In-network stochastic consensus for WSN surveillance applications
Cristian Drăgana, Viorel Mihai, Grigore Stâmâtescu, Dan Popescu · 2017
Surveillance applications require reliable monitoring system architectures and cost-efficient innetwork data processing mechanisms able to provide effective information extraction for event tracking. Wireless Sensor Networks (WSNs) appear to be the most suitable technology due to some well known benefits and the rapid development of embedded devices. Current research is mostly focused on improving and developing new decentralized sensor fusion schemes able to overcome the limitations introduced by the battery operated sensor nodes, but most of the time evaluating the proposed algorithms is performed with a serious lack of precise channel modeling. To close this gap, we provide a stochastic consensus implementation for lossy wireless networks comprising fixed and mobile sensor nodes, as a heterogeneous surveillance system based on ground sensor nodes and UAVs (Unmanned Aerial Vehicles). We developed a simulation framework tailored for fixed and mobile sensor nodes and we are able to evaluate the performance of consensus algorithms from a comparative standpoint, considering deterministic and probabilistic packet propagation models described by common models such as Free Space and Nakagami models. In this paper, we consider a complex in-network adaptive estimation using the distributed least-mean square algorithm tailored for WSN. We perform a mean-square error (MSE) performance analysis for both deterministic and probabilistic propagation models.