Use of bicausal weighting sequences in least squares identification of open-loop unstable dynamic systems

Identification of weighting sequences rather than transfer functions offers certain advantages. In open-loop unstable systems such sequences are divergent and thus cannot form the basis of an identification procedure. The paper circumvents this difficulty by proposing the use of bicausal weighting sequences. It also considers the use of weighting sequences in the derivation of transfer function models and proposes a systematic means of determining model order. The results are illustrated by means of numerical simulation.