Density estimation in large-scale wireless sensor networks

Eroğlu, Alperen
Density estimation is a significant problem in large-scale wireless ad-hoc networks since the density drastically impacts the network performance. It is crucial to make the network adaptive in the run-time to the density changes that may not be predictable in advance. Local density estimators are required while taking run-time control decisions to improve the network performance. A wireless node may estimate the density locally by measuring the received signal strength (RSS) of packets sent by its neighbours. In this thesis, RSS-based individual and cooperative density estimators are validated by controlled field experiments conducted in the FIT IoT-LAB test-bed, in France. According to the experiments these methods cannot be used as accurate density estimators in practice. The success of the individual density is significantly affected by the position of the estimating node and the number of its neighbours. Also, the cooperative density estimator is affected negatively by correlated data. Hence, a new fusion approach is proposed as a new density estimator. New method is more accurate than the two other density estimators. However, it should be considered that the RSS is prone to large- and small-scale fading, and this phenomenon negatively affects the accuracy of density estimators.