Adaptive Neural Network Based Variable Stiffness Control of Uncertain Robotic Systems Using Disturbance Observer

Abstract

The variable stiffness actuator (VSA) has been equipped on many new generations of robots because of its superior performance in terms of safety, robustness, and flexibility. However, the control of robots with joints driven by VSAs is challenging due to the inherited highly nonlinear dynamics. In this paper, a novel disturbance observer based adaptive neural network control is proposed for robotic systems with variable stiffness joints and subject to model uncertainties. By utilizing a high-dimensional integral Lyapunov function, adaptive neural network control is designed to compensate for the model uncertainties, and a disturbance observer is integrated to compensate for the nonlinear VSA dynamics, as well as the neural network approximation errors and external disturbance. The semiglobally uniformly ultimately boundness of the closed-loop control system has been theoretically established. Simulation and extensive experimental studies have also been presented to verify the effectiveness of the proposed approach.