Cooperative SLAM using MM-Space representation of linear features

This paper presents a multi-robot simultaneous localization and map building (SLAM) algorithm, suitable for environments which can be represented in terms of lines and segments. Linear features are described by adopting the recently introduced MM-Space representation, which provides a unified framework for the parameterization of different kinds of features. The proposed solution to the cooperative SLAM problem is split into three phases. Initially, each robot solves the SLAM problem independently. When two robots meet, their local maps are merged together using robot-to-robot relative range and bearing measurements. Then, each robot starts over with the single-robot SLAM algorithm, by exploiting the merged map.The proposed map fusion technique is specifically tailored to the adopted feature representation, and takes into account explicitly the uncertainty affecting both the maps and the robot mutual measurements. Numerical simulations and experiments with a team composed of two robots performing SLAM in a real-world scenario, are presented to evaluate the effectiveness of the proposed approach.

► SLAM algorithm for multi-robot systems in indoor environments.
► MM-Space representation of line and segments.
► Map fusion technique from robot-to-robot measurements.
► Experimental validation with real robots.