Degree Name

Doctor of Philosophy


School of Electrical, Computer and Telecommunications Engineering - Faculty of Informatics


An ad hoc network is a collection of self-configured nodes that are dynamically and arbitrarily located in such a manner that the interconnections between nodes are capable of changing on a continual basis. In order to facilitate communication within the network, a routing protocol plays an important role in discovering and defining the routes between nodes. To date, a considerable amount of research effort has focused on the design of ad hoc network routing protocols. However, relatively little of this work examines the stability of routes in such networks. There are several reasons to believe that the routes in ad hoc networks will be unstable. One reason is that the majority of routing messages are broadcast packets which tend to be unreliable and have lower service quality at lower layer. The routing traffic usually shares resources such as bandwidth and buffer space with data traffic. The data traffic constitutes a larger portion of traffic mix. Furthermore, data traffic comprises mostly of unicast transmission because it offers more reliable delivery and higher service quality for user applications. Thus, the routing packets will suffer from a higher degree of loss when a network contains both routing and data traffic. The loss of routing packets can cause peering sessions to fail and consequently leads to routing instabilities. Due to the layered network architecture, the routing protocol designs often overlook any severe packet loss that is induced not only by the surrounding wireless traffic but also by their own traffic. Therefore, the communication link quality varies significantly even in a static network topology. This dissertation examines the stability and robustness properties of ad hoc routing protocols when the routing control traffic is not isolated from data traffic. The research begins with an analytical model to examine the subtle interactions between delivery of routing (broadcast) and data (unicast) packets in wireless networks. The model highlights the deficiencies of existing MAC protocols in delivering routing packets in congested networks, which can be detrimental for the routing protocols. The study continues with simulation study of routing behaviour in such condition. Two representative routing protocols, Ad hoc On-Demand Distance Vector (AODV) and Optimised Link State Routing (OLSR), are chosen to examine the stability performance of reactive and proactive routing protocols respectively. Through a series of analytical and simulation studies, this study identifies a number of routing pathologies for both reactive (AODV) and proactive (OLSR) routing protocols. Three different approaches have been proposed to address routing instabilities from different network layers. The first approach presents Out-of-Band Routing (OBR) which evaluates the feasibility of isolating routing and data traffic using different radio interfaces. OBR ensures that a channel is used exclusively for routing traffic to reduce the chance of losing important routing packets. Through extensive simulations, OBR is shown to improve the stability of end-to-end sessions by 30% for AODV and 50% for OLSR in a congested network. The second approach aims to improve the stability of AODV by exploiting the interaction between routing and underlying MAC protocols. Two protocol techniques known as Adaptive Bulk Trigger (ABT) and Dynamic Window Selection (DWS) are proposed to tolerate a small amount of packet loss and assign higher priority to nodes with critical routing demand. Simulation results show that combining ABT and DWS increases the duration of end-to-end sessions by 30% while reducing route recovery latency by 50% in a congested scenario. The third and final approach introduces Optimised Link State Routing with Reactive Route Repair (OLSR-R3). The reactive route repair compliments the traditional OLSR protocol by providing reactive routing when routing packets are lost. Simulation results show OLSR-R3 improves the end-to-end session duration by 30% and reduces the route recovery latency by 20% over OLSR by avoiding the erratic routing behaviour caused by incomplete link-state knowledge.