In recent years, wireless sensor networks have emerged as one key technology with the advantages of being low-cost, low-profile, and easy to deploy. These attractive advantages, however, imply that resources available to individual nodes are severely limited. In particular, it is highly likely that energy will continue to be a system bottleneck, as more transistors indicate more power consumption. On the other hand, there is a growing need for long-term, sustainable deployments of sensors to reduce operational costs and ensure service continuity. This research presents an operating system based approach and its underlying foundations for this problem, by providing energy virtualization in long-term unattended sensor network applications. The foundation of this work is the LiteOS operating system: a Unix-like operating system for wireless sensor networks. The centerpiece of this project is operating system based energy virtualization, which represents a comprehensive framework for accounting, reservation, and isolation of energy resource for applications sharing the same platform. The deliveries of this project are (i) an improved, highly reliable LiteOS operating system with built-in support for energy virtualization, (ii) architecture principles, design methodologies, programming APIs, and adaptive communication protocols for software development based on energy virtualization, (iii) results from theoretical and algorithmic investigations of energy virtualization, and (iv) educational testbeds, curriculum and laboratory designs for undergraduate and graduate courses. Our findings will have pervasive impact on applications that are in need of long-term and remote sensing, such as infrastructure protection, assisted living, smart buildings, factory monitoring, and numerous military applications.
