On simplification of Dual-Youla approach for closed-loop identification
This work addresses the practical barrier of coprime factorization in the Dual-Youla method, making it more accessible for industrial applications in system identification.
This paper simplifies the Dual-Youla method for closed-loop identification, enabling direct plant identification without coprime factorization. The simplified method retains the original's benefits, including accuracy regardless of noise models, handling unstable plants, and robustness to controller knowledge uncertainty.
The dual Youla method for closed loop identification is known to have several practically important merits. Namely, it provides an accurate plant model irrespective of noise models, and fits inherently to handle unstable plants by using coprime factorization. In addition, the method is empirically robust against the uncertainty of the controller knowledge. However, use of coprime factorization may cause a big barrier against industrial applications. This paper shows how to derive a simplified version of the method which identifies the plant itself without coprime factorization, while enjoying all the merits of the dual Youla method. This simplified version turns out to be identical to the stabilized prediction error method which was proposed by the authors recently. Detailed simulation results are given to demonstrate the above merits.