Overview
Learn subspace-based identification techniques like N4SID, MOESP, and CVA for fitting dynamical models to noisy data in this 19-minute technical tutorial. Explore how subspace methods relate to Dynamic Mode Decomposition (DMD) while working with real-world data containing noise and disturbances. Follow along with practical demonstrations covering data analysis through spectra examination, frequency-domain estimation, and subspace estimation implementation. Master key concepts including residual analysis, singular value spectrum interpretation, simulation validation, and Bode plot generation. Compare results between noisy and noise-free scenarios, and understand the differences between subspace methods and Prediction Error Methods (PEM). Access the complete tutorial notebook on JuliaHub and utilize tools like JuliaSim and ControlSystemIdentification for hands-on practice. Gain practical experience with system identification techniques while leveraging the Julia programming language's capabilities for dynamical model estimation.
Syllabus
Subspace id intro
The noisy data
Spectra of data
Frequency-domain estimate
Subspace estimation
Residual analysis
Singular value spectrum
Simulation
Bode plots
Try without noise
Comparison to PEM
Taught by
JuliaHub