Improvement of Classical Evolutionary Programming Using State Feedback Controller

In evolutionary programming, each parent has two pieces of information: location and cost. The cost of parent specifies whether its location is suitable for breeding offspring or not. If the parent’s cost is an acceptable value, producing offspring (or even steering other offspring) in the parent’s area is advisable. This information is used in estimating the region of global minimum; then, using the state feedback controller, the offspring is steered to the optimal region. In the proposed method, the cost and coordination of parents have been used for breeding more elite individuals. Many (sixty-five) wellknown cost functions have been selected from different references to reveal the pros and cons of our algorithms. In the first stage, the proposed algorithm has been compared inside the EP family. This stage shows promising results for the proposed algorithm. In the second stage, comparison has been performed out of the EP frontiers in which algorithms are state-of-the-art in the optimization field, and are well known inside and outside of their own families. The statistic test has been performed among the algorithms. CPU time and its sensitivity to variable bounds, population and cost function dimensions have been studied. Finally, the proposed method is used in designing the nonlinear minimum variance controller for CSTR (Continuous Stirred-Tank Reactor) benchmark system.