Optimal lost-link policies for unmanned aircraft

Loss of communication links between unmanned aircraft and its pilot on the ground have resulted crashes over the years. This paper presents a method for computing optimal lost-link policies for unmanned aircraft conducting surveillance alongside manned aircraft in a wildfire scenario. In this setting, the unmanned aircraft must determine how to trade off the reward gained by continuing the surveillance mission against the risk of a malfunction or collision with the manned aircraft. We extend the model by allowing for the possibility of communication restoration. The unmanned aircraft pilot is modeled as a boundedly rational decision maker to consider possible human error in the future. We compute optimal policies using Monte Carlo simulation. To enhance convergence of the Monte Carlo estimate, a multilevel splitting method is introduced, which results in a 25% higher rate of convergence.