Abstract

The new wireless era in the next decade and beyond would be very different from our experience nowadays. The fast pace of introducing new technologies, services, and applications requires the researchers and practitioners in the field be ready by making paradigm shifts. The stringent requirements on 5G networks, in terms of throughput, latency, and connectivity, challenge traditional incremental improvement in the network performance. This urges the development of unconventional solutions such as network densification, massive multiple-input multiple-output (massive MIMO), cloud-based radio access network (C-RAN), millimeter Waves (mmWaves), non-orthogonal multiple access (NOMA), full-duplex communication, wireless network virtualization, and proactive content-caching to name a few.

Ultra-Dense Network (UDN) is one of the preeminent technologies in the racetrack towards fulfilling the requirements of next generation mobile networks. Dense networks are featured by the deployment of abundant of small cells in hotspots where immense traffic is generated. In this context, the density of small cells surpasses the active users’ density providing a new wireless environment that has never been experienced in mobile communication networks. The high density of small cells brings the serving cells much closer to the end users providing a two-fold gain where better link quality is achieved and more spatial reuse is accomplished.

In this thesis, we identified the distinguishing features of dense networks which include: close proximity of many cells to a given user, potential inactivity of most base stations (BSs) due to lack of users, drastic inter-cell interference in hot-spots, capacity limitation by virtue of the backhaul bottleneck, and fundamentally different propagation environments. With these features in mind, we recognized several problems associated with the performance evaluation of UDN which require a treatment different from traditional cellular networks. Using rigorous advanced mathematical techniques along with extensive Monte Carlo simulations, we modelled and analytically studied the problems in question. Consequently, we developed several mathematical frameworks providing closed-form and easy-computable mathematical instruments which network designers and operators can use to tune the networks in order to achieve the optimal performance. Moreover, the investigations performed in this thesis furnish a solid ground for addressing more problems to better understand and exploit the UDN technology for higher performance grades.

In Chapter 3, we propose the multiple association in dense network environment where the BSs are equipped with idle mode capabilities. This provides the user with a “data-shower,” where the user’s traffic is split into multiple paths, which helps overcoming the capacity limitations imposed by the backhaul links. We evaluate the performance of the proposed association scheme considering general fading channel distributions. To this end, we develop a tractable framework for the computation of the average downlink rate.

In Chapter 4, we study the downlink performance of UDNs considering Stretched Exponential Path-Loss (SEPL) to capture the short distances of the communication links. Considering the idle mode probability of small cells, we draw conclusions which better reflect the performance of network densification considering SEPL model. Our findings reveal that the idle mode capabilities of the BSs provide a very useful interference mitigation technique. Another interesting insight is that the system interference in idle mode capable UDNs is upper-bounded by the interference generated from the active BSs, and in turn, this is upper-bounded by the number of active users where more active users is translated to more interference in the system. This means that the interference becomes independent of the density of the small cells as this density increases.

In Chapter 5, we provide the derivation of the average secrecy rate in UDNs considering their distinct traits, namely, idle mode BSs and LOS transmission. To this end, we exploit the standard moment generating function (MGF)-based approach to derive relatively simple and easily computable expressions for the average secrecy rate considering the idle mode probability and Rician fading channel. The result of this investigation avoids the system level simulations where the performance evaluation complexity can be greatly reduced with the aid of the derived analytical expressions.

In Chapter 6, we model the uplink coverage of mMTC deployment scenario considering a UDN environment. The presented analysis reveals the significant and unexpected impact of the high density of small cells in UDNs on the maximum transmit power of the MTC nodes. This finding relaxes the requirements on the maximum transmit power which in turn allows for less complexity, brings more cost savings, and yields much longer battery life. This investigation provides accurate, simple, and insightful expressions which shows the impact of every single system parameter on the network performance allowing for guided tunability of the network. Moreover, the results signify the asymptotic limits of the impact of all system parameters on the network performance. This allows for the efficient operation of the network by designing the system parameters which maximizes the network performance.

In Chapter 7, we address the impact of the coexistence of MTC and HTC communications on the network performance in UDNs. In this investigation, we study the downlink network performance in terms of the coverage probability and the cell load where we propose two association schemes for the MTC devices, namely, Connect-to-Closest (C2C) and Connect-to-Active (C2A). The network performance is then analyzed and compared in both association schemes.

In Chapter 8, we model the uplink coverage of HTC users and MTC devices paired together in NOMA-based radio access. Closed-form and easy-computable analytical results are derived for the considered performance metrics, namely the uplink coverage and the uplink network throughput. The analytical results, which are validated by extensive Monte Carlo simulations, reveal that increasing the density of small cells and the available bandwidth significantly improves the network performance. On the other side, the power control parameters has to be tuned carefully to approach the optimal performance of both the uplink coverage and the uplink network throughput.