Extending the AI in a Commercial Game Engine

Exhibitor: Matthew Ota

Supervisor: Marcus Gallagher

Research Group: Complex and Intelligent Systems

Industry Sector: Media / Entertainment

Companies often release the source code for large-scale commercial game engines publicly over the Internet. This source code is usually in C++, written with Microsoft Visual Studio. It is most often intended to be used by hobbyist players who wish to make mods (modifications) to the game. By law these mods must be non-profit, so players release them for free over the Internet for other players to enjoy. This can extend the shelf-life of the game, which is why many game companies support and encourage mods. Another alternative to extending a game is via an in-built scripting language, if it has one.

The first part of this thesis involved an evaluation of 17 commercial game engines. They were evaluated for the ease in which their AI source code could be understood and extended by an average programmer. The chosen game engine was Unreal Tournament 2003.

The second part was to extend the AI portion of the source code for the chosen game engine. The extension would allow all players and bots (computer-controlled opponents) to fly. This involved manipulating the physics code portions of the engine, in addition to fixing AI pathfinding flaws and AI decision logic flaws that were associated with flying. These flaws were inherent in the game engine. The modification process and lessons learned on the operation of the game's AI code were documented so that others could learn from it.

Thesis Document (PDF)

Other Related Files