An approach to system identification based on risk minimization and behaviours

Abstract

The authors propose and investigate a principle for system identification in the behavioral framework. The first step in the approach is to identify the deterministic part of the observed trajectories. This is done by using risk minimization methods. The identified trajectories are then used in an exact modeling procedure given by A.C. Antoulas and J.C. Willems (1991). It is noted that the risk minimization approach makes it possible to identify worst-case guarantees on the achievable performance in a dynamic system identification context. >