Abstract

A large number of methods are known for system identification, which are used both in the time domain and in the frequency domain. In particular, genetic algorithms are increasingly being used today in order to determine the parameters of a model on the basis of measurements. In this article, the related method 'hill climbing' is used together with the least square criterion in order to correctly identify models of small order on the basis of measured step re-sponses in the time domain. It is shown that the algorithm converges well for many starting values and that this method can be applied very well and efficiently for the topic of system identification.