Marcin Filo

The concept of Ultra Dense Networks (UDNs) is often seen as a key enabler of the next generation mobile networks. In contrast to the traditional cellular networks, it is foreseen that the UDNs will be in many cases installed without traditional RF planning and proper site selection. One of the main characteristics (and at the same time challenge) of the UDNs is therefore excessive inter-cell interference. Additionally, small cells as one of the main components on UDNs are foreseen to operate in licensed as well as license-exempt frequency bands, and therefore inter-system and inter-tier interference become major concerns. Widely used and well established systems such as LTE, or WiFi are often proposed by the industry and academia to be reused in the context of UDNs. However, as these systems were not designed to deal with problems caused by the UDN deployment, a significant effort is currently being undertaken to adapt them and enable their operation in dense and ultra-dense environments. Despite this tremendous effort, due to the need for backward compatibility, the proposed updates and patches usually provide sub-optimal gains and often lead to significant signaling overheads. In this thesis we highlight some of the main challenges and requirements related to UDNs and then provide an extensive review of state-of-art UDN performance analysis and approaches to medium access control (MAC) design for UDNs. Then we investigate performance limits of regular and irregural UDNs. More specifically, we examine the impact of the relative antenna height between BS and UE antennas on the performance of UDNs. Based on our study, we found that regular networks share many of the same performance behaviour as irregular network. In partivular, we showed that by decreasing the relative antenna heights across the network we can counter the decay of per cell average achievable rate. We explicitly derived the relationship between BS density and relative antenna height and confirmed that both regular and irregular networks share this property. Despite the pessimistic conclusion related to the per cell performance found in the literature, in this work we also show that area spectral efficiency does not necessarily decay to zero as BS density approaches infinity. In terms of the benefit of proper BS site selection, we compare the average per cell rate of regular networks and that of the irregular networks, and we find that proper BS deployment may improve network performance to some extent. Finally, based on the lessons learned, we present and discuss a novel MAC protocol designed for 5G UDN deployments. In contrast to other candidates considered by the industry for UDN deployment, the proposed MAC provides a number of built-in features which improve its efficiency in dense and ultra dense deployments. The multi-channel operation along with the dynamic channel selection constitutes the core of the proposed MAC, limiting performance degradation resulting from high level of inter-cell interference and simplifying network planning. The proposed MAC design is further evaluated through simulations for outdoor deployments in non-coexistence and coexistence scenarios. Our results reveal that the proposed MAC is capable of operating effectively in highly dense deployment scenarios when tuned appropriately. In case of the coexistence capabilities of the investigated design, we show that coexistence with LBT-based systems such as WiFi is also possible, but requires additional tuning to maintain fair channel access for all systems. Lastly,we show that the proposed MAC design outperforms WiFi and LTE (which are commonly considered for UDN deployment) in all considered scenarios. More specifically, our results indicate that area spectral efficiency for the proposed MAC is approximately 500% higher compared to WiFi (IEEE 802.11ac), and 40% higher compared to LTE (excl. CA and MIMO), with improved performance for cell-edge users.