Taking Turns in Complete Coverage for Multiple Robots

Coverage is a canonical task where a robot or a group of robots are required to visit every point in a given work area, typically within the shortest possible time. Previous work on offline coverage highlighted the benefits of determining a circular coverage path, divided into segments for different robots (if more than one). This paper contributes a number of significant improvements to the planning and utilization of circular coverage paths with single and multiple robots. We focus on circular paths that exactly decompose the environment into cells, where each obstacle-free cell is covered in a back-and-forth movement. We show that locally changing the coverage direction (alignment) in each cell can improve coverage time, and that this allows for merging bordering cells into larger cells, significantly reducing the number of turns taken by the robots. We additionally present a novel data structure to compactly represent all possible coverage and non-coverage paths between cells in the work area. Finally, we discuss the complexity of global multi-robot assignment of path segments, and present greedy polynomial-time approximations which provide excellent results in practice.