Control-oriented model validation

Model validation is a means of assessing model quality with respect to experimental data. In a robust control context this amounts to determining whether or not a perturbation model is consistent with observed data. T his paper offers a survey of recent results on model validation for H-infinity compatible perturbation models. Several model paradigms and experiment frameworks are considered. The first is a frequency-domain setting for the data. More recent results deal with discrete-time models and time-domain experimental data. The most relevant framework , continuous-time systems and models, with discrete-time sampled data, is dealt with last. Each of the model invalidation tests offered involve tractable convex optimization problems.ISSN:1474-667