In-network data processing approach for heterogeneous wireless sensor networks

A wireless sensor network (WSN) is a set of specialized devices that commonly monitor environmental and physical conditions. A critical aspect of applications with WSNs is their limited resources especially in multivariate sensor features when transmitting large amount of data from the nodes to the base station. The aim is then to optimize power consumption during data transmission by using data reduction methods. In this article, we study multivariate data reduction at node's level. We propose a new efficient model based on reducing collected data by aggregation and polynomial regression. We evaluate and compare our method with existing data aggregation techniques, and with the following well-known compression techniques (xz, bzip2, brotli and gzip). The simulation results show that our approach outperforms the existing methods and offers a good approximation of data quality with small approximation errors.