Abstract
Wireless sensor networks (WSN) composed of large numbers of small devices, are applicable to a wide range of mission-critical applications, including firefighting and emergency response, infrastructure monitoring, military surveillance, and medical applications. For these systems, reliability is the most common and important requirement. However, it is extremely challenging to achieve high-level reliability in these mission-critical applications because of resource-limited sensor devices, high-volume and burst traffic, wireless noise and interference, and unpredictable environment change and human intervention.
In this dissertation, we focus on developing new methodologies to address reliability issues in both networking and application levels. In the networking level, we propose to develop new protocols with two advanced networking techniques to improve communication reliability. The first technique is multi-channel communication. We propose a set of novel multi-channel protocols, which exploit simultaneous transmissions on different channels to reduce interference and then improve communication reliability. The second technique is Active Collision Recovery (ACR). The proposed protocol, ACR, can efficiently recover corrupted packets when collisions occur. In the application level, our work focuses on how to guarantee and maintain application-level operation correctness in mission-critical WSNs over long lifetimes. The core of our proposed solutions is a new design principle, run-time assurance (RTA). RTA requires that a system can be validated and demonstrated (periodically or on demand) that it is capable of operating correctly at run time. We develop a highly automated methodology and novel runtime framework that self-tests WSNs for RTA. In this framework, rigorous specification of both the functional logic of the application as well as the runtime assurance requirements are used to automatically generate the code for both the system implementation as well as the runtime assurance tests. The WSN then performs self-tests to validate its functions at run time. When tests fail, the WSN can automatically send alarms, and collect traces for future diagnosis and repairs.
The work in this dissertation paves the way for networking tiny and resource limited sensor nodes into high-confidence systems for mission-critical applications. Models and protocols we develop in the networking layer lead to more efficient and reliable wireless communication so that high-volume and burst data streams can be efficiently transferred across networks in a timely manner. New design principles, methodologies and a highly automated framework for runtime assurance help the designer and users verify an application's integrity at runtime, and build reliable and trustful WSN systems. In general, by developing novel protocols and methodologies in networking and application levels, an overall reliability enhancement of wireless sensor systems is achieved, which sets up a solid basis for developing wireless sensor networks technology to more mission-critical applications.
