Model quality evaluation in set membership identification (Q1362433)

The authors consider identification of causal, single-output, linear time-invariant, discrete-time systems, which do not necessarily belong to the model class used. This case leads to a nonstandard set membership identification problem. The quality evaluation of different model classes is measured by the conditional radius of information, which is a generalization of the radius in the standard set membership identification theory. Upper and lower bounds on the conditional radius are derived. In particular, explicit formulas for the upper and lower bounds are given for the cse of the Hardy space \(H_2\) identification. An almost-optimal identification is derived, giving identification errors within the range of these bounds. A numerical example shows that the obtained bounds may be sufficiently tight to allow their practical use.