We investigate a class of problems in which a defender dynamically allocates its surveillance assets in anticipation of an attack. The surveillance asset can be an unmanned aerial vehicle (UAV), a radar mounted on a small nonrigid airship, or a patrol oﬃcer.
We study a patrol problem where the defender chooses a patrol sequence among possible attack locations to maximize the probability of detecting an attack. The adversary chooses the attack location to minimize this detection probability. We examine patrol policies that utilize Gittins indices to determine the next location to visit during a patrol. This research leverages rigorous analytical tools from game theory, optimization, stochastic modeling, and simulation, to ultimately generate recommendations for real-time allocation of surveillance assets. Ongoing research will develop robust methods to counter the enemy’s strategy no matter what the enemy does.
Collaborative research with Kyle Lin and Michael Atkinson