Sonar Localisation for Mobile Robots: A Model-Based Approach
Résumé
We advocate a probabilistic, model-based approach to sensing, using the sonar localisation of an autonomous robot vehicle as an example. Our method stresses the thorough mathematical modelling of the sensing process, and in particular the careful statistical modelling of likely error sources and failure modes. We claim that reliable sonar localisation can be achieved in this way, by combining sensor physics with robust probabilistic data fusion techniques.