, M pieces of trial data D2j corresponding to the M sets of segment positions and speed target values are stored.

  Next, in step S205, the learning processing unit 115 calculates an evaluation value for each of the M trial data based on the constraint condition. The calculated evaluation value is stored. Next, in step S206, the learning processing unit 115 obtains a segment position and a speed target value corresponding to the trial data having the best evaluation value among the M trial data. Next, in step S207, the learning processing unit 115 compares the best evaluation value among the newly obtained M evaluation values with the evaluation value obtained in the past, and the evaluation value becomes the best. It is determined whether or not the result has converged. If it has converged, the process proceeds to step S209, a process for completing the adjustment is performed, and the adjustment of the operation command value is completed. When the adjustment of the operation command value is completed, the segment position and the speed target value at which the best evaluation value is obtained become the adjustment result of the operation command value. On the other hand, if not yet converged, the process proceeds to step S208.

  Next, in step S208, the learning processing unit 115 newly defines M sets of segment positions and speed target values, and updates the segment positions and speed target values. The segment positions or speed target values of the M sets of segment positions and speed target values are different from each other. That is, in step S208, the learning processing unit 115 newly sets M sets of operation command values. Each of the M sets of operation command values has a segment position and a speed target value corresponding to each segment position as parameters. In each set of operation command values, there is one more segment position than the number of segments, and there are the same number of speed target values as the number of segments. When the process of step S208 is completed, the process returns to step S201.

  As described above, in the learning process, the robot system 100 performs M trial operations based on the set segment positions and the speed target value set for each segment. The M trial operations are performed under different conditions of the segment position or the speed target value. Each time M trials are completed, the learning processing unit 115 updates the position of the dividing point for each section and the speed target value for each section.

  FIG. 18 is a diagram showing an example of a speed pattern at the time of trial in the robot system 100 according to the third embodiment of the present invention. Moreover, FIG. 19 is a figure which shows an example of the force information acquired at the time of the trial in the robot system 100 by Embodiment 3 of this invention. 18 and 19, P0 to P3 are positions of division points, and S1 to S4 represent four sections. In FIG. 18, V1 to V4 represent speed target values in each section. FIG. 19 shows force information acquired in the trial using the speed pattern shown in FIG.

  In the assembling work as shown in FIG. 14, the reaction force between the end effector 130 holding the first part 210 and the second part 310 is near the limit value near the position where the contact between the parts occurs. May grow. In this case, the amount of force exceeding the limit value can be evaluated as the limit excess amount. In FIG. 19, in section S2, the magnitude F of the force detected by the force sensor 143 exceeds the limit value L0. The limit excess amount DH is obtained by the difference between the detected force magnitude F and the limit value L0 when the detected force magnitude F exceeds the limit value L0. When there is a section where the limit excess amount DH is larger than the set threshold, it is necessary to adjust the speed target value of the section.

  In FIG. 19, the force F detected in the section S2 is large. Therefore, with respect to the speed pattern shown in FIG. 18, the learning processing unit 115 adjusts the speed pattern so that the speed target value V2 in the section S2 becomes smaller. Furthermore, the learning processing unit 115 also adjusts the positions of the dividing points P1 and P2 that are both ends of the section S2. In the speed pattern shown in FIG. 18, the division point P1 is a point at which the speed target value starts to be lowered, and the division point P2 is a point at which the speed target value starts to be raised. That is, the learning processing unit 115 also adjusts the position of the point where the change of the speed target value is started. These adjustments are made based on constraint conditions.

  For example, when the limit value L0 is set for the force magnitude F as a constraint condition, the evaluation value for the force magnitude F is evaluated to be 0 for trials that do not exceed the upper limit value L0. Define a function. When the evaluation value regarding the magnitude F of force does not become 0, the learning processing unit 115 continuously updates the positions of the speed target value V2 and the division points P1 and P2 and adjusts the operation command value. At the same time as this adjustment, an evaluation function can be defined so that work can be performed as fast as possible. In FIG. 19, in the sections S1, S3, and S4, the detected force magnitude F has a margin DL with respect to the limit value L0. Here, it is assumed that the margin amount DL is an index of the amount up to the limit value L0 or the amount up to the limit value L0. The amount up to the limit value L0 is defined by the difference between the limit value L0 and the detected force magnitude F. When the margin amount DL is larger than 0, the speed target value can be adjusted and updated in the direction of increasing. By such adjustment, it is possible to perform adjustment so as to perform work as fast as possible.

  These adjustments are performed in steps S205 and S206 in FIG. At this time, machine learning or an optimization method using an evaluation function can be applied to obtain the position and speed target value of the dividing point Pi that makes the evaluation value the best. For example, methods such as reinforcement learning, Bayesian optimization, and particle swarm optimization are exemplified. By these methods, an operation command value that makes the evaluation value the best can be set. For example, it is assumed that the evaluation function Fq represented by the equation (1) using the force F (t) detected at each time point during work and the work time T is defined. The learning processing unit 115 can obtain an operation command value that reduces the force F (t) and the work time T by adjusting the operation command value so that the evaluation value calculated by the evaluation function Fq is small. it can. As shown in FIG. 17, the adjustment is completed when the evaluation value obtained by the evaluation function has converged.

  The motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment are configured as described above. According to the motion adjustment device 112, the motion control system 110, and the robot system 100 of the present embodiment, the motion of the robot 120 is adjusted for each section. Therefore, the operation of the robot 120 can be adjusted so as not to unnecessarily slow down the operation of the entire work and an excessive load is not applied to the work target 200 or the surrounding environment 300. Adjustment can be facilitated. Furthermore, if it is configured so that the operation of the section in which the detection value of the force sensor 143 is smaller than a predetermined value is adjusted so that the operation becomes faster, the operation of the entire work can be made faster.

  As described above, according to the motion adjustment device 112, the motion control system 110, and the robot system 100 according to the present embodiment, it can be realized by conventional adjustment by learning and updating the optimum motion command value for each section. It was possible to design detailed operation command values that were not present, and as a result, high-speed and high-quality robot work could be realized. Specifically, according to the motion adjustment device 112, the motion control system 110, and the robot system 100 of the present embodiment, the reaction force between the mating components is suppressed in the component fitting operation, the connector insertion operation, and the like. Work time can be shortened.

Embodiment 4 FIG.
FIG. 20 is a block diagram illustrating a configuration example of the motion adjustment device 112c according to the fourth embodiment of the present invention and peripheral blocks. In the motion adjustment device, the motion control system, and the robot system of the present embodiment, other configurations are the same as those shown in FIG. FIG. 20 shows a part of the configuration of the robot system 100 extracted. The motion adjustment device 112c of the present embodiment includes a command value learning unit 113b and a command value sorting unit 116.

  The command value classification unit 116 receives an operation command value before update from the robot control device 111, sensor information that is a detection value of the sensor 140 from the sensor 140, and a constraint condition from the outside. The command value classifying unit 116 divides the operation command value with respect to these inputs using the position of the end effector 130 or the like or the command value progress rate (i = 0, 1, 2,...). N + 1) is defined and output as segment information. The command value learning unit 113b is the same as that shown in FIG.

  The motion adjustment device 112c according to the present embodiment determines the space to be divided by applying, for example, machine learning using the feature amount and the constraint condition of the sensor information, and class information on the feature amount space divided here. Is used to generate the current dividing point Pi. The motion adjustment device 112c performs preprocessing and learning processing similarly to the processing shown in FIG. FIG. 21 is a flowchart showing an example of the flow of preprocessing performed by the operation adjustment device 112c according to the fourth embodiment of the present invention. FIG. 22 is a flowchart showing an example of the flow of learning processing performed by the motion adjustment device 112c according to Embodiment 4 of the present invention.

  The pre-process shown in FIG. 21 differs from the process shown in FIG. 16 in that the number of sections is defined in addition to the section position in step S106b. For example, sections can be automatically generated based on waveform characteristics. For example, regarding the position data, velocity data, force data, and force change rate data acquired in time series as the waveform characteristics, the maximum value or frequency distribution of Tsmp data at regular intervals is input, and clustering is performed based on the input To do. For clustering, a break can be defined for each characteristic history of the waveform using a clustering technique such as the k-means method which is a kind of machine learning. For example, assume that X types of waveform features are defined.

  The original data can then be labeled based on the acquired cluster. For example, for each of the X clusters, the target input similarity S (i) (where i = 1, 2, 3,..., X) is defined, and the attribute characteristics of which group Can be expressed as a percentage. In that case, it can be labeled as the group with the highest percentage. The time t is defined as a label L (t) for each time defined as a variable. In step S106b, the number of divisions and the division position can be defined as breaks in all or some parts where the label changes.

  On the other hand, the learning process shown in FIG. 22 differs from the process shown in FIG. 17 in three processes of step S211, step S212, and step S213. In the learning process shown in FIG. 22, in step S211, the first evaluation value is obtained using an evaluation function for learning the number of divisions and the division position based on the sensor information, the operation command value, the control parameter, and the constraint condition. . In the learning process shown in FIG. 22, the number of divisions and the division position are learned and updated based on the first evaluation value in step S212. Furthermore, the learning process shown in FIG. 22 obtains a second evaluation value for learning the operation command value in step S213. Therefore, the learning process shown in FIG. 22 learns the operation command value after learning the number of divisions and the division position.

  By including the above processing, a framework for automatically learning the category information is added, and it becomes unnecessary to design the category information in advance using prior knowledge, which has the special effect of shortening the design time. Can be obtained.

Embodiment 5 FIG.
FIG. 23 is a block diagram illustrating a configuration example of the motion adjustment device 112d according to the fifth embodiment of the present invention and peripheral blocks. In the motion adjustment device, the motion control system, and the robot system of the present embodiment, other configurations are the same as those shown in FIG. FIG. 23 shows a part of the configuration of the robot system 100 extracted. The motion adjustment device 112d of the present embodiment includes a command value sorting unit 116 and a motion learning unit 117. The command value sorting unit 116 is the same as that shown in FIG.

  FIG. 24 is a block diagram showing a configuration example of the motion learning unit 117 according to the fifth embodiment of the present invention. The action learning unit 117 includes a command value learning unit 113b and a parameter learning unit 118. The command value learning unit 113b is the same as that shown in FIG. The motion learning unit 117 receives an operation command value and control parameters before update from the robot control device 111. In addition, a constraint condition is input to the motion learning unit 117 from the outside. In addition, sensor information is input from the sensor 140 to the motion learning unit 117. Further, the classification information is input from the command value classification unit 116 to the motion learning unit 117. The input signal is input to the command value learning unit 113b and the parameter learning unit 118.

  The parameter learning unit 118 adjusts the gain, impedance parameter, filter design parameter, and the like of the sensor feedback control system based on the external sensor, not the direct robot behavior such as the position command value, the speed command value, and the acceleration command value. That is, the parameter learning unit 118 adjusts the control parameter of the feedback control system. The parameter learning unit 118 receives the classification information, the sensor information, the constraint condition, the command value, and the control parameter as input, and uses them to update the input control parameter to a control parameter that satisfies the constraint condition. Machine learning can be used to update the control parameters. As an example, the parameter learning unit 118 updates the control parameter so that an evaluation value obtained by a predefined evaluation function becomes large, and repeats the calculation until asymptotically converges. Depending on the defined evaluation function, the parameter learning unit 118 updates the control parameter so that the evaluation value becomes small.

  Here, in FIG. 24, the parameter learning unit 118 is exemplified as a configuration independent of the command value learning unit 113b. However, the parameter learning unit 118 and the command value learning unit 113b do not necessarily need to perform independent processing. For example, the parameter learning unit 118 and the command value learning unit 113b can perform processing simultaneously using one evaluation function. The parameter learning unit 118 adjusts the control parameter for each category. Further, the number of sections used in the command value learning unit 113b and the number of sections used in the parameter learning unit 118 are not necessarily the same. For example, there may be a case where the number of sections used by the parameter learning unit 118 is larger than the number of sections used by the command value learning unit 113b.

  In addition, the parameter learning unit 118 can update not only the control parameters in the sensor feedback control system based on the external sensor 142 but also the control parameters in the feedback control system based on the internal sensor 141. High-speed robot work can be realized.

  FIG. 25 is a block diagram illustrating another configuration example and peripheral blocks of the operation adjustment device 112d according to the fifth embodiment of the invention. FIG. 25 illustrates a configuration example that does not include the command value learning unit 113b. In this configuration example, the motion adjustment device 112d updates only the control parameter without updating the motion command value.

Embodiment 6 FIG.
When adjusting the speed pattern, the motion adjusting device, the motion control system, and the robot system according to the present embodiment determine an upper limit value or a lower limit value for the speed target value in each section Si, and set a search space in each section. The definition is based on the rigidity of the work object and the restrictions on the assembly quality of the work object. According to the motion adjustment apparatus, motion control system, and robot system of the present embodiment, it is not possible to search for motion command values or control parameters that can be realized but cause problems in assembly quality in the search space. Therefore, it is possible to converge to the operation command value or control parameter for realizing high-speed assembly within a range in which the work quality required by the user is defined. As a result, the adjusted robot can obtain a special effect that the work force that does not damage the work target without increasing the reaction force can be ensured.

  DESCRIPTION OF SYMBOLS 100 Robot system, 110 Motion control system, 111 Robot control apparatus, 112, 112b, 112c, 112d Motion adjustment apparatus, 113, 113b Command value learning part, 114 Storage part, 115 Learning processing part, 116 Command value classification part, 117 operation | movement Learning unit, 120 robot, 130 end effector, 140 sensor, 141 internal sensor, 142 external sensor, 143 force sensor, 200 work target, 210 first part, 211 holes, 300 ambient environment, 310 second part, 311 protrusion, 401 processor, 402 memory, 403 data bus.

Claims (19)

A robot motion adjustment device used in a robot system including a robot equipped with an end effector and a robot control device for controlling the motion of the robot, wherein the robot performs work on a work target,
An operation command value transmitted from the robot controller to the robot in order to control the operation of the robot by performing learning using the force acting on the end effector detected by an external sensor included in the robot system as an input. An operation adjustment apparatus comprising a command value learning unit for adjusting   The motion adjustment device according to claim 1, wherein the command value learning unit adjusts the motion command value by performing learning using a range of force acting on the end effector as a constraint condition.   The motion adjustment device according to claim 2, wherein the command value learning unit adjusts the motion command value by performing learning using an upper limit of time required for the work as a constraint.   4. The motion according to claim 1, wherein the motion command value is a speed command value that is a target value of the movement speed of the end effector or a target value of the motion speed of the robot. Adjustment device.   The said command value learning part adjusts the said operation command value with respect to each of the some division | segmentation which divided | segmented from the start to the completion | finish of the said operation | work. The operation adjusting device according to 1. A command value division unit that generates a plurality of divisions by dividing from the start to the end of the work,
The operation adjustment device according to claim 5, wherein the command value learning unit adjusts the operation command value for each of the classifications generated by the command value classification unit.   The operation adjustment device according to claim 6, wherein the command value division unit adjusts a position of a division point for dividing the work into the divisions.   The said command value learning part adjusts the said operation command value by performing learning which used either the upper limit or lower limit of force, moment, torque, or electric current value as a constraint condition. The operation adjusting device according to any one of the above.   The command value learning unit adjusts the operation command value by performing learning using either the position or orientation of the robot or the upper or lower limit of the relative position and orientation relative to the surrounding environment as a constraint. Item 9. The operation adjustment device according to any one of Items 2 to 8.   The operation according to claim 5 or 6, wherein the command value learning unit performs an evaluation based on an evaluation function every M trials (M is a natural number) and adjusts the operation command value. Adjustment device. A parameter learning unit for learning a control parameter in at least one of feedback control based on an inner world sensor and feedback control based on the external sensor provided in the robot system;
The parameter learning unit performs learning based on classification information that is information on the classification and sensor information obtained from the external sensor in a plurality of trials, and updates the control parameter. The operation adjusting apparatus according to claim 5 or 6. A robot motion adjusting device used in a robot system including a robot with an end effector mounted thereon and a robot control device for controlling the motion of the robot, wherein the robot performs work on a work target,
Based on feedback control of the operation of the robot based on the internal sensor provided in the robot system and learning based on the external sensor by performing learning using as input the force acting on the end effector detected by the external sensor provided in the robot system. An operation adjustment apparatus comprising: a parameter learning unit that learns a control parameter in at least one of feedback control of the operation of the robot.   13. The motion adjustment device according to claim 12, wherein the parameter learning unit adjusts the control parameter by performing learning using a range of force acting on the end effector as a constraint condition.   The operation adjustment apparatus according to claim 13, wherein the parameter learning unit adjusts the control parameter by performing learning using an upper limit of time required for the work as a constraint.   The said parameter learning part adjusts the said control parameter with respect to each of the some division | segmentation divided | segmented from the start to the completion | finish of the said operation | work, The any one of Claim 12 to 14 characterized by the above-mentioned. Motion adjustment device. A command value division unit that generates a plurality of divisions by dividing from the start to the end of the work,
16. The motion adjustment device according to claim 15, wherein the parameter learning unit adjusts the control parameter for each of the classifications generated by the command value classification unit.   The operation adjustment device according to claim 16, wherein the command value division unit adjusts a position of a division point for dividing the work into the divisions. The motion adjustment device according to any one of claims 1 to 17,
A robot control device that controls the operation of the robot based on the motion command value or the control parameter adjusted by the motion adjustment device. The operation control system according to claim 18;
A robot system comprising the robot controlled by the motion control system.

Images