Online complete coverage path planning for mobile robots based on linked spiral paths using constrained inverse distance transform

This paper presents a sensor-based online coverage path planning algorithm guaranteeing a complete coverage of unstructured planar environments by a mobile robot. The proposed complete coverage algorithm abstracts the environment as a union of robot-sized cells and then uses a spiral filling rule. It can be largely classified as an approximate cellular decomposition approach as defined by Choset. In this paper, we first propose a special map coordinate assignment scheme based on active wall-finding using the history of sensor readings, which can drastically reduce the number of turns on the generated coverage path. Next, we develop an efficient path planner to link the simple spiral paths using the constrained inverse distance transform that we introduced the first time. This planner selects the next target cell which is at the minimal path length away from the current cell among the remaining non-contiguous uncovered cells while at the same time, finding the path to this target to save both the memory and time which are important concern in embedded robotics. Experiments on both simulated and real cleaning robots demonstrate the practical efficiency and robustness of the proposed algorithm.