End-to-End Design and Evaluation of mmWave Cellular Networks

The next generation of cellular networks (5G) is being designed to provide unprecedented performance in mobile scenarios, with an increase in capacity, ultra-low latency and a massive number of connections. This will require the integration of novel technologies in more advanced and complex networks. Millimeter wave (mmWave) communications are considered as a key enabler for ultra-high datarates and low latency, thanks to the massive amount of available bandwidth at such high frequencies. Nonetheless, there are a number of challenges that must be solved before this technology can be deployed, mainly related to the high propagation loss, the need for directional communications and the blockage. This thesis provides system-level solutions to make mmWave mobile networks more reliable, robust and better performing. Notably, we consider mmWave links as a part of more complex, end-to-end networks, in which the quality of experience that the end user perceives is the result of the interaction among the variability and unreliability of the mmWave channel, the full protocol stack, and the deployment strategy of the wireless network. To this end, we develop and describe a tool for the simulation of end-to-end mmWave cellular networks that, combined with analysis and experimental results, makes it possible to consistently evaluate how these systems behave in their entirety. The main research areas that this thesis explores are the design and evaluation of (i) architectures for mmWave systems, in terms of mobility and beam management, and wireless backhaul solutions; (ii) protocols for end-to-end connectivity over mmWave networks; and (iii) intelligent and data-driven optimizations in cellular networks. Among other results, we highlight the importance of multi connectivity for mmWave systems, in the access network and at the transport layer, discuss the tradeoffs of beam management in 3GPP NR, propose how to update protocols at the transport layer for an improved end-to-end performance, and evaluate practical approaches for the integration of intelligent techniques in 5G networks