Developing a Virtual Interactive Tennis Game by Applying Learning Algorithms

Computer games are one of the most popular entertainments. More than a good graphics or sound, it is the intelligent behavior of game characters which makes them closer to reality. In this project, we have tried to make behavior of game characters close to human behavior by applying different approaches of artificial intelligence. For this, the environment of tennis game is used to simulate and test intelligent algorithms. Tennis game environment is a continuous 3D environment which makes decision making of the intelligent agents more complicated. In order to get close to behavior of a professional tennis player, offline and online learning algorithms are used for training intelligent agents. In the proposed method, neural networks are used for offline learning and a combinative method of learning automata and RBF network has been used for online learning. As reported in simulation results, the intelligent agent can improve game level during each match and learns from eachgame result for future matches.