Behavior tree
A hierarchical model for game AI where decisions are built from sequence and selector nodes, ticked each frame, returning success, failure or running.
The tree runs top-down every frame. Sequence nodes run their children in order until one fails; selector nodes try children until one succeeds; decorators modify them (invert, repeat, time out) and the leaves are concrete conditions or actions. The running return value lets an action span several frames.
Damian Isla popularised the technique with his talk on the AI in Halo 2 (2005) as an answer to how finite state machines explode combinatorially as the number of states grows. Behavior trees are modular and reusable — a subtree can be dropped into another character — and have been the industry standard ever since. The movement itself is often delegated to steering behaviors.