Spatially- and temporally-adaptive communication protocols for zero-maintenance sensor networks relying on opportunistic energy scavenging

Wireless sensor networks allow scientists to gather data from remote, difficult to access, and dangerous locations. However, maintenance of aging networks and removal of obsolete or inactive nodes containing toxic materials is expensive and time consuming. Moreover, node lifespan is generally constrained by the reliability of the batteries used in most deployments, especially in the presence of extreme variation in environmental conditions such as temperature and humidity. We consider the problem of designing wireless sensor networks capable of indefinite deployment periods measured in decades, not months. We describe the architectural and capability implications of eliminating batteries from sensor networks and instead relying on opportunistic energy scavenging. Sensor nodes using ambient energy sources become temporarily active at unpredictable but possibly correlated times. In this paper, we use wind power as an example of such a power source, which we model using temporally and spatially correlated random processes. Such models can be built using historical measurements over a geographical range. We describe a method to use energy models in the design of latency-optimized and cost-constrained battery-less wireless sensor networks, and explain the required changes to network architecture, communication protocol, and node hardware. In the context of environmental monitoring applications, we compare the performance of a network designed and managed using our techniques with that of existing design styles.