Understanding dynamic environments with fuzzy perception

Abstract

This paper addresses the problem of dealing with different kinds of dynamic obstacles influencing a place recognition task. We improve an existing approach using independent Marcov chain grid maps (iMac). Furthermore, we add a fuzzy classification to exploit the iMac estimation to refine the likelihood field estimation. We can show that the proposed method increases the performance of place recognition, while still being a compact, interpretable framework.