Predictor antennas in action

Abstract

Connected vehicles in large numbers will be expensive in terms of power and bandwidth unless advanced transmit schemes are employed. These would rely on channel state information at transmitter (CSIT), which rapidly becomes outdated for fading vehicular channels. We here evaluate the predictor antenna concept, that solves this problem by using antennas on the outside of vehicles, with one extra antenna in front of the others. Its estimated channel is a scaled prediction for the channels encountered by rearward antennas when they reach that position. We evaluate this concept on a large set of channel sounding measurements from an urban environment. Recent investigations of the correlations of these measurements indicate that the average normalized mean squared errors (NMSEs) of the complex valued channel predictions should be around -10 dB for prediction horizons in space of up to 3 wavelengths. This represents an extension of the attainable prediction horizon by an order of magnitude, as compared to Kalman or Wiener extrapolation of past channel measurements. It represents an accuracy that would enable e.g. accurate massive multiple input multiple output (MIMO) downlink beamforming to vehicles. We here perform predictions on a subset of the measurements with good channel-to-estimation error power ratio (SNR). The approximate true channels are here available and we evaluate the performance on a validation data set. The results confirm that the distribution of the NMSE, over all investigated propagation environments, is close to that obtained by correlation-based models and outperforms the use of outdated channel measurements.