IDENTIFICATION OF CONTINUOUS-TIME MIMO SYSTEMS VIA TIME MOMENTS

The traditional approach of estimating parameters of rational transfer functions by minimizing the equation error gives erroneous results wh en the equation error is not white. when applied to MIMO systems, this formulation inflates the unknown parameter vector leading to reducibl e estimates. The present work aims at tackling these problems, and a n ew approach towards consistent irreducible model identification of con tinuous-time MIMO systems is proposed. A new class of models based on time moments of system impulse response are proposed for this purpose. To eliminate the undermodeling error resulting from the use of such m odels, we propose an iterative, but time-recursive algorithm, in which the estimation error is reparametrized by an iterative filtering of i nput measurements. Relevant implementation issues are discussed, and c onditions for convergence of this algorithm are derived. Numerical sim ulation experiments are carried out, and results are presented to subs tantiate the proposals.