A Self-Organising Distributed Location Server for Ad Hoc Networks

Wireless networks allow communication between multiple devices (nodes) without the use of wires. Range in such networks is often limited restricting the use of networks to small offices and homes; however, it is possible to use nodes to forward packets for others thereby extending the communication range of individual nodes. Networks employing such forwarding are called Multi-Hop Ad Hoc Networks (MANETS) Discovering routes in MANETS is a challenging task given that the topology is flat and node addresses reveal nothing about their place in the network. In addition, nodes may move or leave changing the network topology quickly. Existing approaches to discovering locations involve either broadcast dissemination or broadcast route discovery throughout the entire network. The reliance on the use of techniques that use broadcast schemes restricts the size of network that the techniques are applicable to. Routing in large scale ad hoc networks is therefore achieved by the use of geographical forwarding. Each node is required to know its location and that of its neighbours so that it may use this information for forward packets. The next hop chosen is the neighbour that is closest to the destination and a number of techniques are used to handle scenarios here the network has areas void of nodes. Use of such geographical routing techniques requires knowledge of the destination's location. This is provided by location servers and the literature proposes a number of methods of providing them. Unfortunately many of the schemes are limited by using a proportion of the network that increases with size, thereby immediately limiting the scalability. Only one technique is surveyed that provides high scalability but it has a number of limitations in terms of handling node mobility and failure. Ad hoc networks have limited capacity and so the inspiration for a technique to address these shortcomings comes from observations of nature. Birds and ants are able to organise themselves without direct communication through the observation of their environment and their peers. They provide an emergent intelligence based on individual actions rather than group collaboration. This thesis attempts to discover whether software agents can mimic this by creating a group of agents to store location information in a specific location. Instead of requiring central co-ordination, the agents observe one another and make individual decisions to create an emergent intelligence that causes them to resist mobility and node failures. The new technique is called a Self Organising Location Server (SOLS) and is compared against existing approaches to location servers. Most existing techniques do not scale well whereas SOLS uses a new idea of a home location. The use of this idea and the self organising behaviour of the agents that store the information results in significant benefits in performance. SOLS significantly out performs Terminode home region, the only other scalable approach surveyed. SOLS is able to tolerate much higher node failure rates than expected in likely implementations of large scale ad hoc networks. In addition, SOLS successfully mitigates node mobility which is likely to be encountered in an ad hoc network.