Research On Lidar-based Hazard Detection Algorithms For Soft Planetary Landing

Landing unmanned automated spacecraft on the planetary body for in-situ analysis or return to Earth laboratories is significant for human's planetary exploration. It's expected future planetary exploration missions will aim at landing a spacecraft in hazardous regions with high scientific potential, thereby requiring the ability to autonomously detect and avoid surface hazards. This paper focuses on hazard detection algorithms using LIDAR sensors.A high-fidelity LIDAR model is firstly built. In order to rectify the terrain data obtained by LIDAR, an approach to construct a symmetrically girded digital elevation map (DEM) from LIDAR terrain data is introduced.To detect hazards of large scales, such as craters, hillocks and so on, the flatness-based regional hazard detection (RHD) algorithm is advanced by establishing a non-linear mapping of the local variance and a flatness factor. To overcome the disadvantage of the flatness-based RHD algorithm that the landing window may be located in local flat regions such as the bottom of a large crater, a local-slope based RHD algorithm is proposed. In this approach, region segmentation techniques are used to extract the flat regions outside all regional hazards as safe regions, while regions surrounded by hazards are marked off as part of the hazards.To meet the real-time requirement of the descend phase, the plane-fitting based local hazard detection algorithm is advanced. In our approach, a modified RANSAC algorithm is used to extract the reference surface from the LIDAR DEM, instead of the LMS algorithm since it is relatively consuming. On another hand, a cockle stairs searching strategy in which the hazard detection processing and landing site searching processing are executed in parallel is applied so that the nearest safe landing site could be output as soon as possible.To validate the proposed hazard detection algorithms, high resolution digital elevation maps of Mars with different terrains and synthetic rocky terrain maps are used to test the detection algorithms. Testing results suggest the proposed algorithms are reliable for regional and local hazard detection.