Robotic Peg-in-hole Assembly Algorithm Research Based On Compliance Control

Robot plays an pivotal role in the manufacturing industry and is widely applied in the assembly process of product manufacturing.However,the environment of the industrial production site is complex and the assembly objects are changeable.How to efficiently complete the assembly task in the unstructured environment is still a difficulty in research.This paper focuses on the robot peg-in-hole assembly in the unstructured environment,designs a robot peg-in-hole assembly system,realizes the entire assembly process from spatial hole finding to insertion,and proposes a Deep Deterministic Policy Gradient(DDPG)-based variable parameter admittance control algorithm integrated with fuzzy reward mechanism.The algorithm proposed in this paper could improve the assembly efficiency in the insertion stage and the generalization ability of complex states.The main work of this paper is as follows:The robotic peg-in-hole assembly mechanism model is established to provide a theoretical basis for the position control and compliance control in the robotic assembly process.The mapping model between joint angles of UR10 actuator and Cartesian space is established by robot kinematics modeling and simulation verification.The compliant assembly strategy is developed by completing the mechanical modeling and simulation analysis on the insertion stage.The holes finding algorithm based on hole parameter identification is proposed to solve the problems of accurate-positioning in space and efficient hole search.An ellipse parameter identification algorithm based on multi-target detection is designed to realize the parameter identification of multiple assembly holes.Experimental results show that the proposed algorithm has good generalization ability in different photo angles and illumination.Through calibrations containing camera and eye-inhand,the mapping model between the pixel plane and the UR10 actuator is established.The spiral search strategy is introduced to solve the problem that the vision system is difficult to search for holes when there is a large positioning error,and to improve the adaptability of the hole search algorithm for complex industrial environments.A DDPG-based variable parameter admittance control algorithm integrated with fuzzy reward mechanism is proposed to realize the robot’s compliant peg-in-hole assembly operation and improve the assembly speed and success rate on the insertion stage.The DDPG algorithm is used to extract and fuse the contact force information and the robot’s position information to realize the online identification of the admittance controller parameters.By designing a reward system fusing fuzzy mechanism builds a more comprehensive evaluation system of the actions derived from reinforcement learning.It avoids accurate and complex mathematical modeling and further improves the training efficiency of DDPG.By gravity compensation of force sensor,the compliance control algorithm is given accurate force perception ability to improves the assembly quality in the assembly.Experimental platform for robotic peg-in-hole assembly is established,and experiments are designed to evaluate the assembly performance of the DDPG-based variable parameter admittance control algorithm integrated with fuzzy reward mechanism.Experiments show that the proposed algorithm has the advantages of high success rate,fast speed and strong generalization ability in the assembly tasks with different hole diameters.It can effectively reduce the number of steps required for the insertion and improve the assembly efficiency and generalization ability for different pose errors between the peg and hole.