Game Development Reference
In-Depth Information
FIG 5.16 Classical sets have distinct
boundaries. Fuzzy sets do not.
A fuzzy rule can be written as
if x is A
then y is B
where x and y , known as linguistic variables , represent the characteristics
being measured (temperature, speed, height, etc.) and A and B , known as
linguistic values , are the fuzzy categories (hot, fast, tall, etc.). A fuzzy rule can
also include AND and OR statements similar to those in Boolean algebra. The
following are examples of fuzzy rules:
if temperature is hot
or UV_index is high
then sunburn is likely
if temperature is warm
and humidity is low
then drying_time will be quick
Given a set of fuzzy rules and a number of inputs, an output value can be
deduced using fuzzy inferencing. There are two common ways to inference on
fuzzy rules: Mamdani style and Sugeno style, both named after their creators.
The Mamdani style is widely accepted for capturing human knowledge and
uses a more intuitive form of expressing rules.
The Mamdani style consists of four steps: fuzzification, rule evaluation,
aggregation, and defuzzification. Consider the following rules by which
an NPC may operate:
Rule 1:
if health is good
and shield is adequate
then mood is happy
Rule 2:
if health is bad
or shield is moderate
then mood is indifferent
Rule 3:
if health is bad
and shield is inadequate
then mood is angry
Fuzzification takes inputs in the form of discrete values for the NPC's health
and shield strength and fuzzifies them. To fuzzify them, we simply pass them
through the respective fuzzy sets and obtain their degrees of membership.
Search Nedrilad ::




Custom Search