Real-World Sensor Network Framework for Long-Term Volcano Monitoring
Wireless sensor network, composed of tens or hundreds of low-power and resource-constrained devices, has the potential to greatly enhance the understanding of volcanoes by permitting distributed deployments of sensor nodes at scales difficult to achieve with traditional instrumentation. An active volcano provides a challenging environment to examine and advance sensor network technology. One of the key challenges is to improve the resource utilization efficiency and best direct those limited network resources to deliver the most valuable data for volcano studies in response to volcano activities. This dissertation studies managing resource usage to improve the data quality provided by volcano sensor networks.We present the following core contributions. Firstly, we designed an energy-efficient remote network management system allowing command & controls to be reliably injected into the network in real time. In this way, the sensor network can adapt to volcano activity changes and deliver the most valuable data. Secondly, we designed a light-weight adaptive linear predictive compression algorithm ALFC to reduce bandwidth demands without compromise of data fidelity. ALFC can adapt to a dynamically changing source such as seismic data and achieve good compression performance. Thirdly, we developed a localized TDMA MAC protocol TreeMAC to address fair bandwidth distribution in multi-hop data collection sensor networks. Driven by the many-to-one network traffic pattern, TreeMAC assigns bandwidth proportional to the demands of sensor nodes to reduce network congestion and maximize network throughput. Furthmore, inspired by TreeMAC we design a QoS-aware robust TDMA MAC protocol RoMac for dynamic wireless networks. It can achieve persistent QoS as well as resilience to network dynamics with very little control message overhead. Finally, we present the design, deployment and evaluation of a real-world sensor network framework for remote volcano monitoring employing those techniques to improve data quality. The successful deployment has demonstrated that a low-cost sensor network system can provide real-time continuous monitoring in harsh environments and greatly promoted the confident use of sensor networks.