The feasibility of improving the performance of an on-line hazard estimation system by modifying the descent trajectory during lunar landing is demonstrated. A previously developed technique for inferring the size of rocks based on the measured size of their shadows is the primary hazard information source which feeds an Extended Kalman Filter (EKF) for recursive estimation of discrete hazard objects through the course of the landing. A dense, probabilistic hazard map of the landing zone is created by using the output of the EKF. An overview is given of the algorithmic steps required to extract hazard information from images, estimate the hazard and spacecraft state in the EKF, and create the hazard probability map. Results from a comparison of two trajectories over simulated terrain is given to show that a change in trajectory can produce a reduction in hazard map uncertainty. As a consequence of lower map uncertainty, a greater fraction of the landing zone can be identified as clear of hazards and so potentially suitable for touchdown.

No abstract