Energy-Efficient Algorithms and Access Schemes for Small Cell Networks

Dense deployment of small base stations (SBSs) brings new challenges such as growing energy consumption, increased carbon footprint, higher inter-cell interference, and complications in handover management. These challenges can be dealt with by taking advantage of sleep/idle mode capabilities of SBSs, and exploiting the delay tolerance of data applications, as well as utilizing information derived from the statistical distributions of SBSs and user equipment (UE)-SBS associations. This dissertation focuses on the formulation of mathematical models and proposes energy efficient algorithms for small cell networks (SCN). It is shown that delay tolerance of some data applications can be taken advantage of to save energy in SCN. This dissertation introduces practical models to study the performance of delayed access to SCNs. Operational states of SBS are modeled as a Markov chain and their probability distributions are analyzed. Also, it argues that SCN can be operated to save energy during low traffic periods by taking advantage of user equipments' (UEs) delay tolerance in SCN while providing high access probability within bounded transmission range. Dense deployment of SCNs cause an increase in overlapping SBS coverage areas, allowing UEs to establish communication with multiple SBSs. A new load metric as a function of the number of SBSs in UE's communication range is defined, and its statistics are rigorously analyzed. Energy saving algorithms based on aforementioned load metric are developed and their efficiencies are compared. Besides, UE's delay tolerance allows establishing communication with close-by SBSs that are either in fully active mode or in sleeping mode. Improvements in coverage probability and bitrate are analyzed by considering different delay tolerance values for UEs. Key parameters such as UE's communication range are optimized with respect to SBS density and delay tolerance. The fundamental problem of local versus remote edge/fog computing and its inherent tradeoffs are studied from a queuing perspective taking into account user/SBS density, server capacity and latency constraints. The task offloading problem is cast as an M/M/1(c) queue in which CPU intensive tasks arrive according to Poisson process and receive service subject to a tolerable delay. The higher the proportion of locally computed tasks, the less traffic SCN handles between edge processor and UE. Therefore, low utilization of SCN can be interperted as increased spectral efficiency due to low interference and close UE-SBS distance. Tradeoff between delay dependent SCN utilization and spectral efficiency is evaluated at high and low traffic loads.