S U M a N N a T H Dissertation: " Exploiting Redundancy for Robust Sensing " Dissertation Research Honors and Awards

Exploiting Redundancy for Robust Sensing In my thesis research, I have shown that several unique properties of hybrid sensing services (built on embedded sensors like motes or more capable sensors like webcams) can be efficiently exploited to make the services highly robust. For wireless sensor networks, I have shown how to exploit the underlying broadcast medium for robust computation of aggregate queries. To achieve this, I have proposed Synopsis Diffusion, a framework for in-network aggregation over arbitrary topology. It enables in-network aggregation over energy-efficient multi-path routing which can be significantly more robust than, but as energy efficient as, existing techniques. I have established the formal foundation of the framework and shown how to make the aggregation topology self-adaptive to cope with node-and link-failures. For wide-area sensing, I extensively studied the failure characteristics of wide-area systems. I have shown how the weak data consistency requirements of sensing services can be exploited to tolerate failures that traditional techniques fail to cope with. I have shown the feasibility of my proposed techniques by incorporating them within IrisNet, a wide-area sensing service infrastructure that I designed and implemented as part of my thesis.