Industrial manipulator control method based on command filtering compensation

Abstract

The invention discloses an industrial manipulator control method based on command filtering compensation, which belongs to the field of manipulator control and comprises the following steps of: firstly, estimating uncertain disturbance or an unmodeled function in an industrial manipulator dynamic system by using a fuzzy logic system; secondly, designing a Levant command filter between every two subsystems, and further designing a corresponding filtering compensation system according to a filtering error generated by the Levant command filter; and finally, constructing a virtual control signal and an actual control signal for driving a terminal joint to move in combination with the dynamic surface frame. According to the invention, on one hand, a filtering compensation system is adopted to weaken filtering errors, so that the tracking precision of a traditional dynamic surface control algorithm is further improved; and on the other hand, the weight parameters in the fuzzy logic system are updated by adopting the minimum parameter estimation mechanism, so that the calculation load required for training the weight parameters is remarkably reduced, and the real-time performance of the control algorithm is improved.

Description

Technical Field

[0001] The present invention belongs to the field of robot control, and in particular relates to an industrial robot control method based on command filtering compensation. Background Art

[0002] In the field of automated trajectory control of industrial robots, traditional dynamic surface control algorithms face significant technical bottlenecks when dealing with complex working conditions: the inherent highly nonlinear characteristics and unmodeled dynamics of the industrial robot's dynamic system can easily lead to insufficient trajectory tracking accuracy, especially in scenarios such as high-precision assembly and flexible processing, which can easily cause cumulative errors; at the same time, the existing fuzzy approximation algorithm requires online updating of all membership function parameters, resulting in an exponential increase in computational complexity, which seriously restricts the real-time response capability of the control system.

[0003] Although the existing industrial robot control system overcomes the "differential explosion" problem by introducing a dynamic surface framework, the coupling effect of unmodeled dynamics and filter errors will still reduce the robustness of the system. In addition, there is a lack of an effective estimation efficiency improvement mechanism in the parameter estimation process, making it difficult to balance control accuracy and computational efficiency. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides an industrial robot control method based on command filtering compensation, comprising:

[0005] Construct a self-learning fuzzy logic system based on the uncertain disturbances or unmodeled functions in the dynamic system of the industrial robot;

[0006] Construct a Levant command filter including a fourth-order filter error compensation system;

[0007] Introducing the self-learning fuzzy logic system and the Levant command filter into a dynamic surface control framework to generate an actual torque control signal for driving the terminal joint motion;

[0008] The manipulator is controlled based on the actual torque control signal driving the terminal joint movement.

[0009] Optionally, the process of constructing the self-learning fuzzy logic system includes:

[0010] The sub-dynamic systems of the industrial robot dynamics model were sorted out to separate the uncertain parts that are difficult to model;

[0011] Using multiple fuzzy logic systems to perform approximate modeling processing on the uncertain parts respectively;

[0012] A low-load adaptive update law model is designed, and the optimal weight parameters in the self-learning fuzzy logic system are updated in real time based on the low-load adaptive update law model.

[0013] Optionally, the process of designing a low-load adaptive update law model includes:

[0014] Construct the coupling relationship between the estimated variables and weight parameters;

[0015] introducing membership function information of a plurality of self-learning fuzzy logic systems into the coupling relationship;

[0016] A self-learning fuzzy logic system is constructed to approximate and model the uncertain components in each sub-dynamic.

[0017] Optionally, the process of constructing a Levant command filter including a fourth-order filtering error compensation system includes:

[0018] A total of three Levant command filters with two adjustment parameters are designed between the sub-dynamic systems of the industrial robot.

[0019] A fourth-order filter error compensation system is constructed for the filter errors generated by the Levant command filters of the two adjustment parameters.

[0020] Optionally, the expression of the Levant command filter is:

[0021] ;

[0022] in, , are the state variables of the filter, and Respectively and The derivative of Indicates the virtual control signal that needs to be designed later. and Both represent positive design parameters, Represents a symbolic function.

[0023] Optionally, the expression of the fourth-order filtering compensation system is:

[0024] ;

[0025] in, , , , , , Both represent positive design parameters, and Both represent the state variables of the filter compensation system; Represents state variables The derivative of Represents state variables The derivative of Indicates compensation for tracking error.

[0026] Optionally, the process of generating an actual torque control signal for driving the terminal joint to move includes:

[0027] Design virtual control signals for low-order subsystems that do not contain actual control signals;

[0028] Calculating a derivative of a previous virtual control signal using a Levant command filter according to the virtual control signal, and using the previous virtual control signal in a design process of a next virtual control signal;

[0029] An actual output torque control signal is generated according to the virtual control signal and the state variable of the Levant command filter.

[0030] Optionally, the process of generating an actual torque control signal for driving the terminal joint to move further includes:

[0031] Design virtual control signals for the first three order sub-dynamic systems of the flexible manipulator;

[0032] Design actual control signals for the last-order subsystem;

[0033] The virtual control signal and the actual control signal are combined with the Levant command filter and the fourth-order filter error compensation system to form a complete control signal chain.

[0034] On the other hand, the present invention further provides an electronic device, comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the method is implemented when the processor executes the computing program.

[0035] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the method when executed by a processor.

[0036] Compared with the prior art, the present invention has the following advantages and technical effects:

[0037] The present invention introduces a Levant command filter between each subsystem, which can preprocess the virtual control signal of the previous subsystem and then generate a high-precision reference signal to input into the subsequent subsystem. At the same time, the constructed filtering compensation system can dynamically compensate for the residual filtering error. This filtering double-closed-loop correction mechanism effectively suppresses the influence of the filtering error on the tracking accuracy of the controller.

[0038] This paper innovatively proposes a fuzzy approximation mechanism based on the minimum weight parameter update criterion. By reconstructing the parameter update rules of the fuzzy logic system, this mechanism reduces the dimensionality of the weight matrix, which traditionally requires full online parameter adjustment, and adaptively updates only the maximum bi-norm of the weight parameters that dominate the dynamic characteristics of the control system, significantly reducing the amount of online computation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0040] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention;

[0041] Figure 2 A comparison diagram of the angular displacement of an embodiment of the present invention and a reference trajectory;

[0042] Figure 3 A tracking error comparison diagram according to an embodiment of the present invention;

[0043] Figure 4 A graph showing changes in estimated values ​​of weight parameters according to an embodiment of the present invention;

[0044] Figure 5 This is a graph showing changes in joint angular velocity and drive motor angular velocity according to an embodiment of the present invention;

[0045] Figure 6 This is a diagram showing the angular displacement change of the drive motor according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0047] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0048] Example 1

[0049] like Figure 1 As shown, this embodiment provides an industrial robot control method based on command filtering compensation, including:

[0050] Construct a self-learning fuzzy logic system based on the uncertain disturbances or unmodeled functions in the dynamic system of the industrial robot;

[0051] Construct Levant command filter based on fourth-order filter error compensation system;

[0052] Generate an actual torque control signal for driving the terminal joint motion based on the self-learning fuzzy logic system and the Levant command filter;

[0053] The manipulator is controlled based on the actual torque control signal driving the terminal joint movement.

[0054] Specifically:

[0055] Step 1: Build a self-learning fuzzy logic system and use it to estimate the uncertain disturbance or unmodeled function in the industrial robot dynamic system;

[0056] Step 2: Design a Levant command filter and its filter error compensation system;

[0057] Step 3: Introduce the self-learning fuzzy logic system and Levant command filter into the dynamic surface control framework to design the actual torque control signal that can be used to drive the terminal joint motion.

[0058] The specific method of constructing a self-learning fuzzy logic system and using it to estimate uncertain disturbances or unmodeled functions in the dynamic system of an industrial robot is as follows:

[0059] First, we need to determine which data items in the dynamic system are difficult to accurately model, so that they can be used as the targets that the fuzzy logic system needs to estimate. For example, consider the following dynamic model of an industrial robot:

[0060]

[0061] in, , They represent the angular displacement and angular velocity of a connecting rod rotating around its own joint; , Indicates the angular displacement and angular velocity generated by the drive motor installed at the joint during operation; The torque transmission flexibility parameter inherent in the connection between the drive motor and the revolute joint; express The derivative of ; Represents the random uncertain noise caused by the sensor feedback angular displacement, It represents the random uncertain noise caused by the sensor feedback of the angular displacement and angular velocity generated by the drive motor installed at the joint during operation; Indicates the mass of the connecting rod; represents the acceleration due to gravity; Indicates the connecting rod length; Represents the friction model at the revolute joint, which is related to the angular velocity of the joint; It represents the torque that needs to be provided to each joint, which can be regarded as the control signal that needs to be designed; represents the natural damping parameter of the drive motor; represents the moment of inertia of the entire joint; Indicates the output signal.

[0062] On the one hand, the relevant sensors will inevitably bring random uncertain noise when feeding back information such as joint angle and motor output angle; on the other hand, the friction force at the joint will also be affected by many factors such as load and lubrication, so it is difficult to obtain an accurate mathematical model of friction force. Therefore, it is necessary to adjust the original dynamic model of the manipulator. Further sorting is done to distinguish the certain part from the uncertain part. The certain part will be used as available information in controller design; the self-learning fuzzy logic system will make an approximate estimate of the uncertain part. It is further organized into the following formula (2):

[0063]

[0064] in, , Both represent the uncertain components in each sub-dynamic system that are difficult to be accurately modeled. Expressed as follows:

[0065]

[0066] However, in actual control environments, it is often difficult to obtain Precisely expressed Therefore, it is proposed to use self-learning fuzzy logic system to approximate modeling of these uncertain components. Then, these uncertain components Nature is regarded as an approximation object of the self-learning fuzzy logic system.

[0067] Next, to approximate the uncertain part As an example, the design process of the self-learning fuzzy logic system is further explained. First, it is necessary to construct A fuzzy logic system with fuzzy rules:

[0068]

[0069] in represents the input signal vector, is the output signal; ,and is an important component of fuzzy sets. Please note that the fuzzy logic system formula The input signal is the uncertain part Input variables Therefore, it is necessary to summarize through experiments which input variables are related to the uncertain parts of each sub-dynamic system. Similarly, designing fuzzy logic systems for other uncertain parts only requires replacing the input information and determine the total number of fuzzy rules Next, the output signal of the fuzzy logic system can be It is further expressed as follows:

[0070]

[0071] in, represents the optimal weight parameter vector, Represents the activation function vector. The activation function at each node It can be expressed as:

[0072]

[0073] in, express The corresponding member function, Indicates the kth input variable, when k=1 , k=2 is , k=3 is , represents the total number of fuzzy rules, Indicates the symbol for cumulative multiplication;

[0074] According to the above design steps, the mathematical models of the other three fuzzy logic systems can be expressed as follows:

[0075]

[0076] in, express The corresponding output signal, express The corresponding fuzzy logic system is express The corresponding output signal, express The corresponding fuzzy logic system is express The corresponding output signal, express The corresponding fuzzy logic system.

[0077] It can be found that the above four fuzzy logic systems correspond to different optimal weight parameters ( , , , ). However, the traditional weight parameter solution is to design adaptive parameter update laws to update the weight parameter values ​​in each fuzzy logic system, which will inevitably cause a huge computational load. Therefore, from the perspective of reducing the computational load, this embodiment further constructs a single estimation variable The coupling relationship expression with the weight parameters in all fuzzy logic systems is:

[0078]

[0079] in, represents the optimal weight parameter The second norm of .

[0080] That is, the estimated variable An equivalent estimate is made for the maximum value of the two norms of all weight parameters, such as Next, this embodiment estimates the variable An online self-learning update law is designed:

[0081]

[0082] in, , , Both represent the input signal vector, The definition of is shown in the formula Below; represents the compensation tracking error; , , All represent positive design parameters; express The derivative of . Therefore, the dynamic model Each uncertain component in can be approximated and modeled using the following self-learning fuzzy logic system:

[0083]

[0084] The specific method of designing a Levant command filter and its fourth-order filtering error compensation system is as follows:

[0085] The classic process of dynamic surface controller design framework is to design virtual or real control signals for each subsystem, and then use the state variables of the filter to connect the control signals of the subsystems together to form the entire control system. However, the first-order subsystem has a clear control goal, which is to make the output signal Real-time tracking of reference trajectory .then, It can be manually set according to the actual work task and can therefore be regarded as a known quantity. However, there is no clear tracking target for the second-order to the last-order subsystem. Therefore, the solution provided in this embodiment is to use the state variable of the Levant command filter as the tracking target to achieve cascade control. Therefore, it is necessary to design a Levant command filter between the first to second, second to third, and third to fourth order subsystems respectively:

[0086]

[0087] in , Both represent the state variables of the filter; and Respectively and respective derivatives; Indicates the virtual control signal that needs to be designed later; and All represent positive design parameters. represents the symbolic function, which is defined in detail as:

[0088]

[0089] State variables in the Levant command filter Can be regarded as the The reference signal that each subsystem needs to track, When stable, it can be regarded as An estimate of the derivative of . Will be the first This design eliminates the need for virtual control signals in a subsystem. The complex derivation operation solves the "differential explosion" problem of backstepping control.

[0090] The solution of using the state variables of the Levant command filter to approximate the virtual control signal and its derivative will inevitably produce approximation errors. Therefore, in order to reduce the impact of the filtering error on the tracking accuracy of the closed-loop system, this embodiment further designs the following fourth-order filter compensation system:

[0091]

[0092] in , , , , , Both represent positive design parameters, and Both represent the state variables of the filter compensation system; Represents state variables The derivative of Represents state variables The derivative of .

[0093] The self-learning fuzzy logic system and Levant command filter are introduced into the dynamic surface control framework to design an actual torque control signal that can be used to drive the terminal joint motion. The specific method is as follows:

[0094] The main function of the control system in industrial robots is to enable the terminal to track the specified reference trajectory signal. The dynamic surface controller design framework refers to the design of virtual control signals for the first three sub-dynamic systems of the flexible manipulator. , and and design the actual control signal for the last-order subsystem The functions of these virtual control signals are reflected in two aspects: first, to ensure the stability of each sub-dynamic system, for example Mainly to ensure the first-order sub-dynamic system The stability of the second is the Levant command filter formula and filter compensation system formula Provide input signals. Next, the specific design process of these three virtual control signals and actual control signals is introduced.

[0095] To quantize the output signal With the reference trajectory signal The gap between them can be defined as the following tracking error:

[0096]

[0097] The filter compensation system designed in this embodiment can further reduce the tracking error Therefore, in order to quantitatively evaluate the effect of the state variables of the filtering compensation system on the weakening of the tracking error, we can focus on the first-order sub-dynamic system of the industrial manipulator ( ) is defined as follows to compensate for the tracking error :

[0098]

[0099] in, is the state variable of the filtering compensation system.

[0100] Next, you need to compensate for the tracking error and self-learning fuzzy logic systems For virtual control signals Therefore, this embodiment will Designed as follows:

[0101]

[0102] in , , , and All represent design parameters greater than zero; Represents the state variables corresponding to the robust compensation system; Indicates the reference trajectory signal The derivative of . Introducing these two The purpose is to further improve the anti-interference performance of the controller.

[0103] In order to ensure the state variables in the second-order dynamic subsystem of the industrial robot The boundedness of , we can define the following tracking error:

[0104]

[0105] That is, let the virtual control signal Driver state variables Tracking the state variables corresponding to the bounded Levant command filter Of course, the state variable corresponding to the second subsystem in the fourth-order filter compensation system is Also used to reduce tracking error Therefore, for the second-order sub-dynamic system of the industrial robot ( ) defines the following compensation tracking error:

[0106]

[0107] Next, you need to compensate for the tracking error and self-learning fuzzy logic systems For virtual control signals Therefore, this embodiment will Designed as follows:

[0108]

[0109] in , , and are design parameters greater than zero. Note that The state variable of the Levant command filter Therefore, first Medium robust compensation term of The derivative of the state variable of the filter commanded by Levant Instead, then introduce new design parameters and , thereby obtaining the virtual control signal The robust compensation term in .

[0110] In order to ensure the state variables in the third-order dynamic subsystem of the industrial robot The boundedness of , we can define the following tracking error:

[0111]

[0112] That is, let the virtual control signal Driver state variables Tracking the state variables corresponding to the bounded Levant command filter Of course, the state variable corresponding to the third subsystem in the fourth-order filter compensation system is Also used to reduce tracking error Therefore, for the third-order sub-dynamic system of the industrial robot ( ) defines the following compensation tracking error:

[0113]

[0114] Next, you need to compensate for the tracking error and self-learning fuzzy logic systems For virtual control signals Therefore, this embodiment will Designed as follows:

[0115]

[0116] in , , and All represent design parameters greater than zero.

[0117] Since the actual torque control signal has appeared in the fourth-order dynamic subsystem , so we can The state variable corresponding to the Levant command filter is further designed. Provides a reference trajectory signal for the fourth-order dynamic subsystem, so the following tracking error can be defined:

[0118]

[0119] Of course, the state variable corresponding to the fourth subsystem in the fourth-order filter compensation system is Also used to reduce tracking error Therefore, for the fourth-order sub-dynamic system of the industrial robot ( ) defines the following compensation tracking error:

[0120]

[0121] Next, you need to compensate for the tracking error and self-learning fuzzy logic systems For actual control signal Therefore, this embodiment will Designed as follows:

[0122]

[0123] in , , and In addition, it is necessary to provide state variables in real time at each time of adoption. Therefore, this embodiment designs the following robust update law:

[0124] =

[0125] in, and are design parameters greater than zero, is a state variable The derivative of .

[0126] The actual control signal with the filter compensation system designed in this embodiment is input into the constructed industrial robot simulation model. Figure 2 Demonstrates that the robot end can quickly track the specified reference trajectory signal with an initial angular displacement difference of 0.1rad In order to further illustrate the positive effect of the filter compensation system on the tracking accuracy of the control algorithm, the simulation results of the traditional dynamic surface control algorithm are given, such as Figure 3 As shown. The traditional dynamic surface control algorithm lacks a filtering compensation system, so its filtering error cannot be effectively suppressed, resulting in a decrease in tracking accuracy. The tracking error of the controller described in this embodiment has been reduced to within the range of 0.005865 after the control system enters the stable period, as shown in Figure 2. Figure 3 Shown as solid line. Figure 3 The midpoint solid line shows the tracking error of the traditional dynamic surface control algorithm, whose amplitude can only be controlled within the range of 0.02026, so the tracking accuracy is reduced by about . Figure 4 Adaptive parameter estimates are given From the real-time change curve, we can see that regular changes have occurred after 8 seconds. Figure 5 The curves of the joint angular displacement and the drive motor angular displacement are shown respectively. In order to drive the joint to rotate, the drive motor has a higher speed at the initial moment. Figure 6 The curve of the driving motor angular displacement changing with time t is given.

[0127] On the other hand, this embodiment further provides an electronic device, comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the method is implemented when the processor executes the computing program.

[0128] On the other hand, this embodiment further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the method when executed by a processor.

[0129] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An industrial robot control method based on command filtering compensation, characterized in that: include: Construct a self-learning fuzzy logic system based on the uncertain disturbances or unmodeled functions in the dynamic system of the industrial robot; Construct a Levant command filter including a fourth-order filter error compensation system; Introducing the self-learning fuzzy logic system and the Levant command filter into a dynamic surface control framework to generate an actual torque control signal for driving the terminal joint motion; The manipulator is controlled based on the actual torque control signal driving the terminal joint movement.

2. The method according to claim 1, characterized in that The process of constructing the self-learning fuzzy logic system includes: The sub-dynamic systems of the industrial robot dynamics model were sorted out to separate the uncertain parts that are difficult to model; Using multiple fuzzy logic systems to perform approximate modeling processing on the uncertain parts respectively; A low-load adaptive update law model is designed, and the optimal weight parameters in the self-learning fuzzy logic system are updated in real time based on the low-load adaptive update law model.

3. The method according to claim 2, characterized in that The process of designing a low-load adaptive update law model includes: Construct the coupling relationship between the estimated variables and weight parameters; introducing membership function information of a plurality of self-learning fuzzy logic systems into the coupling relationship; A self-learning fuzzy logic system is constructed to approximate and model the uncertain components in each sub-dynamic.

4. The method according to claim 1, wherein The process of constructing a Levant command filter including a fourth-order filtering error compensation system includes: A total of three Levant command filters with two adjustment parameters are designed between the sub-dynamic systems of the industrial robot. A fourth-order filter error compensation system is constructed for the filter errors generated by the Levant command filters of the two adjustment parameters.

5. The method according to claim 4, characterized in that The expression of the Levant command filter is: ； in, , are the state variables of the filter, and Respectively and The derivative of Indicates the virtual control signal that needs to be designed later. and Both represent positive design parameters, Represents a symbolic function.

6. The method according to claim 5, characterized in that The expression of the fourth-order filtering error compensation system is: ； in, , , , , , Both represent positive design parameters, and Both represent the state variables of the filter compensation system; Represents state variables The derivative of Represents state variables The derivative of Indicates compensation for tracking error.

7. The method according to claim 1, characterized in that The process of generating the actual torque control signal for driving the terminal joint movement includes: Design virtual control signals for low-order subsystems that do not contain actual control signals; Calculating a derivative of a previous virtual control signal using a Levant command filter according to the virtual control signal, and using the previous virtual control signal in a design process of a next virtual control signal; An actual output torque control signal is generated according to the virtual control signal and the state variable of the Levant command filter.

8. The method according to claim 1, characterized in that The process of generating an actual torque control signal for driving the terminal joint movement further includes: Design virtual control signals for the first three order sub-dynamic systems of the flexible manipulator; Design actual control signals for the last-order subsystem; The virtual control signal and the actual control signal are combined with the Levant command filter and the fourth-order filter error compensation system to form a complete control signal chain.

9. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein: When the processor executes the computing program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images