Co-evolution of Antagonistic Intelligent Agents Using Genetic Algorithms

The aim of this paper is to attest the improvement on strategies of intelligent adaptive agents created using genetic algorithms in electronic games. We present an experiment on the use of genetic algorithms to create intelligent adaptive agents which iterates upon the opponent strategy. A predatory food chain was simulated, containing carnivores, herbivores and plants. This simulation uses the approach of a co-evolved asymmetric antagonistic agent population. Because they use each other as part of their environment, they are also able to learn from exhibited behavior after their evolution. Agents are expected to show a satisfactory evolution, analogous to the learning process of an intelligent being.