Conclusions
This work proposes a clustering protocol that addresses the Control Channel Problem in C-WMNs, called CONNOR. CONNOR allows to SUs to discover each other and to exchange control information to form clusters in which a SU is chosen to be the CHN. Further, our protocol allows to cluster’s members to dynamically choose the control channel used to exchange control information within the cluster, adapting to the time and space variability of channels. To avoid that the control information sent from the CHN to the furthest SU come to it when they are already obsolete, we limit the distance from the CHN and the furthest SU (Delay Cluster Formation Constraint).
CONNOR has also other two characteristics. It avoids the frequent
re-clustering problem, when a PU appears on a cluster’s control channel. This is
done using at least two common channels between all cluster’s members. The first one is actually the control channel used, while the second one is a backup
control channel on which cluster’s members switch when they sense a PU on
the used control channel (Re-Clustering Cluster Formation Constraint). The switching of the cluster’s members on another channel, preserving the formed cluster, is better than a complete cluster breaking. Further, CONNOR does not use any synchronization among SUs, avoiding all synchronization man-aging complexity.
Testing our protocol based on some metrics we had shown that CONNOR
Conclusions
is able to form less clusters than the comparison algorithm in many cases, and in many others to reach the same number of formed clusters, without an a priori synchronization among CMDs. The re-clustering constraint limits CONNOR to form the same number of clusters as the comparison algorithm in cases with a small number of channels and larger transmission ranges, because the re-clustering constraint reduces the probability to find common channels between clusters’ members. But the re-clustering constraint allows to CONNOR to avoid complete clusters’ breaking when PUs appear on the used control channel. The comparison algorithm, instead, suffers of a com-plete cluster breaking, that leads it to reach again the minimum number of clusters with more time than CONNOR. Further, we had shown that the comparison algorithm forms a small number of clusters also with a small number of channels and larger transmission ranges because it uses a syn-chronization among CMDs that always allows to it to find the channel with the higher connectivity level. But the synchronization used by the the com-parison algorithm is also its main drawback, because it employs more time than CONNOR to reach the minimum number of clusters.