Design and implementation of an on-demand ad-hoc routing algorithm for a positional communication system.

Abstract

A mobile ad-hoc network is an autonomous network of mobile devices that are connected via wireless links. In such networks there is no pre-existing infrastructure and nodes are free to move in a random fashion. Due to this mobility mobile ad-hoc networks have dynamic topologies. A host in the network typically has limited bandwidth and energy resources. Routing is a major challenge in the development of such systems and there have been many solutions proposed in the recent past. The aim of this work is to design and implement a routing scheme for a Positional Communication System (PCS). The PCS is a network of mobile handheld pocket PCs connected via wireless interfaces. The system allows voice and data communication between nodes in the network. This dissertation addresses the process of designing a routing protocol for an ad-hoc network. There have been many proposed algorithms that solve the routing problem in a mobile ad-hoc network. It is a difficult task to compare the performance of'these protocols qualitatively as there are many parameters that affect network performance. Various simulation packages for networks of this type exist. One such package is the Network Simulator (NS-2). It is a discrete time event simulator that can be used to model wired and wireless networks. This dissertation presents NS-2 simulations that compare four recently proposed routing algorithms. From this comparison study it is shown that on-demand algorithms perform best in a mobile ad-hoc environment. The dissertation then describes the design of a novel on-demand routing algorithm. The ondemand algorithms proposed thus far use a blind flooding technique during the route discovery process. This method is inefficient and creates excessive routing overhead. The routing protocol proposed in the dissertation implements a query localization technique that significantly reduces the network traffic. The protocol also introduces a load checking metric in addition to the metric used by most on-demand schemes, namely hop count. Simulation results show that such a scheme makes the on-demand routing algorithm more efficient and scalable than existing ones. It is widely believed that prior to implementing a routing protocol in real world systems it is essential that it is tested and validated on a test-bed. The dissertation presents the implementation of an on-demand routing algorithm in a Positional Communication System test-bed, where each handheld PC in the network runs an embedded Linux operating system.