Mobile Robot Planning in Dynamic Environments and on Growable Costmaps

Abstract

We present two applications of path planning oncostmaps: (i) the Probabilistic Navigation Function uses a smoothlyvarying co-occurrence estimation to trade-off collision risk versusdetour lengths, and (ii) a navigation system for exploration ofunknown environment using growable costmaps, interweaving mapping,replanning, and control.By relying on costmaps as a general basis for planning and pathtracking as a generic motion control interface, our approach andimplementation covers a wide range of planners and controllers. Weachieve a relatively general purpose system and introduce a limitedamount of well-defined user-definable heuristics that allow users toadapt the system to a given application. System integration andgenericity is demonstrated by providing three specific implementationsof the planner and controller components, all working withinthe same framework.