Stable Approximations of Unstable Models

This paper considers the following problem: starting from noisy data a model for the transfer function is estimated. For a finite number of data points, the model is not guaranteed to be stable, even when the true linear system is known to be stable. D'haene et al. [2006] describes an iterative procedure for calculating a stable approximation of an unstable model. This paper presents an improved algorithm resulting in significantly smaller approximation errors. In addition stable minimax solutions can also be generated.