In this paper, our purpose is to estimate time-varying Rayleigh fading channels in Orthogonal Frequency Division Multiplexing (OFDM) mobile systems. When the fading channel is approximated by an AutoRegressive (AR) process, the direct estimation of the model parameters from the noisy observations available at the receiver may yield biased val-ues. To avoid this drawback, the joint estimation of both the channel and its AR parameters must be addressed. Existing solutions to this dual estimation issue require Kalman filter-ing. This kind of filtering is optimal in the H2 sense provided that the underlying state-space model is accurate. Moreover, the initial state, the driving process and the measurement noise must be independent, white and Gaussian. However, in real cases, these assumptions may no longer be satisfied. To relax them, we propose to consider a structure based on two cross-coupled H∞ filters. This method makes it possible to provide robust estimation of the fading channel and its AR parameters