Research On Robotic Fasten Assembly Technology Under Visual Guidance

In recent years,the degree of automation in the manufacturing industry has been increasing,and robots have been used more and more in the field of assembly,but most of them are based on position control.When the assembly task of the assembly object is complicated and the process is deformed,it cannot be completed with high quality.Based on deep reinforcement learning,image processing technology and force control technology,this paper takes low-voltage circuit breaker as the research object and conducts in-depth research on the snap-fit assembly of the upper and lower covers.The robotic snap fit is achieved based on the Vision Guided Rough Positioning and Deep Deterministic Policy Gradient(DDPG)algorithm.The main work of this paper includes:Firstly,design the upper and lower cover snap-fit assembly system for small low-voltage circuit breakers.Based on the D-H parameter table of the seven-joint arm KUKA iiwa7,a kinematic model was established to analyze the assembly process of the miniature circuit breaker.The mechanism of the snap-fit assembly of the upper and lower covers is studied,and the method of combining the force and pose to describe the contact state is proposed,which lays a foundation for the subsequent design of the algorithm structure and strategy.Secondly,according to the structural characteristics of the small circuit breaker lower cover,a visual guidance system with "eyes outside the hand" is designed.First,calibrate the camera,assembly target,and robot arm.Then,by analyzing the internal structure of the circuit breaker,the relative positional relationship between the four riveting holes and the lower cover is obtained.Finally,based on image preprocessing.feature extraction and other algorithms to obtain the pose of the workpiece,and guide the robot arm to the target position,to achieve visual guidance coarse positioning.Then,the depth deterministic strategy gradient algorithm is studied and a flexible snap-fit assembly method based on DDPG is proposed.Combined with the snap-fit assembly process of the circuit breaker,the principle and training process of the DDPG algorithm are mainly studied.The strategy network structure,value network structure,reward function,action constraints and assembly success criteria of the algorithm are designed.A complete snap-fit assembly model is constructed and the model network parameters are optimized through experimental testing.Finally,perform functional realization and performance testing of the robotic snap-fit assembly system under visual guidance.Based on HIKVISION's monocular industrial camera,KUKA iiwa 7 industrial robot,claw hand,workbench and other related components,a video-assembly and DDPG-based snap-fit experimental platform was built and algorithm verification was performed.The results show that the industrial robot can be guided to the target position for the workpiece to be loaded in any pose.The established DDPG network learns assembly skills online and is continually optimized to achieve flexible snap-fit assembly.