Evolutionary strategies for function optimization, fuzzy modeling and control

Solving complex problems with evolutionary approaches has attracted increasing attention in recent years. The main advantage of such problem solving strategy is its robustness and ability to adapt to an unknown environment without explicit assumptions.

The aim of this thesis is to investigate evolutionary computation techniques and their applications in addressing three different problems - global function optimization, fuzzy modeling and control. We start by introducing the pertinent necessary background of evolutionary algorithms, and this is followed by a discussion of the adaptation mechanism in evolutionary algorithms. An adaptive mutation scheme based on the gradient decent approach is then presented for non -constrained function optimization and an adaptive penalty function method is proposed for handling constraints in constrained function optimization. We will then examine the conventional fuzzy models and propose several adopted fuzzy rule based systems. The evolutionary approach for the construction and identification of fuzzy rule base is investigated and applied to the dynamic reconstruction and fuzzy gain scheduling controller design. The performances of these proposed methods are examined through the empirical studies of benchmark problems, and simulations are given to demonstrate the possibility of integrating fuzzy systems with evolutionary algorithms in nonlinear dynamic system modeling and control.

The thesis concludes with discussions on the contributions made and suggestions of direction for further research.