Wireless sensor networks (WSNs) offer a powerful combination of distributed sensing, computing and communication, which enable a broad spectrum of applications and, at the same time, lead numerous challenges due to their distinctiveness, primarily the non-negligible power consumption from especially radio activities and stringent energy constraints to which sensor nodes are typically subjected. The distinguishing traits of sensor networks have a direct impact on their protocol design at each layer, especially at the Medium Access Control (MAC) layer since it manages transmission scheduling as well as duty cycling for energy conservation. To maximize energy efficiency of WSNs, my thesis studies duty cycling in time and frequency domains for both MAC schedulers and applications. The first part of the thesis focuses on energy efficiency at the MAC layer, including modeling, evaluating and designing MAC schedulers with duty cycling; the second part of the dissertation investigates energy efficiency in applications, which introduces two duty-cycled sensor applications that are deployed in a large building.
In the first part of this dissertation, I begin by studying the impact of perfect duty cycling, in addition to perfect transmission scheduling, on the capacity of random wireless networks with single and multiple channels. The analysis of the duty-cycled throughput reveals nontrivial scaling gains resulting from the ability to avoid interference by spreading interferers to mutually exclusive times, which corroborates the importance of efficient co-scheduling of both transmissions and duty cycling for energy efficiency. Since duty cycling and transmission scheduling are controlled by MAC schedulers, I analytically quantify the gap between the duty-cycled throughput with an optimal scheduler and with existing MAC schedulers.
In order to characterize energy efficiency achieved by existing MACs, I classify CSMA-based MAC protocols in terms of critical MAC-design factors into four classes and introduce an analytical framework for performance modeling of each class as a function of key protocol parameters. I instantiate the framework to evaluate various performance metrics of MACs across the configuration space. A surprising finding is that one MAC class consistently achieves the best or close-to-the-best performance across much of the configuration space. Moreover, via the analytical model, I discover a distributed way of adapting duty cycle at the MAC layer to changing traffics for optimality of performance. In terms of energy efficiency in the frequency domain, I propose Chameleon, which is a light-weight MAC protocol that maximizes energy efficiency over the spectrum by scheduling traffics across multiple frequencies with duty cycling.
In the second part of this dissertation, I present two long-lived sensor networks deployed in a large building. Towards duty cycling Heating, Ventilation, and Air Conditioning systems of large buildings such that comfort and efficiency can be maintained simultaneously, I described ThermoNet, which is a system for temperature monitoring in large buildings. Access to fine grain information reveals temporal and spatial dynamics that help quantify the level of (non-)compliance with the building's thermal comfort standards and identify ill-conditioned rooms that need maintenance. For another application, to increase the battery life of the elevator network from a couple of days to several years, I introduce a self-stabilizing token-ring protocol that maintains duty-cycle coordination across the partitions of a static network of nodes.