MANET routing with prediction.

Abstract

Route stability of Mobile Ad-Hoc Networks (MANETs) is one of the major problems in defence tactical wireless networks. The dynamic nature of MANETs may cause the network topology to change frequently as a result of unstable links, which may result in frequent route changes. Unstable routes may cause retransmissions and drop outs. Therefore, the network can experience heavy traffic overload and high packet losses. Many network applications rely on a stable and reliable route. Hence, it is important for the military to have a reliable network that allows effective communications amongst various platforms to effectively perform the tasks they have been assigned. For this reason, the route’s stability in MANETs needs to be understood. However, many existing MANET routing protocols are not explicitly designed for route stability. It is expected that prediction can assist in increasing a MANET’s route stability. This thesis explores the potential benefits and the trade-offs in the use of prediction with the Ad-hoc On-demand Distance Vector (AODV) routing protocol. In the context of using prediction in routing, research has shown that using “accurate” predictions can improve MANETs’ routing performance. However, Chapter 3 shows that it is difficult to achieve accurate predictions. To the author’s knowledge, very little work has been attempted to analyse the routing performance with reduced prediction accuracies, and the effects of having inaccurate prediction. Thus more specifically, this thesis examines the robustness of using link duration prediction with various accuracies for MANETs, and identifies the conditions for which predictions can improve routing performance. This is achieved by first examining how using perfectly accurate link duration prediction can improve routing performance. For this purpose, a new routing protocol, Ad-hoc On-demand Distance Vector with Perfect Prediction (AODV-PP), has been created to propagate link duration prediction information for route establishment. The OPNET simulator was used to simulate network scenarios with AODV and AODV-PP for analysis, and the routing performance of the two protocols have been compared. The thesis later explores how inaccurate link duration prediction affects routing performance. However, the AODV-PP protocol does not inform the source about the change in predicted link duration. This can cause delays in route re-establishment and high packet loss. Hence, AODV with Prediction Update (AODV-PU) has been proposed to allow link duration prediction updates to be sent to the source for route maintenance. Network scenarios with AODV-PU were simulated to analyse and compare its routing performance with AODV and AODV-PP. This thesis shows stable routes can be found with perfect prediction, which reduces packet loss and routing overhead. However, it also indicates that it is difficult to use link duration prediction to find a more stable route with inaccurate long-term predictions. Nevertheless, link duration prediction can be useful for route updates and route re-establishments, which only requires short-term predictions, to allow more seamless route transitions and to reduce packet loss. The trade-off being that more control traffic is required for route maintenance. This in turn creates a more robust platform for the military applications that require this type of network.