White Noise vs. Wiener Process Models in Filtering and Control.

摘要

Continuous time stochastic system models involving a Wiener process model for the additive noise on the observed process are inappropriate in actual use when nonlinear Ito integrals in the observed data have to be instrumented, and thus much of current nonlinear filtering theory is of questionable practical utility. White noise models would appear to be more appropriate and this is illustrated by a canonical problem in system identification, both in terms of band-limited approximation of Ito integrals and in terms of direct white noise integrals.