Design of optimal GBN sequences for identification of MIMO systems

This paper presents a systematic approach to the design of optimal Generalized Binary Noise (GBN) sequences as excitation inputs for control relevant identification of MIMO systems. The aim of this research is to propose a method to design the GBN signals, which are able to extract the main characteristics of a MIMO dynamic system in frequency bandwidth that is important for control purpose. The frequency interval is specified based on prior knowledge of the system, which is obtained through preliminary tests such as step tests. The optimal GBN sequences are then designed by driving the signal´s power spectrum to maximize the signal energy in the frequency bandwidth. To validate the improvement of the proposed method as compared to experience-based approaches, a case study on a real time MIMO system is investigated. Results from applying the proposed method of designing GBN sequences for identification and Model Predictive Control (MPC) of the system show that the proposed method is attractive choice for control relevant system identification.