Learning the Voronoi Centers of a Mobile Robot’s Configuration Space

Partitioning the configuration space of a mobile robot is essential for the robot path planning task. However, most existing techniques either rely upon precise geometrical descriptions of the environment, or are static by nature. In this paper we propose a method by which the robot dynamically builds a Voronoi tessellation of its configuration space. In order to do this, we apply an adaptive -means clustering algorithm to the robot’s free space, while letting the robot explore its environment. The cluster centers, corresponding to the centers of the respective Voronoi cells, are connected into a mesh by using the Delaunay triangulation. Then, we show how the basic robot tasks can be integrated on the resulted graph.