Creating an Architecture for Wireless Sensor Networks

Wireless Sensor Networks (WSNs) have the potential to benefit society in a myriad of ways, such as accelerating scientific research, increasing productivity, and enhancing security. WSNs also pose many fascinating scientific challenges, ranging from device physics to encoding techniques to distributed algorithms. There is a large, diverse, and rapidly increasing network literature in this area. Unfortunately, much of this work has been done in isolation; all too often individual components are crafted and evaluated without an overall vision or a context for deployment. Consequently, these components cannot be reliably composed to form a coherent system. The few successful prototype deployments resulted from complete ``vertical integrations," in which research groups created their own suite of components. The result is a lack of synergy, both between research groups and between components. The barriers to prototyping complete systems are far too high and, for those focusing on individual components, there is little guidance as to how to build them so that they would be usable by others.
A systems architecture, encompassing general design principles, broad functional decompositions, and detailed interfaces, would provide guidance about what components are necessary and how they should fit together. If the architecture is successful, components would be largely reusable across applications, and between research and development efforts. Moreover, the resulting designs could more comfortably accommodate new generations of devices. The central tenet of this proposal is that the primary factor limiting progress in WSNs is not any particular scientific challenge but is instead the lack of an overall sensor network architecture. This proposal is dedicated to creating such an architecture for WSNs.
The ``narrow waist" of this architecture -- playing the role of IP in the Internet architecture -- is the Sensor-net Protocol (SP), which is a best-effort single-hop broadcast. The proposed work is to create a coherent and flexible framework for components above and around SP that can be used in a wide variety of settings. It should embrace heterogenity and allow for application-specific optimization within a consistent framework.
This will require identifying a set of basic building blocks.
Components to support address-free protocols, address-based protocols,
and in-network storage will certainly be required, as will components
devoted to cross-layer tasks such as security, systems management, power management, and time synchronization.
Intellectual Merit: An architecture is neither a single working design nor a a rigorously proven theorem and can only be evaluated by whether, over time, it guides the development of a wide array of systems and thereby leads to substantial synergy and interoperability. Finding the right compromise between generality and performance, and articulating where the essential interfaces should be drawn, poses a very significant intellectual challenge that will require distilling all of the community's WSN experience into a widely applicable framework, and then testing this framework through subsequent implementations and deployments.
The team assembled has broad experience in both sensor networks and
Internet architecture, and thus is ideally suited to this task.
Broader Impact:The creation of a successful architecture for WSNs would allow far great reuse of existing components. The resulting reduction in development delays and costs would greatly broaden the sphere of potential applications. Thus, an architecture would break sensor networks out of the confines of the few very high payoff/budget applications like national security, and would enable their use in less lucrative but extremely valuable areas such as scientific research and ecosystem management. In so doing, the fruits of this proposal would have a significant beneficial impact on society.
The work in this proposal would also greatly enhance student education. An architecture is a powerful pedagogical tool. The Internet is best presented in classrooms not as a random collection of technologies but as a coherent architecture. Such an approach would also benefit sensor networks, thereby better training the next generation of researchers in this field.