Analysis and performance enhancement of network coding for wireless networks
Mane Pallavi Ramesh · INFLIBNET · 2014
Keywords: Network Coding, Wireless Networks, Random Linear Network Coding, Throughput, Bandwidth Utilization, Reed Solomon Codes, Packet delivery delay, Packet loss, Nakagami-m fading channel. In recent years, wireless network has become a necessity rather than a luxury. A wireless network plays a significant role to meet our basic need of providing connectivity with mobility. Current wireless networks struggle to meet throughput and reliability demands placed on them by today’s wireless applications. Network Coding is proposed to be a solution to these problems. This dissertation is rooted in this new area of Network Coding. It explores two types of network coding: centralized network coding and decentralized network coding. Centralized network coding requires knowledge of network topology and needs to use resources to store and update the network dynamics. Decentralized network coding works independent of the network topology and is considered to be robust to network changes and channel conditions. The contributions of this thesis are multifold. It develops a model to study network coding benefits for unicast flows in Wireless Local Area Network (WLAN) with wheel topology based on queuing theory. It provides an analysis of a centralized network coding technique with relevant comparison over the currently used store and forward (SF) technique at saturated and unsaturated traffic conditions. Also packet losses due to channel impairments are taken into consideration in the analysis Later, the work focuses on decentralized network coding. It investigates the cardinality aspects of coding vector set used in Random Linear Network Coding (RLNC) over a finite field for a typical single broadcast network. Probability of selecting a valid coding vector from a set of vectors defined over a finite field is also computed. It highlights the advantages of modifying the coding vector generation strategy in RLNC by using a Vandermonde matrix. It discusses the cardinality aspects of coding vector set of this modified RLNC technique (RLNCV) as well as probability of selecting valid coding vectors and provides appropriate comparisons between the RLNC and RLNCV. Decentralized network coding for data dissemination to multiple users in wireless networks by modeling the channel as Nakagami-m fading in the presence of Additive White Gaussian Noise (AWGN) is also investigated.