ABSTRACT
A behavior tree in game artificial intelligence (AI) is used to model the behaviors that an agent can perform. The tree structure allows elemental actions to be combined to create a higher level behavior. This higher level behavior can then be treated as an elemental behavior and used to compose even higher level behaviors. A planner in game AI is used to create a sequence of elemental actions to achieve some goal, given the world state. This sequence of actions is called a plan. The planner maintains a model of the world state, a collection of all elemental actions available to an AI, and a goal heuristic. The basic premise of the hybrid approach is simple: combine the strengths of behavior trees and planners to produce an AI system that is flexible and durable in the face of design changes while allowing the designers full control over the structure of the actions available to the AI.
