Research On Compliance Peg-in-hole Assembly Method Of Industrial Robot

With the development of robot technology,industrial robots have been greatly used in the assembly tasks.However,due to the very complicated assembly process,the actual environmental noise interference,the robot's own positioning accuracy and the limitations of visual technology,etc.,the robot's high-precision compliance assembly is still challengeable.Taking the typical assembly scenario of peg-in-hole assembly as an example,this thesis explores the challenges during peg-in-hole assembly tasks.This thesis mainly analyzes the search and insertion phases of peg-in-hole assembly,and proposes some algorithms for peg-in-hole tasks.The main contributions of this thesis are as follows:1)A robust and stable compliant assembly experiment platform based on force-sensing feedback is established,and a force sensor calibration method that comprehensively considers sensor zero drift,sensor installation angle,load gravity,load center of mass,etc.is proposed.Use the sensor data of more than 3 robot poses identify the above parameters and eliminate the effect of the sensor zero and load gravity on the force perception so that the accurate forces and moments on the end of the robot can be obtained.Finally,The accuracy of the calibration method was verified by a robot following motion experiment.This experiment also proved the stableness of the assembly platform and the validity of the compliance control method.2)Aiming at the problem of relative position and orientation deviation between peg and hole due to the limitation of positioning accuracy during the assembly of peg-in-hole,a search strategy based on the combination of multi-layer perceptron(MLP)and force-position hybrid control was designed.This method is mainly divided into two parts:the top search hole trajectory planner and the bottom force-position hybrid controller.The top-level trajectory planner is a multi-layer perceptron model,which is used to determine the direction of the next search hole movement.The bottom force-position hybrid controller is used to receive the upper-level instructions and generate a behavior that interacts smoothly with the environment to ensure the safety and stability of the search process.The experimental results show that the search strategy can effectively reduce the number of search steps and improve assembly efficiency.3)Aiming at the problem that there will still be a slight attitude deviation between the peg and hole after the hole search phase is completed,firstly the ordinary impedance control is used to complete the peg-in-hole task,and then aiming at the problem that the common impedance control is sensitive to the initial position,a peg-in-hole assembly optimization algorithm based on fuzzy Q-learning is proposed.This algorithm can make the underlying impedance controller learn the optimal impedance parameters and then face the complex dynamic assembly process.Experimental result shows that the surface-optimized algorithm can complete the assembly task with a smaller number of assembly steps and smaller assembly contact forces,and it shows a certain degree of robustness and adaptability to the initial position.