Evaluating 3D Polygon Maps for Mobile Robot Localisation

Real time localisation of mobile robots within dense 3D polygon maps is computationally expensive. Computation can be reduced by localising against an appropriate sub-set of polygons, and methods have been suggested as to how to select particular sub-sets. In the absence of detailed experimental results it can be difficult to determine if a sub-set is useful for localization. In this study we propose a method for evaluating sets of polygons for their localization utility within an environment map, or a sub region thereof, based on the rate of the change in expected sensor measurements against change in position. The measure is described, a localization "map" of an example environment presented, and the measure is used to compare various polygon sub-sets. Finally predicted localisation utility is evaluated against actual localisation results.