Doing contract AI work for Redshirt, a game developed by Mitu Khandaker was a great experience. The game is now available for purchase on Steam, GOG and direct from the developer. Redshirt has received some really positive reactions from players and press alike, and I’m very proud to have been involved in the project.

The cornerstone of my research, the Integrated Influence Architecture (or I2A) is an endeavour to marry long-term deliberative reasoning with quick and easily calculated reactive reasoning. These two paradigms of AI have traditionally been at opposite ends of the spectrum, but they both had their advantages. As we move towards thinking about our NPCs being smarter, we need them to act in rational and realistic manners that respect their long term plans, but we need to do that using as little processor time as possible.

“Easy Money?” was the first full game I worked on, with my frequent partner-in-crime Heather Decker-Davis. I was responsible for the majority of the gameplay coding, helping to realise Heather’s initial concept and then working with her to tune the design over time. I’m most proud of the procedural maze generation algorithm we created, which was both elegant and very scalable. Although Heather has had to leave the project, it is back in active (slow) development now with Verena Riedl as collaborator. An Alpha build is available here.

The Strathclyde Poker Research Environment (SPREE) was a project I supervised in 2010. There had been significant research into AI for playing poker but from a machine learning point of view, a lot of the work was based on very questionable data collected from online casinos. By its nature, this data was very lacking in detail, so we conceived of SPREE as both a system by which we could gather more complete data, and also use as a testbed for AI players. In both cases it was very successful. The system itself is available on Sourceforge, and a paper about the work was published at the 2011 AISB Symposium on AI and Games.

The Primal project was based around the Brood Wars API, and the spin-off JNIBWAPI projects. It was an attempt to produce a wrapper for some of the more common procedures used when playing Starcraft, so that future AI projects could focus on the high-level strategy of the game, without being concerned with reimplementing basics such as base building, resource collection and so forth. It was also envisioned as a way to abstract a lot of the gameplay for a human player, who could equally focus on calling the shots at a strategic level.