Coverage Path Planning for Autonomous Road Sweepers in Obstacle-cluttered Environments

Autonomous road sweepers are an enabling technology for addressing cleaning tasks in public open spaces. In this paper, we investigate the sweeping of an obstacle-cluttered environment (e.g., a parking lot) using autonomous road sweepers. The challenge of this problem lies in the fact that the sweeper usually has a large turning radius and a small coverage radius. With a small coverage radius, the sweeper must operate near obstacles to maximize its coverage, making the coverage path sensitive to the mapping uncertainty. Due to a large turning radius, enough buffer should be left around obstacles, and the planned coverage paths should be dynamically feasible. To this end, we propose a novel coverage path planning (CPP) algorithm that explicitly accounts for the mapping uncertainty and vehicle dynamics, which includes three steps. First, the map is post-processed with morphological operations and convexification to reduce uncertainty in the map. Based on the post-processed map, the boustrophedon cellular decomposition is then modified to generate path segments that cover the sweeping area considering vehicle dynamics. Finally, a generalized traveling salesman problem is formulated and solved to connect the path segments for a CPP path with the minimum length. Through experiments in a parking lot at Isuzu Technical Center of America using an Isuzu VL22 truck, the effectiveness of the proposed approach is verified.