Overhead, uncertainty, and interference in wireless networks
Date
2014
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
In general, the performance of many wireless systems is approaching the fundamental limits on transmission capacity. For example, current commercial wireless standards such as 3GPP LTE-A and IEEE 802.11ac have a near-optimal physical layer. In order to meet the ever growing demand for capacity, other directions for improving network performance must be found. In most existing research on wireless networks, overhead, the "non-data'' portion including coordination, control signaling and other costs of serving different purposes, is assumed to be negligible. However, the final application throughput could be much lower than the theoretical bounds as a result of overhead, especially in large and dynamic networks. Therefore, it is critical to quantitatively analyze the overhead in wireless networks, which could provide clear insights on the performance in practical systems and could help to identify opportunities for improvements in their designs. Surprisingly, the fundamental limits on overhead are largely unknown, and the framework needed to design overhead-aware systems has not been adequately investigated. In addition, interference is one of the main performance-limiting factors in most future wireless applications. Conventional "interference avoidance'' techniques might not be feasible because the degrees of freedom (for example, bandwidth, number of orthogonal codes, and time) might be limited. Although the interference can be mitigated quite efficiently with centralized control, existing approaches are usually very sensitive to channel uncertainties; if the knowledge of the channel state information is imperfect, the system performance could be severely degraded. Also, collecting accurate information incurs a significant amount of overhead due to the time-varying nature of the wireless medium. Thus, it is imperative to jointly consider overhead, uncertainty, and interference. In this dissertation, we investigate practical and overhead-aware designs that can achieve better performance in a realistic networking context. We start with a simple, single-user, two-hop cooperative relaying network model. For this model, we first prove that M -group cooperation is the optimal distributed space-time block coding strategy when neither central control nor inter-relay communications is permitted. Then, we consider the relay selection problem where a small and acceptable amount of overhead is allowed. The tradeoff between the feedback overhead and the performance is investigated via rate distortion theory. Compared to existing research, which is usually highly dependent on the specific implementation approaches, the analysis presented here addresses the fundamental tradeoff of a general network. Using our theoretical results, we also compare practical centralized and decentralized relay selection schemes in terms of spectral efficiency. Then, interference-limited networks with multiple concurrent transmissions are studied. We analyze and compare the performance of cooperative and non-cooperative schemes. Although cooperation among relay nodes increases the reliability of point-to-point transmission, it also produces a higher level of interference and degrades the overall performance of a multi-user network. The tradeoff between cooperative gain and the additional interference is investigated, and a criterion which determines whether we should cooperative or not is derived. We next focus on multi-hop linear networks, which have one or more intermediate nodes along the path that receive and forward information via wireless links. Instead of assuming equal hop distances, we propose a novel model that permits randomness in the node locations, and then we determine the optimum number of hops for maximizing the end-to-end spectral efficiency. Then, for a multi-hop linear network with cooperative relays, a relay deployment strategy is proposed and studied. After that, for downlink multi-user networks, we present a novel quantization technique, sparse coding quantization (SCQ), which is an extension of classic vector quantization (VQ) and provides a balance between performance and complexity. In particular, the computational complexity of conventional VQ can be significantly reduced by applying SCQ, with a negligible reduction in performance. Comparisons among different quantization techniques are also provided. Beside considering specific quantization schemes, we also study the overhead-performance tradeoff for general MU-MIMO systems by applying a rate distortion framework. Finally, we investigate robust a user pairing problem for a heterogeneous network in the presence of channel uncertainty. Different definitions of robustness and uncertainty are considered to formulate the corresponding optimization problems. We develop an algorithm that is robust to uncertainty in channel measurement and thereby performs well in practical systems. Simulation results validate the robustness of the proposed method.