“…To be more precise, the bound is either a bound on the 2-norm of E for the case where the structure of E is unknown (unstructured uncertainty) or a bound on the maximal perturbation in the entries of the uncertain matrix for the case where E linearly depends on parameter variations (structured parametric uncertainty). If those bounds are not exceeded and if A is already D-stable, then the eigenvalues of A ‡ E keep lying in D. The bounds, which are called robustness bounds or robust D-stability bounds, are deduced from the various NSCs for nominal matrix root clustering mentioned above, derived from the Lyapunov approach (Patel and Toda 1980, Sezer and S Ï iljak 1989, Yedavailli 1993b, Luo et al 1993, Bakker et al 1993, Chouaib and Pradin 1995, from the logarithmic approach (only available for unstructured uncertainty (Wang and Lin 1992, Chouaib and Pradin 1994 or from the LMI approach (Bachelier and Pradin 1998 b). Also, especially for structured uncertainty, the reader can also consult the results proposed by Luo et al (1996) as well as the good work achieved by Yedavalli (1993 a) and improved by Gardiner (1997).…”