Heterogeneous wireless sensor networks for computational partitioning of Markov parameter-based system identification

Abstract

Embedded computation in wireless sensor networks (WSN) can extract useful information from sensor data in a fast and efficient manner. Embedded computing has the benefit of saving both bandwidth and power. However, computational capability often comes at the expense of power consumption on the wireless sensor node. This is an especially critical issue for battery powered wireless sensor nodes. By developing a hybrid network consisting of wireless units optimized for sensing interspersed with more powerful computationally focused units, it is now possible to build a network that is more efficient and flexible than a homogeneous WSN. For this project, such a network was developed using Narada units as low-power sensing units and iMote2 units as ultra-efficient computational engines. In order to demonstrate the capabilities of such a configuration a network was created to extract structural modal parameters based on Markov parameters. This paper validates the performance of the heterogeneous WSN using a laboratory structure tested under impulse loading.