The growth of Wireless Sensor Network (WSN) technologies have spawned promising applications such as military surveillance and medical monitoring. However, due to the limited power resource, energy efficient networking has always remained a major challenge in making large scale WSN deployments viable.
The difficulty in achieving energy efficient networking depends heavily on the amount of network traffic. In this dissertation, we present techniques centering the Medium Access Control (MAC) and routing layers to enhance energy efficiency across a wide range of traffic: from when traffic is ultra-low and interference-free communication can be safely assumed, to when traffic becomes non-trivial and the explicit consideration of interference becomes necessary, and to finally override our mind set on interference by exploiting interference rather than avoiding it.
We first empirically investigate the efficiency of different classes of low-power scheduling protocols under different traffic situations. Existing knowledge regarding low-power MAC scheduling across different traffics is fragmented. To unify, we categorize a wide range of low-power MAC protocols according to the centricity and synchrony of their scheduling and group them into four scheduling classes. We extensively evaluate representative MAC protocols from each class under a wide range of traffic, showing that our implementation almost always outperforms those from the other classes.
A key observation from our empirical study above is that certain classes of MAC protocols can approximate interfere-free scheduling when traffic is low. From this observation, we developed two protocols that optimize the energy efficiency of convergecast via joint MAC and routing control. When traffic is ultra-low and interference-free scheduling can be approximated regardless of routing, we present a distributed and self-stabilizing protocol that minimizes power consumption. When traffic is not ultra-low but there exists some routing tree in which interference-free scheduling can be effectively approximated, we prove that the optimization problem is a linear one and present a centralized solution.
In cases where interference is no longer negligible and warrants explicit consideration, we revisit efficiency optimization by explicitly modeling interference using an Signal to Interference plus Noise Ratio (SINR) model. We solution consists of two steps: first, given a set of minimum link rate constraints, our technique optimizes the MAC layer by simultaneously controlling transmission scheduling and reception scheduling in order to minimize power consumption; secondly, we propose an routing metric that takes into account both transmission and reception energy expense.
In our last piece of work, we transition from managing and avoiding interference to actively exploiting interference. A common pattern in sensor network applications is the collection of up-to-date neighborhood metrics to perform local or distributed decision making. We may define this pattern in terms of a local state predicate over the sensor values or other local variables at each neighboring node. We present two primitives that exploit simultaneous communications to enable a polling node to calculate the number (or set) of its neighbors where some state predicate of interest holds.