Computation of a Database of Trajectories and Primitives for Decision-Based Contingency Management of UAVs over Congested Areas

We present a novel approach to enable provably safe contingency path planning in complex high-risk environments within a strict regulatory framework. We decompose the flight planning task between a network of vertiports into an offline and an online phase. This work focuses on the offline phase covering the flight planning, which can be validated to fulfill regulatory boundary conditions and target safety levels before flight. In the online phase the planning task is collapsed to a trajectory selection problem, which is low dimensional and provides only solutions with guaranteed feasibility. The proposed method computes a tree of trajectories with constant time separation of branches and leading to both the destination and diversion vertiports from each fork point. We further extend the planning tree with motion primitives to enable transitions between flight levels and local holding patterns for temporal deconfliction without a change of the lateral flight profile. We demonstrate the method in a realistic urban scenario and provide metrics for memory and time complexity of the algorithm.