In the context of this PhD Thesis defense, Biljana Bojovic presents her work realized at CTTC.
Abstract:
This PhD thesis has the characteristic to span over a long time because while working on it, I was working as a research engineer at CTTC with highly demanding development duties. This has delayed the deposit more than I would have liked. On the other hand, this has given me the privilege of witnessing and studying how wireless technologies have been evolving over a decade from 4G to 5G and beyond. When I started my PhD thesis, IEEE and 3GPP were defining the two main wireless technologies at the time, Wi-Fi and LTE, for covering two substantially complementary market targets. Wi-Fi was designed to operate mostly indoor, in unlicensed spectrum, and was aimed to be a simple and cheap technology. Its primary technology for coexistence was based on the assumption that the spectrum on which it was operating was for free, and so it was designed with interference avoidance through the famous CSMA/CA protocol. On the other hand, 3GPP was designing technologies for licensed spectrum, a costly kind of spectrum. As a result, LTE was designed to take the best advantage of the spectrum while providing the best QoE in mainly outdoor scenarios. The PhD thesis starts in this context and evolves with these two technologies. In the first chapters, the thesis studies radio resource management solutions for standalone operation of Wi-Fi in unlicensed and LTE in licensed spectrum. We anticipated the now fundamental machine learning trend by working on machine learning-based radio resource management solutions to improve LTE and Wi-Fi operation in their respective spectrum. We pay particular attention to small cell deployments aimed at improving the spectrum efficiency in licensed spectrum, reproducing small range scenarios typical of Wi-Fi settings. IEEE and 3GPP followed evolving technologies over the years: Wi-Fi has grown into a much more complex and sophisticated technology, incorporating the key features of cellular technologies, like HARQ, OFDMA, MU-MIMO, MAC scheduling, and spatial reuse. On the other hand, since Release 13, cellular networks have also been designed for unlicensed spectrums. As a result, the two last chapters of this thesis focus on coexistence scenarios, in which LTE needs to be designed to coexist with Wi-Fi fairly, and NR, the radio access for 5G, with Wi-Fi in 5 GHz and WiGig in 60 GHz. Unlike LTE, which was adapted to operate in unlicensed spectrum, NR-U is natively designed with this feature, including its capability to operate in unlicensed in a complete standalone fashion, a fundamental new milestone for cellular. In this context, our focus of analysis changes. We consider that these two technological families are no longer targeting complementarity but are now competing, and we claim that this will be the trend for the years to come. To enable the research in these multi-RAT scenarios, another fundamental result of this PhD thesis, besides the scientific contributions, is the release of high fidelity models for LTE and NR and their coexistence with Wi-Fi and WiGig to the ns-3 open-source community. ns-3 is a popular open-source network simulator, with the characteristic to be multi-RAT and so naturally allows the evaluation of coexistence scenarios between different technologies. These models, for which I led the development, are by academic citations, the most used open-source simulation models for LTE and NR and have received fundings from industry (Ubiquisys, Wi-Fi Alliance, SpiderCloud, Interdigital, Facebook) and federal agencies (NIST, LLNL) over the years.
Links