Exploiting Network Coding in Lossy Wireless Networks

Abstract

Users increasingly depend on Wireless LANs (WLANs) for business and entertainment. It is well-known that wireless links are error-prone and require retransmissions to recover from errors and packet losses. Previous work proposed to use network coding (NC) for more efficient MAC-layer retransmissions in WLANs. However, their design is independent from underlying physical layer and MAC capabilities and algorithms. These independent design may result in inefficient utilization of network coding gain, or even impair system performance.This dissertation presents a practical network coding-aided MAC layer retransmission scheme, namely XOR Rescue (XORR). Unlike previous independent network coding design, XORR provides a global approach by integrating the utilization of network coding, adaptation to time-varying wireless channel, fairness, and multi-rate transmission in wireless networks. The main characteristic of XORR is opportunism: each node relies on local information to detect the best transmission/retransmission and exploits the benefits provided by both network coding and multiuser diversity whenever they arise.The contributions of this dissertation are multifold. First, it builds a practical link layer retransmission architecture by integrating network coding and wireless physical and MAC design. Specifically, the system presented in this dissertation is the first to accommodate network coding into complex wireless environments, e.g. time-varying link quality. Second, the work presents novel algorithms and introduces new concepts which may be applicable to other wireless network coding protocols. A Bayesian-learning based estimation scheme for evaluating reception status can providing substantially coding opportunities without extra overheads. A framework of an network coding aware fair opportunistic scheduling is designed with the objective of maximizing the system goodput as well as maintaining fairness.A new coding metric, namely expected goodput, is devised for exploiting the gain of network coding and multiuser diversity. The concept of network coding fairness is proposed, where not only the fair resource share is guarantee but also the performance for every wireless station is improved compared to non-coding scheme. Finally, we present theoretical analysis and extensive simulations. Our results show that XORR outperforms the non-coding fair opportunistic scheduling and 802.11 by 25% and 40%, respectively.