An Experimental Investigation of Cooperative Slam

This paper considers the problem of cooperative navigation and mapping by a heterogeneous team of multiple autonomous underwater vehicles (AUVs). We address a form of cooperative Simultaneous Localization and Mapping (SLAM) in which only one vehicle is responsible for maintaining estimates of the map and poses for each robot. By combining inter-vehicle measurements with observations of the environment made by each vehicle, the result is a better knowledge of the poses of each robot in the group. We present results of an experiment in which three land robots perform cooperative SLAM in a large, ambiguous environment. Comparisons with ground truth are provided.