Monolith Productions’ 2005 shooter F.E.A.R. featured enemy AI that many developers still cite as a high-water mark for tactical combat: soldiers who took cover, laid suppressing fire while teammates flanked, dove through windows, and used grenades to flush the player out. At GDC 2006, lead AI programmer Jeff Orkin explained how it worked in the paper “Three States and a Plan: The A.I. of F.E.A.R.,” and the surprising answer was real-time planning.
Rather than scripting each situation, F.E.A.R. used Goal-Oriented Action Planning (GOAP). Orkin shrank the character finite state machine to “only three states” - Goto, Animate, and UseSmartObject - and moved all the decision logic into a planner that “most closely resembles the STRIPS planning system from academia,” searching for action sequences with A*. As he put it, “a planning system tells the A.I. what his goals and actions are, and lets the A.I. decide how to sequence actions to satisfy goals.”
The motivation was lean staffing: F.E.A.R. had “only one A.I. programmer, but there are lots of A.I. characters,” and the combinatorial complexity of layering behaviors was “getting unmanageable.” Orkin explicitly tied his work to Damian Isla’s GDC 2005 Halo 2 talk, calling planning “our attempt at solving the problem” of managing that complexity. GOAP went on to become a standard technique, and the F.E.A.R. paper one of the most-cited primary documents in game AI.