Management of moving network clusters: Master's thesis

This work presents a clustering protocol that is able identify and manage moving network clusters. The clusters are detected based on the nodes’ proximity with an associativity-based mechanism. After the cluster is set up, individual nodes can join and leave the cluster, and two clusters can be merged as well. The protocol also features a mechanism for reducing the forwarding of broadcast control messages. The protocol’s functionality and performance was evaluated using simulations. In addition to the basic functionality, the simulation scenarios tested the overhead of the protocol’s control traffic and its effect on the application data flow. The results show that the clustering protocol achieves its goals and is able to identify clusters and perform the basic management operations. It was discovered that the associativity-based neighbour discovery mechanism induces only a little overhead, but the real challenge is in choosing the right parameters, which in turn depend on the use environment and radio technology characteristics. Based on the results, it can be concluded that the presented protocol is well suited for its tasks, and it also looks promising for future work. It is advisable to continue research especially on scalability, routing optimisations, and enhancing the neighbour discovery method.