AA 273 Final Project

We study the SE-Sync algorithm and how it can be utilized for pose-graph SLAM. We begin by presenting a historical review of the general SLAM problem and the optimization based paradigm of pose-graph SLAM. We then proceed to carefully define the pose-graph SLAM problem and present an intuitive but still sufficiently complete treatment of the SE-Sync algorithm and how it is applied to solve the corresponding optimization problem.

Our main contribution is an implementation of SE-Sync in a ROS compatible SLAM system obtained by replacing the back-end optimizer of an existing open source implementation. Utilizing a realistic simulation environment, we compare its performance with the original implementation in four different mapping environments of varying size and complexity, and find that SE-Sync performs equally well or slightly better in all cases.

We finally conclude that application of the SE-Sync algorithm in practical SLAM systems seems promising, but that further work is necessary to determine whether or not the improvement is of significant practical importance.