Abstract

Wireless networks consist of a number of nodes, which communicate with each other over wireless channels. Unlike wired-networks, wireless networks have limited bandwidth, and are much more susceptible to environmental effects such as wireless interference. As a result, it is difficult to transmit information reliably at high data rates. The problem is further compounded by the quality of service (QoS) requirements such as minimum delay and maximum throughput imposed by current and future applications. That said, recent advances in coding techniques, communication protocols and architectures give the promise of future wireless networks that will proliferate high quality wireless applications.
Network coding (NC) and successive interference cancellation (SIC) have been shown to improve the throughput of multi-hop wireless networks (MWNs). NC enables a node to transmit multiple packets concurrently as a single coded packet, while SIC allows multi-packet reception (MPR) by removing interference. However, emphasis of the work done so far has been determining maximum throughput of such networks without giving consideration to QoS requirements. Maximization of the throughput may lead to paths in the network that experiences very high packet delays. The objective of this thesis is the minimization of average packet delay in a MWN for a given traffic demand matrix with joint application of NC and SIC techniques.
We formulate a cross-layer optimization that performs scheduling, routing, and more importantly capacity allocation in a way that the average packet delay is minimized. Our optimization model considers thoroughly all feasible NC and MPR opportunities in the network and allows nodes to encode up to 4 packets together.
We consider a network that uses conflict-free scheduling and has multi-path routing capability. The method is valid both in the presence and the absence of opportunistic listening on any wireless network topology and any pattern of traffic. We present numerical results to evaluate the performance of the proposed scheme. The results are also compared to that of the previous studies that treat NC and SIC separately. Our findings indicate that significant throughput improvement can be achieved by a winning combination of NC and SIC techniques.