Mobile robot localization in dynamic environments using dead reckoning and evidence grids

Abstract

Dead reckoning provides a simple way to keep track of a mobile robot's location. However, due to slippage between the robot's wheels and the underlying surface, this position estimate accumulates errors over time. In this paper, we introduce a method for correcting dead reckoning errors by matching evidence grids constructed at different times. A hill-climbing algorithm is used to search the space of possible translations and rotations used to transform one grid into the other. The transformation resulting in the best match is used to correct the robot's position estimate. This technique has been tested on a real mobile robot and has demonstrated robustness to transient changes (moving people) and lasting changes (rearranged obstacles) in dynamic environments.