Actionable topological mapping for navigation using nearby objects

In this paper, we propose a mapping and navigation method for mobile robots with low computational resources and limited memory capacity. The proposed navigation method is based on topological mapping of visual features for its compact representation and robustness in localization. In order to improve the localization accuracy and minimize the memory requirement, we propose the use of visual features from nearby objects. This paper presents a new topological map, called an actionable topological map, which is constructed using visual features from nearby objects and motion information between places in the map. By incorporating motion information, we make it easier for a robot to navigate using the compact representation of a topological map. The proposed method is suitable for light-weight, low-cost robotic platforms due to its low computational and space requirements. We demonstrate the effectiveness of our approach in experiments using an inexpensive, off-the-shelf robotic platform.