Comparison of methods to efficient graph SLAM under general optimization framework

Abstract

Simultaneous Localization and Mapping(SLAM) algorithms can infer the robot's trajectory as well as the map under unknown environment. Robust and time-efficient optimization methods are important requirements for SLAM. There are many algorithms designed for the graph optimization. However, it is hard to select an appropriate algorithm and corresponding software library, due to the difficulty of evaluating algorithms' adaptabilities under various situations. In this paper, we summarize these algorithms under general optimization framework, conduct several sets of experiments to compare these algorithms in three software libraries, and give some suggestions to choose algorithms.