Beam alignment of arrays at this scale is challenging due to the overhead in determining the optimal transmit and receive beams. Existing beam alignment schemes proposed for mmWave applications are largely based on beam sweeping and make use of hierarchical beam codebook for efficient algorithm design, which requires the use of wide beams at the start. Such wide beams suffer low antenna gains. They are also more susceptible to Doppler spread, which is another challenge for beam sweeping based method. It should be noted that these solutions do not make use of any side information. In vehicular environments, a large amount of side information can be available from various sources, e.g., automotive radars, visual cameras, LIDARs, or even DSRC devices. 

Motivated by this fact, we are developing a low overhead beam alignment algorithm leveraging the side information obtained from various sources using a learning based approach. The side information from sensors also can be used to predict possible millimeter wave signal blockages in advance and allow to have continuous, reliable data transmissions.

Mmwave Fingerprinting