Resilience of cellular networks: An analysis of multi-connectivity, beamforming and infrastructure sharing

In the last decades, wireless communication has become one of the critical infrastructures - especially in sectors such as the power grid, critical manufacturing and finance. With an increasing likelihood of (natural) disasters due to climate change, ensuring a reliable and fast internet connection is crucial. As wireless technology advances to accommodate novel applications and meet the increasing demand, it is important to understand and analyse network performance to enhance the resilience of networks in case of disruptions to the network.
In this thesis, we explore how various approaches in wireless communications can improve the resilience and performance of cellular networks. Our focus is on multi-connectivity and beamforming, introduced in 5G technology, as well as on infrastructure sharing at a national level. We analyse how these different approaches can improve network resilience and show how we can effectively model these networks.

One method to improve resilience and increase performance in terms of reliability and throughput is multi-connectivity, where users connect to multiple base stations at once. In Chapter 3 we propose a cellular network model based on the AB geometric random graph to examine the structure of cellular networks with multi-connectivity, where users connect to their closest k base stations. Given this model, we derive the degree distribution for two different base station deployments. Continuing with the developed graph model for multi-connectivity, we investigate in Chapter 4 the performance implications of multi-connectivity, in particular on per-user throughput and outage probability. Contrary to previous findings, our analysis reveals that increasing the degree of multi-connectivity generally decreases per-user throughput. However, multi-connectivity improves network resilience against failures and contributes to a fairer distribution of channel capacity. To exploit the benefits of multi-connectivity we recommend a targeted multi-connectivity approach, for example for users who are likely to benefit most, such as those located at the edges of the network. We explore these association schemes with multi-connectivity in Chapter 5, in a mmWave network with beamforming antennas at the base station. We formulate an optimization problem to determine an optimal degree of multi-connectivity per user. The numerical analysis indicates that the optimal degree of multi-connectivity depends on user density, rate requirements, and beam configurations. Based on these observations, we propose the heuristic user association scheme beam-align, which operates locally at each base station and only uses only local signal quality information. Our heuristic performs close to optimal compared to more complex schemes, and is robust to imperfect conditions such as rain, making it a practical solution for mmWave networks.

Another approach to improve resilience is national roaming, where mobile network operators collaborate to extend coverage and capacity, both in normal operations and under disasters. Chapter 6 investigates to what extent national roaming can improve resilience. By analysing data from the Netherlands, we find that full national roaming significantly enhances both coverage and capacity. We highlight the substantial benefits of operator collaboration and show that shared infrastructure can greatly improve network performance and resilience across various regions and operators. However, data on base station locations in the Netherlands revealed that base stations of different operators are often located on the same cell-tower. Chapter 7 investigates the effect of these co-located base stations on the benefits of infrastructure sharing. We develop a model to understand how co-location influences the optimal operational radius of a network and the associated sharing gains and find that co-location is a crucial factor in maximizing the benefits of resource sharing. While sharing benefits smaller operators consistently, larger operators may not always gain from co-location, especially in cases of unequal base station and user densities between operators.


Finally, Chapter 8 addresses the problem of graph reconstruction, a useful technique for understanding network topology. In this chapter, we extend the simple algorithm to handle geometric random graphs and prove that reconstruction of these graphs can be done almost edge-optimal. Additionally, we show that only a linear order of queries (i.e., asking the network whether a connection exists between two nodes) is sufficient to detect a large fraction of non-edges in a network.