Gradient Guided Search for Autonomous Contingency Landing Planning

The growing reliance on autonomy in uncrewed aircraft systems (UAS) necessi-tates a real-time solution for assured contingency landing management during in-flight emergencies. This paper presents a novel gradient-guided search algorithm for risk-aware emergency landing trajectory generation with a wing-lift UAS loss of thrust use case. This framework integrates a compact four-dimensional discrete search space with aircraft kine-matic and ground risk cost. A multi-objective cost function is employed, combining flight envelope feasibility, optimal descent, and overflown population risk terms. To ensure discrete search convergence, a constrained hypervolume definition is introduced around the destination. A holding pattern identification algorithm is defined to minimize risk during the necessary flight path angle constrained descent to final approach. Planner effectiveness is validated through randomly generated case studies over a region of Long Island, NY under steady wind conditions. Benchmark comparisons with a 3D Dubins solver demonstrate the approach’s improved risk mitigation and acceptable real-time com-putation overhead. Future development will focus on integrating collision avoidance into the discrete search-based landing planner.