Chapter 30. Fuzzy Logic
Classical logic can only model crisp values: The enemy is either dead or not. Probabilities express the likelihood of a crisp variable having a specific value: This door has an 80-percent chance of being open. Fuzziness is conceptually different because it models degrees of truth: The enemy is mostly dead, and this door is nearly shut. At times, crisp AI suffers from clear-cut decisions and very robotic control. The fuzzy approach resolves the issue by providing shades of gray between black and white, allowing animats to attack with moderation (decision making) or turn left slightly (smooth control).
This form of knowledge representation is very intuitive, and comes surprisingly close to modeling human thoughts. Using a model closer to linguistic definitions can simplify the design of systems, enabling humans (not only experts) to add knowledge to the system.
This chapter covers the following topics:
The theory from this chapter can be applied to a wide variety of problems in game development, providing levels of realism that other crisp techniques struggle to reach. Fuzzy techniques in the next chapter help the animats climb ladders and use platforms.