Night-Rider

Night-Rider: Nocturnal Navigation of Autonomous Robots in Street-light Maps

Visual SLAM with common visible-light cameras (e.g., ORB-SLAM, DVO, DSO, etc.) mostly fails in the night environment due to the degenerated illumination condition. To enable a robot to autonomously drive at night, we investigate exploiting streetlights for robot navigation considering that streetlights are almost ubiquitous in the cities. Our framework consists of two modules,  street-light map creation, and visual localization in the created street-light map.  The key issues to be solved include how to create an accurate street-light map (what kind of landmarks can be potentially useful?), how to do the correspondence between two consecutive camera frames in the low-light environment, how to do the data-association between the online visual frames and the offline stored data, etc.

Reference paper: