Hephaestus: A multisensor data fusion algorithm for multiple applications on wireless sensor networks

Hephaestus is proposed, an entropic information fusion algorithm, which uses Mean, Kurtosis and Skewness to apply a heuristics that divides the dataset into multiple features, for multiple applications, and achieved high accuracy while incurring in low overhead for the resource constrained devices of WSNs.

Abstract

Wireless sensor networks (WSN) are core components of the Internet of Things paradigm. By applying techniques such as sensor virtualization, the wireless sensor network infrastructure can be shared by a set of applications. On such scenario, the massive amount of data produced by the widely spread sensors produces a value-added information for the end user. By sharing the same infrastructure with multiple users, the set of applications in execution may change quickly, according to the users' necessity. In this case, the application requirements (for instance, data intervals or the potential events of interest) may not be known a priori. In this work, we aim to provide a solution to properly integrate data from multiple applications without the knowledge about specific application requirements, and unconver useful information from such data. We propose Hephaestus, an entropic information fusion algorithm, which uses Mean, Kurtosis and Skewness to apply a heuristics that divides the dataset into multiple features, for multiple applications. Hephaestus achieved high accuracy while incurring in low overhead for the resource constrained devices of WSNs.