Cartographie de l'environnement et localisation robuste pour la navigation de robots mobiles

This work presents a deep investigation on the conception of a mobile robot mapping and localization system for indoor environments. Among the contributions, we propose a new robust estimator using nonlinear optimization and robust statistics for mobile robot pose estimation. Further, a decoupled approach for stochastic environment mapping using constraints to reduce the divergence risk has been developed. It aims at respecting the theoretical necessary conditions for consistent map building. The proposed solutions have been validated experimentally on a mobile platform equipped with two exteroceptive sensors: a video camera and a laser rangefinder.