Dynamically expanding occupancy grids

Abstract

Previous evidence grid based robot architectures have used static arrays for representing the environment map. We present a method for dynamically sizing the environment map. Basically, the mobile robot explores and as it explores the map grows in a non-rectilinear fashion. Thus, the area formed as a result of this method may not be a rectangular structure. This approach is flexible and imposes no size or shape constraints on the Occupancy Grid. We are interested only in mapping here, but the architecture we are proposing can also be used for localization and navigation. This method should work well with Markov and Monte Carlo Localization methods, and will have significant advantages over static Occupancy Grid representation methods.