Learning-based hybrid control of closed-kinematic chain robotic end-effectors

Abstract

The authors propose a control scheme that combines the concepts of hybrid control and learning control for controlling force and position of a six-degree-of-freedom robotic end-effector with a closed-kinematic chain mechanism, which performs repeatable tasks. The control scheme consists of two control systems: the hybrid control system and the learning control system. The hybrid control system is composed of two feedback loops, a position loop and a force loop, which produce inputs to end-effector actuators, based on errors in position and contact forces of selected degrees of freedom. The learning control system, consisting of two proportional-derivative type learning controllers also arranged in a hybrid structure, provides additional inputs to the actuators to improve the end-effector performance after each trial. Experimental studies performed on a two-degree-of-freedom end-effector show that the control scheme provides path tracking with satisfactory precision while maintaining contact forces with minimal errors after several trials.<>