Behavior Tree Overview
- Behavior Tree graph of nodes; Blackboard asset stores Blackboard Keys the tree uses for decisions.
- Execute left-to-right, top-to-bottom. Decorators (Conditionals) attached to Composite nodes; Tasks are actions (leaves).
- Pawn + AI Controller; controller runs Behavior Tree on possess. NavMeshBoundsVolume needed for navigation.
- Event-driven: passively listen for events rather than checking every frame; event updates Blackboard key → lower-priority task aborted for higher-priority branch. Better performance and debugging (history shows only relevant changes).
- Conditionals are Decorators not leaf Tasks: readability; all leaves are actions; decorators can act as observers.
- No full parallel: Simple Parallel (main single Task + background subtree: "While doing A, do B"), Services (attached to Composite; callbacks every specified seconds; active while execution in subtree), Observer Aborts on Decorators (abort Self etc.).