A new neural iterative learning control approach for position tracking control of robotic manipulators: Theory, simulation, and experiment

Abstract

This paper presents a new effective iterative learning control method for repetitive motion-tracking control problems of robotic manipulators. The controller is comprised of two control loops. In the inner loop, a simple proportional-derivative signal is adopted to stabilize the closed-loop system that facilitates design of the outer loop. Tracking control mission is mainly achieved by a novel iterative control signal designed in the outer loop. The effectiveness of the proposed control method is resulted in from a new iterative design where the iterative signal is flexibly structured from both the current and previous information on the iterative axis. To this end, a neural network is developed to estimate the iterative disturbances using information synthesized from the past and present iterations. A proper inherent function is then employed to connect the iterative-based and time-based control signals. Stability of the overall system is analyzed using absolution regression series criteria. The effectiveness and feasibility of the proposed controller are intensively discussed based on the comparative simulation results and real-time experiments obtained from a 6 degree-of-freedom robot.