Organizational Unit:
Institute for Robotics and Intelligent Machines (IRIM)
Institute for Robotics and Intelligent Machines (IRIM)
2006-03
,
Serrano, Navid
,
Bajracharya, Max
,
Howard, Ayanna M.
,
Seraji, Homayoun
This paper presents a novel, tiered sensor fusion methodology for real-time terrain safety assessment. A combination of active and passive sensors, specifically, radar, lidar, and camera, operate in three tiers according to their inherent ranges of operation. Low-level terrain features (e.g. slope, roughness) and high-level terrain features (e.g. hills, craters) are integrated using principles of reasoning under uncertainty. Three methodologies are used to infer landing safety: fuzzy reasoning, probabilistic reasoning, and evidential reasoning. The safe landing predictions from the three fusion engines are consolidated in a subsequent decision fusion stage aimed at combining the strengths of each fusion methodology. Results from simulated spacecraft descents are presented and discussed.