Source metadata

  • Type: Worked-example game AI textbook
  • Author: Mat Buckland
  • Published: 2005, Wordware

Key takeaways

  • Starts with a practical math and physics primer, then moves into full AI systems rather than disconnected algorithms.
  • Gives a strong worked-example treatment of finite state machines, including messaging between agents and reusable machine infrastructure.
  • Covers classic steering behaviours, their combinations, spatial partitioning, and group behaviours in a way that is still highly teachable.
  • Uses graphs and A* in a practical path-planning context rather than only as abstract search.
  • Extends beyond reactive AI into goal-driven behaviour, scripting, and fuzzy logic.

Notable claims

  • The book treats FSMs as a sturdy backbone that can later be combined with other techniques such as fuzzy logic and goal arbitration.
  • Steering is presented as a distinct layer from locomotion, making behaviour reuse easier across different movement implementations.
  • Goal arbitration is framed as desirability-driven, which makes it a useful stepping stone toward more formal utility and planning systems.

Relevance

Directly informs:

Open questions raised

  • Which of Buckland’s worked examples are worth recreating in modern Unity/C# as teaching code?
  • How should the wiki present the relationship between Buckland’s goal-driven behaviour and more formal GOAP?

ai-state-machine-pattern · steering-behaviours · pathfinding-algorithms · goal-oriented-action-planning · fuzzy-logic-for-games · overview-artificial-intelligence-in-games