Challenge Team Interim Report
Team Number: 027
School Name: Clovis High School
Area of Science: Artificial Intelligence
Project Title: Making Them Smart
Progress has been slow, but we have decided the way in which we will achieve the intelligence within our AI engine. We plan to integrate Bayesian Belief Networks into our AI engine along with a few other ideas that are designed to work with and modify the Bayesian Belief Networks as the AI executes. We chose to use Bayesian Belief Networks, mostly, because we thought them more powerful and more useful to our application than other methods, but we partly chose Bayesian Belief Networks because we felt that most other methods are overused. We, however, have very little information gathered on Bayesian Belief Networks. This is due to the lack of good information for free on the internet; so few sites are even up to date or stay in existence long enough to be useful. But this is not the largest problem standing in our way. The largest problem is finding a C++ API that will run these Bayesian Belief Networks. The only C++ APIs I have found on the net actually are not there (bad links). We have, however, found the next best thing: a Windows DLL made by Microsoft themselves. This is a problem since it would only run on a Windows platform. All of these problems, however, will not stop us from trying.
And now about our plans about how to design our AI engine. We plan to set up an engine in which a history of actions is kept for the AI engine's reference, but the end begins to go vaguer and vaguer as the list gets longer, much like a person's memory is more accurate when dealing with the immediate past than the distant past. From this history list, the probability of certain actions of the AI engine change. It will respond to repeated cycles and will try to set up a long term strategy to deal with the player, based partially on the AI personality of the enemy and partially on what the player does. The AI, like the player, has a limited number of actions that it can do at a time. We will be using the Bayesian Belief Networks to produce the actions of the AI based upon these things.