On a least-squares-based algorithm for identification of stochastic linear systems. (Q2724229)

A direct bias-eliminated least-squares algorithm for the identification of parameters of a linear time-invariant discrete-time system is presented and analysed. A key feature of the algorithm is that a consistent estimate of the system parameters can be obtained in the presence of arbitrarily colored noise, without prefiltering observed data and without introducing a higher-order augmented system. Hence the proposed routine has a simple structure and requires a moderate amount of computations, as compared with other methods of such kind. The efficiency of the approach is illustrated by means of simulation examples.