Our AI agents are created to fulfill the role of autonomous agents for playing fighting games (Currently, all our AIs are configured for Street Fighter Zero 3 played on Visual Boy Advance). They are capable of learning and improving themselves (each type of AI has its own learning style) in real-time while playing the game. They utilize the experience they receive to improve their performance in a similar way human players improve themselves.
Normally, most fighting games' AI usually utilizes scripting. Although scripted AI can provide some level of challenges for human players, many players prefer playing against a human opponent because scripted AI is too predictable and repetitive. A human player can provide a greater variety of challenges. Most fighting games in the market now come with an online mode, allowing players to find another human opponent with ease. Slow internet connection, however, still limits the online play. With fighting games, precise timing is crucial. An internet lag can cause the online mode to be unplayable. Therefore an AI still has its place in providing satisfying experience for players.
Our research goal is to create an advanced AI agent for fighting games which can overcome the scripted-AI's drawback. We developed several kinds of AIs. The links on the left provides further details and useful information on AI development and environment used in our research.