Game Development Game AI Syllabus

Game AI syllabus — full chapter and topic list

Expand any chapter to see its topics and sub-topics. This is the whole examinable outline for Game AI in Game Development, not a summary of it.

1. 1

   Introduction to Game AI

   3 topics

   • History of Game AI
   • Basic Concepts and Definitions
   • Applications of Game AI
2. 2

   Fundamental AI Techniques

   4 topics

   • Finite State Machines (FSM)
   • Decision Trees
   • Behavior Trees
   • Utility Systems
3. 3

   Pathfinding and Navigation

   5 topics

   • A* Algorithm
   • Dijkstra's Algorithm
   • Navigation Meshes
   • Path Smoothing
   • Local Avoidance
4. 4

   Movement and Steering Behaviors

   6 topics

   • Seek and Flee
   • Arrival and Departure
   • Pursuit and Evasion
   • Wander
   • Obstacle Avoidance
   • Path Following
5. 5

   Advanced AI Techniques

   5 topics

   • Machine Learning in Games
   • Neural Networks
   • Genetic Algorithms
   • Reinforcement Learning
   • Monte Carlo Tree Search (MCTS)
6. 6

   AI for Different Game Genres

   5 topics

   • AI in Strategy Games
   • AI in First-Person Shooters
   • AI in Role-Playing Games
   • AI in Sports Games
   • AI in Puzzle Games
7. 7

   AI Optimization and Performance

   4 topics

   • Optimizing Pathfinding
   • Reducing Computational Overhead
   • Balancing AI Complexity
   • Multithreading and Parallelism
8. 8

   AI Tools and Frameworks

   4 topics

   • Unity ML-Agents
   • Unreal Engine AI Tools
   • OpenAI Gym
   • TensorFlow and PyTorch
9. 9

   Case Studies and Real-World Examples

   3 topics

   • AI in Classic Games
   • AI in Modern AAA Titles
   • AI in Indie Games

Game AI flashcards for Game Development

22 of 50 cards from the Game AI deck — real questions with worked answers.

1. In what decade did academic and commercial game AI begin, and which early game is credited as one of the first to use AI to control an opponent?

   Game AI dates to the 1950s-1970s; Nim-playing machines (e.g., Nimrod, 1951) and especially Pac-Man (1980), whose four ghosts each used distinct pursuit behaviors, are landmark early examples of AI-controlled opponents.

2. What distinguishes the goal of "game AI" from the goal of academic/general AI?

   Academic AI aims to maximize optimal/intelligent behavior, while game AI aims to create the illusion of intelligence that is fun, believable, and challenging for the player — it prioritizes player experience over true optimality.

3. Define an "agent" in the context of game AI.

   An agent is an autonomous entity that perceives its environment through sensors, makes decisions, and acts upon the environment through actuators to pursue goals — in games, typically an NPC (non-player character).

4. What are the three broad layers commonly used to structure game AI?

   Movement (how an agent moves, e.g., steering/pathfinding), Decision-making (what to do, e.g., FSM/behavior trees), and Strategy (group/team-level coordination and planning).

5. Name four common applications of AI within video games.

   Non-player character (NPC) behavior, pathfinding/navigation, procedural content generation, difficulty balancing/adaptation, and player modeling/opponent tactics.

6. What is a Finite State Machine (FSM) in game AI?

   A model of computation consisting of a finite set of states, one active state at a time, and transitions triggered by conditions or events; the agent's behavior is defined by the currently active state.

7. List the core components that define any Finite State Machine.

   A set of states, an initial (start) state, a set of input conditions/events, and a transition function mapping (current state, input) to a next state, plus the actions performed in/on each state.

8. What is a key limitation of standard FSMs that motivated hierarchical FSMs?

   State explosion — the number of transitions grows combinatorially as states increase (up to $n(n-1)$ transitions for $n$ states), making large FSMs hard to design and maintain; hierarchical FSMs group states to reduce this.

9. What is a Decision Tree in game AI, and how is it evaluated?

   A tree of decision (condition) nodes and action (leaf) nodes; evaluation starts at the root, and at each internal node a boolean/attribute test selects a branch, recursing until a leaf action is reached.

10. How does a Decision Tree differ structurally from a Behavior Tree?

    A decision tree's internal nodes are pure conditional tests that route to a single leaf action each traversal, whereas a behavior tree has composite control nodes (sequence/selector) that can execute multiple actions over time and return running/success/failure statuses.

11. What are the three standard return statuses a Behavior Tree node can report?

    Success, Failure, and Running (in-progress).

12. In a Behavior Tree, what is the behavior of a Sequence node?

    It runs children in order and returns Success only if all children succeed; it returns Failure as soon as any child fails (logical AND), and Running while a child is still executing.

13. In a Behavior Tree, what is the behavior of a Selector (Fallback) node?

    It runs children in order and returns Success as soon as any child succeeds; it returns Failure only if all children fail (logical OR), and Running while a child is executing.

14. What are decorator and parallel nodes in a Behavior Tree?

    A decorator has one child and modifies its result or repetition (e.g., inverter, repeater, until-fail); a parallel node runs all children simultaneously and returns success/failure based on a configured policy (e.g., succeed when N children succeed).

15. What is a Utility System (utility-based AI) in game AI?

    A decision-making approach that assigns each possible action a numeric utility score computed from world state via utility/consideration functions, then selects (or weight-randomly picks) the action with the highest score.

16. How does a Utility System typically choose an action, and what advantage does this give over an FSM?

    It scores every action with a utility function and picks the maximum (or samples proportionally); this scales gracefully and produces smooth, context-sensitive, nuanced decisions without hand-authoring every transition as an FSM requires.

17. What problem does the A* algorithm solve, and what category of algorithm is it?

    A* solves the single-source shortest-path problem to a specific goal on a weighted graph; it is an informed (heuristic) best-first search algorithm.

18. State the core evaluation function A* uses to order nodes.

    $f(n) = g(n) + h(n)$, where $g(n)$ is the cost of the path from the start to node $n$ and $h(n)$ is the heuristic estimate of the cost from $n$ to the goal.

19. What property must an A* heuristic have to guarantee an optimal (shortest) path?

    It must be admissible — it never overestimates the true cost to the goal: $h(n) \leq h^{*}(n)$ for all $n$, where $h^{*}(n)$ is the actual optimal remaining cost.

20. What is the difference between an admissible and a consistent (monotone) A* heuristic?

    Admissible: $h(n) \leq h^{*}(n)$. Consistent (stronger): for every edge $n \to n'$ with cost $c$, $h(n) \leq c(n,n') + h(n')$ and $h(\text{goal})=0$. Consistency implies admissibility and lets A* avoid re-expanding nodes.

21. Give the two most common admissible heuristics used on a grid, and when each applies.

    Manhattan distance $h = |x_1 - x_2| + |y_1 - y_2|$ for 4-directional movement, and Euclidean distance $h = \sqrt{(x_1-x_2)^2 + (y_1-y_2)^2}$ (or Chebyshev/octile) for movement allowing diagonals.

22. What does A* reduce to when the heuristic is $h(n) = 0$ for all nodes?

    It reduces to Dijkstra's algorithm — an uninformed uniform-cost search that expands nodes purely by accumulated path cost $g(n)$.

See more Game AI flashcards →

Planning Game AI for Game Development

Game AI (Game Development) FAQ