A Resource-Centric Design Paradigm for Scalable Sensor Networks

To my parents. For constantly reminding me that my best isn’t good enough. ii iii Wireless sensor networks offer a promising way to enable a new class of highly pervasive sens-ing and actuation applications. These devices are intended to be low-cost and often battery-operated making them highly constrained computing platforms. In this dissertation, we con-sider the use of resource management across critical infrastructure elements to improve fun-damental aspects of the network including clock synchronization, CPU scheduling, network scheduling, and transducer control. By making all primitives in the system resource-aware, we build infrastructure elements capable of providing Quality of Service (QoS) properties to a wide variety of sensor networking applications. We substantiate our resource-centric design paradigm through the development of a new hardware platform, a real-time operating system and a range of communication protocols and network services. Finally, the system’s function-ality is evaluated as part of an extensible Internet-scale sensing and actuation infrastructure called Sensor Andrew. i