Automatic generation of real time strategy tournament units using differential evolution

Abstract

This paper demonstrates the research results obtained for the application of Differential Evolution (DE) algorithm in a well known real time strategy game, namely Warcraft 3. The DE algorithm is one of the global optimizers that commonly used in solving real-time problems. In this work, the DE algorithm is combined with the conventional feed-forward artificial neural network in optimizing the solutions. The DE acts as an optimization technique used during evolution whilst the neural network operates as the controller in deciding the unit that should be spawned for mimicking the computer AI. The experimentation results show a group of mixed randomized opponent from a larger food limit can be defeated by the generated AI army using DE. Thus, it proofs that the DE used has successfully tuned the neural weights which acts as controllers in this tournament game. Furthermore, the generated controllers could decide the best units that should be spawned in defeating the opponent.