Gradient-free sensor-based navigation of a nonholonomic robot for source seeking in cluttered environments

Abstract

A nonholonomic Dubins-car like mobile robot travels with a constant speed in a plane cluttered with arbitrarily shaped (possibly maze-like) obstacles. The workspace hosts an unknown scalar field; the sensors give access only to the field value at the robot's current location and to the current distance from the robot to the nearest obstacle. A new navigation strategy is proposed that autonomously drives the robot to the field maximizer through the obstacle-free part of the plane. This is demonstrated by a mathematically rigorous global convergence result and is confirmed via extensive computer simulations. The proposed navigation method does not employ gradient estimation and is non-demanding with respect to both computation and motion.