Preference Oriented Multi-Objective Optimization for Tuning of Controllers: A Reference Point Based Approach

Abstract

This paper proposes a reference point based controller tuning approach to include the designer's preference in multi-objective optimization (MOO) procedure. As the MOO procedure gives a Pareto-optimal set of solutions, the designer preferred reference points are considered to include the required specifications. The reference point based nondominated sorting genetic algorithm (R-NSGA-II) is applied to find the set of solutions near those preferred points only. Extensive studies have been conducted on various types of linear and nonlinear systems with and without delay for fractional order PID (FOPID) controllers. The results obtained demonstrate the superiority of proposed approach to achieve robust controller design. Eventually, better set point following, disturbance rejection, and variable input following are observed.