Localization and Mapping of Rover by 3D Laser Scanner and Omnidirectional Camera

Abstract

Hazard detection, simultaneous localization and mapping (SLAM), and terrain classification are crucial for autonomous navigation of the Moon or remote planets. In this paper, we integrate a 360-degree laser sensor (LIDAR) and an omni-directional camera in the co-axial arrangement to achieve the above three goals. The point-cloud information from the LIDAR is used to detect hazardous objects, create a high-precision map of the surrounding environment, and then localize the rover. The visual images from the omni-directional camera are used to add semantic information of the detected objects through the mask recurrent convolutional neural network (Mask R-CNN) technique. The performance of the integrated system was evaluated in a simulated planetary field.