Optimal Real-Time Strategy AI

Within my final year of University, we had to work on a Final Year Project, with the aim to solve a problem. It was a project we had the entire year to work on, with there being a considerable amount of research, implementation, and testing to come up with a solution for the chosen problem. I knew that I wanted to work on a problem related to Artificial Intelligence (AI) as that is a field I am heavily interested in. I decided to combine that desire with one of my favourite game genres: Real-Time Strategy (RTS), where I would develop a project on developing an optimal Real-Time Strategy Artificial Intelligence, which was already quite a big field for AI research.

The project involved a large amount of research, looking into the different areas of AI development and the algorithms that go into them. I researched algorithms from pathfinding to tactical and strategic decision making. The goal was to take a few of these algorithms and see what combination create the best AI’s. I decided to do this testing within StarCraft 2, a heavily popular RTS game which already had a quite popular AI API that could be used to create bots within the game itself.

The whole project was quite an interesting experience to say the least. I do not believe that it was up to a standard I would of liked, as I never did really solve the issue of creating an optimal RTS AI. I still believe it was quite a good project though! With the report definitely being a useful resource for learning about the techniques that go into an RTS AI that someone could easily further expand on.