| Summary: | 碩士 === 淡江大學 === 資訊工程研究所 === 83 === It is difficult for a fuzzy logic controller designer to
determine the universe of discourse of the input and output
linguistic variable, the shape of the membership function and
the fuzzy control rule of the controller. In general,these
parameters of the controller is based on the expert knowledges
and the operator''s experiences. It needs a very time-consuming
trial-and- error procedure to finely tune these parameters. In
this thesis, we propose a rule mapping fuzzy controller. It is
constructed by the method of rule mapping and Genetic
Algorithms under the innocences of the controlled process. The
system response can be described by the transient and steady
state characteristics. Various system needs different
performance. To meet the different specifications of the
controlled system, we propose a weighted type of the fitness
function in Genetic Algorithms for various different purposes
so that a satisfactory performance (fast rise time, small
maximum overshoot and small integral of absolute error) in the
step response can be obtain. Simulation results of the inverted
pendulum system demonstrate the efficiency of the proposed
control scheme. On the other hand, the most effective way to
improve the performance of a fuzzy controller is to optimize
its fuzzy control rule. If the control rules can change
according to the states of system''s response,then the system
can obtain a comparatively better performance.Since the control
rules of the original rule mapping fuzzy controller are decided
by a regulating factor, we employ a Rule Self-Regulating
Mechanism to change the regulating factor for different states
of system responses. That is fuzzy control rules can be
regulated on line. The simulation shows that the performance of
self-regulating fuzzy controller can be improved further than
that of a non-self -regulating fuzzy controller.
|
|---|
