My Master's Thesis

Abstract

Wireless sensor networks comprise anywhere from a few dozen to many thousands of nodes collaborating on a common sensing task. In such networks all nodes must generally assist with data communication by participating in message forwarding. Over the course of the last fifteen years, a large number of novel approaches to solving the challenging problem of routing in ad-hoc and sensor networks have been proposed. Thus, a comprehensive catalog of performance comparison studies is needed.
While some number of performance studies exist, the majority of these are simulation-based and it is therefore believed that real-world experiments are still needed. Furthermore, very few existing studies directly compare non-localized, topology-based protocols against localized protocols based on the geographic position of individual nodes.
The work presented in this thesis, compares Dynamic Source Routing (DSR) against two variants of Greedy Forwarding; one distance-based (Greedy-dist) and one based on Link Quality Indication (Greedy-lqi). Single source, single destination experiments were performed on a sensor network testbed deployment based on battery-powered Tmote Sky nodes arranged in grid layouts with sizes ranging from 2x2 to 7x7. From the results, a complex picture emerges with no clear overall winner:
In networks with a static topology and in highly dynamic scenarios (where each sent message requires a separate routing task), DSR outperforms both Greedy variants in terms of average end-to-end delay of sending a message from source to destination. Similarly, in terms of data rate, DSR achieves the highest values when the communication path is saturated and the network is above a certain size. If the communication path is under-saturated, both Greedy variants outperform DSR and both achieve comparable data rates. In terms of the data packet delivery ratio, the stable links employed by Greedy-lqi, ensures that this protocol outperforms both DSR and Greedy-dist in all investigated scenarios, except when using link layer acknowledgements in static network topologies. In this situation DSR and Greedy-dist achieve delivery ratios that are comparable to that of Greedy-lqi. Finally, if path length in terms of number of hops, is the most important performance metric, DSR and Greedy-dist are the preferred choices.