SLAM based shape adaptive coverage control using autonomous vehicles

Abstract

The complete coverage problem requires the full exploration of the entire area, with real-world applications like floor cleaning, lawn mowing, search and rescue, etc. These tasks often do not have the exact a priori knowledge of the target area (e.g., exact shape of the lawn or oil spill area). Thus it is essential that the autonomous vehicle uses on-board sensor feedbacks for exploration so as to: i) dynamically build the a priori unknown environment, and ii) adapt its path in situ. In this regard, it is desired that the autonomous vehicle not only adapts to the obstacles (e.g., landmarks) but also to the shape of the target area (e.g., the lawn) to save time and energy. Since, GPS may not be accessible in all environments, this paper presents a SLAM-based shape adaptive coverage algorithm which assumes that the exact a priori information of the desired workspace is either unknown or only partially known. This algorithm integrates the online information of obstacle and boundary detection with the navigation control. The algorithm is built upon a discrete event supervisory controller which utilizes the concept of multi-resolution navigation to prevent the autonomous vehicle from getting stuck into any local minimum. The efficacy of the algorithm has been validated in a lawn mowing example on the high-fidelity Player/Stage simulator.