A Methodology to Develop Survivability Maps for Autonomous Aerial Vehicles

As envisioned, autonomous vehicles will operate in very dense, highly uncertain, and increasingly complex environments. Given the risks associated with operating in such a volatile environment, an autonomous vehicle (AV) will not be allowed to operate unless it has a high probability of reaching its destination and has an ironclad guarantee that it will survive regardless of what might happen along the way. To this end, we investigate the underlying principles and mechanisms that contribute to autonomous vehicle survivability, which we define as the likelihood of an AV having at least one trajectory to a valid landing site. We observe that survivability depends on the number of feasible trajectories and the reliability of those trajectories, and propose a back-tracking algorithm to compute feasible trajectories from a given state, as well as a Monte Carlo simulation to assess the reliability of a trajectory in the presence of air traffic uncertainties. This is one of the first studies to develop a comprehensive survivability design framework and visualize the impacts of vehicle lifetime and maneuverability, airport topology, and air traffic on the survivability of an AV in a map. The paper's findings will be useful for strategic and tactical contingency planning, as well as for improving aviation safety.