It has been pointed out over recent years that so much of AI decision-making is dependent on the information that the AI has available to reason with. This is especially true with regards to spatial awareness. This year, Nick Bamford (Slant Six Games) shows some of the ways that this was addressed in Resident Evil: Operation Raccoon City.
From the GDC site description:
Decision-making in game agents can only be as good as the underlying data used to make those decisions. If the knowledge of the situation is incomplete or erroneous, behavior may look inconsistent or even illogical. In tactical shooters, much of the information that an agent needs involves the terrain and its features. However, processing information about the battlefield can be a difficult and time-consuming process. By showing examples from the upcoming Resident Evil: Operation Raccoon City, this lecture will show how tactical behavior can be improved by better modeling an agent’s understanding of the combat space.
Nick Bamford, Resident AI Expert, Slant Six Games
Nick is a career-long AI engineer with 15 years of industry experience on a multitude of titles of almost every genre possible. He has been impassioned with AI since the 16-bit generations where he did early stand-out work on the use of genetic algorithms in the sports AI of Gremlin Interactive, and flock dynamics on Core Design’s Herdy Gerdy, through to recent work on the first western developed Resident Evil title, Resident Evil: Operation Raccoon City. He graduated with a B.Sc from Nottingham University with post graduate work in group dynamics. He is currently residing in beautiful British Columbia, working as an AI domain expert at Slant Six Games on the Resident Evil franchise and the companies ‘Hexane’ game engine.