System Identification
Table of contents
Sparse Linear Regression
Spare Linear Regression follows the concept of parsimonuous modeling to identify nonlinear dynamics with reduced model complexity. It is implemented via an l1-regularized regression. For the identification of nonlinear dynamical systems the approach was proposed under the name SINDy. The chair has focused on further developing the idea to allow its application to industrial problems with limited measurements.
Selected Publications
- E.Burgin, F. Thiele, C. Dehombreux, B. Garnier, X. Manuel-Juanpere, H. Pfifer: Identification of Nonlinear Sloshing Dynamics Using Operational Manoeuvres
in IfacsOnline2026 - C. Dehombreux, F. Thiele, E. Burgin, B. Garnier, X. Manuel-Juanpere, H. Pfifer, P. Acquatella: Real-Time System Identification for Complex Disturbance rejection and Versatile Missions
Adaptive Methods
Adaptive methods are suitable for identifying unknown and potentially nonlinear dynamics and external disturbances based on limited measurements taken during normal operations of the spacecraft or aircraft. The identified dynamics can then be used directly to augment the existing control laws to further increase the system’s performance. Currently, the chair is focused on the identification of periodic disturbances (vibrations) by combining spectral analysis for frequency estimation with a gradient descent algorithm for identification and control augmentation.
Selected Publications:
- F. Thiele, I. Fernandez-Imana, X. Manuel-Juanpere, H. Pfifer: Adaptive Control for Vibration Attenuation of a Laser Communication Terminal at ESA GNC 2023
- F. Thiele, I. Fernandez-Imana, X. Manuel-Juanpere, H. Pfifer : Robust Attenuation of Micro-Vibrations in Laser Communication Using Adaptive Control in CEAS Space Journal (Preprint)