Author: Yuyang Wang

Publisher:

Published: 2020

Total Pages: 346

ISBN-13:

Millimeter-wave (MmWave) vehicular communication enables massive sensor data sharing and various emerging applications related to safety, traffic efficiency and infotainment. Estimating and tracking beams in mmWave vehicular communication, however, is challenging due to the use of large antenna arrays and high mobility in the vehicular context. Fortunately, wireless cellular communication systems have access to vast data resources, which can make beam training more efficient. Data-driven approaches are able to leverage side information and underlying channel statistics to optimize link configuration in mmWave vehicular communication with negligible overhead. In the first part of this dissertation, we develop a situational awareness-aided beam alignment solution using machine learning. Situational awareness, defined as the locations and shapes of the receiver and its surrounding vehicles, can be obtained from sensors to extract environment information and retrieve good beam directions. We formulate mmWave beam selection as a multi-class classification problem, based on hand-crafted features that capture the situational awareness in different coordinates. We provide a comprehensive comparison among the different classification models and various levels of situational awareness. To demonstrate the scalability of the proposed beam selection solution in the large antenna array regime, we propose two solutions to recommend multiple beams and exploit an extra phase of beam sweeping among the recommended beams. In the second part of this dissertation, we develop mmWave vehicular beam alignment solutions with relaxed requirements of connected vehicles and sensor information sharing. The proposed model focuses on designing compressive sensing techniques that leverage the underlying channel angular statistics in site-specific areas using fewer channel measurements. We investigate the problem from an online learning-based approach that optimizes the sensing matrix on the fly and an offline approach that designs the compressive sensing framework using a convolutional neural network. We incorporate hardware constraints of the phased array in the sensing matrix optimization. We investigate structures in frequency-domain channels and propose solutions to optimize power allocated for different subcarriers. Numerical results show that data-driven approaches can achieve accurate link configuration for mmWave vehicular communication with negligible training overhead