Efficient Coverage of Unstructured Environments.

In this paper, we present a novel solution to mobile robot coverage of unstructured environments. We apply boustrophedon-based planning and use a heuristic to make the minimum sum of altitudes (MSA) decomposition, which computes an optimal exact cellular decomposition, applicable to more complex environments. Contrary to previous approaches, our technique explicitly takes into account different entry and exit points for the obtained cells and hence allows for minimizing inter-region distances in the corresponding traveling salesman problem (TSP) formulation. This is a highly important factor for unstructured environments, which heavily influences the quality of the final plan. Furthermore, we show how our method is applicable to coverage with finite resources. We implemented our planner in ROS and performed extensive experiments in a V-REP simulation environment in various scenarios. Comparisons with a state-of-the-art boustrophedon-based method show that our approach has a significantly lower total coverage time. Additionally, we demonstrate that our system is capable of performing online recharging and replanning in dynamic, crowded environments while obtaining a high coverage percentage. The results of this work are relevant for a variety of real-world applications such as autonomous floor cleaning.