Robot Navigation with PolyMap, a Polygon-Based Map Format

Abstract

Vector-based map formats such as PolyMap are more compact in terms of memory footprint compared with regular occupancy grid map formats. The sparse nature of a PolyMap along with the explicit representation of traversable space and frontiers makes it suitable for mono and multi-robot autonomous exploration. In this paper, we propose a PolyMap-based navigation stack suitable for autonomous exploration. This navigation stack internally relies on a topological graph that is automatically computed from a PolyMap. This graph is critical to significantly reduce the search space when computing a global path compared with traditional occupancy grid maps. We also discuss two optimizations to speed up path planning computation on this graph. Our experiments demonstrate the efficiency of our PolyMap-based navigation stack.