Abstract
This paper presents a new framework for closed-loop identification of linear continuous-time systems directly from the sampled I/O data based on trial iterations. The method achieves identification through ILC (iterative learning control) concepts without any knowledge of compensators in the loop. The robustness against measurement noise is achieved through both projection of continuous-time I/O signals onto a finite dimensional space and noise tolerant learning algorithms. Its effectiveness is demonstrated through numerical examples for a linear, non-minimum phase and unstable plant.
