Since the biases caused by noise can be eliminated, then the problem of set-membership identification turns to be a problem of a model fitting, the same as for a deterministic linear system. If we increase the degree of the nominal part of the system and at the same time weaken the unmodeled dynamics, then the estimated bound of the parameters set will be reduced, and a more accurate estimation of the system can be reached. It is worthy to notice that the method presented here needs no any previous knowledge of the noise model. REFERENCES [l] E. Fogel, "System identification via membership set constrains with energy constrained noise," IEEE Tmns. Auromat. Contr., vol. AC-24, [2] R. L. Kosut, M. K. Lau, and S. P. Boyd, "Set-membership identification of systems with parametric and nonparametric uncertainty ," IEEE Trans. Automat. Contr., vol. 37, pp. 929-941, 1992. [3] R. C. Younce and C. E. Rohrs, "Identification w i t h nonparametric uncertainty," [4] C.-B. Feng and W.-X. Zheng, "On-line modified least-squares parameter estimation of hear systems with input-output data polluted by measurement noise," in Pmc, 8th IFAC Symp. Identification, System and Parameter Esrimation, Absftaet-h this paper, the expansion of det( A + B + C) is given. The robustness problem for discrete-time state-space systems with uncertain parameters is investigated by applying the Sehw stability theory of interval polynomials. A stability robustness criterion for the uncertain systems is derived. Illustrative exampks are presented.