Optimal Predictive Guidance for Autonomous Hazard Detection and Avoidance
Kento Tomita, Koki Ho · 2024
This paper formulates the guidance optimization problem for autonomous hazard detection and avoidance (HD\&A) for planetary landing. With the physically limited accuracy of terrain sensing and associated uncertainty, future observation and trajectory must be optimized to maximize the probability of safe landing. We propose two reachability-based guidance algorithms based on different approximations. The inner-loop optimization for the reachability computation is bypassed with a neural network-based reachable set evaluator. The numerical experiments with simulated Mars landing scenarios are presented.