Neural network adaptive impedance control of constrained robots

Abstract

Traditional impedance control requires exact dynamic modelling of robots and constraint environments. Although some robust or adaptive control methods are used to handle uncertainties in constrained robot systems, very few of them consider uncertainties of both the constraint and the robot. The desired impedance is also fixed regardless of the types of the constraints. In this article, a model-free neural-network-based adaptive impedance control scheme is developed considering uncertainities of both the robot and the constraint. The neural network and the desired impedance are tuned directly with the impedance tracking and the position tracking errors respectively. The controller guarantees that the desired impedance is achieved asymptotically with good position/force-tracking performances. Simulation results are provided to verify the effectiveness of the scheme.