Localization and Autonomous Exploration for Lunar Rover with Only RGBD Sensors

This study proposes a systematic control framework of lunar rover for localization and path planning to complete extraterrestrial exploration tasks using only RGBD cameras, with lunar Digital Elevation Map (DEM) available beforehand. The framework contains two main blocks, the path planning algorithm and the localization algorithm. Global path and motion instructions are generated by the path planning part. A* algorithm is first used to obtain a global path that keeps an appropriate distance to the impassable area, whereas local trajectory planning generates motion instructions to avoid any collision. The estimation of relative motion is achieved by fusing the pose estimation values of the two RGBD cameras (front and rear ones). To improve the localization accuracy, an absolute localization method is devised using the idea of matching the elevation map projection curve with the real-time captured image curve. Finally, the effectiveness of the algorithm is verified through simulations in Webots.