Guidance augmentation for reducing uncertainty in vision-based hazard mapping during lunar landing

A new guidance augmentation scheme, which generates information-seeking trajectory adjustments, is shown to produce improvements in hazard mapping during autonomous lunar landing. This approach utilizes previously developed techniques for detecting hazard objects from images and an Extended Kalman Filter recursive estimation framework in order to create an occupancy grid representation of the hazards in the landing area. The guidance augmentation is driven by a model-predictive scheme which uses predictions of map entropy and fuel usage in order to generate information-seeking acceleration commands which are combined with the targeting capability of Modified Apollo Guidance. An overview of the algorithmic steps required to adjust the trajectory and predict mapping performance and fuel costs are presented. Hazard maps generated using the online information-seeking trajectory adjustments show significant improvement over un-adjusted trajectories where the hazard image data collected is only incidental.