A Cooperative SLAM Framework with Efficient Information Sharing over Mobile Ad Hoc Networks

Book excerpt: As the cost of simple robotic platforms plummets, interest in solutions that explore the cooperation between robotic agents grows. Cooperative techniques enjoy a multitude of advantages over their individual counterparts, namely concerning the solution of easily divisible problems. Simultaneous Localization and Mapping (SLAM) is one of the most researched topics in Robotics. It deals with the problem of endowing a mobile robotic platform with the ability to, with no a priori knowledge of the environment, build a representation of its surroundings and, simultaneously, localize itself in it. The exploration of unknown environments seems to be a particularly and intuitively divisible problem, at least at a human level: we simply divide the team into smaller groups, or even into individuals, which explore different areas in the environment. After exploring, the team reassembles and builds a joint representation of the region. However, this approach gives rise to a new set of problems, such as communication, synchronization and information fusion. This work focuses in the solution of these problems, known collectively as Multi-Robot SLAM. We propose a novel software solution that seeks to provide any system capable of performing single robot SLAM with the ability to efficiently communicate with its teammates and to build a global representation of the environment based on the information it exchanges with its peers. This solution was validated through experiments conducted over real-world data, both on- and off-line, i.e., both using prerecorded and data gathered during the experiment. It has shown acceptable performance, both in scalability and communication efficiency.