UCLA

Distributed Clustering Algorithm ensures that all nodes form clusters. Also uses LCC algoritm to minimize reclustering required. All nodes within a cluster share same CDMA code and communication between clusters require gateways to resynchronize with other clusterhead. All intercluster packets pass through these gateway nodes. Bandwith is controlled by a PRMA (packet reservation media access) style polling scheme. Delay increases dramatically if packets traverse other clusters, while within a cluster there exist excellent connectivity.
The clusterheads are responsible for synchronization, fowarding packets within the cluster and through the cluster via gateways. Routing is done by maintaining two types of tables. The first is a cluster member and the destiation clusterhead table at each node. This table is used to map destination addresses to the destination clusterhead address. This is broadcasted peridoically and a node will update its table when it receives a new one from its neighbor. Sequence numbers are attached to the updates as in DSDV to avoid stale tables. The second table is used to select the next node needed to reach the destination cluster. It is constructed by a minimum hop to a clusterhead criteria. Packets are sent first to a clusterhead and then get routed appriately to the correct gateway. This gateway then forwards it the adjacent clusterhead and this process can repeat many times until the final clusterhead forwards the packet to the final destination. The exact method used to distribute the hop information varies slightly from paper to paper. DSDV style updates are sometimes used to distribute the minimum hop to clusterhead infomation. Sometimes all nodes participate in these updates and sometimes only gateways and clusterheads.
Note that most of the results that come from the UCLA project is through simulation.