FAST MOBILE ROBOT LOCALIZATION USING LOW COST SENSORS

Bayesian filtering is a well known probabilistic filtering method. Its applications to mobile robot localization are very popular, but an active approach to the problem of localization was never presented. An interesting question is: what is the best action that the robot should choose to localize itself in the minimum number of steps? This paper presents the Fast Particle Filtering (FPF) algorithm to select the best action that allows a fast global localization using particle filtering. The appropriateness of our approach is demonstrated empirically using a mobile robot equipped with low cost sonar sensors in a structured office environment. Comparisons with classical Bayesian filtering approaches are also presented to demonstrate the better performances and the lower computational cost of the FPF algorithm.