Performance of a real-time filter-based hazard detection algorithm

Landing hazard detection is desirable for science and exploration missions to enable access to more sites of interest and to lower the risk of mission failure at landing. This paper explores the performance of a real-time filter-based hazard detection algorithm developed for use with three dimensional surface data. Test surfaces and flash LIDAR data are used to verify algorithm accuracy relative to known terrain. The current best method found in the literature is a plane fitting method, and this is used as a point of comparison for the filter based method. Both methods produce good results when lander size and acceptable slope and bump height are properly correlated. Computation time of the hazard detection code was measured for both methods, with the filter-based method executing in 34% the time required by the plane fitting method for a realistic 1800×1800 pixel surface. The filter-based hazard detection method was implemented in C++ on a ruggedized commercial off the shelf (COTS) embedded processor and incorporated into a hazardous landing demonstration system. The demonstration system runs in real-time (10 Hz), using either real data from a flash LIDAR system or simulated data, and allows a human pilot to guide a virtual vehicle to a safe landing with the guidance of the hazard detection algorithms.