Posture Estimation and Adjustment for Robotic Precise Peg-in-Hole Based on Fuzzy Variable Admittance Control

Posture adjustment is a key step in robotic precise peg-in-hole assembly. The processes of the rough and fine adjustment are used to achieve the posture adjustment. Firstly, during the robot brings the peg approaching the hole, the force-angle samples of the robot end effector are taken in different states. Multi-layer perceptron (Mlp) model is employed to train the samples. Mlp associates the load force with the deflection angle of the peg from the z-axis of the tool frame efficiently. The robot is guided to rough adjust the peg and make the peg in contact with the workpiece. Secondly, for the peg posture adjustment in uncertain contact states, a fuzzy variable admittance control (FVAC) model is proposed for posture accurate estimation and adjustment. The FVAC exactly extrapolates the contact state of the peg from the workpiece and fulfills the end-surface fitting of them. Finally, comprehensive experiments are conducted with the unknown initial posture of the peg. Results are analyzed through comparison with other common methods regarding adjustment accuracy, operation time, success rate, and so on.