Research On Robot Polishing Technology Based On Constant Force Control By Reinforcement Learning

Polishing is a very important processing procedure,and the quality of polishing directly influences the product quality of the workpiece.The traditional hand polishing has some problems such as poor accuracy and low efficiency,so it is necessary to introduce automated methods to complete the polishing work.With the development of robotics and artificial intelligence technology,robots have assisted or replaced humans in many fields.Using robots for polishing has broad application prospects.In order to solve the problem of poor environmental adaptability and dependence on environmental parameters in conventional robot constant force control algorithms,a deep reinforcement learning method is introduced to improve the stability and robustness of the force control algorithm.The content of this paper is as follows:1.The material removal model is analyzed and the polishing depth model is established.After analyzing the material removal model,the velocity distribution model and the pressure distribution model in polishing area are established.Based on the velocity and pressure model,the polishing depth of straight and curved polishing paths are modeled,and the optimal polishing path spacing is obtained by simplified model.2.Robot polishing system has been established.The structure of robot polishing system is introduced in detail,and the coordinate frame of robot polishing system is established.Based on the D-H parameter model,the forward kinematics model of the robot is established.The inverse kinematics analysis is carried out and the inverse kinematics model of the robot is established.Then the forward and inverse kinematics model of the robot is verified in the simulation software.Besides,the communication mode between each module in the robot polishing system is detailly introduced.3.The force signal is processed,including the processing of force sensor signal and gravity compensation.In order to obtain the accurate signal of the force sensor,the signal of the force sensor is filtered,and the force model of the force sensor is established.The gravity compensation of polishing tool is carried out,and the effect of the gravity compensation method based on posture of tool is compared with that based on neural network.4.The contact force constraint algorithm in the transition stage and the constant force control algorithm in the polishing process are studied.To address the issue of overshoot of contact force during the transition phase,the contact force is modeled and analyzed,and the contact force is constrained from two aspects:contact velocity and relative contact position.For constant force control,this paper uses SAC algorithm,a deep reinforcement learning method,to set PID parameters in real time.During the constant force control process of the robot,the SAC-PID algorithm outputs the position increment of the robot in the force control direction.The position control is taken as the inner loop and the force control is taken as the outer loop to control the polishing force in real time,so as to realize constant force control in the polishing process.5.The host computer software is designed and the experimental platform is built to carry out experiments on the algorithms in this paper.The interfaces and functions of the host computer software are designed,and the communication modes between the host computer software and other modules are debugged and designed.The gravity compensation algorithm and contact force constraint algorithm are verified on the experimental platform.The SAC-PID algorithm is compared with fuzzy PID algorithm in robot constant force control.The experimental result shows that SAC-PID algorithm has better control precision.