| One of the key differences of Wireless Sensor Networks compared to traditional networks is the inclusion of energy consumption as a first-order optimization goal. As a result, significant research effort has been channeled towards creating low power hardware platforms. Currently, wireless sensor network devices fall into two main categories. The first category, motes, includes tiny, inexpensive, resource-constrained devices that operate for long periods of time on battery power. The second category, microservers, is comprised of devices that provide increased computational and networking capacity compared to motes, at the cost of higher power consumption. To address the fundamental tension between the need for ample numbers of nodes, and ample resources per node, the sensor network community has been increasingly exploring heterogeneous tiered architectures, where the system is composed of a mixture of both platform categories.;In this dissertation, we advocate exploiting heterogeneity for routing in Wireless Sensor Networks, by taking advantage of the different capabilities of platforms in a heterogeneous system. We demonstrate via two separate protocol designs---one at each tier---that our approach can often lead to performance improvements.;On the mote tier, we implement CentRoute, an on-demand mote routing protocol that exploits the increased memory and computational resources of the microserver to centralize routing decisions. CentRoute avoids routing instabilities and inconsistencies that sometimes occur in distributed designs and also incurs low control overhead as paths are maintained only when data flows through them. Simulations and testbed experiments show that CentRoute provides better than 99% network connectivity in medium and high densities with 60% less control overhead compared to distributed protocols.;On the microserver tier, we approach the problem of establishing on-demand multihop end-to-end paths in a network that is duty-cycled in order to conserve energy. We use a second, low-bandwidth mote-class radio which can be found in newer generations of microservers to selectively wake up nodes along the required path. Numerical models and initial testbed experiments show that our proposed mechanism uses up to 60% less energy compared to alternative approaches while incurring only 9--12 seconds of extra latency. |
