A Probabilistic Technique for Simultaneous Localization and Door State Estimation with Mobile Robots in Dynamic Environments

Dzintars Avots, Edward Lim, Romain Thibaux, and Sebastian Thrun

Virtually all existing mobile robot localization techniquesoperate on a static map of the environment. When the environment changes (e.g., doors are opened or closed), there is an opportunity to simultaneously estimate therobot's pose and the state of the environment. The resulting estimation problem is high-dimensional, render-ing current localization techniques inapplicable. This paper proposes an efficient, factored estimation algorithm for mixed discrete-continuous state estimation. Our al-gorithm integrates particle filters for robot localization, and conditional binary Bayes filters for estimating the dynamic state of the environment. Experimental resultsillustrate that our algorithm is highly effective in estimating the status of doors, and outperforms a state-of-the-artlocalizer in dynamic environments.

The full paper is available in gzipped Postscript and PDF