A switched systems approach to vision-based tracking control of wheeled mobile robots

Abstract

Conventional methods for image-based guidance, navigation, and control of a wheeled mobile robot (WMR) require continuous, uninterrupted state feedback at all times. However, tracked features may be lost due to occlusions or the trajectory of the WMR. In this paper, a set of dwell time conditions that can be used for trajectory design have been developed to relax the constant visibility constraint, while maintaining the ability to self-localize and track a desired trajectory. The use of a predictor for state estimates when landmark features are not visible helps to extend the time before image feedback of landmark features is required. Using Lyapunov-based switched systems analysis methods, maximum and minimum dwell time conditions are derived for periods when features are visible or not. A simulation is performed with a trajectory formed by Bézier splines to demonstrate a globally uniformly ultimately bounded trajectory tracking result despite intermittent measurements.