Turn-based games present a unique set of challenges for an AI, but also open up the door for new techniques that are too expensive in real-time games. GREED CORP in particular is a fun and challenging strategy game on a dynamically collapsing landscape, which made the AI particularly interesting to build.
In this presentation at the Paris Game/AI Conference 2011, you'll hear about the many iterations that the opponent AI went through — ranging from neural networks to minimax, including hand crafted utility networks. You'll also see the tools used on the game that helped polish the decision-making and its impact on the player's experience.
The files used during the presentation are available here:Turn-Based AI in GREED CORP from Neural Networks to Minimax Format: Mp4 Turn-Based AI in GREED CORP from Neural Networks to Minimax Format: MOV (QuickTime) Turn-Based AI in GREED CORP from Neural Networks to Minimax Format: MP3
Giliam is a Senior programmer at Vanguard Games where he designed and implemented the opponent AI for GREED CORP, and is currently working on the graphics, physics and networking for the upcoming GATLING GEARS.