Autonomous Target Allocation Recommendations

We consider the problem of land vehicles under attack from a number of unmanned aerial systems. As the number of unmanned aerial systems increase, it may become difficult for human operators to coordinate actions across vehicles in a timely manner. In this paper, we study a number of algorithms designed to recommend actions to operators that will maximise the survivability of the vehicle fleet. We present a comparison of several assignment approaches including evolutionary strategies, genetic algorithms, multi-armed bandits, probability trees and basic heuristics. The performance of these algorithms is analysed across six different simulated scenarios. Our findings indicate that while there was no single best approach, Evolution Strategies, Ensemble and Genetic Algorithms were the strongest performers. It was also seen that a number of heuristic algorithms and the multi-armed bandits approach offered reliable performance in a number of scenarios without the need for any training.