Monte Carlo Localization Robust against Successive Outliers

We propose new methods of localization for a robot from surround views and dead reckoning data. Localization is one of very important techniques for autonomous robots, e. g. in RoboCup (autonomous robot succor league). Recently a resetting Monte Carlo localization (ML) method was proposed. But the method cannot deal with successive outliers well. The methods we proposed in this paper are improvements of the resetting ML method and good at dealing with successive outliers.