Channel Prediction in Rapidly Time-Varying OTFS Systems Using FAR Models

Although channel information can be estimated in the delay–Doppler (DD) domain using orthogonal time frequency space (OTFS) modulation, traditional channel estimation techniques often provide outdated information due to rapid movements. In this letter, we formulate the channel prediction problem as a functional-coefficient autoregressive (FAR) model and propose a debiased estimator, to capture the functional dependence of the channel coefficients across time steps, as well as the innovation variance, aiming to predict future coefficients prior to their realization. Simulation results demonstrate that the proposed method follows the channel dynamics and achieves superior estimation performance compared with state-of-the-art methods.