Application-driven data processing in wireless sensor networks

Abstract

Wireless sensor networks (WSNs) are composed of spatially distributed, low-cost, low-power, resource-constrained devices using sensors and actuators to cooperatively monitor and operate into the environment. These systems are being used in a wide range of applications. The design and implementation of an effective WSN requires dealing with several challenges involving multiple disciplines, such as wireless communications and networking, software engineering, embedded systems and signal processing. Besides, the technical solutions found to these issues are closely interconnected and determine the capability of the system to successfully fulfill the requirements posed by each application domain. The large and heterogeneous amount of data collected in a WSN need to be efficiently processed in order to improve the end-user comprehension and control of the observed phenomena. The thesis focuses on a) the development of centralized and distributed data processing methods optimized for the requirements and characteristics of the considered application domains, and b) the design and implementation of suitable system architectures and protocols with respect to critical application-specific parameters. The thesis comprehends a summary and nine publications, equally divided over three different application domains, i.e. wireless automation, structural health monitoring (SHM) and indoor situation awareness (InSitA). In the first one, a wireless joystick control system for human adaptive mechatronics is developed. Also, the effect of packet losses on the performance of a wireless control system is analyzed and validated with an unstable process. A remotely reconfigurable, time synchronized wireless system for SHM enables a precise estimation of the modal properties of the monitored structure. Furthermore, structural damages are detected and localized through a distributed data processing method based on the Goertzel algorithm. In the context of InSitA, the short-time, low quality acoustic signals collected by the nodes composing the network are processed in order to estimate the number of people located in the monitored indoor environment. In a second phase, text- and language-independent speaker identification is performed. Finally, device-free localization and tracking of the movements of people inside the monitored indoor environment is achieved by means of distributed processing of the radio signal strength indicator (RSSI) signals. The results presented in the thesis demonstrate the adaptability of WSNs to different application domains and the importance of an optimal co-design of the system architecture and data processing methods.