Robot circle-enclosing motion control method based on vibration suppression

Abstract

The application discloses a robot circle motion control method based on vibration suppression. Firstly, the application carries out fast initial positioning through multi-scale FFT phase correlation, and combines an improved RANSAC algorithm to robustly fit a target circle track under strong interference, and outputs a circle center, a radius and an inner point proportion in real time. Secondly, the application uses an end IMU to monitor vibration, extracts vibration energy and a dominant frequency, and dynamically adjusts a motion angular velocity according to the vibration energy and the dominant frequency, so as to actively avoid vibration from a planning level. Finally, a compound controller is designed, which fuses a coupled dynamics feedforward compensation, a perception quality adaptive sliding mode control and a frequency perception active damping injection. Through deep closed-loop fusion of visual perception, vibration monitoring and adaptive control, the method effectively overcomes the strong interference and mechanical vibration coupling problem under a complex working environment, and significantly improves the track precision, motion stability and overall system robustness of a related process.

Description

Technical Field

[0001] This invention belongs to the field of robot motion control, specifically relating to a method for controlling the circular motion of a robot based on vibration suppression. Background Technology

[0002] Circular motion of robots is a core motion paradigm in many high-end manufacturing and precision operations, not only limited to circular welding, but also widely present in scenarios of high-precision circumferential trajectory tracking and continuous processing. For example, in precision circumferential grinding, polishing, cutting, gluing, and inspection operations in aerospace, automotive manufacturing, and other fields, the robot's end-effector is required to perform continuous, stable, and high-precision movements along a preset or sensor-recognized circular or arc-shaped trajectory.

[0003] These types of circular motion control generally face three common challenges, severely restricting the improvement of work quality: First, strong interference with the perception environment. In scenarios such as grinding, polishing, and welding, pervasive dust, high-speed flying debris, coolant atomization, and sparks from the machining process itself severely contaminate the field of view of the vision sensor, causing visual tracking methods based on traditional feature matching or edge detection to easily fail. Second, complex dynamic disturbances. Mechanical vibrations induced by the machining process itself (such as intermittent contact force between the tool and the workpiece, and imbalance of high-speed spindle rotation) are coupled with structural resonance and centripetal force excited when the robot performs circular motion, resulting in unpredictable vibrations and speed fluctuations in the end effector, directly affecting the surface quality of the machined material (such as producing vibration marks) or process consistency (such as uneven glue lines). Third, insufficient adaptive capability of the control system. Existing solutions mostly use preset trajectories combined with fixed-gain PID control or simple filtering, which cannot simultaneously cope with time-varying visual interference, dynamic mechanical vibration, and nonlinear system dynamics. They suffer from problems such as response lag, limited vibration suppression effect, or poor robustness, making it difficult to meet the stringent requirements of high-precision circular motion for trajectory smoothness, continuity, and stability. Summary of the Invention

[0004] This invention discloses a method for controlling the circular motion of a robot based on vibration suppression, comprising the following steps:

[0005] S1. Visual Servo Trajectory Recognition: Images of the robot's work scene are acquired, and the real-time images are quickly initially matched with the reference template using the multi-scale FFT phase correlation method; edge feature points are extracted, and the improved RANSAC algorithm is used to calculate the real-time geometric parameters of the circular trajectory, while outputting the proportion of inliers representing the confidence of the current fit.

[0006] S2. Vibration monitoring and trajectory adaptation: Acceleration signals are collected, filtered, and separated to calculate real-time vibration energy indices and dominant vibration frequencies; based on the vibration energy indices and their changing trends, the angular velocity commands used for planning circular motion are dynamically adjusted.

[0007] S3. Adaptive Composite Control: Generate a reference trajectory using the real-time geometric parameters and angular velocity commands obtained in the preceding steps; design a composite controller whose total output torque includes a feedforward term, an adaptive sliding mode control term, and an active damping term.

[0008] Specifically, the feedforward term is calculated based on the robot dynamics model and includes a coupling compensation term for the additional dynamic effects caused by the dynamic adjustment of angular velocity.

[0009] Specifically, the adaptive sliding mode control term is τ smc =J a T [K(t)·σ(s)+D s ·s],σ(s)=tanh(s / ζ), where s is the sliding surface, ζ is the boundary layer thickness, and D s Let K(t) be the sliding mode damping matrix, and J be the sliding mode gain matrix. a It is a Jacobian matrix.

[0010] Specifically, the active damping term is ,Da(t,f v )=D0+K d ·V(t)·I n×n ·β(f v ), Where D0 is the basic damping matrix, I n×n K is the identity matrix. d Let f0 be the damping gain, f0 be the system's natural frequency, and σ be the damping gain. f For frequency selection, the bandwidth parameter is chosen, γ is the weighting coefficient, V(t) is the vibration energy index, and f v The dominant vibration frequency.

[0011] Specifically, in step S1, the improved RANSAC algorithm includes dynamically adjusting the number of samplings based on the proportion of interior points.

[0012] Specifically, in step S2, the calculation formula for the dynamic adjustment angular velocity command is as follows: , where ω d The rated angular velocity is V(t), the vibration energy index is K. ω V is the angular velocity gain coefficient. thr is the vibration threshold, and sig is the Sigmoid function.

[0013] Specifically, the sliding mode gain matrix K(t) = diag(k1(t), ..., k n (t)), whose diagonal element k i (t) represents the gain of the i-th control channel, which satisfies γi (t)=γ i0 ·(1+β γ ·(1-η inliers )), where s i For the i-th component of the sliding surface, δ i γ is the attenuation coefficient. i0 As the baseline adaptive gain, β γ η is the visual quality adjustment factor. inliers The ratio is the ratio of interior points.

[0014] Specifically, the damping gain is K d0 Based on the damping gain, β d This is the vibration energy adjustment coefficient.

[0015] Specifically, the coupling compensation term τ coupling The calculation formula is: Where q is the joint position vector, K c For scaling gain, M is the robot inertia matrix, and J is the scaling gain. a Let R be the Jacobian matrix, R be the fitting radius, and ω be the Jacobian matrix. a For adaptive angular velocity, θ a (t) is the adaptive phase, Δω(t) = ω d -ω a .

[0016] Beneficial Technical Effects: The solution of this invention significantly improves the robustness and versatility of circular trajectory sensing under strong interference environments, achieving stable and accurate identification of various circular trajectories such as workpiece edges, preset marking lines, or actual processing paths, providing reliable position feedback for control; it realizes active vibration suppression from "sensing" to "planning," innovatively using vibration monitoring information for online optimization of circular motion trajectories, dynamically adjusting angular velocity through vibration energy and trends, reducing the excitation of system resonance from the source of motion commands; it designs a universal controller integrating feedforward compensation, adaptive sliding mode, and frequency damping, introducing sensing quality as the basis for adjusting adaptive gain, automatically enhancing control robustness when sensing signals degrade, and precisely suppressing the resonant frequency identified in real time, effectively addressing structural vibration problems under different processing loads; it effectively balances tracking accuracy, response speed, and motion stability, enabling the system to smoothly suppress high-frequency jitter while ensuring high-precision trajectory tracking, achieving a unity of accuracy, robustness, and stability. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the robot's circular motion.

[0018] Figure 2This is a flowchart of a robot circular motion control method based on vibration suppression according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Figure 1 The core components and spatial relationships of the robot's circular motion are clearly presented. The robot body drives the end effector (such as welding gun, grinding tool, glue spray nozzle, etc.) through a multi-joint linkage structure. The motion trajectory of the end effector is a preset or visually recognized target circular trajectory. The motion plane can be set according to the actual operation scenario (such as weld path, processing surface).

[0021] like Figure 2 As shown, this invention discloses a method for controlling the circular motion of a robot based on vibration suppression, which specifically includes:

[0022] I. Visual Servo Trajectory Recognition

[0023] Images of the robot's work scene are acquired, and the Canny edge detection and geometric fitting method is used to output the weld seam trajectory parameters, including the center point p, in real time. c =[x c ,y c ,z c ] T The radius R is used as the input to the control module.

[0024] (I) Multi-scale FFT phase correlation fast localization: A Gaussian pyramid is constructed for the reference template image and the real-time image, with 4 layers and a downsampling factor of 2. The FFT phase correlation is calculated layer by layer from the top to the bottom to obtain a global displacement estimate and initially locate the weld area. This achieves millisecond-level fast and robust initial localization of the weld area, avoiding feature matching failures or excessive time consumption caused by rapid and large-scale image translation during the operation, and providing an accurate search starting point for subsequent fine processing.

[0025] (II) Direct Edge Extraction and Geometric Fitting: Adaptive Canny edge detection is applied to extract the weld contour in the initially located weld area, and a robust circle fitting is performed using an improved RANSAC algorithm. This accurately extracts the true geometric parameters of the weld edge and outputs an index reflecting the current fitting reliability. Specific steps include:

[0026] 1. Sampling strategy: Select a combination of points with uniform spatial distribution;

[0027] 2. Interior point determination: Calculate the geometric distance from all edge points to the current circle model, where the i-th edge point ([x...)...) i ,y i ,z i ] T The geometric distance d i for d i Points smaller than the inner point distance threshold are considered inner points. The inner point distance threshold can be set to 1-2 pixels.

[0028] 3. Adaptive Iteration Count: The number of sampling iterations is dynamically adjusted based on the proportion of interior points: N = log(1-b) / log(1-(1-ε)). 3 ), where b is the confidence level and ε is the outlier rate estimate;

[0029] 4. Model Refinement: Fit the optimal circle parameters using the least squares method with all interior points. ;

[0030] 5. Output the final center point p c =[x c ,y c ,z c ] T And the radius R, and output the final interior point ratio η. inliers , for subsequent control.

[0031] II. Vibration Monitoring and Trajectory Adaptation

[0032] (I) Vibration monitoring and feature extraction

[0033] 1. The end-effector inertial measurement unit (IMU) acquires acceleration signals a m The rigid body acceleration 'a' is obtained through a low-pass filter. lowpass Calculate the vibration energy index:

[0034] , among which, T w The analysis window length is defined by t, where t represents the current time and τ is the integral variable. The vibration energy V(t) simultaneously affects the angular velocity adjustment and damping gain of the subsequent trajectory adaptive module, forming a dual vibration suppression mechanism.

[0035] 2. The dominant vibration frequency f is extracted in real time using short-time Fourier transform. v Specifically, for the preprocessed acceleration residual signal a res =a m -a lowpassA sliding Hanning window is applied, and a fast Fourier transform is performed window by window to calculate the frequency component with the highest energy in the spectrum at each time moment, which is taken as the dominant vibrational frequency f at that moment. v It is used for frequency-weighted control of the subsequent damper.

[0036] The end-effector acceleration signal contains both rigid body motion of the robot and harmful high-frequency vibrations, which need to be separated and quantized. This invention calculates the scalar V(t) characterizing the jitter intensity and the critical frequency f that may induce resonance. v This provides a clear input for subsequent adaptive control.

[0037] (ii) Adaptive adjustment of trajectory shape

[0038] 1. Based on visual output parameterization of desired trajectory p d (Assume the weld is on the XOZ plane, y=0):

[0039] Wherein, where ω d Φ0 represents the rated angular velocity and the initial phase.

[0040] 2. Dynamically adjust the angular velocity according to the vibration index, and the adjusted angular velocity ω d for:

[0041] Among them, K ω V is the angular velocity gain coefficient. thr The vibration threshold is denoted by sig, where sig is the Sigmoid function.

[0042] When the system vibrates, continuing to move at the original constant angular velocity may exacerbate the vibration (resonance). This step intelligently and smoothly adjusts the movement speed according to the vibration trend, actively avoiding vibration from the planning command level, making the movement more stable.

[0043] 3. Generate adaptive trajectory: , where θ a (t) represents the adaptive phase, specifically... .

[0044] Furthermore, when the visual-control bias e vis-ctrl If the angular velocity is too large, it can be further reduced. The safe angular velocity after reduction is ω. a safe (t)=ω a (t)·exp(-λe vis-ctrl ), where λ is the attenuation coefficient, and the vision-control deviation is the deviation between the visual trajectory and the control trajectory. Specifically, it can be represented by the deviation between the center and radius of the visual recognition circle and the center and radius of the control circle, i.e., e vis-ctrl =||p c vis -pc ctr l||+|R vis -R ctrl |

[0045] III. Adaptive Composite Control

[0046] This step uses adaptive sliding mode control as its core framework, specifically:

[0047] Sliding surface design:

[0048] 1. Define the task space tracking error: e(t) = p a (t)-p(t), where p a p(t) is the adaptive trajectory, and p(t) is the actual end position.

[0049] 2. Design the sliding surface: , where Λ is a positive definite diagonal matrix that determines the error convergence rate.

[0050] (II) Control Law Design: The control torque consists of three parts τ ff τ smc τ damp Specifically:

[0051] 1. Feedforward term (model compensation):

[0052]

[0053] Where M is the robot's inertia matrix, C is the Coriolis force matrix, g is the gravity vector, q is the joint position vector, and v a For the velocity of the adaptive trajectory, J a Let τ be the Jacobian matrix under the current configuration. coupling For coupling compensation terms, K c For dimensionless scaling gain, Δω(t) = ω d -ω a .

[0054] τ ff It accurately compensates for known model forces, significantly reduces the burden on the feedback controller, and improves tracking accuracy and system bandwidth.

[0055] 2. Adaptive sliding mode control term:

[0056] τ smc =J a T [K(t)·σ(s)+D s ·s], where σ(s) is the boundary layer function, using a hyperbolic tangent function to reduce chattering, σ(s) = tanh(s / ζ), where ζ is the boundary layer thickness, Ds To increase the system's damping characteristics, D is the sliding mode damping matrix. s (t)=D s0 ·(1+α d ·H(e vis-ctrl -e thr )), D s0 The basic sliding mode damping matrix, α d H is the amplification factor, and H(·) is the step function. When the vision-control deviation exceeds the control threshold e thr That is, e vis-ctrl >e thr When the sliding mode damping increases, K(t) is the sliding mode gain matrix, K(t) = diag(k1(t), ..., k n (t)), whose diagonal element k i (t) Adjust online based on error changes: Where k i (t) represents the gain of the i-th control channel, s i The attenuation coefficient δ is the i-th component of the sliding surface. i >0, adjustment rate γ i (t) Adaptively adjusts γ based on visual quality indicators: i (t)=γ i0 ·(1+β γ ·(1-η inliers )), where γ i0 As the baseline adaptive gain, β γ β is the visual quality adjustment factor. γ >0, when visual quality decreases (η) inliers When γ decreases, i (t) increases, thereby accelerating the adaptive gain k i Increase the growth rate of (t) to improve the robustness of the system.

[0057] This adaptive mechanism ensures that the control gain is increased when the model uncertainty and disturbance are large, while the gain is reduced to reduce chattering when the model is stable. This ensures stable tracking under model uncertainty and general disturbances, and automatically enhances the control stiffness to prevent error divergence when the visual quality deteriorates.

[0058] 3. Frequency-adaptive damping term:

[0059] Where the adaptive damping matrix is: Da(t,f) v )=D0+K d ·V(t)·I n×n ·β(f v ), frequency weighting function β(f v Enhance damping near the dominant vibration frequency: Where D0 is the basic damping matrix, I n×n Let f0 be the system's natural frequency and σ be the identity matrix. f Choose the bandwidth parameter for the frequency, where γ is the weighting coefficient.

[0060] Adaptive adjustment of damping gain: K d0 Based on the damping gain, β d β is the vibration energy adjustment coefficient. d >0. When the vibration energy V(t) increases, the damping gain K d (t) increases linearly, thereby enhancing the damping effect.

[0061] Robotic mechanical structures often exhibit resonance points, which traditional fixed damping either fails to adequately suppress or negatively impacts normal motion response. This invention provides additional damping precisely matched to the current vibration intensity and frequency, effectively suppressing high-frequency jitter while minimizing its impact on tracking performance.

[0062] Sliding mode control term τ smc As the primary feedback control, it provides robust tracking capability, with the feedforward term τ ff Compensation for known model dynamics, reducing the burden of sliding mode control, damping term τ damp It is specifically designed to suppress high-frequency vibrations, complementing the low-frequency robustness of sliding mode control. This allows the sliding mode control to automatically adjust under different operating conditions, avoiding chattering or insufficient response caused by fixed gain.

[0063] In the several embodiments provided by this invention, it should be understood that the disclosed methods can be implemented in other ways. For example, the embodiments described above are merely illustrative. It will be apparent to those skilled in the art that this invention is not limited to the details of the above exemplary embodiments, and that it can be implemented in other specific forms without departing from the basic characteristics of the invention. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this invention and not to limit it. Although this invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this invention without departing from the spirit and scope of the technical solutions of this invention.

Claims

1. A method for controlling the circular motion of a robot based on vibration suppression, characterized in that, Includes the following steps: S1. Visual Servo Trajectory Recognition: Images of the robot's work scene are acquired, and the real-time images are quickly initially matched with the reference template using the multi-scale FFT phase correlation method; edge feature points are extracted, and the improved RANSAC algorithm is used to calculate the real-time geometric parameters of the circular trajectory, while outputting the proportion of inliers representing the confidence of the current fit. S2. Vibration Monitoring and Trajectory Adaptation: Acceleration signals are acquired, filtered, and separated to calculate real-time vibration energy indices and dominant vibration frequencies; based on the vibration energy indices and their changing trends, the angular velocity command ω used for planning circular motion is dynamically adjusted. a (t), its calculation formula is: , where ω d The rated angular velocity is V(t), the vibration energy index is K. ω V is the angular velocity gain coefficient. thr The vibration threshold is denoted by sig, where sig is the Sigmoid function. S3. Adaptive composite control: Generate a reference trajectory using the real-time geometric parameters and angular velocity commands obtained in the preceding steps; Design a composite controller whose total output torque includes a feedforward term, an adaptive sliding mode control term, and an active damping term; The adaptive sliding mode control term is τ. smc =J a T [K(t)·σ(s)+D s ·s],σ(s)=tanh(s / ζ), where s is the sliding surface, ζ is the boundary layer thickness, and D s Let K(t) be the sliding mode damping matrix, and J be the sliding mode gain matrix. a It is a Jacobian matrix; The active damping term is D a (t,f v )=D0+K d ·V(t)·I n×n ·β(f v ), Where D0 is the basic damping matrix, q is the joint position vector, and I n×n K is the identity matrix. d Let f0 be the damping gain, f0 be the system's natural frequency, and σ be the damping gain. f For frequency selection, the bandwidth parameter is γ, which is the weighting coefficient, and f v The dominant vibration frequency.

2. The method according to claim 1, characterized in that, The feedforward term is calculated based on the robot's dynamics model and includes a coupling compensation term for the additional dynamic effects caused by the dynamic adjustment of angular velocity.

3. The method according to claim 1, characterized in that, In step S1, the improved RANSAC algorithm includes dynamically adjusting the number of samplings based on the proportion of interior points.

4. The method according to claim 1, characterized in that, Sliding mode gain matrix K(t) = diag(k1(t), ..., k n (t)), whose diagonal element k i (t) represents the gain of the i-th control channel, which satisfies γ i (t)=γ i0 ·(1+β γ ·(1-η inliers )), where s i For the i-th component of the sliding surface, δ i γ is the attenuation coefficient. i0 As the baseline adaptive gain, β γ η is the visual quality adjustment factor. inliers The ratio is the ratio of interior points.

5. The method according to claim 1, characterized in that, Damping gain is K d0 Based on the damping gain, β d This is the vibration energy adjustment coefficient.

6. The method according to claim 2, characterized in that, Coupling compensation term τ coupling The calculation formula is: K c Here, M is the scaling gain, R is the robot inertia matrix, and θ is the fitting radius. a (t) represents the adaptive phase, Δω(t) = ω d -ω a .

Images