Distributed Localization and Submapping for Robot Formations using a prior map

Abstract

In this paper, we present a distributed algorithm which allows a robot formation to jointly improve a prior stochastic map of the environment where it has to accomplish a commanded task. To reduce the computational cost of the global map updates we exploit the fact that a robot formation work in the same map area allowing us to use conditional independence properties over the state distribution. Each robot maintains its own local and global maps which can be improved with the information received when communications among robots take place. Besides, the robots also exchange their positions in order to maintain the structure of the formation. Simulation experiments were conducted showing that, after the synchronization steps, each robot will have exactly the same information about the map and about the location of the robots at its disposal. Our results also demonstrate the achieved precision and efficiency of the proposed distributed algorithm.