In this paper, we are concerned with the solution of optimal L-infinity-mod
el-reduction problems using genetic algorithms (GAs). More precisely, we pr
esent an approach to facilitate using GAs to effectively search optimal red
uced-order models for high-order linear time-invariant systems such that th
e L-infinity-norm of approximation error is minimized. The proposed approac
h has the following distinct features: (i) the parameter space for GA searc
h is bounded- (ii) the finite and infinite zero structures, as well as the
stability of the original system is retained in the reduced-order models; a
nd (ii) the GA solution accuracy can be greatly enhanced. The first two fea
tures are achieved through representing a reduced-order model in a transfer
function parameterized in terms of Schur and anti-Schur polynomials, where
as the third feature is gained by using schemes of region contraction and l
ocal improvement.