UKF-based Active Modeling and Model-reference Adaptive Control for Mobile Robots

Abstract

The Unscented Kalman Filter (UKF) is employed to build an online model for mobile robots by means of joint estimation of states and parameters. Based on this active model, the inverse dynamic control (IDC) is further proposed to make the robot autonomously adaptive to its internal uncertainties. Extensive simulations are conducted with respect to the dynamics of an omni-directional mobile robot. The convergence and tracking ability as well as the uncertainty bound of UKF to estimate time-varying states and parameters are presented. Results of the IDC enhanced by active estimation are also compared with those of a classic PD control with constant gains to demonstrate the effectiveness of the control scheme.