Towards a consequences-aware emergency landing system for unmanned aerial systems

Abstract

We present an algorithm with which an aerial robot is capable of planning a consequences-aware flight path. The robot is capable of reasoning about possible faults, and which emergency actions are available to it. In this paper, we focus on faults that would force the vehicle to land in a short amount of time, and create a system that allows the vehicle to reason about its ability to execute a safe emergency landing from its current state. Such a system may improve the safety and reduce the economic cost of aerial robots, by allowing them to operate more flexibly whilst still achieving suitable safety. Using an onboard camera, the flying robot is able to identify safe landing spots, and attempt to compute trajectories from a future state to the available landing spots. If no such emergency trajectories are found, the robot must return to the last state at which a safe emergency trajectory was available. Initial in-lab experiments are shown, validating the feasibility of the concept.