Agent-Based Computational Architectures for Distributed Data Processing in Wireless Sensor Networks.

Abstract

As the structural health monitoring (SHM) community continues to develop algorithms for monitoring performance and detecting degradation in engineered systems, the importance of pervasive sensing and autonomous data processing methodologies will increase. Fortunately, the emergence of wireless sensor technologies at the forefront of SHM research has provided a platform on which problems related to both sensor density and processing autonomy can be addressed. By utilizing wireless communication links instead of expensive data cables, wireless monitoring systems can be deployed with much greater sensor density and at significantly lower costs than traditional SHM systems. Perhaps more importantly, because wireless sensing units typically integrate a traditional sensor with a low-power microprocessor, analog-to-digital converter, and wireless transceiver, wireless sensing networks (WSNs) have shown great promise in their ability to process sensor data in-network (i.e., without the need for a centralized data center).

Over the past decade, the wireless SHM community has shown that it is possible to minimize problems associated with power efficiency, data loss, and finite communication range by processing data before transmitting it to a central repository. Recently, in an effort to further improve the efficiency and capability of in-network computation, researchers have started to move away from centralized processing frameworks (where no data is shared between nodes) towards more hierarchical data processing architectures. However, work to date in this area has yet to fully leverage the computational advantages provided in large networks of wireless sensors.

In this dissertation, several distinct agent-based architectures are developed for distributed data processing in WSNs. Each of these agent-based architectures leverages the ad-hoc communication and pervasive nature inherent to wireless sensing technology, and can be viewed as a parallel computing system with an unknown and possibly changing number of processing nodes. As such, sophisticated data analysis can be performed while maintaining a scalable environment that is not only resistant to sensor failure, but that also becomes increasingly efficient at higher nodal densities. These agent-based architectures represent a significant step towards the creation of a fully autonomous WSN for application to SHM.