Action selection based on localizability for active global localization of mobile robots

An actions selection mechanism for mobile robots is proposed to accelerate convergence of global localization without increasing computational complexity. In this mechanism, the localizability matrix of every pose is calculated off-line over the apriori probabilistic grid map, and during the phase of global localization with particle filtering, robot actions are selected actively according to maximization of localizability distinctness for all particles. Real robot and real world experiments are implemented for verifying the proposed approach. The experimental results show that the localizability matrix is able to represent the probability distribution of possible observations in the given map accurately, and the actions selection strategy improves the efficiency compared with standard global localization method.