A method for offline automatic tuning of controller parameters for industrial robotic arms

Abstract

The present invention discloses an offline automatic tuning method for controller parameters, comprising: applying feedback control and setting initial values ​​of feedback controller parameters; inputting an expected motion instruction to a driving joint, obtaining a following error, a calculated error index, and a relative error of the error index; if the relative error does not meet a preset convergence precision, inputting the following error as a motion instruction to the driving joint, obtaining an actual rotation angle of a motor in a second experiment; obtaining a parameter increment until the relative error meets the convergence precision; applying feedforward control and feedback control, and setting initial values ​​of feedforward controller parameters; inputting an expected motion instruction to the driving joint, obtaining a following error, an actual rotation angle of the motor, an error index, and a relative error of the error index; if the relative error does not meet the preset convergence precision, obtaining a parameter increment using the actual rotation angle of the motor, until the relative error meets the convergence precision, thereby completing the offline tuning of the controller parameters; the present invention improves the degree of automation of controller parameter tuning.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, in particular to an offline automatic tuning method for controller parameters of an industrial robot arm. Background Art

[0002] Industrial robots generally adopt a single-joint decentralized control strategy that combines feedback control and feedforward control. That is, under the premise of ensuring system stability through feedback controller, the dynamic following accuracy of the driven joint is improved through feedforward control. Figure 1 This is the structural diagram of the existing drive joint control system.

[0003] The tuning of controller parameters is an important guarantee for the transient characteristics and stability of each driving joint of the robot, and is crucial for improving the dynamic accuracy of the robot. At present, the tuning of controller parameters in the industry mainly includes: (1) based on the critical output response characteristics of the controlled object, using empirical formulas for tuning; (2) by identifying the parameter model of the controlled object, using pole configuration method, internal model control method and amplitude margin method and other theories for tuning; (3) by continuously optimizing through intelligent algorithms such as genetic algorithms, reinforcement learning and neural networks, and adjusting the parameters in the algorithm in real time. However, industrial robots are nonlinear multi-input and multi-output time-varying systems, which are difficult to make them work in the critical point state and use empirical formulas for tuning; the mathematical model of the driving joint is complex and difficult to identify accurately; the hyperparameter adjustment in the intelligent algorithm depends on the experience of the staff and requires high computing power to support the real-time operation of the algorithm, which is difficult to apply to actual control. Therefore, there is an urgent need for an offline automatic tuning method for controller parameters to adapt to the control system of complex robots and improve the automation level of controller parameter tuning. Summary of the Invention

[0004] In order to solve the problems existing in the prior art, the purpose of the present invention is to provide a method for offline automatic tuning of the controller parameters of an industrial robot arm. The present invention does not require real-time identification of the parameter model of the controlled object, is suitable for robot-driven joint control systems with repetitive operation characteristics, and improves the degree of automation of controller parameter tuning.

[0005] To achieve the above object, the present invention adopts a technical solution: a method for offline automatic tuning of controller parameters of an industrial robot arm, comprising the following steps:

[0006] Step 1: Apply only feedback control and set the initial values ​​of the feedback controller parameters Where i represents the i-th experiment of feedback controller parameter tuning, and k represents the k-th adjustable parameter of the feedback controller;

[0007] Step 2: Input the desired motion command Θ to the driving joint d , get the following error Calculation error index And the relative error of the error index

[0008] Step 3: If the relative error does not meet the preset convergence accuracy ε fb , then the following error Used as motion command input to drive the joint and obtain the actual rotation angle of the motor in the second experiment

[0009] Step 4: Use the actual motor angle from the second experiment Get parameter increment Until the relative error meets the convergence accuracy;

[0010] Step 5: After completing the feedback controller parameter tuning, apply feedforward control and feedback control at the same time and set the initial value of the feedforward controller parameter Where, j represents the jth experiment of feedforward controller parameter tuning, and l represents the lth tunable parameter of the feedforward controller;

[0011] Step 6: Input the desired motion command Θ to the driving joint d , get the following error Actual motor angle Error index And the relative error of the error index

[0012] Step 7: If the relative error Does not meet the preset convergence accuracy ε ff , then use the actual rotation angle of the motor Get parameter increment Until the relative error meets the convergence accuracy, the controller parameter offline tuning is completed.

[0013] As a further improvement of the present invention, the step 2 specifically includes the following steps:

[0014] Step 2.1: Input the desired motion command Θ to the driven joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment;

[0015] Step 2.2: Subtract the desired motion angle at the corresponding moment from the actual motor rotation angle at each sampling moment to obtain the following error at each sampling moment. Calculation error index And the relative error of the error index

[0016]

[0017] Where n represents the nth sampling moment, Ts represents the sampling period, Represents the following error at the nth sampling moment.

[0018] As a further improvement of the present invention, step 4 specifically includes the following steps:

[0019] Step 4.1: Get the actual motor angle Perform Laplace transform and get

[0020] Step 4.2: Get the gradient of the following error with respect to each parameter of the feedback controller

[0021]

[0022] Among them, C (i) is the transfer function of the feedback controller in the i-th feedback controller parameter tuning experiment;

[0023] Step 4.3, Perform inverse Laplace transform to get

[0024] Step 4.4: Calculate the error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix

[0025]

[0026]

[0027] Step 4.5: Get parameter increment Update the feedback controller parameters:

[0028]

[0029] Step 4.6: Let i = i + 1 and repeat steps 2 to 4 until the relative error meets the convergence accuracy.

[0030] As a further improvement of the present invention, step 6 specifically includes the following steps:

[0031] Step 6.1: Input the desired motion command Θ to the driven joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment Where n represents the nth sampling moment, T s Indicates the sampling period;

[0032] Step 6.2: Subtract the desired motion angle at the corresponding moment from the actual motor rotation angle at each sampling moment to obtain the following error at each sampling moment. Calculation error index And the relative error of the error index

[0033]

[0034] in, Represents the following error at the nth sampling moment.

[0035] As a further improvement of the present invention, step 7 specifically includes the following steps:

[0036] Step 7.1: If the relative error Does not meet the preset convergence accuracy ε ff , the actual motor angle obtained in step 6.2 Perform Laplace transform and get

[0037] Step 7.2: Get the gradient of the following error with respect to each parameter of the feedforward controller

[0038]

[0039] Among them, C (j) is the transfer function of the feedback controller in the jth feedforward controller parameter tuning experiment, F (j) is the transfer function of the feedforward controller in the jth feedforward controller parameter tuning experiment;

[0040] Step 7.3, Perform inverse Laplace transform to get

[0041] Step 7.4: Calculate the error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix

[0042]

[0043] Step 7.5: Get parameter increment Update the feedback controller parameters:

[0044]

[0045] Step 7.6: Set j = j + 1 and repeat steps 6 and 7 until the relative error meets the convergence accuracy, completing the offline tuning of the controller parameters.

[0046] The core of the present invention is to construct the gradient vector and Hessian matrix of the error index with respect to the feedback controller parameters by following the error and the actual rotation angle of the motor, obtain the controller parameter increment by the gradient descent method, and then gradually reduce the value of the error index, and make the relative error of the error index converge to a preset accuracy.

[0047] The beneficial effects of the present invention are:

[0048] The present invention does not require real-time identification of the parameter model of the controlled object, is applicable to a robot-driven joint control system with a repetitive operation property, and improves the degree of automation of controller parameter setting. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is the structural diagram of the existing drive joint control system;

[0050] Figure 2 4 is a flowchart of offline automatic tuning of controller parameters according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0052] Example

[0053] like Figure 2 As shown, a method for offline automatic tuning of controller parameters includes the following steps:

[0054] S1: Only apply feedback control and set the initial value of the feedback controller parameters Where i represents the i-th experiment of feedback controller parameter tuning, and k represents the k-th adjustable parameter of the feedback controller;

[0055] S2: Input the desired motion command Θ to the driving joint d , get the following error Calculation error index And the relative error of the error index

[0056] S3: If the relative error does not meet the preset convergence accuracy ε fb , then the following error of the previous step is Used as motion command input to drive the joint and obtain the actual rotation angle of the motor in the second experiment

[0057] S4: Using the actual rotation angle of the motor in the second experiment Get parameter increment Until the relative error meets the convergence accuracy;

[0058] S5: After completing the feedback controller parameter tuning, apply feedforward control and feedback control at the same time, and set the initial value of the feedforward controller parameter Where j represents the jth experiment of feedforward controller parameter tuning, l represents the lth tunable parameter of the feedforward controller;

[0059] S6: Input the desired motion instruction Θ to the driving joint d , get the following error Actual motor angle Error index And the relative error of the error index

[0060] S7: If the relative error Does not meet the preset convergence accuracy ε ff , then use the actual motor angle of the previous step Get parameter increment Until the relative error meets the convergence accuracy, the controller parameter offline tuning is completed.

[0061] The step S2 comprises:

[0062] S21: Input the desired motion command θ to the driving joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment;

[0063] S22: Subtract the desired motion angle at the corresponding moment from the actual motor rotation angle at each sampling moment to obtain the following error at each sampling moment Calculation error index And the relative error of the error index

[0064]

[0065] The n represents the nth sampling moment, T s represents the sampling period, Represents the following error at the nth sampling moment.

[0066] The step S4 comprises:

[0067] S41: The actual rotation angle of the motor obtained in step S3 Perform Laplace transform and get

[0068] S42: Obtain the gradient value of the following error with respect to each parameter of the feedback controller

[0069]

[0070] The C(i) is the transfer function of the feedback controller in the i-th feedback controller parameter tuning experiment;

[0071] S43: Yes Perform inverse Laplace transform to get

[0072] S44: Calculation error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix

[0073]

[0074]

[0075] S45: Get parameter increment Update feedback controller parameters;

[0076]

[0077] S46: Let i=i+1, and repeat steps S2, S3, and S4 until the relative error meets the convergence accuracy.

[0078] The step S6 comprises:

[0079] S61: Input the desired motion instruction θ to the driving joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment The n represents the nth sampling moment, T s Indicates the sampling period.

[0080] S62: Subtract the desired motion angle at the corresponding moment from the actual motor rotation angle at each sampling moment to obtain the following error at each sampling moment. Calculation error index And the relative error of the error index

[0081]

[0082] The Represents the following error at the nth sampling moment.

[0083] The step S7 comprises:

[0084] S71: If the relative error Does not meet the preset convergence accuracy ε ff , the actual rotation angle of the motor obtained in step S62 Perform Laplace transform and get

[0085] S72: Obtain the gradient value of the following error with respect to each parameter of the feedforward controller

[0086]

[0087] The C (j) is the transfer function of the feedback controller in the jth feedforward controller parameter tuning experiment, and the F (j) is the transfer function of the feedforward controller in the jth feedforward controller parameter tuning experiment.

[0088] S73: Yes Perform inverse Laplace transform to get

[0089] S74: Calculation error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix

[0090]

[0091] S75: Get parameter increment Update feedback controller parameters;

[0092]

[0093] S76: Let j=j+1, repeat steps S6 and S7 until the relative error meets the convergence accuracy, and complete the offline tuning of the controller parameters.

[0094] The above-described embodiments merely represent specific implementations of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.

Claims

1. A method for offline automatic tuning of controller parameters of an industrial robot arm, characterized in that: The following steps are involved: Step 1: Apply only feedback control and set the initial values ​​of the feedback controller parameters Where i represents the i-th experiment of feedback controller parameter tuning, and k represents the k-th adjustable parameter of the feedback controller; Step 2: Input the desired motion command Θ to the driving joint d , get the following error Calculation error index And the relative error of the error index The step 2 specifically includes the following steps: Step 2.1: Input the desired motion command Θ to the driven joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment; Step 2.2: Subtract the desired motion angle at the corresponding moment from the actual motor angle at each sampling moment to obtain the following error at each sampling moment. Calculation error index And the relative error of the error index Where n represents the nth sampling moment, T s represents the sampling period, represents the following error at the nth sampling moment; Step 3: If the relative error does not meet the preset convergence accuracy ε fb , then the following error Used as motion command input to drive the joint and obtain the actual rotation angle of the motor in the second experiment Step 4: Use the actual motor angle from the second experiment Get parameter increment Until the relative error meets the convergence accuracy; Step 5: After completing the feedback controller parameter tuning, apply feedforward control and feedback control at the same time and set the initial value of the feedforward controller parameter Where, j represents the jth experiment of feedforward controller parameter tuning, l represents the lth tunable parameter of the feedforward controller; Step 6: Input the desired motion command Θ to the driving joint d , get the following error Actual motor angle Error index And the relative error of the error index Step 7: If the relative error Does not meet the preset convergence accuracy ε ff , then use the actual rotation angle of the motor Get parameter increment Until the relative error meets the convergence accuracy, the controller parameter offline tuning is completed.

2. The method for offline automatic tuning of controller parameters of an industrial robot arm according to claim 1, characterized in that: The step 4 specifically includes the following steps: Step 4.1: Get the actual motor angle Perform Laplace transform and get Step 4.2: Get the gradient of the following error with respect to each parameter of the feedback controller Among them, C (i) is the transfer function of the feedback controller in the i-th feedback controller parameter tuning experiment; Step 4.3, Perform inverse Laplace transform to get Step 4.4: Calculate the error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix Step 4.5: Get parameter increment Update the feedback controller parameters: Step 4.6: Let i = i + 1 and repeat steps 2 to 4 until the relative error meets the convergence accuracy.

3. The method for offline automatic tuning of controller parameters of an industrial robot arm according to claim 2, characterized in that: The step 6 specifically includes the following steps: Step 6.1: Input the desired motion command Θ to the driven joint d , use the angle encoder to record the actual rotation angle of the motor at each sampling moment Where n represents the nth sampling moment, T s Indicates the sampling period; Step 6.2: Subtract the desired motion angle at the corresponding moment from the actual motor rotation angle at each sampling moment to obtain the following error at each sampling moment. Calculation error index And the relative error of the error index in, Represents the following error at the nth sampling moment.

4. The method for offline automatic tuning of controller parameters of an industrial robot arm according to claim 3, characterized in that: The step 7 specifically includes the following steps: Step 7.1: If the relative error Does not meet the preset convergence accuracy ε ff , the actual motor angle obtained in step 6.2 Perform Laplace transform and get Step 7.2: Get the gradient of the following error with respect to each parameter of the feedforward controller Among them, C (j) is the transfer function of the feedback controller in the jth feedforward controller parameter tuning experiment, F (j) is the transfer function of the feedforward controller in the jth feedforward controller parameter tuning experiment; Step 7.3, Perform inverse Laplace transform to get Step 7.4: Calculate the error index Gradient vector with respect to the feedback controller parameters and the Hessian matrix Step 7.5: Get parameter increment Update the feedback controller parameters: Step 7.6: Set j = j + 1 and repeat steps 6 and 7 until the relative error meets the convergence accuracy, completing the offline tuning of the controller parameters.

Images