Mobile robot localization failure recovery.

摘要

Mobile robot localization is one of the most important problems in robotics. Localization is the process of a robot finding out its location given a map of its environment. A number of successful localization solutions have been proposed, among them the well-known and popular Monte Carlo localization method, which is based on particle filters. This thesis proposes a localization approach based on particle filters, using a different way of initializing and resampling of the particles, that reduces the cost of localization. Ultrasonic and light sensors are used in order to perform the experiments. Monte Carlo Localization may fail to localize the robot properly because of the premature convergence of the particles. Using more number of particles increases the computational cost of localization process. Experimental results show that, applying the proposed method robot can successfully localize itself using less number of particles; therefore the cost of localization is decreased.