Computation of the kernel of linear fractional representations

This paper presents several algorithms for computing the kernel of a linear fractional representation (LFR). This problem is shown to be multi-criteria because, in addition to numerical stability, the complexity of the kernel (/spl Delta/-block size) and its well-posedness radius must also be optimized. Depending on the treated LFR, the proposed algorithms are more or less efficient; their respective advantages are discussed.