A data-driven control method for flexible manipulator vibration suppression

Abstract

The application discloses a data-driven control method for flexible manipulator vibration suppression and relates to the technical field of robot control, and comprises the following steps: a feedback control system for end vibration suppression of a flexible manipulator is constructed, and residual vibration collection information of the end vibration sensor is taken as driving data in a period; a corresponding adjustable feedback controller of the control system is designed; the adjustable feedback controller is optimized through adjustable parameters; the feedback control system for realizing residual vibration minimization is obtained; and an adaptive parameter optimization algorithm is developed to realize the minimization of end residual vibration as a target; the developed algorithm is based on driving data to realize real-time optimization of the feedback controller; the application breaks through the limitation of the existing active vibration control method, only relies on real-time sampling signals of the end vibration sensor of the flexible manipulator as driving data of the control system, and can realize effective suppression of end residual vibration, thereby realizing a completely data-driven flexible manipulator vibration control method.

Description

A data-driven control method for vibration suppression in flexible robotic arms Technical Field

[0001] This invention relates to the field of robot control technology, and more specifically to a data-driven control method for vibration suppression in flexible robotic arms. Background Technology

[0002] The evolution of robotic arms in robotic systems is increasingly focused on higher payload ratios, with lightweight, flexible robotic arms becoming more widely used in the field. However, the insufficient stiffness of flexible robotic arms leads to elastic deformation at the actuator end, inevitably triggering free vibration. Furthermore, the relatively low structural damping of flexible robotic arms results in prolonged residual vibration. For example, in space robots for aerospace engineering, the use of lightweight flexible robotic arms can significantly reduce launch costs. However, due to the near absence of damping in the space environment, the vibration of the flexible robotic arm will persist for extended periods. This intense vibration severely limits the rapid response capability of the end effector, subsequently affecting its motion accuracy and execution stability. Long-term vibration can also cause fatigue damage to the structure, reducing its lifespan.

[0003] Currently, vibration suppression at the end effector of flexible robotic arms can be broadly categorized into passive control and active control methods. Passive control methods optimize the size or shape of the flexible robotic arm itself to achieve vibration suppression, or introduce external constraint structures for vibration suppression. However, for more complex situations, considering that the mass of the end effector changes continuously under different working scenarios, the fundamental frequency of vibration due to the flexibility ratio is uncertain. Passive control methods based on the structural design of the flexible robotic arm struggle to achieve effective vibration suppression optimization when the fundamental frequency is unknown. Furthermore, the introduction of external constraint structures contradicts the lightweight design principle of the flexible arm. Active control methods mostly optimize the control process of the joint drive torque of the robotic arm, rationally plan the motion trajectory, and reduce the overall vibration caused by impacts during joint rotation, thereby indirectly suppressing the vibration at the end effector. However, in high-speed or heavy-load environments, the vibration accompanying joint rotation is difficult to control effectively, resulting in poor indirect suppression of end effector vibration. Therefore, direct, rapid vibration suppression methods for the end effector of flexible robotic arms have emerged.

[0004] The main approach to direct vibration control of flexible robotic arms is to directly drive the arm itself using piezoelectric actuators, coupled with an adaptive active vibration control algorithm to achieve vibration suppression, without being limited by an unknown fundamental frequency. However, existing adaptive active vibration control algorithms for piezoelectric-driven flexible robotic arms are largely designed based on the dynamic model of the flexible robotic arm, such as through theoretical modeling or system identification. When the flexible robotic arm is subjected to constantly changing working scenarios, changes in the mass of the end effector will cause continuous changes in the dynamic model of the flexible robotic arm. The vibration suppression effect of model-dependent adaptive active vibration control algorithms will decrease sharply due to model inaccuracies, and may even completely fail to suppress vibration. Therefore, a data-driven control method for vibration suppression of flexible robotic arms is needed to address these issues. Summary of the Invention

[0005] The purpose of this invention is to provide a data-driven control method for vibration suppression of flexible robotic arms, so as to solve the problems existing in the prior art mentioned in the background.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A data-driven control method for vibration suppression in flexible robotic arms includes the following steps:

[0008] S1: Construct a feedback control system for vibration suppression at the end of a flexible robotic arm, using the residual vibration data collected periodically by the end-effector as driving data;

[0009] S2: Design an adjustable feedback controller corresponding to the control system in S1. Optimize the adjustable parameters of the adjustable feedback controller to obtain a feedback control system that minimizes residual vibration.

[0010] S3: Develop an adaptive parameter optimization algorithm to minimize the residual vibration at the end of S1. The developed algorithm optimizes the feedback controller in S2 in real time based on the driving data.

[0011] S4: Based on the residual vibration signal data collected by the vibration sensor at the end of the flexible robotic arm, the drive feedback control system is equipped with an adaptive parameter optimization algorithm to realize online adjustment of the adjustable feedback controller. Then, through the output of the adjustable feedback controller, the drive feedback control system realizes active control of vibration.

[0012] Preferably, the feedback control system in S1 includes a flexible robotic arm, a piezoelectric actuator, a vibration sensor, an I / O data acquisition card, an embedded controller, a charge / voltage amplifier, and a drive power supply. The charge / voltage amplifier is connected upstream to the vibration sensor and downstream to the I / O data acquisition card.

[0013] The residual vibration signal collected by the vibration sensor is amplified and then transmitted to the embedded controller via the I / O data acquisition card; the upstream of the drive power supply is connected to the embedded controller and the downstream is connected to the piezoelectric actuator; the control signal output by the embedded controller is amplified and then drives the piezoelectric actuator to perform active vibration control.

[0014] Driven by the data of the residual vibration signal at the end of the flexible robotic arm, the data-driven vibration control algorithm on the embedded controller will provide a control signal, which will drive the piezoelectric actuator to suppress the residual vibration at the end.

[0015] Preferably, the adjustable feedback controller used in S2 consists of an IIR filter with free coefficients and a preset weighting function. The adjustment of the feedback control system is achieved by adjusting the parameter vector corresponding to the IIR filter coefficients.

[0016] The corresponding weighting function is a bandpass filter preset according to the possible working conditions of the flexible robotic arm, and the corresponding cutoff frequency design criterion is to cover the frequency band corresponding to the possible working conditions.

[0017] Preferably, the development of the adaptive parameter optimization algorithm in S3 includes the following steps:

[0018] S31: In any iteration loop, the algorithm generates positive and negative random perturbations consisting of ±1 that satisfy the Bernoulli distribution, based on the dimension of the parameter vector corresponding to the IIR filter coefficients.

[0019] S32: During any iteration of the algorithm, within a certain period, the control system, under the action of a positive random disturbance, collects the mean square value of the residual vibration signal corresponding to each iteration within that period as driving data and temporarily stores it.

[0020] S33: During any iteration of the algorithm, within a certain period, the control system, under the action of negative random disturbance, collects the mean square value of the residual vibration signal corresponding to each iteration within that period as driving data and temporarily stores it.

[0021] S34: In any iteration of the algorithm, the gradient estimate for parameter updates is calculated based on the driving data temporarily stored in S32 and S33.

[0022] S35: During any iteration of the algorithm, the adjustable parameters required for the next iteration are updated based on the gradient estimation in S34.

[0023] Repeat steps S31-S35 to update the coefficients of the IIR filter in the control system in real time until the residual vibration signal is minimized.

[0024] Preferably, the specific steps of S31 are as follows:

[0025] In each iteration of the algorithm, an n-dimensional random perturbation vector Δ is generated based on the dimension n of the adjustable parameter. The perturbation is applied to the adjustable parameter θ corresponding to the current IIR filter control to obtain two parameters: positive perturbation θ+cΔ and negative perturbation θ-cΔ, where c is the perturbation step size.

[0026] Preferably, the specific steps of S32 and S33 are as follows:

[0027] The response is evaluated periodically, and the complete active vibration control system is run according to the two parameters of positive disturbance θ+cΔ and negative disturbance θ-cΔ.

[0028] Period 1: Using the positive disturbance θ+cΔ as the coefficient of the IIR filter, the control system executes a cycle with a set number of iterations. The residual vibration signal corresponding to each iteration in this cycle is collected as the driving data, and the corresponding mean square value J is obtained. (θ+cΔ) ;

[0029] Period 2: Using the negative disturbance θ-cΔ as the coefficient of the IIR filter, the control system executes a cycle with a set number of iterations. The residual vibration signal corresponding to each iteration in this cycle is collected as the driving data, and the corresponding mean square value J is obtained. (θ-cΔ) .

[0030] Preferably, the specific steps of S34 and S35 are as follows:

[0031] According to the objective function J corresponding to the random disturbance (θ+cΔ) and J (θ-cΔ) Obtain the updated gradient estimate of parameter θ. And there are

[0032]

[0033] Further adjust the parameter θ according to the gradient direction, as follows:

[0034]

[0035] Where 'a' is the parameter update step size.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] This invention is based on data-driven residual vibration signals at the end of a flexible robotic arm. Without relying on the dynamic model of the flexible robotic arm, it designs a data-driven adaptive active vibration control system for suppressing vibration at the end of a surface-type flexible robotic arm. The model-free adaptive active vibration control method proposed in this invention provides a specific solution to the model uncertainty and unknown vibration problems caused by multiple execution environment changes of flexible robotic arms, effectively supplementing the shortcomings of existing vibration control strategies for flexible robotic arms. Attached Figure Description

[0038] Figure 1 is a schematic diagram of the control system of the present invention;

[0039] The components include: 1. Flexible robotic arm; 2. Piezoelectric actuator; 3. Vibration sensor; 4. Drive power supply; 5. Charge / voltage amplifier; 6. I / O data acquisition card; and 7. Embedded controller.

[0040] Figure 2 is a block diagram of the control system of the present invention.

[0041] Figure 3 is a flowchart of the adaptive parameter optimization process of the present invention.

[0042] Figure 4 shows the simulation test verification data of the present invention. Detailed Implementation

[0043] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0044] Please refer to Figures 1-4. The present invention provides the following technical solutions:

[0045] A data-driven control method for vibration suppression in flexible robotic arms includes the following steps:

[0046] S1: Establish a feedback control system for the flexible robotic arm 1, as shown in Figure 1.

[0047] The feedback control system includes a flexible robotic arm 1, a piezoelectric actuator 2, a vibration sensor 3, an I / O data acquisition card 6, a charge / voltage amplifier 5, a drive power supply 4, and an embedded controller 7.

[0048] Vibration sensor 3 acquires the residual vibration at the end of the flexible robotic arm 1 in real time. The electrical signal acquired by vibration sensor 3 is amplified by charge / voltage amplifier 5 and then acquired by I / O data acquisition card 6, and input to embedded controller 7 as drive data for the control system. At the same time, embedded controller 7 outputs drive signal according to adaptive parameter optimization algorithm, which is amplified by drive power supply 4 and drives piezoelectric actuator 2 to act on flexible robotic arm 1. Thus, a complete closed-loop control system is formed to effectively suppress the vibration at the end of flexible robotic arm 1.

[0049] S2: Design an adjustable feedback controller for embedded controller 7

[0050] The adjustable feedback controller is integrated into the embedded controller 7 in Figure 1, and includes an adjustable IIR filter and a deterministic weighting function.

[0051] As shown in Figure 2, the control response process from the piezoelectric actuator 2 to the flexible end is taken as the controlled object of the corresponding flexible robotic arm 1. An adjustable IIR filter with a weighting function is used as an adjustable feedback controller. The developed adaptive parameter optimization algorithm is used as the adjustment algorithm of the feedback controller. The inertial force at the end of the flexible robotic arm is used as the excitation interference of the end vibration, and the corresponding control loop is established.

[0052] To improve the robust stability of the closed-loop system under conditions of adjustable feedback controller tuning and parameter perturbations involving the controlled object, the following weighting function is designed:

[0053]

[0054] Equation (3) is designed as a bandpass filter with a bandwidth covering the vibration frequencies of all possible working conditions of the flexible robotic arm. Within the passband, the signal required for vibration interference suppression is allowed, so that the feedback controller generates a high-gain anti-interference signal at the vibration frequency. Within the stopband, the signal is significantly attenuated to avoid slow oscillations caused by low frequency bands and amplification of measurement noise in high frequency bands.

[0055] To achieve online adjustability of the adjustable feedback controller, and to suppress vibration using fewer adjustable parameters for complex control systems, an IIR filter is adopted as follows:

[0056]

[0057] The coefficients of the IIR filter in equation (4) The flexible robotic arm is freely adjustable and can achieve online optimization of parameter θ in the feedback controller through the developed adaptive parameter optimization algorithm, thereby minimizing the vibration of the flexible robotic arm.

[0058] S3: Design an adaptive parameter optimization algorithm based on synchronous disturbance random approximation for an adjustable feedback controller. The specific process is shown in Figure 3. The adaptive parameter optimization algorithm mainly includes:

[0059] S31: Generate positive and negative random perturbations; in specific implementation, generate the random perturbation corresponding to the k-th iteration.

[0060] In the k-th iteration of the algorithm, according to the parameter θ in step two... k The dimension n = n A +nB Generate an n-dimensional random perturbation vector Δ k And the random perturbation vector Δ k It consists of ±1 that satisfy a Bernoulli distribution; for the k-th iteration, the adjustable parameter θ of the control system is... k Apply a perturbation to obtain a positive perturbation θ k +cΔ k and negative perturbation θ k -cΔ k Two parameters, where c is the preset perturbation step size.

[0061] S32: Response evaluation under positive random perturbation; in specific implementation, the positive perturbation response evaluation corresponding to the k-th iteration is performed.

[0062] In t k to t k During the +lT time period, with positive perturbation θ k +cΔ k As a coefficient execution control system for an IIR filter, the residual vibration signal e corresponding to each iteration within a period of lT is acquired. k As the driving data, the corresponding mean square value is obtained. Where T is the sampling period.

[0063] S33: Response evaluation under negative random perturbation; in specific implementation, the response evaluation of the negative perturbation corresponding to the k-th iteration is performed.

[0064] In t k +lT to t k During the +2lT time period, with negative perturbation θ k -cΔ k As the coefficients of the IIR filter are controlled, the residual vibration signal e corresponding to each iteration within a period of lT is acquired. k As the driving data, the corresponding mean square value is obtained.

[0065] S34: Calculate the gradient estimate for parameter updates; in practice, calculate the parameter θ. k The gradient estimate is updated in the k-th iteration.

[0066] In t k At time +(2l+1)T, based on the random perturbation θ k +cΔ k and θ k -cΔ k The corresponding objective functions and Calculate parameter θ k Updated gradient estimate as follows:

[0067]

[0068] S35: Update the adjustable parameters; in practice, update the adjustable parameters corresponding to the k-th iteration.

[0069] Gradient estimation Direction online update parameter θ k ,as follows

[0070]

[0071] Where 'a' is the preset parameter update step size.

[0072] At this point, based on the driving data e k Based on the adaptive parameter optimization algorithm in step three, the updated θ k+1 As the coefficients of the IIR filter in the (k+1)th iteration, the above S1-S5 process is repeated to update the coefficients of the IIR filter in the control system in real time until the residual vibration signal is minimized.

[0073] This invention uses a piezoelectric actuator 2 to directly drive a flexible robotic arm 1, and develops an adaptive feedback control method that does not rely on the dynamic model of the flexible robotic arm 1. It only uses the detection signal data of the residual vibration at the end of the flexible robotic arm 1 to drive the optimization of the controlled parameters through real-time online learning, thereby forming a data-driven adaptive active vibration control system for the flexible robotic arm.

[0074] This invention uses an IIR filter as an adjustable feedback controller and employs an SPSA-based adaptive optimization algorithm to update the IIR filter in real time. This method can effectively suppress the vibration of the flexible robotic arm, thereby improving the motion accuracy of the flexible robotic arm.

[0075] The effectiveness of the invention was verified through simulation.

[0076] Simulation verification adopts As a controlled flexible robotic arm, y(t)=2sin(20πt)e -0.9t As a residual vibration interference signal, white noise is superimposed to simulate basic test interference, and the control effect shown in Figure 4 is obtained. The data-driven flexible manipulator vibration suppression method developed in this invention does not require a dynamic model based on the flexible manipulator, nor does it require prior measurement of interference information. It only relies on the measurement signal e of the vibration sensor 3 at the end of the flexible manipulator to drive the data, which can effectively suppress the residual vibration at the end. It is suitable for scenarios with constantly changing working conditions.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data-driven control method for vibration suppression in flexible robotic arms, characterized in that, Includes the following steps: S1: Construct a feedback control system for vibration suppression at the end of a flexible robotic arm, using the residual vibration data collected periodically by the end-effector as driving data; S2: Design an adjustable feedback controller corresponding to the control system in S1. Optimize the adjustable parameters of the adjustable feedback controller to obtain a feedback control system that minimizes residual vibration. The adjustable feedback controller used in S2 consists of an IIR filter with free coefficients and a preset weighting function. Adjustment of the parameter vector corresponding to the IIR filter coefficients achieves the adjustment of the feedback control system. The corresponding weighting function is a bandpass filter preset based on the possible operating conditions of the flexible robotic arm, and the corresponding cutoff frequency design criterion is to cover the frequency band corresponding to the possible operating conditions. S3: Develop an adaptive parameter optimization algorithm to minimize the residual vibration at the end of S1. The developed algorithm performs real-time optimization of the feedback controller in S2 based on the driving data. The development of the adaptive parameter optimization algorithm in S3 includes the following steps: S31: In any iteration loop, the algorithm generates a parameter vector that satisfies a Bernoulli distribution based on the dimension of the parameter vector corresponding to the IIR filter coefficients. The algorithm generates positive and negative random disturbances; S32: In any iteration cycle, within a certain period, the control system, under the influence of positive random disturbances, collects the mean square value of the residual vibration signal corresponding to each iteration within that period as driving data and temporarily stores it; S33: In any iteration cycle, within a certain period, the control system, under the influence of negative random disturbances, collects the mean square value of the residual vibration signal corresponding to each iteration within that period as driving data and temporarily stores it; S34: In any iteration cycle, the algorithm, based on the driving data temporarily stored in S32 and S33... Data is used to calculate and obtain gradient estimates for parameter updates; S35: During any iteration of the algorithm, the adjustable parameters required for the next iteration are updated based on the gradient estimate in S34; Steps S31-S35 are repeated to update the coefficients of the IIR filter in the control system online in real time until the residual vibration signal is minimized; S4: Based on the residual vibration signal data collected by the vibration sensor at the end of the flexible robotic arm, the adaptive parameter optimization algorithm on the feedback control system is driven to realize the online adjustment of the adjustable feedback controller, and then the output of the adjustable feedback controller drives the feedback control system to achieve active vibration control.

2. The data-driven control method for vibration suppression of flexible robotic arms according to claim 1, characterized in that, The feedback control system in S1 includes a flexible robotic arm, a piezoelectric actuator, a vibration sensor, an I / O data acquisition card, an embedded controller, a charge / voltage amplifier, and a drive power supply. The charge / voltage amplifier is connected upstream to the vibration sensor and downstream to the I / O data acquisition card; it amplifies the residual vibration signal collected by the vibration sensor and then transmits it to the embedded controller via the I / O data acquisition card. The drive power supply is connected upstream to the embedded controller and downstream to the piezoelectric actuator. After amplifying the control signal output by the embedded controller, the piezoelectric actuator is driven to perform active vibration control. Under the data-driven effect of the residual vibration signal at the end of the flexible robotic arm, the data-driven vibration control algorithm on the embedded controller will provide a control signal, which drives the piezoelectric actuator to suppress the residual vibration at the end.

3. The data-driven control method for vibration suppression of flexible robotic arms according to claim 1, characterized in that, The specific steps of S31 are as follows: In each iteration of the algorithm, an n-dimensional random perturbation vector is generated based on the dimension n of the adjustable parameter. Adjustable parameters corresponding to the current IIR filter control Apply a perturbation to obtain a positive perturbation. and negative disturbance Two parameters, where c is the perturbation step size.

4. The data-driven control method for vibration suppression of flexible robotic arms according to claim 1, characterized in that, The specific steps of S32 and S33 are as follows: Response evaluation is performed periodically, based on the positive disturbance. and negative disturbance The two parameters operate separately to form a complete active vibration control system; Period 1: with positive disturbance As coefficients of the IIR filter, the actuator control system performs a cycle with a set number of iterations, collects the residual vibration signal corresponding to each iteration within this cycle as driving data, and calculates the corresponding mean square value. Period 2: with negative perturbation As coefficients of the IIR filter, the actuator control system performs a cycle with a set number of iterations, collects the residual vibration signal corresponding to each iteration within this cycle as driving data, and calculates the corresponding mean square value. 。 5. A data-driven control method for vibration suppression of flexible robotic arms according to claim 1, characterized in that, The specific steps of S34 and S35 are as follows: based on the objective function corresponding to the random perturbation and Get parameters Updated gradient estimate And there are (1) Further adjust the parameters according to the gradient direction ,as follows: (2) Where a is the parameter update step size.

Images