End Contact Force Control And Peg-in-hole Assembly Of Robot Without Force/Torque Sensor

Robots are widely used in industrial production and are gradually replacing workers in traditional processing operations such as grinding,welding,and assembly.During assembly,the robot end effector clamps the workpiece and makes rigid contact with the external environment,thereby forming a closed-loop structure and generating interaction forces.Excessive contact force will cause damage to the robot,workpiece and the external environment.Therefore,it is necessary to control the contact force at the end of the robot during processing.The classic force control schemes rely on force/torque sensors to obtain the force information at the end of the robot.Since the force/torque sensor is expensive,the integrated application costs of these solutions are relatively high.For a class of applications that require relaively low force control accuracy,in order to reduce the cost of robot integrated application and enhance the usability of robots,this paper studies the robot end contact force control scheme without force/torque sensor,and applies it to non-precision peg-in-hole assembly.The key to the success of this scheme is to design a robot end contact force estimation algorithm with a certain accuracy to replace the force/torque sensor.This requires a well understanding of the changing law of the joint torque when the end of the robot is subjected to force,that is,the robot dynamic model needs to be studied in depth.Based on the accurate and reliable force estimation algorithm and the analysis of peg-in-hole assembly mechanism,a better assembly strategy and the appropriate position and orientation adjustment control algorithm can be proposed,which finally help realize the robot peg-in-hole assembly without force/torque sensor.Therefore,this paper studies the following five aspects in turn.Firstly,in order to obtain the changing law of the joint torque during the robot movement,the robot dynamic modeling is carried out by Newton-Euler method.The identification of dynamic minimum parameter set is completed with the help of Fourier series excitation trajectory and least square method,and a joint torque Kalman filter method is proposed to remove part of the noise.The detailed derivation of the robot minimum dynamics parameter set is given,the method of simplifying the dynamic equation is discussed,and the calculation efficiency of the dynamic equation before and after simplification is compared.Furthermore,the inertia matrix,Coriolis force and centrifugal force matrix,and the gravitational torque term are sorted out from the linear dynamic equation about the minimum parameter set,and the intuitive expression of the robot rigid body dynamic equation is obtained.Secondly,chaos theory,regression fitting and neural network methods are combined to compensate the robot's low-speed motion nonlinear dynamics,and the correction of the dynamic equation of the low-speed robot is completed.With the help of chaos theory,the nonlinearity of the low-speed single joint rotation experimental data is analyzed,and the feature quantities related to the law of joint torque fluctuation are determined.By combining the phase space reconstruction method with curve fitting and neural network,a phase space reconstruction neural network is proposed to compensate the robot's low-speed motion nonlinear dynamic characteristics,which makes the dynamic equation suitable for the low-speed motion scene of contact and peg-in-hole assembly.Then,based on the generalized momentum theory,a disturbance observer is proposed to estimate the force acting on the end of robot,and the force estimation algorithm is calibrated and compensated with the help of machine learning theory.Since there are unmodeled dynamics stimulated by the contact between the robot end and the external environment,the disturbance observer has estimation error.In order to improve the accuracy of the force estimation algorithm,a convolutional neural network supervised learning method with robot joint variable matrix as input is proposed for the force estimation algorithm calibration.Next,a hybrid sliding mode impedance control method is proposed,whose stability is proved by a Lyapunov function and the torque adjustment is converted into position adjustment by numerical solution.The proposed force controller is combined with the force estimation algorithm and iterative control method to realize the constant force tracking control on the plane workpiece and the curved workpiece,which verified the feasibility of the robot end contact force control without force/torque sensor.Finally,the assembly mechanism of the clearance fit between the ball-end round shaft and the chamfered round hole is analyzed in detail,and the assembly strategy and fuzzy PD shaft orientation adjustment algorithm are proposed to realize the robot peg-in-hole assembly without force/torque sensor.The geometry and force relationship of the contact between shaft and hole are discussed,and the two-point contact extreme state of the shaft and hole is focused on,and the force and moment equilibrium equation is derived.A theoretical study is made on the two-point contact state where the axis projection of the shaft does not fall on the diameter of the hole.The relationship among the shaft and hole's tolerance,their center distance and the angle between their axes is obtained by simulation and numerical analysis.According to the assembly mechanism,a peg-in-hole assembly strategy is established and a shaft orientation adjustment algorithm based on fuzzy PD control is proposed.Combined with the re-calibrated force estimation algorithm in the peg-in-hole assembly application scenario,the robot peg-in-hole assembly without force/torque sensor is finally realized.