Models and Predictions of Scattered Radio Waves on Rough Surfaces

Scattering of radio waves on rough surfaces is investigated using ray tracing techniques, which results in a sinusoidal model with time varying amplitudes. An AR(d) model with nonzero mean is proposed to characterize and predict the time variation of the amplitudes. A covariance sequence based method is proposed to estimate the autoregressive coefficients from the channel observations. An adaptive channel predictor using a Kalman filter is proposed to predict the complex amplitudes of the scattering signal. The proposed method outperforms other sinusoidal modeling based channel predictors and linear predictors with single scattering scenarios in simulations.