Coevolution with Danger Zone Levels Strategy for the Weapon Target Assignment Problem

To address the challenging problem of defending a fleet of military vehicles against a swarm of drones, this paper presents an improved coevolutionary algorithm for a simulated counter-UAS game. The algorithm introduces a new chromosome representation that considers the state of the agents in the game based on danger zones, thus making decisions more wisely than the earlier representation: Ordered Preferred Action List (OPAL). In addition, a better configuration of the memory mechanism is used for adding to and choosing from the Hall-of-Fame. Furthermore, using different opponents to evaluate each strategy in the proposed algorithm is explored. Finally, an experimental analysis is carried out, including different scenarios and benchmark algorithms. The results are analysed using four different methods, and the positive effects of the new features are highlighted.