Vision based approach for active selection of robot’s localization action

The paper presents a mobile robot localization system that integrates Monte-Carlo localization (MCL) with an active action-selection approach based on an aliasing map. The main novelties of the approach are: the off-line evaluation of the perceptual aliasing of the environment; the use of this knowledge to actively perform the localization processes; the use of an improved SIFT feature extractor to aliasing map evaluation and to measure image similarity. Results, obtained in a real scenario using a real robot, show improved performances in the number of steps needed to correctly localize the robot and in the localization error, compared with the classic MCL approach. Also better performances in computational time due to improvements in the vision system are shown.