Beru Learning

The enemies and other non-player characters (NPCs) in video games aren't truly 'thinking.' Most of the time, their seemingly intelligent behavior is powered by a remarkably simple and effective logic model: the Finite State Machine (FSM). It’s one of the oldest and most fundamental tools for programming game AI.

Understanding the concept of an FSM demystifies how game characters work. It’s a simple but powerful system for defining behavior, and once you grasp the idea, you'll start seeing its structure in almost every game you play.

The Core Components: States and Transitions

A Finite State Machine operates on two very simple concepts. At its heart, an FSM is a character that can only be in one 'state' at any given moment, from a finite (limited) list of possible states.

States: A state is a specific behavior the character is currently performing. Common states for an enemy AI might include Patrolling, Chasing, Attacking, or Fleeing. The character can only be doing one of these things at a time.
Transitions: These are the rules that govern how the character moves from one state to another. A transition is triggered by a condition. For example, if the character is in the Patrolling state and the condition 'Player is visible' becomes true, it will transition to the Chasing state.

That's it. The entire 'brain' of the AI is just a list of possible behaviors and the specific rules that make it switch between them.

A Practical Example: The AI of a Simple Guard

Let's design the FSM for a classic video game guard. Our guard will have four states: Patrol, Investigate, Chase, and Attack.

State: Patrol. The guard walks a predefined path. The main transition rule is: if the guard 'hears a sound,' transition to Investigate. If the guard 'sees the player,' transition to Chase.
State: Investigate. The guard moves to the location of the sound. If they find nothing after a few seconds, they transition back to Patrol. If they 'see the player' while investigating, they transition to Chase.
State: Chase. The guard runs directly at the player. If the player 'gets into attack range,' transition to Attack. If the guard 'loses sight of the player' for a set amount of time, they might transition back to Investigate at the player's last known location.
State: Attack. The guard performs their attack animation. Once the attack is complete, if the player is still in range, they attack again. If the player 'moves out of range,' they transition back to Chase.

By linking these simple states and rules, you get behavior that feels responsive and logical to the player.

The Strengths and Limits of FSMs in Modern Games

FSMs are popular because they are easy to design, debug, and understand. They are perfect for characters with simple, predictable behaviors. However, they can become messy as the number of states and transitions grows. A very complex character might require dozens of states with hundreds of transitions, creating a tangled web that is hard to manage.

For this reason, more advanced games often use other AI models like 'Behavior Trees' or utility-based AI for their most complex characters. But even in these games, FSMs are still widely used for simpler NPCs, background characters, and even managing systems like game state or UI.

Frequently Asked Questions

Is a Finite State Machine a type of machine learning?

No, not at all. An FSM is a hand-scripted logic model. Every state and transition is explicitly programmed by a developer. It does not learn or adapt on its own.

Are FSMs only for enemy AI?

No, they can be used to control any entity that has a limited number of states. This could be a friendly NPC, a piece of interactive machinery, or even the state of a quest in the game's code ('Not Started', 'In Progress', 'Completed').

What is a Behavior Tree?

A Behavior Tree is a more advanced AI model that is better at handling complex decision-making. It organizes tasks in a hierarchical structure, which can be more flexible and scalable than the 'flat' structure of an FSM.

Key Takeaways

A Finite State Machine (FSM) is a simple logic model used to program AI behavior in games.
It consists of 'states' (behaviors) and 'transitions' (rules for switching between behaviors).
An AI character can only be in one state at any given moment.
FSMs are easy to implement and understand, making them ideal for characters with predictable routines.
While modern games use more advanced techniques for complex AI, FSMs remain a fundamental and widely used tool in game development.

How Finite State Machines Power Simple Video Game AI