Multiple hypothesis tracking for mobile robot localizaiton

Global localization has been studied as a challenging problem in mobile robotics for decades. This paper presents an efficient solution to global localization which includes hypothesis generation, tracking, and elimination process. A combination of a line feature and a point feature can specify the hypotheses of the robot's pose in the map. We keep track of the hypotheses by Monte Carlo localization in which a hypothesis is represented by a set of particles. The false hypotheses are removed when the new hypotheses are generated. Simulations were conducted to prove its performance. The simulation results show that our method outperforms the former methods.