Channel Prediction with Cascade AR Modeling

Abstract

Cascade autoregressive (AR) model with three parameters for any channel prediction order p is investigated. Although higher order AR model is better to get more precise future channel prediction profile, as the order of channel prediction goes higher, the number of channel prediction parameters we have to know is increased. Besides that, the pole locations may be extremely sensitive function of the coefficients for high order filter. If we use the cascade form of lower order filters which is already used as one of fading generation methods, we can get more stable AR model for higher order channel prediction. Moreover, due to a Doppler spectrum characteristic which has only two peaks at the positive and negative maximum Doppler frequency, we just have to know the second and the first order filter to make cascade AR model of higher order, that is, we can make higher AR model to predict future channel with only three parameters.