F. Cheong and R. Lai (Australia)
Fuzzy Logic Controller, Evolutionary Algorithms, Differ
ential Evolution, Control System, ATM
We describe a method for tuning Fuzzy Logic Controllers
(FLCs) which is not based on expert knowledge or trial and
error. The method consists of generating one or more con
trol tables and using an Evolutionary Algorithm (EA) to
optimize only the membership functions as it is straightfor
ward to map the entries of the control table to a rule base.
We tested our approach on two different controllers: a con
troller which is not related to the family of PID controllers
and a PD-like Mamdani FLC where the inputs are the error
and the change in error. In both cases, our approach was
successful as both controllers performed very well.