Multiobjective design of low complexity digital controllers

Abstract

For the practical implementation of controllers, it is frequently necessary to design controllers with low complexity. There is usually a trade-off between the controller complexity and closed loop performance, thus the problem of low-complexity controller design is multiobjective. Kolmogorov-Chaitin computational complexity provides a fundamental concept of complexity. In this paper, fixed point digital controller design is performed for the European Journal of Control restricted complexity benchmark problem. The complexity measure used is based on Kolmogorov-Chaitin computational complexity, and is hence more specifically aimed to practical implementation in hardware and software. The design problem is solved using multiobjective genetic algorithms.