A packet encoding scheme for wireless network coding

Abstract

Network coding for wireless networks is a promising technique to improve network throughput. As a new packet forwarding architecture, COPE is the first to bring a network coding layer to the conventional TCP/IP network stack. However, COPE’s packet encoding scheme does not consider optimality. This thesis describes Local Optimal Search (LOS), a new packet encoding scheme for COPE. We present a formal description of the packet encoding problem related to wireless network coding. We propose the concept of coding graph to model the relations among packets related to a packet encoding problem. With the assistance of coding graph, we design LOS, which always searches for a local optimal set of packets to be encoded together for each transmission. We also design a new metric, Local Coding Gain (LCG), to measure the effectiveness of a set of selected packets. To set up an evaluation environment, we revise and implement COPE in NS2. We adjust the original NS2 network stack for 802.11 wireless networks and insert a new network coding layer between LLC sublayer and MAC sublayer. NS2 simulation results on this coding layer are close to the experiment results of a wireless mesh network testbed. We prove that LOS always finds the largest number of packets to be encoded for a single transmission. We conduct simulations on four basic coding scenarios in NS2. Compared with COPE, the simulation results show that the number of link level transmissions is reduced up to 6.19% for delivering a certain number of packets in the transport layer.