Channel Prediction for Millimeter-Wave V2V Communication Using Autoregressive Models

Vehicle-to-vehicle (V2V) communications using millimeter-wave (mmWave) technologies can offer gigabit-per-second date rates, thus attracting large research interest. One of the challenges, however, is the potentially severe channel aging effect due to high mobility of both the transmitter (TX) and the receiver (RX). This effect may lead to an unaffordable channel estimation overhead and a degradation of system performance. In order to combat channel aging, we propose channel prediction, i.e., future channel state is predicted based on the estimated but outdated channel state information (CSI). In the paper, classical and low-complexity autoregressive (AR) models are used to predict CSI. We evaluate the prediction scheme based on measured mmWave V2V channels at 41 GHz, where TX-RX relative velocities were as high as 120–180 km/h. Simulation results show that the predicted CSI is much more accurate than the aged one, with normalized prediction errors around -13 to -30 dB. Specifically, channel components or delay taps with strong energy are relatively easy to predict, while for weak components caused by e.g., scattering or diffused reflection, it is very difficult to obtain a low prediction error. The research indicates that channel prediction using AR models can effectively overcome the channel aging problem in mmWave V2V communications. Hence, the proposed scheme has a potential of being used in practical systems to ensure high-mobility system performance.