Hybrid Learning-based Control of Closed-Kinematic Chain Mechanism Robot Manipulators

This paper presents a novel hybrid learning-based control scheme for position control of robot manipulators whose structure is based on a closed-kinematic-chain mechanism (CKCM). The developed control scheme integrates two complementary control components: The Feedback Controller and the Learning Controller. The Feedback Controller is designed using linearization about a desired trajectory and a PID control law whose gains are select-ed by a tuning algorithm to guarantee semi-global stability of the closed-loop system. The Learning Controller incorporates PID-type iterative learning strategy to generate additional control inputs to compensate for modeling uncertainties and unmodeled dynamics. By up-dating the control input iteratively from trial to trial, the Learning Controller progressively improves the overall control performance. The effectiveness of the developed control scheme is demonstrated through computer simulations conducted on a six-degree-of-freedom CKCM robot manipulator. Simulation results are presented and discussed to evaluate the tracking accuracy and robustness of the developed approach.