Identification algorithm of nonminimum phase system based on t-distribution assumption

This paper presents a novel method for robustly estimating nonminimum phase time-invariant systems. A nonlinear weighting function that is based on assuming that the residual is λ degree t(λ)-distributed. The conventional approach is the special case of t(λ)-distribution; i.e. when λ=∞. When small λ is used, a large weighting function is used when the residual is small. On the other hand, when the residual is large, a small weighting function is utilized. By doing so, the effect of large residual parts are reduced. Simulation results show that by applying small λ, we can obtain smaller variance estimation results than that when large λ is utilized. Furthermore, by applying small λ we can achieve smaller average estimation error than the estimation result when large λ is applied.