|Abstract:||Wireless sensor networks promise an unprecedented potential for observing the physical world. Their battery life, however, is usually the bottleneck that limits application lifetime, a problem that is exacerbated by the disparity between the rapidly growing processing speed and the slowly improving battery capacity. We observed the importance of this problem in our first-hand experiences: during our involvement in two large-scale sensor network deployments, VigilNet, and EnviroMic, energy is the primary concern in the development and deployment of applications. Our experiences with these projects motivate us to consider the central role that energy plays in the design and implementation of software stacks for wireless sensor networks, including not only the design choices of protocols but also the overall system architecture. In this dissertation, we present an energy-centric software architecture, LiteArch, that addresses the following three problems.
First, we address the problem of energy conservation, where we consider a stack of protocols that extends the lifetime of sensor networks through duty scheduling, and provide an analysis on the quantitative relationship between energy consumption and surveillance performance. Second, we address energy awareness, where we design energy-efficient communication protocols, whose performance usually dominates the overall energy efficiency of applications. Finally, we address energy isolation, where we implement an energy reserve abstraction that virtualizes energy sources, giving each application sharing the same platform an illusion of having its own guaranteed energy source.
Because these three problems are closely related to the role played by energy, we call the overall LiteArch architecture energy-centric. The immediate implications of LiteArch is that it can improve the performance of the motivating projects, VigilNet and EnviroMic. In the long term, the impact of LiteArch will be manifested through our continuing development of LiteOS, an operating system on which we have implemented several representative protocols in LiteArch. Our solutions reflect the profound impact of energy on the design and implementation of the sensor network software stack, and shed light on the general principles that help design and implementation of future sensor network platforms.