Vision based Hazard detection and obstacle Avoidance for planetary landing

Current lunar and planetary missions aim for pinpoint landing accuracy. Landing a set of science instruments at the edge of a particular crater or critical life-supporting supplies in proximity of a lunar outpost is not possible with the kilometer level precision of past. To achieve this level of precise landing, accurate identification of potential hazards (like craters, boulders, steep slopes, etc.) is necessary. In this paper, a robust algorithm for autonomous hazard detection and avoidance (HDA) for a low cost, pinpoint planetary lander, Magnolia-1 is proposed. The algorithm requires no input from classical inertial sensors. The hazard assessment and navigation is done purely by using imaging camera. The primary algorithm incorporates different functions which include (a) Feature selection and hazard identification, (b) Feature transformation and hazard mapping, (c) Safe site selection and (d) Site tracking and retargeting. Implementation of each functional block and simulation results for the algorithm are presented in this paper.