This paper presents an efficient implementation of constraint optimum input design for on-line system identification for systems exhibiting actuator saturation. The constraint optimal input is calculated recursively based on the imminent available information content in the inverse correlation matrix of the data. The new input is computed one step ahead of time with a predictive filter so that it will increase the information content in the inverse correlation matrix. The information content is maximized with help of a simple genetic algorithm. A numerical example indicates superiority of the proposed method over the traditional method where white gaussian noise is used as the input.

This content is only available via PDF.
You do not currently have access to this content.