SLAM Based Topological Mapping and Navigation

Abstract

Simultaneous localization and mapping (SLAM) is getting more and more popular in modern robotic navigation systems. Grid map provided by LiDAR SLAM can represent reliable traversable space for global path planning. However, map points of visual SLAM are sparse and noisy, which cannot represent traversable spaces reliably for path planning. This paper proposes a novel and efficient topological mapping approach based on modern SLAM for global path planning. Our approach utilizes not only map points but also trajectories of SLAM to build the topological map. Mapping experiments demonstrate that the topological map is free from the sparsity and outlier problems. Moreover, a navigation system integrating our topological map with a local planner passes all navigation tests with additional obstacles changing original environment.