Yeongkwon Choe, Yejun Kim, Jaehyuck Cha, and Chan Gook Park, Department of Aerospace Engineering / Automation and Systems Research Institute, Seoul National University, South Korea

View Abstract Sign in for premium content

Abstract:

This study demonstrates the feasibility of a loosely coupled LiDAR-visual-inertial positioning system for the approach phase of a lunar landing sequence through an experiment on a scaled lunar terrain model. A digital elevation model obtained in advance by lunar reconnaissance orbiters is generally of lower resolution than the point clouds obtained by a lander-mounted LiDAR (light detection and ranging) when close to the ground. The resolution mismatch gets worse as the lander approaches the ground. To solve the problems of initial registration with a sparse map and the deepening of resolution mismatch, we have developed a LiDAR-based absolute positioning system using an ensemble Kalman filter. However, since a flash LiDAR, commonly used in lander navigation, is available at relatively very low altitudes, an additional navigation system that can be used before reaching the altitude is required. Therefore, this paper proposes a loosely coupled system that combines the LiDAR-based positioning system with a visual-inertial odometry system to alleviate error drift. This paper also conducts experiments using a 1:200 scale terrain model to confirm the feasibility of the proposed approach. The experimental results based on the actual lunar digital elevation model, NASA/JAXA SLDEM2015 (60 m-resolution), show that the proposed system is capable of positioning with an error of 5.21 m (1-sigma).