Research of peg - in - hole assembly for a redundant dual-arm robot based on neural networks

Abstract

Considering that the actual assembly work demands high precision and flexibility for the dual-arm robot at high speed. This paper introduces the design of the dual-arm robot architecture from two aspects of hardware and communication. The work space and constraint relationship of the dual-arm robot during the assembly process are analysed, and the force feedback loop control based on neural networks was proposed to realize the on-line error correction during the assembly process. The force data collected by the six dimensional force sensor will be input to the trained neural network controller. The controller will transform the force error into the position error in the end of the robot arm and output a revised position value. Finally, this paper verifies that the control strategy have higher precision and flexibility than traditional methods through simulation and assembly experiment.