Performance Analysis of Cellular and Ad-Hoc Sensor Networks: Theory and Applications

Fifth-generation (5G) mobile networks have three main goals enhanced mobile broadband (eMBB), massive machine-type communication (mMTC) and ultra-reliable low latency communication (URLLC). The performance measures associated with these goals are high peak throughput, high spectral efficiency, high capacity and mobility. Moreover, achieving ubiquitous coverage, network and device energy efficiency, ultra-high reliability and ultra-low latency are associated with the performance of 5G mobile networks. One of the challenges that arises during the analysis of these networks is the randomness of the number of nodes and their locations. Randomness is an inherent property of network topologies and could occur due to communication outage, node failure, blockage or mobility of the communication nodes. One of the tools that enable analysis of such random networks is stochastic geometry, including the point process theory. The stochastic geometry and Poisson point theory allow us to build upon tractable models and study the random networks, which is the main focus of this dissertation. In particular, we focus on the performance analysis of cellular heterogeneous networks (HetNet) and ad-hoc sensor networks. We derive closed-forms and easy-to-use expressions, characterising some of the crucial performance metrics of these networks. First, as a HetNet example, we consider a three-tier hybrid network, where microwave (µWave) links are used for the first two tiers and millimetre wave (mmWave) links for the last tier. Since HetNets are considered as interference-limited networks, therefore, we also propose to improve the coverage in HetNet by deploying directional antennas to mitigate interference. Moreover, we propose an optimisation framework for the overall area spectral and energy efficiency concerning the optimal signal-to-interference ratio (SIR) threshold required for µWave and mmWave links. Results indicate that for the µWave tiers (wireless backhaul) the optimal SIR threshold required depends only on the path-loss exponent and that for the mmWave tier depends on the area of line-of-sight (LOS) region. Furthermore, we consider the average rate under coverage and show that the area spectral and energy efficiency are strictly decreasing functions with respect to the SIR threshold, thereby concluding that the lowest possible SIR threshold available in the system is optimal.
Second, in ad-hoc sensor networks, coverage probability is usually defined according to a fixed detection range ignoring interference and propagation effects. Hence, we define the coverage probability in terms of the probability of detection for localisability. To this end, we provide an analysis for the detection probability and S-Localisability probability, i.e. the probability that at least S sensors may successfully participate in the localisation procedure, according to the propagation effects such as path-loss and small-scale fading. Moreover, we analyse the effect of the number of sensors S on node localisation and compare different range based localisation algorithms.