In this lesson, you will learn how to make your AI reactive by using Blackboard Observer Aborts to interrupt ongoing tasks the moment specific data changes.
Understanding Observer Aborts
By default, a Behavior Tree node finishes its task before checking its conditional Decorator again. This creates "stale" AI behavior where, for example, a guard might continue walking to a patrol point even after spotting a player because the "MoveTo" task hasn't finished yet. Observer Aborts allow you to break that cycle by forcing the tree to re-evaluate conditions immediately.
Configuring Aborts
When you select a Decorator (like a Blackboard condition) in your Behavior Tree, you will find the Flow Control section in the Details panel. The Observer Aborts property determines how the tree responds when the Blackboard value changes:
- None: The default. The condition is only checked when the execution flow reaches that node.
- Self: If the condition for the current node becomes false, the node itself (and its children) will abort immediately.
- Lower Priority: If the condition for a node becomes true, any nodes currently running on a "lower priority" branch (to the right in your tree) are aborted, allowing the AI to switch tasks instantly.
- Both: A combination of the two, ensuring the most responsive behavior possible.
A Concrete Example
Imagine an IsTargetVisible decorator attached to an "Attack" task. If the player runs behind a wall, the Blackboard key TargetVisible flips to false. By setting the Observer Abort to Self, the "Attack" task will receive an immediate Abort signal the moment the Blackboard key updates. The AI won't finish its current "shoot" animation or wind-up; it will terminate the task instantly, allowing the tree to move to a "Search" or "Idle" state.
Using these settings effectively transforms your AI from a static script into a responsive agent that feels alive and aware of its environment.
Key Takeaway: Observer Aborts provide the "interrupt" signal needed to make your AI transition between behaviors instantly based on real-time data changes, rather than waiting for current tasks to complete.