Chapter 2. LITERATURE REVIEW
2.1 AI Planning and GOAP
AI planning is a field of classical AI that has been around for quite some time and has been extensively researched in academia.
GOAP is based on a modification of the STRIPS language which is defined using goals and operators. Dr. Bridge outlines the basics of the STRIPS language and some of the issues faced by AI planners in the field of classical AI planning in his lecture slides (Bridge, 2007). STRIPS goals describe some desired state of the world to reach, and operators are defined in terms of preconditions and effects. An operator may only execute if all of its preconditions are met, and each operator changes the state of the world in some way through its effects which are carried out by the add and delete lists. A classical AI planning problem known as the Sussman Anomaly is also discussed along with how it can be overcome using partial-order planners or total-order planners.
A game produced by Monolith productions, First Encounter Assault Recon or F.E.A.R., is the first AAA commercial title known to use a real-time planner system in games and won several awards due its ground-breaking AI. The lead architect of the AI system, Jeff Orkin, has written several articles that describe the agent and system architecture used in F.E.A.R.. His work and publications provided the inspiration for this dissertation as GOAP is such a new AI technology and the challenge of creating a state-of-the-art system whilst discovering its possible benefits over existing traditional technologies was quite appealing from the outset.
A presentation given by Orkin at the Game Developers Conference 2006 (Orkin, 2006) provoked the first interest in the field of AI planning in games. The difficulty in
managing the complexity of FSMs was cited as the primary reason as to why GOAP was 3