Abstract:
The most important aspect of human behaviour is learning. One of the learning
methodologies applied by humans is learning through iteration. This human capability
has recently been used by control engineers to design Iterative Learning Controls (ILC).
The problem with ILC is that it is designed for a specific system and a specific desired
response. Moreover, the number of iterations is high, especially if the system dynamics
are not known. Our research work aims at reducing the number of iterations for
convergence and evolving a design mechanism that can adapt for changing systems and
varying desired responses, without the need to redesign the ILC. This thesis develops a
number of Iterative Learning Controllers to meet these requirements. Stability and
convergence criteria of these controllers are also established.
Fuzzy control is another emerging control methodology focusing on human
perception and fuzzy thinking. The problem with fuzzy design is the uncertainties
associated with the design of membership functions and rule base. Moreover, controlled
design requirements are generally given in the form of steady state error, percentage over
shoot etc. These requirements need to be translated into fuzzy design. The work also
establishes a number of fuzzy controllers combined with ILC to over come these short
comings in fuzzy design. The designs are tested through simulations and practical setups.
For the practical setups, a Six Degree of Freedom Hexapod, a DC motor kit by Quanser
and a custom built Two Degree of Freedom Tracker were used. Stability and convergence
of these Iterative Learning Fuzzy Controllers are also discussed.
The research concludes that in order to reduce uncertainties associated with fuzzy
logic based design we have to incorporate learning. This hybrid approach can open up a
new era of controller design.
