Koopman Explicit Model Following Control and Koopman Inverse Dynamics Control for Nonlinear Systems
IEEE Transactions on Industrial Informatics, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1109/tii.2026.3705896
- Dergi Adı: IEEE Transactions on Industrial Informatics
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Compendex, INSPEC, Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: Data-driven control, explicit model following control, extended dynamic mode decomposition, inverse dynamics control, Koopman operator theory, least-squares methods, nonlinear control systems
- Hacettepe Üniversitesi Adresli: Evet
Özet
Nonlinear industrial systems are traditionally controlled using dynamic inversion-based methods such as explicit model-following control and inverse dynamics control to achieve high-performance tracking through analytical model inversion. However, a precise analytical model is one of the major constraints of these methods in complex or uncertain scenarios. This article proposes a data-driven dynamic inversion framework that extends classical nonlinear control allocation principles using Koopman operator theory. A lifted linear predictor identified from data enables closed-form computation of control inputs without requiring analytical dynamics or online optimization. Two complementary frameworks are developed: Koopman Explicit Model Following Control), which combines Koopman-based feedforward allocation with stabilizing feedback, and Koopman Inverse Dynamics Control, where control allocation is performed directly inside the feedback loop via lifted state shaping. The resulting controllers preserve the structure and interpretability of classical inversion-based control while enabling data-driven implementation. Stability analysis establishes input-to-state stability under Koopman approximation errors. Simulation studies on Duffing and Van der Pol oscillators demonstrate effective regulation and numerically well-conditioned control input generation. The proposed framework provides a computationally efficient alternative to optimization-based Koopman control suitable for real-time industrial applications.