Coevolutionary Procedural Generation of Battle Formations in Massively Multiplayer Online Strategy Games

This paper presents a coevolutionary genetic algorithm for the development of battle formations in Massively Multiplayer Online Real-Time Strategy games. We consider the context of the game Call of Roma, where the battles are turn based and two sides fight each other, each side involving one or more players. The coevolutionary genetic algorithm takes as input a predetermined battle formation, presented by the defense side, and returns a battle formation adapted for the Attack side. The algorithm aims to maximize the battle performance, which is a subjective concept that varies from player to player. The individual encodes the various characteristics of the fighting heroes. We tested the algorithm over a cloud-computing platform, considering test cases modeled by several active players. The results illustrate that the proposed algorithm is able to find a victorious solution for the Attack team, even when it is under unfavorable conditions.