Control device, robot control device and control method

Abstract

The learning-based vibration reduction effect is improved during production on the user side. The control device for generating correction amounts for controlling robot movements includes: a learning control unit having parameters used in the learning control for generating correction amounts; a parameter storage unit storing parameters set before shipment; and a parameter adjustment unit that adjusts the parameters stored in the parameter storage unit and sets them in the learning control unit during robot production. The parameter adjustment unit adjusts the parameters, for example, based on the reciprocal of the robot's frequency response characteristics. Alternatively, the parameter adjustment unit may use a genetic algorithm for parameter adjustment.

Description

Technical Field

[0001] This invention relates to control devices, robot control devices, and control methods, and particularly to control devices, robot control devices, and control methods for generating correction quantities for controlling the movements of robots. Background Technology

[0002] Reducing vibrations during robot movements to achieve higher speeds or improve trajectory accuracy directly impacts production efficiency and quality. Therefore, it is desirable to minimize vibrations or trajectory deviations generated during robot actions.

[0003] In response to this hope, Patent Document 1 proposes the following method: Install sensors such as accelerometers at the parts where vibration needs to be eliminated or where high-precision trajectories need to be achieved, measure the vibration in the robot's movements through the sensors, and perform learning control to reduce vibration.

[0004] Specifically, Patent Document 1 describes a robot comprising: a robot mechanism having sensors at a location designated for position control; and a control device for controlling the movement of the robot mechanism. The control device includes: a normal control unit for controlling the movement of the robot mechanism; and a learning control unit that causes the robot mechanism to move via a work program and learns by calculating learning correction amounts to make the position of the robot mechanism detected by the sensors approach a target trajectory or position assigned by the normal control unit.

[0005] Existing technical documents

[0006] Patent Literature

[0007] Patent Document 1: Japanese Patent Application Publication No. 2011-167817 Summary of the Invention

[0008] The problem that the invention aims to solve

[0009] The learning control unit generated before the robot leaves the factory does not take into account the robot's operational state (during production) on the user side (e.g., the robot's tools and posture). In learning control units generated based on the frequency response characteristics of standard load states before the robot leaves the factory, optimal performance is sometimes not achieved during user-side production.

[0010] In order to improve the effect of learning-based vibration reduction in the production state on the user side, a control device, a robot control device, and a control method of the control device are required, which have a learning control unit that takes into account the frequency response characteristics during production.

[0011] Methods for solving problems

[0012] (1) The control device of the first aspect of this disclosure is a control device for generating correction amounts for controlling the movements of a robot, the control device having:

[0013] The learning control unit has parameters used in the learning control for generating the correction amount;

[0014] A parameter storage unit stores the parameters set before leaving the factory;

[0015] The parameter adjustment unit adjusts the parameters stored in the parameter storage unit and sets them in the learning control unit during the production process performed by the robot.

[0016] (2) The robot control device of the second aspect of the present disclosure includes: the control device described in the above technical solution (1); and a motion control unit that receives a correction amount from the control device for controlling the motion of the robot and controls the motion of the robot.

[0017] (3) The third control method of this disclosure is a control method for making a control device for controlling the correction amount of the robot's movements.

[0018] Read the parameters used in the learning control for generating the calibration amount from the parameter storage unit before shipment.

[0019] During production by the robot, the parameters stored in the parameter storage unit are adjusted according to the reciprocal of the robot's frequency response characteristics.

[0020] Effects of the Invention

[0021] According to the method disclosed herein, the effect of learning-based vibration reduction can be improved in the production state on the user side. Attached Figure Description

[0022] Figure 1 This is a structural diagram illustrating a structural example of a robot system according to an embodiment of the present invention.

[0023] Figure 2 yes Figure 1 The diagram shows the structure of the robot's mechanism.

[0024] Figure 3 This is a block diagram representing the structure of a robot control device.

[0025] Figure 4 This is a Bode plot representing an example of the frequency characteristics of input-output gain and phase delay.

[0026] Figure 5 This means that through learning, the actual trajectory y is made... i A diagram illustrating the case where r(t) approaches the ideal trajectory.

[0027] Figure 6 It is a characteristic graph representing the reciprocal of the two frequency characteristics of the input / output gain and phase delay before leaving the factory, and the frequency characteristic of the learning control unit.

[0028] Figure 7 It is a characteristic diagram representing the reciprocal of seven frequency characteristics of the user-side production, input / output gain, and phase delay, as well as the frequency characteristics of the learning control unit.

[0029] Figure 8 It is a block diagram showing the structure of the motion control unit.

[0030] Figure 9 It is a flowchart representing the actions of the robot control method of the robot control device in the user-side production environment. Detailed Implementation

[0031] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0032] Figure 1 This is a structural diagram illustrating a structural example of a robot system according to an embodiment of the present invention. Figure 2 yes Figure 1 The diagram shows the structure of the robot's mechanism. Figure 3 This is a block diagram representing the structure of a robot control device.

[0033] like Figure 1 As shown, the robot system 10 includes a spot welding robot 100 and a robot control device 200 for controlling the movements of the robot 100. The robot 100 and the robot control device 200 are interconnected via cables.

[0034] The robot 100 includes: a robot mechanism 101, a spot welding gun 102 mounted on the front end of the robot mechanism 101, and sensors 110 such as an acceleration sensor mounted on the spot welding gun 102. The spot welding gun 102 serves as the robot's position detection unit. When the sensor 110 is wired, it is connected to the robot control device 200 via a cable; when the sensor 110 is wirelessly connected, it communicates wirelessly with the robot control device 200.

[0035] In this embodiment, the case where sensor 110 is an acceleration sensor is described, but sensor 110 can also be other sensors, such as a gyroscope sensor, an inertial sensor, a force sensor, a laser tracker, a camera, or a motion capture device.

[0036] like Figure 2As shown, the robot mechanism 101 has six joint axes 1011 to 1016, each of which is equipped with a motor. The robot 100 defines a world coordinate system fixed in space and a mechanical interface coordinate system located at the flange position of the robot 100.

[0037] like Figure 3 As shown, the robot control device 200 includes a frequency generation unit 210, a frequency characteristic measurement unit 220, a control unit 230, and a motion control unit 240. The control unit 230 includes a parameter adjustment unit 231, a parameter storage unit 232, and a learning control unit 233. The control unit 230 corresponds to the control device. Furthermore, one or more of the frequency generation unit 210, the frequency characteristic measurement unit 220, and the control unit 230 may be provided within the motion control unit 240. One or both of the frequency generation unit 210 and the frequency characteristic measurement unit 220 may also be provided within the control unit 230.

[0038] The frequency generation unit 210 outputs a sine wave signal as an action command to the frequency characteristic measurement unit 220, the control unit 230, and the action control unit 240 while changing the frequency.

[0039] The frequency response measurement unit 220 uses the operation command (sine wave) generated by the frequency generation unit 210 as the input signal and the detection position (sine wave) output from the sensor 110 as the output signal to measure the amplitude ratio (input-output gain) and phase delay of the input signal and output signal at each frequency specified by the operation command. The frequency response measurement unit 220 outputs the measured frequency characteristics (frequency response characteristics) of the input-output gain and phase delay to the parameter adjustment unit 231 of the control unit 230.

[0040] Before leaving the factory, the frequency characteristic measurement unit 220 measures the frequency characteristics of the input-output gain and phase delay of the robot 100 when it moves in various postures (inertia) under maximum load on the test workpiece and under no load, and outputs the results to the parameter adjustment unit 231 of the control unit 230. Figure 4 This is a Bode plot representing an example of the frequency characteristics of input-output gain and phase delay.

[0041] In addition, the frequency characteristic measurement unit 220 measures the frequency characteristics of the input-output gain and phase delay when the robot 100 is active in the production environment on the user side, and outputs them to the parameter adjustment unit 231 of the control unit 230.

[0042] Here, "production time" refers to the learning process before actual production on the user side, where the environment during production is set to the same tools and postures as during actual production. However, when the parameter adjustment unit 231 adjusts the parameters of the learning control unit 233 during actual production, "production time" refers to the actual production time. Hereinafter, "production time" refers to either the learning process before actual production on the user side or the actual production time on the user side.

[0043] The parameters of the learning control unit 233 are the internal parameters of the transfer functions of the structural elements constituting the learning control unit 233. When the learning control unit 233 has a bandpass filter Wt, a low-pass filter Q, a learning controller L, and a high-pass filter Wp, the parameters of the learning control unit 233 are the internal parameters of the transfer functions of each of the bandpass filter Wt, the low-pass filter Q, the learning controller L, and the high-pass filter Wp. The internal parameters of the transfer functions of each of the bandpass filter Wt, the low-pass filter Q, the learning controller L, and the high-pass filter Wp will be described later.

[0044] Production is not limited to factory-centered manufacturing activities; it refers to all activities in which robots can use resources, services, or generate added value. For example, production includes activities performed by agricultural robots such as fruit-picking robots, forestry robots such as pruning robots, and commercial robots such as material handling robots.

[0045] Before leaving the factory, the parameter adjustment unit 231 calculates the reciprocal of the frequency response characteristics (frequency response characteristics) of the input / output gain and phase delay output from the frequency response measurement unit 220, adjusts the parameters of the learning control unit 233 based on the reciprocal, determines the optimal parameters, stores the parameters in the parameter storage unit 232, and sets them in the learning control unit 233. In this way, a learning control unit is generated before leaving the factory.

[0046] During user-side production, the parameter adjustment unit 231 reads parameters from the parameter storage unit 232 and sets them as initial parameters. It calculates the reciprocal of the frequency characteristic output from the frequency characteristic measurement unit 220, adjusts the initial parameters of the learning control unit 233 based on this reciprocal, determines the optimal parameters, and sets these parameters in the learning control unit 233. This generates the learning control unit for user-side production.

[0047] The learning control unit 233 calculates a correction amount to make the detected position output from the sensor 110 close to the target position of the motion command. In addition, the learning control unit 233 performs an inverse transformation on the calculated correction amount to obtain the correction amount corresponding to the joint axes 1011 to 1016, and outputs the calculated correction amount for each joint axis to the motion control unit 240.

[0048] The motion control unit 240 uses the correction values ​​output from the learning control unit 233 to control each joint axis of the robot 100. By correcting the vibration of each joint axis, the vibration in the world coordinate system is corrected.

[0049] The parameter adjustment unit 231, the learning control unit 233, and the motion control unit 240 will be described in more detail below.

[0050] Learning Control Department

[0051] The learning control unit 233 uses iterative learning control to update the correction amount in order to make the detected position output from the sensor 110 approach the target position of the action command. Iterative learning control is described, for example, in “Survey Of Iterative Learning Control: A Learning-Based Method for High-Performance Tracking Control”, Bristow, DA, Tharayil, M., & Alleyne, AG (2006), IEEE Control Systems, 26(3), 96-114. and “Iterative Learning Control Analysis, Design, and Experiments”, Mikael Norrlof, Department of Electrical Engineering Linkopings University, SE-581 83 Linkoping, Sweden Linkoping 2000.

[0052] The output (correction amount) from the learning control unit 233 to the motion control unit 240 is updated according to mathematical formula 1 (hereinafter formula 1). In mathematical formula 1, u i+1 (t) represents the output (correction amount) to the motion control unit 240 during the (i+1)th learning iteration, u i (t) represents the output (correction amount) to the motion control unit 240 during the i-th learning iteration, e i (t) represents the trajectory error during the i-th learning iteration, Q represents the transfer function of the low-pass filter included in the learning control unit 233, and L represents the transfer function of the learning controller (digital filter) included in the learning control unit 233.

[0053] Using y iLet r(t) represent the robot's actual trajectory during the i-th learning iteration, and let r(t) represent the ideal trajectory (command trajectory). Then, let equation 2 (Equation 2 below) represent the trajectory error e. i (t). The robot's actual trajectory y i (t) is equivalent to the detection position output from sensor 110, and the ideal trajectory r(t) is equivalent to the action command (position command) output from frequency generation unit 2100.

[0054] [Formula 1]

[0055] u i+1 (t)=Q(u i (t)+Le i (t))

[0056] [Formula 2]

[0057] e i (t)=r(t)-y i (t)

[0058] Figure 5 This indicates that through learning, the actual trajectory y i A diagram illustrating the case where r(t) approaches the ideal trajectory.

[0059] The following two properties of the Learning Control Department 233 are important.

[0060] (1) Stability

[0061] Output (correction amount) to motion control unit 240 u i (t) does not diverge, as shown in mathematical formula 3 (hereinafter formula 3), and converges to a bounded value u. ∞ (t) is important.

[0062] [Formula 3]

[0063]

[0064] (2) Monotonic decreasing property

[0065] With trajectory error e i (t) Temporarily increasing such a shift and achieving convergence is generally unacceptable in practice. Therefore, as shown in Equation 4 (hereinafter Equation 4), the trajectory error e i (t) converges uniformly to some bounded value e. ∞ The monotonic reduction is important.

[0066] [Formula 4]

[0067] γ<1, ||e i+1 -e ∞ ||≤γ||ei -e ∞ ||

[0068] The correction amount u in the i-th learning iteration i (t) and trajectory error e i (t) and the correction amount u at the (i+1)th learning iteration. i+1 (t) and trajectory error e i+1 The relationship between (t) is expressed by mathematical formula 5 (hereinafter Formula 5) and mathematical formula 6 (hereinafter Formula 6). In mathematical formulas 5 and 6, Q is the transfer function of the low-pass filter, L is the transfer function of the learning controller, and P is the transfer function from the correction input of the motion control unit 240 to the output to the robot 100.

[0069] [Formula 5]

[0070] ||u i+1 (t)-u ∞ (t)||2≤||Q(1-LP)|| ∞ ||u i (t)-u ∞ (t)||2

[0071]

[0072] [Formula 6]

[0073] ||e i+1 (t)-e ∞ (t)||2≤||Q(1-LP)|| ∞ ||e i (t)-e ∞ (t)||2

[0074]

[0075] In mathematical formulas 5 and 6 above, if mathematical formula 7 (hereinafter formula 7) is the same as or smaller than 1, then the correction amount and the trajectory error converge to u. ∞ (t), e ∞ (t), stability and monotonic reduction are maintained.

[0076] [Formula 7]

[0077] ||Q(1-LP)|| ∞

[0078] The above are the conditions for maintaining the stability and monotonicity of the learning control unit 233.

[0079] When the learning control unit 233 has a bandpass filter Wt, a low-pass filter Q, a learning controller L, and a high-pass filter Wp, the parameter adjustment unit 231 designs the internal parameters of the transfer function of the learning control unit 233 (described later) to minimize mathematical formula 8 (hereinafter formula 8).

[0080] [Formula 8]

[0081] ||WtQ(1-LWpP)|| ∞

[0082] The transfer function of the bandpass filter Wt has the passband (wt) and the gain in the passband (dcwt) as internal parameters. The transfer function of the low-pass filter Q has the filter order (Nq) and cutoff frequency (wn) as internal parameters. The transfer function of the learning controller L has the value of the oscillation before which sampling is used (N_ILC) and its order (No) as internal parameters. The transfer function of the high-pass filter Wp has the cutoff frequency (wp), the filter gain (dcwp), and the filter order (wpNo) as internal parameters.

[0083] <Parameter Adjustment Section>

[0084] Before leaving the factory, the parameter adjustment unit 231 obtains multiple frequency characteristics of the robot 100's input-output gain and phase delay when the robot is under maximum load on a test workpiece and under no-load conditions, moving in various postures (inertia), from the frequency characteristic measurement unit 220, and calculates the reciprocals of these multiple frequency characteristics. Furthermore, the parameter adjustment unit 231 changes at least one of the nine internal parameters of the transfer functions of the bandpass filter Wt, low-pass filter Q, learning controller L, and high-pass filter Wp, for example, two internal parameters of the transfer function of the learning controller L, so that the frequency characteristics of the learning control unit 233 fall between the reciprocals of the multiple frequency characteristics.

[0085] The frequency characteristics of the learning control unit 233 are preferably close to the reciprocals of the measured frequency characteristics of the input / output gain and phase delay. The reason for this is that since the frequency characteristics of the learning control unit 233 are consistent with the reciprocals of the measured frequency response, the correction amount made by the learning control unit 233 operates in the opposite direction to the vibration setting, thereby being able to counteract the vibration of the robot 100.

[0086] Figure 6 This is a characteristic graph representing the reciprocals of the two frequency characteristics of the input / output gain and phase delay before leaving the factory, and the frequency characteristic of the learning control unit. Figure 6In the diagram, the dashed lines represent the reciprocals of multiple frequency characteristics of input / output gain and phase delay, while the solid lines represent the frequency characteristics of the learning control unit 233. Both the frequency characteristics shown by the dashed lines (the reciprocals of multiple frequency characteristics of input / output gain and phase delay) and the frequency characteristics shown by the solid lines represent the frequency characteristics before learning.

[0087] exist Figure 6 In the diagram, the two dashed lines represent the reciprocals of the frequency response of the robot 100 when it moves in two postures (inertia). By adjusting the parameters mentioned above relative to the reciprocals of the two frequency characteristics of the input-output gain and phase delay of these two postures (inertia), the frequency characteristics of the learning control unit 233 shown by the solid line are changed, thereby suppressing vibrations in the two postures.

[0088] However, when the robot 100 moves in other postures (inertia) in the production environment on the user side, the pre-shipment parameters may not be able to suppress vibration.

[0089] Therefore, the parameter adjustment unit 231 adjusts the internal parameters to be the reciprocal of the frequency characteristics of the input / output gain and phase delay measured under the conditions of production on the user side. At this time, the parameter adjustment unit 231 reads the parameters from the parameter storage unit 232 and sets them as initial parameters. Based on the reciprocal of the frequency characteristics output from the frequency characteristic measurement unit 220, the initial parameters of the learning control unit 233 are adjusted to determine the optimal internal parameters, and the optimal internal parameters are set in the learning control unit 233.

[0090] Figure 7 This is a characteristic graph representing the reciprocals of seven frequency characteristics (input / output gain and phase delay) during user-side production, as well as the frequency characteristics of the learning control unit. Figure 7 In the diagram, thick single-dash lines, thin single-dash lines, thick double-dash lines, thin double-dash lines, thick dashed lines, thin dashed lines, and wide-spaced dashed lines represent the reciprocals of the seven frequency characteristics of input / output gain and phase delay, while solid lines represent the frequency characteristics of the learning control unit 233. Figure 7 The reciprocals of the multiple frequency characteristics shown, including input / output gain and phase delay, and the frequency characteristics shown by the solid line, all represent the frequency characteristics before learning.

[0091] The frequency characteristics of the learning control unit 233, shown by the solid line, are changed by adjusting the reciprocal of the seven frequency characteristics relative to the seven postures (inertia), thereby suppressing vibrations in the seven postures.

[0092] In addition to attitude, inertia also varies depending on the load (such as a servo welding gun mounted on the front of the robot).

[0093] The method by which the parameter adjustment unit 231 searches for the optimal internal parameters is not particularly limited; for example, the genetic algorithm described below can be applied.

[0094] (1) Prepare two sets in advance, each containing N individuals (N being a natural number greater than 2). Hereinafter, these two sets will be referred to as the "current generation" and the "next generation". Each individual has information about the nine internal parameters that have already been described.

[0095] (2) Randomly generate N individuals in the current generation. Each individual is randomly generated within the range of possible internal parameters.

[0096] (3) Calculate the fitness of each individual in the current generation using the evaluation function. The smaller the value of the mathematical formula 8 in the evaluation function, the higher the fitness. Fitness can be used to determine whether each individual (a combination of parameters) is a good learning controller.

[0097] (4) Perform one of the next three actions with a certain probability and save the result to the next generation.

[0098] A. Select two individuals for crossover.

[0099] Inherit a parameter from one of the current generation's individuals into the next generation's individual.

[0100] B. Select an individual to undergo mutation.

[0101] Randomly change some or all of the parameters of the selected individuals.

[0102] C. Select an individual and copy it directly.

[0103] (5) Repeat the above steps (4) until the number of individuals in the next generation is N.

[0104] (6) When the number of individuals in the next generation is N, transfer all the content of the next generation to the current generation.

[0105] (7) Repeat the actions after (3) until the maximum number of generations is G (G is a natural number greater than 2), and finally output the individual with the highest fitness in the "current generation" as the "solution".

[0106] <Motion Control Department>

[0107] Figure 8 This is a block diagram showing the structure of the motion control unit 240. The motion control unit 240 is provided corresponding to the six joint axes 1011 to 1016, but in the following description, the motion control unit 240 controls the motor 1020 of the joint axis 1011 of the robot 100.

[0108] like Figure 8As shown, the motion control unit 240 includes: a subtractor 2401, an adder 2402, a position control unit 2403, a subtractor 2404, a speed control unit 2405, a subtractor 2406, a current control unit 2407, an amplifier 2408, a differentiator 2409, and a correction unit 2410.

[0109] Subtractor 2401 calculates the difference between the command position of the motion command and the position feedback value output from the position detector such as the rotary encoder of the motor 1020 of the robot's joint axis, and outputs the difference as the position deviation to adder 2402.

[0110] The adder 2402 adds the position deviation output from the subtractor 2401 and the correction amount output from the correction unit 2410, and outputs the corrected position deviation to the position control unit 2403.

[0111] The position control unit 2403 generates a speed command value based on the corrected position deviation and outputs it to the subtractor 2404.

[0112] Subtractor 2404 calculates the difference between the speed command value output from position control unit 2403 and the speed feedback value from differentiator 2409, and outputs this difference as speed deviation to speed control unit 2405.

[0113] The speed control unit 2405 generates a current command value based on the speed deviation and outputs it to the subtractor 2406.

[0114] The subtractor 2406 calculates the difference between the current command value output from the speed control unit 2405 and the current feedback value from the amplifier 2408, and outputs this difference as a current deviation to the current control unit 2407.

[0115] The current control unit 2407 generates a torque command value (current value) based on the current deviation and outputs it to the amplifier 2408.

[0116] Amplifier 2408 calculates the desired power based on the current value from current control unit 2407 and inputs it into motor 1020 of joint axis 1011 of robot 100.

[0117] Differentiator 2409 differentiates the position feedback value and outputs it to subtractor 2404.

[0118] The correction unit 2410 stores the correction amount output from the learning control unit 233 and outputs the correction amount to the adder 2402.

[0119] In order to achieve Figure 1The robot control device 200 shown is a functional block. The robot control device 200 can be constructed from a computer equipped with a CPU (Central Processing Unit) or other arithmetic processing units. In addition, the robot control device 200 also includes auxiliary storage devices such as HDDs (Hard Disk Drives) for storing application software or various control programs such as operating systems (OS), and main storage devices such as RAMs (Random Access Memory) for storing data temporarily needed based on the execution of programs by the arithmetic processing unit.

[0120] Furthermore, in the robot control device 200, the arithmetic processing unit reads application software or operating system from the auxiliary storage device, expands the read application software or operating system in the main storage device, and performs calculations based on the application software or operating system. Additionally, the arithmetic processing unit controls various hardware components of the robot control device 200 based on the calculation results. Thus, the functional blocks of this embodiment are implemented. In other words, this embodiment can be implemented through hardware and software cooperation.

[0121] Regarding the learning control unit 233, when the computational load accompanying learning is high, for example, by equipping a personal computer with GPUs (Graphics Processing Units), a technology called GPGPU (General-Purpose computing on Graphics Processing Units) can be used to perform high-speed processing of computational tasks accompanying machine learning. Furthermore, for even faster processing, multiple computers equipped with such GPUs can be used to construct a computer cluster, and the multiple computers within this cluster can perform parallel processing.

[0122] Next, the operation of the robot control device 200 in the production environment on the user side will be explained.

[0123] Figure 9 This is a flowchart illustrating the actions of the robot control method of the robot control device 200 in the production environment on the user side.

[0124] exist Figure 1 The parameter storage unit 232 shown stores the optimal parameters of the learning control unit 233 before it leaves the factory. The parameter adjustment unit 231 adjusts the parameters of the learning control unit 233 based on the reciprocal of the frequency characteristics (frequency response characteristics) output from the frequency characteristic measurement unit 220 before it leaves the factory, thereby determining the optimal parameters.

[0125] In step S10, in the production environment on the user side, the frequency generation unit 210 outputs a sine wave signal as an action command to the motion control unit 240, and the motion control unit 240 causes the robot 100 to perform an action.

[0126] In step S11, the frequency characteristic measurement unit 220 uses the action command (sine wave) generated by the frequency generation unit 210 and the detection position (sine wave) output from the sensor 110 to measure the frequency characteristics of the amplitude ratio (input-output gain) and phase delay of the input signal and the output signal.

[0127] In step S12, the parameter adjustment unit 231 reads the parameters of the control unit 230 before leaving the factory from the parameter storage unit 232 as the initial parameters.

[0128] In step S13, the parameter adjustment unit 231 calculates the reciprocal of the frequency characteristic output from the frequency characteristic measurement unit 220, adjusts the initial parameter of the learning control unit 233 based on the reciprocal, determines the optimal parameter, and sets the parameter in the learning control unit 233.

[0129] In step S14, the learning control unit 233 calculates a correction amount in order to make the detection position output from the sensor 110 close to the target position of the motion command, and the motion control unit 240 uses the correction amount to control each joint axis of the robot 100.

[0130] The control unit (called the control device) of the above-described embodiment can improve the effect of learning-based vibration reduction during production on the user side.

[0131] Furthermore, since the control unit uses initial parameters to adjust and learn the control unit's parameters during user-side production, significant parameter changes are unnecessary, allowing for fewer trial runs. As a result, user-side setup time is reduced.

[0132] Furthermore, if the control unit does not use the initial parameters before delivery during production on the user side, but adjusts the parameters of the learning control unit based on the frequency characteristics output from the frequency characteristic measurement unit, it may become a learning control unit that causes vibration divergence when the frequency characteristics change. However, by using the initial parameters before delivery, the divergence can be suppressed.

[0133] The embodiments of the present invention have been described above, but the various structural components of the control device, robot control device, and control method of this embodiment can be implemented by hardware, software, or a combination thereof. For example, they can also be implemented by electronic circuits. Furthermore, the image processing method performed through the cooperation of the aforementioned structural components can also be implemented by hardware, software, or a combination thereof. Here, implementation by software means implementation by loading and executing a program into a computer.

[0134] Programs can be stored and provided to a computer using various types of non-transitory computer-readable recording media. Non-transitory computer-readable recording media include various types of tangible storage media. Examples of non-transitory computer-readable recording media include magnetic recording media (e.g., hard disk drives), optical-magnetic recording media (e.g., optical discs), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash memory ROMs, and RAMs (random access memory)).

[0135] The above-described embodiments are preferred embodiments of the present invention, but the scope of the present invention is not limited to the above-described embodiments. Various modifications can be made without departing from the spirit of the present invention.

[0136] For example, in the above embodiments, the robot system 10 is configured such that the robot 100 and the robot control device 200 are separate, but it can also be configured such that the robot control device 200 is included in the robot 100.

[0137] Furthermore, in the above embodiments, a robot for spot welding was described as the robot, but other robots may also be used, such as a painting robot, an assembly robot, or a handling robot.

[0138] The control device, robot control device, and control method disclosed herein include the embodiments described above, and can be implemented in various ways having the following structures.

[0139] (1) A first aspect of this disclosure is a control device (e.g., control unit 230) that generates correction amounts for controlling the movements of a robot (e.g., robot 100), wherein,

[0140] The control device has:

[0141] A learning control unit (e.g., learning control unit 233) has parameters used in the learning control for generating the correction amount;

[0142] A parameter storage unit (e.g., parameter storage unit 232) stores the parameters set before leaving the factory;

[0143] The parameter adjustment unit (e.g., parameter adjustment unit 231) adjusts the parameters stored in the parameter storage unit and sets them in the learning control unit during the production process performed by the robot.

[0144] According to this control device, the effect of learning-based vibration reduction can be improved during production on the user side.

[0145] (2) According to the control device described in (1) above, the parameter adjustment unit adjusts the parameters according to the reciprocal of the frequency response characteristics of the robot.

[0146] (3) The control device according to (1) or (2) above, wherein the parameter adjustment unit adjusts the parameters by a genetic algorithm.

[0147] (4) The second aspect of this disclosure is a robot control device, having:

[0148] The control device described in any one of (1) to (3) above;

[0149] The motion control unit receives correction values ​​from the control device to control the robot's motion and controls the robot's motion.

[0150] According to this robot control device, the effect of learning-based vibration reduction can be improved during production on the user side.

[0151] (5) The robot control device according to (4) above, wherein the robot control device comprises:

[0152] The frequency generation unit generates signals that change frequency.

[0153] The frequency response measurement unit measures the frequency response characteristics of the robot based on the signal and the output signal from the sensor installed on the position detection part of the robot.

[0154] (6) The robot control device according to (5) above, wherein the sensor is any one of an accelerometer, a gyroscope, an inertial sensor, a force sensor, a laser tracker, a camera, and a motion capture device.

[0155] (7) A third aspect of this disclosure is a control method for manufacturing a control device (e.g., control unit 230) for generating a correction amount for controlling the movements of a robot (e.g., robot 100), wherein,

[0156] Read from the parameter storage unit (e.g., parameter storage unit 232) the parameters used in the learning control for creating the calibration amount before leaving the factory.

[0157] During production by the robot, the parameters stored in the parameter storage unit are adjusted according to the reciprocal of the robot's frequency response characteristics.

[0158] According to this control method, the effect of learning-based vibration reduction can be improved under the production conditions on the user side.

[0159] Explanation of reference numerals in the attached figures

[0160] 10 Robotic Systems

[0161] 100 robots

[0162] 101 Robotics Department

[0163] 102 Spot Welding Gun

[0164] 110 sensor

[0165] 200 Robot Control Device

[0166] 210 Frequency Generation Unit

[0167] 220 Frequency Response Measurement Unit

[0168] 230 Control Department

[0169] 231 Parameter Adjustment Section

[0170] 232 Parameter Storage Unit

[0171] 233 Learning Control Department

[0172] 240 Motion Control Unit.

Claims

1. A control device comprising a correction amount for controlling the motion of a robot, characterized in that, The control device has: The learning control unit has parameters used in the learning control for generating the correction amount; The parameter storage unit stores the parameters set before leaving the factory; The parameter adjustment unit adjusts the parameters stored in the parameter storage unit and sets them in the learning control unit during the production process performed by the robot. The parameter adjustment unit adjusts the parameters based on the reciprocal of the measured frequency response characteristics of the robot.

2. The control device according to claim 1, characterized in that, The parameter adjustment unit adjusts the parameters using a genetic algorithm.

3. A robot control device, characterized in that, have: The control device according to claim 1 or 2; The motion control unit receives correction values ​​from the control device to control the robot's motion and controls the robot's motion.

4. The robot control device according to claim 3, characterized in that, The robot control device has the following features: The frequency generation unit generates signals that change frequency. The frequency response measurement unit measures the frequency response characteristics of the robot based on the signal and the output signal from the sensor installed in the position detection part of the robot.

5. The robot control device according to claim 4, characterized in that, The sensor is any one of an accelerometer, gyroscope, inertial sensor, force sensor, laser tracker, camera, or motion capture device.

6. A control method for manufacturing a control device for generating correction quantities for controlling the movements of a robot, characterized in that, Read the parameters used in the learning control for generating the calibration amount from the parameter storage unit before shipment. During production by the robot, the parameters stored in the parameter storage unit are adjusted based on the reciprocal of the robot's frequency response characteristics as measured.

Images