Research On Accurate Force/Pose Control Methods For Robotic Compliant Manipulation

Robotic compliant manipulation is a hot issue in current research and application fields.Since the contact force may change with the position during the manipulation process,the robot needs to have the control ability of the specified force at the specified position(i.e."specified force at specified position").Obviously,the realization of "specified force at specified position" depends on the accurate control of the force and pose precision of the robot.However,there are some problems in the force and pose control of the robot:(1)when using the contour error to quantitatively evaluate the pose accuracy of the robot,the control effect of the contour error is affected by its estimation accuracy and controller performance;(2)the contouring control based on the motor position feedback is difficult to overcome the influence of the deformation of joints and links on the pose accuracy,so a full closed-loop control of end pose is needed;(3)in the impedance force control,the environment parameters are unknown,and the impedance parameters are difficult to select,and the influence of the residual position tracking error on the force tracking accuracy is less considered,so the force control performance is limited.In view of the above problems,under the robotic control framework with dual force/contour loop,this paper follows the research idea of "semi-closed inner loop contouring control ? full-closed inner loop visual servoing ? outer loop impedance force control".First,the contouring control strategies based on high accuracy contour error estimation,and the visual servoing strategy integrating feature depth estimation and image error metric are proposed to improve the pose accuracy of the robot.Then,on the basis of on-line estimation of environment parameters and adaptive change of impedance parameters,the high accuracy contact force control is realized through the impedance strategy corrected by the residual position tracking error.The main research work of this paper is listed below:(1)Dual sliding mode contouring control based on high accuracy contour error estimation.The contour error of six-axis robot is defined,and a high accuracy contour error estimation method based on second-order Taylor expansion and linear ratio is proposed.On this basis,a dual sliding mode contouring controller is proposed.Experiments on an open six-axis robot show that,compared with the existing best method,the proposed estimation method can reduce the position and orientation contour errors by about 39% and 73%,respectively,and reduce the calculation amount by about 65%;the proposed controller can effectively reduce the control signal chattering,and compared with the traditional sliding mode contouring controller,the position and orientation contour errors can be reduced by about 25% and 27%,respectively.(2)Interpolation-based contour error estimation and component-based contouring control.A method of position contour error estimation based on quadratic polynomial interpolation is proposed,and then the orientation contour error is estimated by using the position estimation results and linear interpolation.By summing the tracking errors and the weighted contouring error components in task space,and assigning the second-order dynamic characteristics to the newly constructed error vector,a component-based contouring controller is proposed.Experiments on an open six-axis robot show that,the proposed estimation method does not need any other curve information except for the reference pose,and compared with the method in(1),the estimation accuracy is similar,but the calculation amount can be reduced by about 37%;the proposed controller can effectively avoid the problem that the contour error does not decrease but increase in the local region of the trajectory in the existing contouring controller.(3)Visual servoing integrating feature depth estimation and K-L divergence-based image error metric.By analyzing the classical perspective camera model,the dynamic equation of the perspective system is obtained.A structure of logarithmic reduced-order feature depth observer is proposed.A visual servoing method using the K-L divergence-based image error metric is proposed.MATLAB/Simulink simulations show that,compared with the existing typical observers,the proposed observer has the comprehensive advantages of global asymptotic convergence,faster convergence rate than exponential structure,less restrictive observability condition and stronger noise robustness;biaxial motion experiment shows that,compared with Kinect V2 sensor,the maximum estimation error of the proposed observer is less than 0.05188 m.Vi SP simulation shows that the proposed servoing method does not need feature extraction,tracking and matching process,and compared with the histogram-based method,the convergence rate can be increased by more than 40%.(4)Adaptive impedance force control considering residual position tracking error.Under the traditional position-based impedance control framework,related factors affecting the force tracking accuracy are analyzed,and then the residual position tracking error of the robot is considered to modify the impedance strategy.The existing adaptive method is adopted to estimate the environmental stiffness and position parameters.On this basis,according to the model reference adaptive theory,the real-time adjustments of the impedance stiffness and damping parameters are achieved.MATLAB/Simulink simulation shows that,the proposed adaptive impedance control method can effectively reduce the force oscillation caused by sudden change of the contact force;contact experiment of the six-axis robot shows that,compared with the case without considering the residual position tracking error,considering the residual position tracking error can reduce the force tracking error by more than 80%.Through the application of robotic grinding and polishing of blade,the effectiveness of the proposed force control method under the actual working condition is further verified.