Abstract |
In wireless ad hoc networks nodes that do not belong in the same transmission range communicate with each other using intermediate nodes as relays. Traffic is routed from each node to the next relay until it reaches the final destination, which may be located far away from the source. In this case, every node must behave not only as a source or a destination, but as a router as well. Since wireless networks have a limited battery life, it is possible for a node not to wish to accept all the incoming relay requests. But, if all nodes behave in this selfish way, there will be a dramatic drop of the network data throughput. The purpose of this master thesis is to suggest and evaluate two algorithms that create motives between the nodes of wireless ad hoc networks, in order to avoid the bad scenario of misbehaviour just described above. The first algorithm presented is based in some previous related work, which dictates that each node is ready to help, i.e. to accept a relay request, only if it has been already helped analogously by the whole network. However, the presence of selfish nodes affects badly the entire network data throughput. We suggest an extension of the idea described above that aims to deal better with selfishness. The past behaviour of each neighbour of each relay node is recorded separately. The new algorithm dictates that each node is ready to help another one, only if it has been helped analogously by this specific node during the past. The second algorithm we present is based on a leaky bucket approach. It maintains the difference of the given and received help amounts of each node within specific limits, which are determined by the size of a bucket. Both algorithms can be expanded to a version that uses different weights for the equalization of given and received help amounts. This version is recommended when most of the network traffic is destined to one or more access points. The basic experiment results showed that the first algorithm manages to reduce the geographic boundaries of the bad effects of selfish behaviors and improve the total throughput, while the second algorithm deals better with burst traffic scenarios.
|