Data driven simulation with applications to system identification
| Authors: | Markovsky Ivan, K.U.Leuven, Belgium
Willems Jan C., K.U.Leuven, Belgium
Rapisarda Paolo, University of Maastricht, Netherlands
De Moor Bart L. M., K.U.Leuven, Belgium |
|---|
| Topic: | 1.1 Modelling, Identification & Signal Processing |
|---|
| Session: | Modelling and Interconnection of Open Dynamical Systems |
|---|
| Keywords: | Deterministic identification, impulse response estimation, oblique projection. |
|---|
Abstract
We present an iterative algorithm for computation of linear time-invariant system responses directly from an exact finite input/output trajectory of that system. In particular, the impulse response can be obtained from data, which opens the possibility to use realization theory in order to retrieve the system parameters. The oblique projection, used in subspace identification, computes free responses which initial conditions form a state sequence. We show that the oblique projection corresponds to a block version of the data driven simulation algorithm. The iterative algorithm, however, has computational advantages over the block algorithm, which makes it attractive for application in subspace algorithms.
