Decoupled Dual-Regularization for Phase-Wise Identification of Linear Time-Periodic Systems Doǧrusal Zamanla Periyodik Sistemlerin Faz-Bazli Tanilanmasinda Ayriştirilmiş Çift Düzenlileştirme
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Türkiye, 7 - 10 Temmuz 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/siu71813.2026.11636630
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: decoupled regularization, Periodic systems, system identification, Tikhonov regularization
- Hacettepe Üniversitesi Adresli: Evet
Özet
This paper considers the phase-wise identification problem of linear time-periodic (LTP) systems. In conventional single-coefficient regularization approaches, the phase-to-phase variations of the state matrix (A) and the input matrix (B) are modeled under the same smoothness assumption. However, this assumption may limit estimation performance in hybrid dynamical systems where these two parameter groups exhibit physically different variation characteristics. To address this issue, a decoupled dual-regularization approach using independent penalty coefficients for the state and input matrices is proposed. The proposed method is evaluated on a synthetic system excited by a chirp signal, using separate datasets of 10 cycles each for training, validation, and testing. Numerical results show that the proposed approach provides a clear improvement over unregularized least-squares estimation and a consistent improvement over standard Tikhonov regularization.