Cloud-Cluster Architecture for Detection in Intermittently Connected Sensor Networks

We consider a centralized detection problem where sensors experience noisy measurements and intermittent connectivity to fusion center. The collaborate locally within predefined sensor clusters fuse their data reach common local estimate of the detected event in each cluster. cluster is depends on available communication opportunities Upon receiving estimates from all connected center fuses received make final determination regarding occurrence across deployment area. refer this hybrid scheme as cloud-cluster architecture. propose method for optimizing decision rule analyzing expected performance resulting our scheme. Our tractable addresses high computational complexity caused by heterogeneous sensors’ clusters’ quality, heterogeneity opportunities, non-convexity loss function. analysis shows that clustering provides resilience noise case low probability with cloud. For larger clusters, steep improvement possible even using