An improvement of Particle Swarm Optimization and its application to a model-free PIλDμ tuning problem

Abstract

Particle Swarm Optimization (PSO) is an easily applicable population-based stochastic optimization technique which does not require much knowledge about the problem at hand. However, in many cases there is some a priori knowledge available that can be used to improve the optimization process. In this contribution a novel framework is proposed that allows a combination of the classical PSO algorithm with a method for exploiting available a priori knowledge. This so-called Knowledge Supported PSO (KS-PSO) method is applied to a specific optimization problem, namely the model-free tuning of a fractional order PID controller.