Energy efficiency optimization in millimeter wave backhaul heterogeneous networks

Abstract

Within the last few years, there has been a massive growth in the number of wireless devices and internet connections. This is expected to continue during the next few years. To satisfy the resulting high data traffic demands, dramatic expansion of network infrastructures as well as fast escalation of energy demands are expected. Meanwhile, there has been a growing concern about the energy consumption of wireless communication systems and their global carbon footprint. To that end, future wireless systems must satisfy three main requirements. Firstly, they must provide users with very high throughput. Secondly, they must be able to provide seamless connectivity as well as ubiquitous access to the expected enormous number of users. Finally, they must achieve the first two points with less energy consumption. The requirements can be summarized into the joint optimization of energy efficiency (EE), user association and backhaul (BH) flow assignment, which remains a fundamental objective in the design of next generation networks. This thesis consists of two studies on EE maximization in heterogeneous networks (HetNets). In the first study, it is assumed that each user has already been associated to a single base station (BS). Under this setting, We consider enforcing a strict throughput demand on all user equipment (UEs), called joint EE, power, and flow control (JEEPF), versus allowing an acceptable range of demands for each, called joint EE, power, flow control, and throughput (JEEPFT). This minor change causes a drastic difference in the formulation of both problems. JEEPF is convex while JEEPFT is quasiconvex, for which we propose a bisection method-based approach. In the second study, the problem of user association is added to the joint optimization of EE, power and BH flow control, and an energy efficient user association, power and flow control (EEUAPF) algorithm is proposed. The original EEUAPF optimization problem is a non-convex mixed integer programming problem, and therefore NP-hard. We show how this non-convex problem can be tailored into a form that can be approached using a classical mathematical programming technique called column generation and convex programming to derive the optimal solution with a low complexity. Simulation results are used to demonstrate the EE gains of the proposed approaches in both studies.