Nonlinear control using evolutionary fitness functions based on scaling transformations

In evolutionary computation, for such applications as intelligent systems, it is especially important to improve the evolution level by establishing a self-adaptive fitness function expression that can maintain group diversity and convergence properties while enhancing the evolutionary speed and performance. This paper employs a simple model for the fitness function, based upon scaling transformations, and applies the approach to the design of nonlinear controllers, and in particular, the generalized ANFIS control structure. The scaling transformation uses the minimum, maximum and average value of the fitness function at each generation in tuning the transformation parameters. Results indicate that the fitness function based on these transformations is an attractive approach for improving system performance, even under plant parameter variations and system noise.