Our approach to HDA will use monocular vision to collect hazard information. Cameras are ideally suited for a small spacecraft as it is a low mass and power sensing solution. Computer vision algorithms can detect and match features through the descent image stream, which enables lander localization and can provide a motion constraint on the lander trajectory to improve its navigation. Rock height can also be estimated directly given lander altitude and a known angle to the sun. These two techniques will be extended and adapted for the particular challenges associated with small landers, namely to small hazards that have been of less concern in prior work in this area. Previous work in small hazard detection has been conducted using very high resolution sensors, such as LIDAR, that are too large and power intensive to be flown easily on a small lander with a large payload fraction; work using lightweight sensors has, in general, been more focused on descent navigation relative to large terrain features rather than identifying small hazards.
New research is being conducted in the area of dual-control to improve small-scale mapping accuracy, landing site evaluation, and retargeting performance. Our first goal is to extract two sources of hazard data from the same image: rocks identified by shadow-based detection methods and point features matched through multiple images. By using both information sources, we can exploit the lowest covariance measurements derived from each method to decrease our overall hazard map uncertainty. In the current state-of-the-art HDA systems, the lander trajectory can be recomputed during descent to land at a safer site based on hazard information. However, the data collected for use in this retargeting scheme are largely incidental as the lander trajectory is only dependent on the current target site. We will improve the quality of the hazard data gathered by flying a slightly modified trajectory, designed to not only get the lander to the chosen landing site, but also to actively collect information about the terrain in order to aid in the selection of the best site.
Last modified Tue, 10 Feb, 2015 at 10:32