AI Behavior Trees
Personal Project
The Need This Code Fulfills
Behavior trees help organize complex AI decision-making in a clear and structured way, making it easier to control how NPCs or enemies react to different situations. They allow for more modular and reusable AI logic, so you can create smarter characters without writing messy code. This is useful for making AI behaviors easier to manage, debug, and expand over time. Additionally it allows non-programmers to create many different types of AI behavior without having to alter code.
Behavior Trees Using Scriptable Objects
Behavior Tree Nodes
I built a behavior tree system made up of different types of nodes, each with a specific function. The most basic nodes are leaf nodes, which perform actions like moving toward the player or check conditions like detecting if the player is nearby. Composite nodes control how multiple child nodes are executed. Sequence nodes run its children in order and fail if any fail, while selector nodes run their children until one succeeds. These nodes allow me to build complex AI behaviors by combining simple decision-making steps.
AI Blackboard
The AI blackboard is a shared memory system for storing important shared information like the player's location. It inherits from the ScriptableObject class to allow different blackboards to be applied for different types of AI when necessary. Instead of each node trying to figure out where the player is, they can all pull from this shared blackboard, making AI decision-making more efficient and allowing for coordinated AI behaviors. When a node needs data, like checking if the player is in range, it reads from the blackboard instead of calculating it from scratch. This allows different AI agents to use the same behavior tree while keeping their own unique state and decision-making processes.
ScriptableObject-Based Behavior Trees
To make the behavior tree flexible and reusable, I used ScriptableObjects to define the nodes and the tree structure. Each node type is stored as a separate ScriptableObject, allowing non-programmers to modify behaviors directly in Unity’s editor instead of writing new scripts for every AI. The behavior tree itself is also a ScriptableObject, holding a reference to the root node, which then calls its child nodes in order. This modular approach makes it easy to tweak AI behavior without touching code.
AI Controller
The AI controller acts as the bridge between the AI’s decision-making using the behavior tree, shared data in the blackboard, and the data which is specific to that particular NPC or enemy itself such as the NavMeshAgent. Each frame, the behavior tree is processed to determine what action the AI should take, using the data on the blackboard and the AI Controller’s personal data. It could be further optimized for performance by lowering the tick rate at which it processes the behavior tree.