Method and system for parameterizing a regulator for a robotic process

Abstract

A method of parameterizing a regulator (11) for a robot process, comprising the steps of providing a first time history of a first state variable of the robot process for a first value of a first regulator parameter of the regulator of the robot process, providing at least one further time history of the first state variable for a further value of the first regulator parameter, determining a quality value for the provided time histories by means of an at least partially machine learning based data processing (20), and optimizing (S50) the value of the first regulator parameter based on the values of the first regulator parameter and the quality values determined for the time histories of the values of the first regulator parameter by means of an optimizer (30). The invention also relates to a system or computer program (product).

Description

Technical Field

[0001] The present invention particularly relates to a method and system for parameterizing a regulator for a robotic process, and a computer program or computer program product for performing the method. Background Technology

[0002] To date, regulators used in robotic processes have typically been manually parameterized in practice, which is costly, difficult to reproduce, and requires experienced professionals. Summary of the Invention

[0003] The purpose of this invention is to improve the parameterization of regulators used in robotic processes and / or improve the execution of robotic processes.

[0004] The object of this invention is achieved by a method for parameterizing a regulator used in a robotic process. This invention also relates to a system, computer program, or computer program product for performing the methods described herein.

[0005] According to one embodiment of the present invention, a method for parameterizing or pre-given parameters (“regulator parameters”) of a regulator for a robotic process includes the following steps: - Provide the first time history of the first state variable of the robot process for the first value of the first regulator parameter of the regulator used for the robot process; - (respectively) for another value of the first regulator parameter, provide another time history of the first state variable, preferably providing multiple additional time histories; - For one or more provided time periods, a quality value is determined respectively using preferred, at least partially machine learning-based data processing; and - The values ​​of the first regulator parameters are optimized by means of an optimizer based on the values ​​of the first regulator parameters and the quality values ​​determined by the time history of these values ​​of the first regulator parameters. Preferably, the optimizer provides the values ​​of the first regulator parameters and their respective or corresponding quality values, which have been determined by the time history of each value of the first regulator parameters.

[0006] This is based on the idea of ​​providing the optimizer with quality standard values ​​for different values ​​of regulator parameters, which are determined or have been determined by means of artificial intelligence; in one implementation, the optimizer can be a known and / or numerical optimizer, and in another implementation it can be a gradient-based optimizer, an evolutionary optimizer, an optimizer at least in part based on machine learning, etc.

[0007] Thus, in particular, by means of at least partly machine learning-based data processing to independently determine quality criterion values, where the data processing will be trained or has been trained for this purpose, the parameterization of the regulator can be improved, and in particular, the parameterization of the regulator can be performed (more) quickly, (more) cost-effectively, (more) reliably, consistently and / or without the involvement of experienced professionals, and / or better regulator parameters can be determined.

[0008] In one implementation, the time history previously provided for optimizing the value of the first regulator parameter is the time history of the first state variables of the robot process, which (respectively) correspond to the same value of the second regulator parameter of the regulator and different values ​​of the first regulator parameter, and the method includes the following additional steps: - Provide a second time history of the robot process, or particularly preferably a second state variable, for the optimized value of the first regulator parameter and the first value of the second regulator parameter; - (respectively) provide another time history, preferably multiple additional time histories, for the first or second state variable for another value of the second regulator parameter and the optimized value of the first regulator parameter; - For the one or more provided time histories, quality values ​​are determined respectively by means of preferably the same, at least partially machine learning-based, data processing, particularly preferably the same data processing that has been used to optimize the value of the first regulator parameter or for which quality values ​​have been determined for different values ​​of the first regulator parameter; and - Using an optimizer, particularly preferably the same optimizer already used to optimize the values ​​of the first regulator parameters, the second regulator parameters are optimized based on the values ​​of the second regulator parameters and quality values ​​that have been determined for the time histories of these values ​​of the second regulator parameters, wherein preferably the optimizer provides the values ​​of the second regulator parameters and their respective or corresponding quality values ​​that have been determined for the time histories of each value of the second regulator parameters.

[0009] This is based on the idea that, in a two-stage system, assuming the second regulator parameter is constant, the first regulator parameter is optimized first, and then the second regulator parameter is optimized while the first regulator parameter remains constant and has already been optimized. In one extended embodiment, the first and second regulator parameters are regulator parameters of a cascaded regulator; in another implementation, the first regulator parameter is the regulator parameter of the inner regulation loop (deeper than the second regulator parameter), and the second regulator parameter is the regulator parameter of the outer regulation loop (shallower than the first regulator parameter). In other words, in one embodiment, the cascaded regulators are parameterized or optimized hierarchically and preferably from the inside out.

[0010] This can improve the parameterization of the controller, and in particular, it can improve the parameterization of the controller (more) quickly, (more) costly, (more) reliably and / or without the need for experienced professionals, and / or determine better controller parameters.

[0011] In one implementation, the method includes setting a regulator based on or utilizing or according to the value of a first regulator parameter determined or optimized by an optimizer, and in an extended embodiment, setting a regulator based on or utilizing or according to the value of a second regulator parameter determined or optimized by an optimizer.

[0012] In one implementation, the first state variable includes a first state parameter of the robot process and one or more other state parameters, wherein the time history of the first state variable includes the time history of the first state parameter and the time history of the other state parameters.

[0013] By using state variables with two or more state parameters, the parameterization of the regulator can be improved, and in particular, the parameterization of the regulator can be performed more quickly and / or more reliably, and / or better regulator parameters can be determined.

[0014] In one extended scheme, the first state parameter is a different physical state parameter from the other one or more other state parameters, and in one implementation, they have different physical dimensions.

[0015] Alternatively or particularly preferably additionally, in an extended embodiment, the first state parameter includes posture parameters, particularly Cartesian position and / or orientation and / or drive and / or joint posture, and / or the first and / or higher-order time derivatives of Cartesian position and / or orientation and / or drive and / or joint posture. In one embodiment of this extended embodiment, the one or more additional state parameters (respectively) include load parameters, particularly force and / or torque, and / or the first and / or higher-order time derivatives of force and / or torque.

[0016] In one extended embodiment, the first state parameter includes load parameters, particularly force and / or torque, and / or the first and / or higher-order time derivatives of the force and / or torque. In one implementation of this extended embodiment, the one or more additional state parameters (respectively) include position parameters, particularly Cartesian position and / or orientation and / or drive and / or joint attitude, and / or the first and / or higher-order time derivatives of the Cartesian position and / or orientation and / or joint attitude.

[0017] Alternatively or particularly preferably additionally, in one embodiment, the first and / or at least one other state parameter (respectively) includes the difference between the actual value and the target value, particularly preferably the difference between the actual value and the target value of the previously mentioned attitude parameter or load parameter.

[0018] In one implementation, the time history of the first state variable includes the time history of the first state parameter and the time history of at least one other state parameter that follows it. Preferably, the time histories of the first state parameter and the at least one other state parameter are sequential.

[0019] By using one of these features, or particularly preferably two or more features, the parameterization of the regulator can be improved, in particular the parameterization of the regulator can be performed (more) quickly and / or (more) reliably, and / or better regulator parameters can be determined.

[0020] In one implementation, providing a first time history and at least one additional time history for each value of a first regulator parameter of a first state variable includes: actually or simulatedly executing a robot process; changing the value of the first regulator parameter during the execution of the robot process; and dividing the total time history determined during the execution of the robot process into the first time history and at least one additional time history. In an extended implementation, providing a second time history and at least one additional time history for each value of a second regulator parameter and optimized values ​​of the first regulator parameter includes: actually or simulatedly executing a robot process; changing the value of the first regulator parameter during the execution of the robot process; and dividing the total time history determined during the execution of the robot process into the first time history and at least one additional time history.

[0021] This can improve the parameterization of the controller, especially by performing the parameterization of the controller (more) quickly, (more) costly, (more) reliably and / or without the need for experienced professionals, and / or by determining better controller parameters.

[0022] In one implementation, providing a first time history and at least one additional time history for each value of the first regulator parameter of the first state variable includes: scaling the first time history and at least one additional time history differently. In an extended implementation, providing a second time history and at least one additional time history for each value of the second regulator parameter and each optimized value of the first regulator parameter includes: scaling the second time history and at least one additional time history differently.

[0023] By using one of these features, or particularly preferably two or more features, the parameterization of the regulator can be improved, in particular the parameterization of the regulator can be performed (more) quickly and / or (more) reliably, and / or better regulator parameters can be determined.

[0024] In one implementation, the data processing, at least partially based on machine learning, includes: artificial neural networks and / or decision tree algorithms and / or random decision algorithms and / or training based on manually labeled time histories of first state variables. This data processing determines, with the aid of the data processing, a quality value for optimizing the values ​​of the first regulator parameters for each value of the first state variable, based on the time histories of the first state variables. In an extended embodiment, the data processing determines, with the aid of the data processing, a quality value for optimizing the values ​​of the second regulator parameters for each value of the second regulator parameters, based on the time histories of the state variables. In other words, the method for parameterizing a regulator for a robotic process according to the invention can use data processing that has been accordingly trained, at least partially based on machine learning, and in an extended embodiment includes training based on manually labeled time histories of (at least) the first state variable itself. Manual marking may specifically include the evaluation of different time histories by a human operator, preferably based on experience, and in one implementation, assigning different quality values ​​to the time histories based on or taking into account the actual or simulated execution of the robotic process, for example, a time histories in which (larger) vibrations and / or adjustment errors and / or process deviations and / or disturbances are evaluated as (poor), and so on.

[0025] By using one of these features, or particularly preferably two or more features, the parameterization of the regulator can be improved, in particular the parameterization of the regulator can be performed (more) quickly and / or (more) reliably, and / or better regulator parameters can be determined.

[0026] In one embodiment, the robotic process is one having at least one robotic arm and / or at least one robot-guided, preferably robot-arm-guided, tool, particularly a robot-guided welding clamp; and / or a regulator for controlling at least one robot, particularly at least one robotic arm, and / or at least one robot-guided tool, preferably at least one robot-guided welding clamp. In an extended embodiment, the regulator is one for controlling at least one actuator, particularly a motor, preferably an electric motor, of at least one robot, preferably at least one robotic arm, and / or controlling at least one robot-guided, particularly robot-arm-guided tool, in one embodiment a welding clamp. Alternatively or particularly preferably additionally, in one embodiment, the first regulator parameter includes, and may particularly be, a constant of a proportional regulator and / or an integral regulator and / or a derivative regulator. In an extended embodiment, the second regulator parameter may also include, and particularly may be, a constant of a proportional regulator and / or an integral regulator and / or a derivative regulator.

[0027] For these applications, the present invention is particularly advantageous due to the process conditions.

[0028] According to one embodiment of the present invention, a method for performing a robotic process by means of a regulator includes the following steps: - Parameterize the regulator according to the method described in this article; and - The robot process is executed with the help of a parameterized regulator.

[0029] According to one embodiment of the invention, a system for parameterizing a regulator for a robotic process is specifically designed, in particular, by hardware and / or software techniques, especially programming techniques, to perform the methods described herein, and / or includes: - A means for providing a first time history of a first state variable of a robot process for a first value of a first regulator parameter for a regulator used in a robot process; - A means for providing at least one additional time history of a first state variable for another value of a first regulator parameter; - A means for determining a quality value for a given time history using at least partially machine learning-based data processing, wherein the data processing is preferably the at least partially machine learning-based data processing, and in one embodiment includes the data processing described above; and - An apparatus for optimizing the value of a first regulator parameter by means of an optimizer, preferably the optimizer, based on the value of the first regulator parameter and a quality value determined by a time history of these values ​​of the first regulator parameter.

[0030] The systems and / or apparatuses of this invention can be designed using hardware and / or software techniques, and in particular, include: at least one processing unit, preferably connected to a storage system and / or a bus system via data or signal connections, particularly a digital processing unit, particularly a microprocessor unit (CPU), graphics card (GPU), etc.; and / or one or more programs or program modules. The processing unit can be designed to: process instructions implemented as a program stored in the storage system; acquire input signals from the data bus; and / or send output signals to the data bus. The storage system can have one or more, particularly different, storage media, particularly optical, magnetic, solid-state, and / or other non-volatile media. The program can be implemented to embody or execute the methods described herein, enabling the processing unit to perform the steps of this method, and thus, in particular, to optimize the value of a first regulator parameter, and in an extended embodiment, also optimize the value of a second regulator parameter, in another extended embodiment, parameterize the regulator using these values, and in one embodiment, perform a robotic process by means of the parameterized regulator. In one embodiment, the computer program product may have, and may be particularly computer-readable, non-volatile, storage medium for storing a program or instructions, or a storage medium having a program or instructions stored thereon. In one embodiment, execution of the program or instructions causes a system or controller, particularly a computer or an array of computers, to execute the program or instructions, causing the system or controller, particularly one or more computers, to perform the methods described herein or one or more steps thereof, or the program or instructions are designed for this purpose.

[0031] In one implementation, one or more, particularly all, steps of the method are implemented entirely or partially by computer, or one or more, particularly all, steps of the method are executed entirely or partially automatically, particularly by the system or its apparatus.

[0032] In one implementation, the system has one or more robots, particularly one or more robotic arms and / or robot-guided tools, particularly welding clamps, and / or adjusters. Attached Figure Description

[0033] Further advantages and features will be given in the following description with reference to the embodiments.

[0034] Figure 1 A system for parameterizing regulators for robotic processes to perform a method according to an embodiment of the present invention is illustrated schematically; and

[0035] Figure 2 The illustration schematically shows a method for parameterizing a regulator according to one embodiment of the present invention.

[0036] List of reference numerals

[0037] 10: Controller

[0038] 11: Regulator

[0039] 12: Proportional Components

[0040] 13: Integral Components

[0041] 14: Differential component

[0042] 20: Data Processing

[0043] 30: Optimizer

[0044] 100: Robotic Arm

[0045] 101: Driver

[0046] 110: Welding clamps

[0047] 111: Driver Detailed Implementation

[0048] Figure 1 A system is shown for parameterizing the regulator 11 of the controller 10 used to control the robot arm 100, preferably a welding clamp. The regulator 11 may be, in particular, a regulator for controlling the actuator 101 of the robot arm 100 or the actuator 111 of the robot-guided tool 110.

[0049] In step S10 (see Figure 2 In this process, a data processing unit 20, at least partially based on machine learning, is first trained based on manually labeled time histories of a first state variable, which includes, for example, position and torque traction errors during the robotic process utilizing the welding clamp 110. Here, different processes are executed with varying process conditions and / or regulator parameters, and these processes are manually assigned quality values, for example, processes with greater vibration and / or process deviations and / or process disturbances are assessed as (less) poor. The data processing unit 20 is then trained to assign appropriate quality values ​​to different time histories.

[0050] In step S20, the robot welding process is executed multiple times for each value of the proportional constant of the proportional component 12 and / or each value of the differential constant of the differential component 14 and / or each value of the integral constant of the integral component 13 of the regulator. The regulator may accordingly include, in particular, a P regulator, a PI regulator, a PD regulator or a PID regulator.

[0051] In step S30, these data are prepared, wherein time histories are provided for each value of one or more regulator parameters, including the time history of the position traction error of the regulator parameter(s) value and the time history of the torque traction error of the regulator parameter(s) value immediately following or attached thereto.

[0052] In step S40, optimizer 30 first optimizes, for example, the scaling constant of scaling component 12. To do this, the optimizer obtains corresponding quality values ​​for each value of the scaling constant, which are determined by the data processing 20 through the process between the corresponding values ​​of the scaling constant.

[0053] In step S50, optimizer 30 now optimizes, for example, the differential constant of differential component 14 or the integral constant of integral component 13, using the already optimized proportional constant. To this end, the optimizer obtains corresponding quality values ​​for each value of the differential or integral constant and the optimized proportional constant, which are determined by data processing 20 based on the time history of these regulator parameter values.

[0054] In the PID controller, the third proportional derivative and integral constants can also be optimized in a similar way (step S60).

[0055] Subsequently, in step S70, these values ​​can be used to parameterize the regulator 11; and in step S80, the robot process can be executed using the regulator that has been parameterized in this way.

[0056] In this disclosure, "having X" generally does not represent a final enumeration, but is a shorthand for "having at least X," and also includes "having two or more X" and "having Y in addition to X." Although exemplary embodiments have been set forth in the foregoing description, it should be noted that many variations are possible. Furthermore, it should be noted that these exemplary embodiments are merely examples and should not be construed as limiting the scope of protection, application, or construction. Rather, the foregoing description provides guidance for those skilled in the art to make modifications to at least one exemplary embodiment, wherein various changes, particularly regarding the function and arrangement of the components, can be implemented without departing from the scope of the invention, for example, as can be obtained according to the claims and their equivalent combinations of features.

Claims

1. A method for parameterizing a regulator (11) for a robotic process, wherein the method comprises the following steps: - Provide a first time history of a first state variable of the robot process for a first value of a first regulator parameter of a regulator (11) used for the robot process; - Provide at least one additional time history of the first state variable for another value of the first regulator parameter; - For the provided time history, a quality value is determined by means of data processing (20) that is at least partially based on machine learning; as well as - Using the optimizer (30), the value of the first regulator parameter is optimized (S40) based on the value of the first regulator parameter and the quality value determined by the time history of the value of the first regulator parameter.

2. The method according to claim 1, characterized in that, The time history provided to optimize the value of the first regulator parameter is the time history of the first state variable of the robot process, corresponding to the same value of the second regulator parameter of the regulator and different values ​​of the first regulator parameter, and the method includes the following additional steps: - Provide a second time history of the first state variable or the second state variable of the robot process for the optimized value of the first regulator parameter and the first value of the second regulator parameter; - Provide at least one additional time history of the first state variable or the second state variable for the optimized value of the first regulator parameter and another value of the second regulator parameter; - Determine quality values ​​for the provided time history using data processing that is at least partially based on machine learning; as well as - Using an optimizer, the second regulator parameter is optimized (S50, S60) based on the value of the second regulator parameter and the quality value determined from the time history of the value of the second regulator parameter.

3. The method according to any one of the preceding claims, characterized in that, The first state variable includes a first state parameter of the robot process and at least one other state parameter, wherein the time history of the first state variable includes the time history of the first state parameter and the time history of the at least one other state parameter.

4. The method according to the preceding claims, characterized in that, The first state parameter and the at least one additional state parameter are different physical state parameters; and / or one of the first state parameter and the at least one additional state parameter includes an attitude parameter or a load parameter; and / or the first state parameter and / or the at least one additional state parameter includes the difference between the actual value and the target value; and / or the time history of the first state variable includes the time history of the first state parameter and the time history of the at least one additional state parameter that immediately follows it.

5. The method according to any one of the preceding claims, characterized in that, Providing a first time history and at least one additional time history for each value of the first regulator parameter includes: actually or simulatedly executing the robot process; changing the value of the first regulator parameter during the execution of the robot process; and dividing the total time history determined during the execution of the robot process into the first time history and the at least one additional time history.

6. The method according to any one of the preceding claims, characterized in that, Providing a first time history of the first state variable and the at least one additional time history for each value of the first regulator parameter includes: scaling the first time history and the at least one additional time history differently.

7. The method according to any one of the preceding claims, characterized in that, The data processing, at least in part based on machine learning, includes: artificial neural networks and / or decision tree algorithms and / or stochastic decision algorithms and / or training based on manually labeled time histories of the first state variable, wherein the data processing is used to determine quality values ​​for optimizing the values ​​of the first regulator parameters for each value of the first state variable.

8. The method according to any one of the preceding claims, characterized in that, The robotic process is a robotic process having at least one robotic arm (100) and / or at least one robot-guided tool (110); and / or the regulator is a regulator for controlling at least one robot (100) and / or the robot-guided tool (110), and / or at least one actuator (101) of the robot (100), and / or at least one actuator (111) of the robot-guided tool (110); and / or the parameters of the first regulator include constants of proportional regulators and / or integral regulators and / or derivative regulators (12, 13, 14).

9. A method for performing a robotic process by means of a regulator (11), wherein the method comprises the following steps: - Parameterize the regulator according to any one of the preceding claims; as well as - The robotic process is executed with the help of a parameterized regulator.

10. A system for parameterizing a regulator (11) for a robotic process, designed to perform the method according to any one of the preceding claims, and / or comprising: - A means for providing a first time history of a first state variable of the robot process with a first value of a first regulator parameter for the regulator of the robot process; - A means for providing at least one additional time history of the first state variable for another value of the first regulator parameter; - A means for determining a quality value for a given time history by means of at least partly machine learning-based data processing (20); as well as - An apparatus for optimizing the value of the first regulator parameter by means of an optimizer (30) based on the value of the first regulator parameter and a quality value determined by a time history of the value of the first regulator parameter.

11. A computer program or computer program product, wherein, The computer program or computer program product includes, in particular, instructions stored on a computer-readable and / or non-volatile storage medium, which, when executed by one or more computers or the system according to claim 10, cause the one or more computers or the system according to claim 10 to perform the method according to any one of claims 1 to 9.

Images