Path Planning Using Flexible Region Sampling For Arbitrarily-Shaped Obstacles

In this paper, we propose a path planning algorithm targeting an environment with complex and arbitrarily-shaped obstacles. To tackle this issue, we propose a method that adaptively divides a free region into smaller regions of various sizes. The proposed method uses recursive tree traversal and k-means clustering to create regions of diverse sizes according to the distance of the obstacle. Divided regions represent a large area with a small number of points, unlike the grid map used in the conventional methods. Via two simulations, we validate that the proposed path planning approach is capable of finding a feasible path while substantially improving computational time and memory consumption compare to the existing methods.