Chapter 9: Behavior Trees

In a preceding chapter, we saw a basic but effective way to implement and manage character states and behaviors: finite state machines (FSMs). FSMs are simple to implement and intuitive, but they have a fatal flaw: it is tough to make them scale once there are many states and transitions. For example, imagine a character that behaves differently depending on its health and mana (high, medium, or low). We have a state in which both health and mana are high, one in which health is medium and mana is high, one in which they are both medium, and so on. In total, we have nine states just for those. If we add other conditions (such as player proximity, time of day, equipment, player's score, or whatever you may imagine), the number of states grows exponentially.

Luckily, we have a solution: behavior trees (BTs). In essence, BTs are just another way to visualize complex FSMs, but they are fast, provide reusability, and are easy to maintain. After their introduction...

Technical requirements

For this chapter, you need Unity3D 2022 and the free plugin for Unity, Behavior Bricks. Don't worry, we will see how to install this plugin together. You can find the example project described in this chapter in the Chapter 9 folder in the book repository here: https://github.com/PacktPublishing/Unity-Artificial-Intelligence-Programming-Fifth-Edition/tree/main/Chapter09.

Introduction to BTs

A BT is a hierarchical tree of nodes that controls the AI character's behavior flow. It can also be used to coordinate groups of characters (for example, to model the attack pattern of a small platoon), or even disembodied agents such as an AI story director.

When we execute a BT's node, the node can return three states: success, failure, or running (if the node's execution is spread over multiple frames, for instance, if it plays an animation). When the BT executor runs a tree, it starts from the root and executes every node in order, according to rules written in the nodes themselves.

A node can be of three types:

• A task (a node without children), also called a leaf.
• A decorator (a node with a single child)
• A composite (a node with multiple children)

In general, leaves represent the Action that the characters can do or know (that is why they are commonly called an Action or Task); they may be actions such as GoToTarget...

Implementing a BT in Unity with Behavior Bricks

Behavior Bricks is a robust but free BT implementation for Unity developed by the Complutense University of Madrid in Spain. Using Behavior Bricks, you can start using BTs in your projects without implementing BTs from scratch. It also has a visual editor where you can drop and connect nodes without any additional code.

Follow these steps to install Behavior Bricks:

1. We need to go to the Unity Asset Store by going on the website https://assetstore.unity.com/.
2. Search for Behavior Bricks.
3. Click on Add to My Assets. Once it's done, we can import it into our project.

Figure 9.3 – Behavior Bricks Asset Store main page

4. Go to Package Manager (Window | Package Manager).
5. Go to My Assets.

Figure 9.4 – The Package Manager window

6. Import the Behavior Bricks package by clicking on the Import button.

Figure 9.5 –...

Implementing the nodes

After we have made a plan for our BT, the next step is to check whether our BT implementation of choice (in our case, Behavior Bricks) already includes some of the nodes we need. Of course, we want to reuse as many pre-made nodes as possible. Reading the Behavior Bricks documentation, we can see that it already includes nodes such as IsTargetClose, MoveToGameObject, Wander, and AlwaysTrue, plus, of course, Repeat and Selector.

Therefore, we need to write all the other tasks. Note that Behavior Bricks tasks are not MonoBehaviors; therefore, we do not need to attach them to some object in the scene. We only need to put the scripts in any folder in our project's assets, and we are good. Let's look at a step-by-step process to do this:

1. Let's start with the ShootOnce action by creating a ShootOnce.cs file in the project assets. First, we create a simple Action attribute called ShootOnce that, as the name says, shoots a single bullet:

   using...

Summary

In this chapter, we explored the general background behind any BT implementation. We saw what a BT is, what its basic components are, and how can we use a BT to describe game character behavior. Then, we implemented a demo using a free plugin called Behavior Bricks. In the demo, we created the behavior for a simple scenario: the player and a patrolling robot. We also implemented a day/night cycle to spice up the scenario.

BTs are the cornerstones of modern AI for game characters. Implementation details and deeper examples would require a full book to explain them fully. Luckily, the web is full of resources for the curious reader.

Now, we will take a break from AI character design by looking at a different application of AI in games. In the next chapter, we will look at the fascinating field of procedural content generation.

Further reading

• The official Behavior Bricks project page from the Complutense University of Madrid in Spain: http://gaia.fdi.ucm.es/research/bb/.
• Behavior Bricks official documentation: http://bb.padaonegames.com/doku.php.
• Chris Simpson (the developer of Project Zomboid) wrote a nice explanation of BTs on Gamasutra: https://www.gamasutra.com/blogs/ChrisSimpson/20140717/221339/Behavior_trees_for_AI_How_they_work.php.
• Chapter 6 of GameAI Pro, which explores many implementation details of BTs (in C++), is free and available at the following link: https://www.gameaipro.com/GameAIPro/GameAIPro_Chapter06_The_Behavior_Tree_Starter_Kit.pdf.