Publications
Thesis
Investigations into new tunable fuzzy and modular
neuro-fuzzy controllers
Resume
The chief aim of the work was to bring into sharp focus the utility
of fuzzy logic controllers (FLC) and modular neuro-fuzzy logic
controllers in a wide variety of situations.
A fuzzy logic controller was designed to control a process plant
with widely varying conditions in its operations. The varying
conditions considered were the order, type, time-delay and associated
parameters. Disturbances at the input level and sudden set point
variations were also considered to bring out the features of the FLC
to the fore.
A neuro-fuzzy/fuzzy adaptive controller with four self-learned fuzzy
control rules was also arrived at, to effectively control the above
process plant.
Modern automobile industries are focusing a lot of attention on the
(semi-) active suspension systems for good ride comfort, by bringing
down the vertical acceleration and deflection as far as possible.
A FLC designed for a quarter car system was found to exercise control
in checking its vertical acceleration and deflection to a level, that
of a hypothetical model. The same controller was found to contain a
wide variations in the parameters of the suspension system without
any modification whatsoever.
A modular neuro-fuzzy structure was proposed to position and to balance
the cart-pole system. The two modules were made to learn the control
strategies using temporal back propagation algorithm. The fuzzy de-compositional rule of inferencing has been used to arrive at the
modular structure.
On a similar fashion a modular FLC was designed for the cart-pole
system and its input-output mappings were captured with the help
of a modular neuro controller. This will reduce the computational
burden associated with the FLC, in its implementation stage.
These studies clearly demonstrated the utility, flexibility and
robustness of fuzzy and/or neuro-fuzzy logic controllers in disparate
applications.