Lance: Optimizing high-resolution signal collection in wireless sensor networks

An emerging class of sensor networks focuses on reliable collection of high-resolution signals from across the network. In these applications, the system is capable of acquiring more data than can be delivered to the base station, due to severe limits on radio bandwidth and energy. Moreover, these systems are unable to take advantage of conventional approaches to in-network data aggregation, given the high data rates and need for raw signals. These systems face an important challenge: how to maximize the overall value of the collected data, subject to resource constraints. In this paper, we describe Lance, a general approach to bandwidth and energy management for reliable data collection in wireless sensor networks. Lance couples the use of optimized, data-driven reliable data collection with a model of energy cost for extracting data from the network. Lance's design decouples resource allocation mechanisms from application-specific policies, enabling flexible customization of the system's optimization metrics. We describe the Lance architecture in detail, demonstrating its use through a range of target applications and resource management policies. We present an extensive study driven by both real and synthetic data distributions, through simulations and runs on a large sensor testbed. We show that Lance maximizes the value of the collected data under a range of resource constraints, achieving near-optimal allocation of radio bandwidth and energy. Finally, we present results from a real sensor network deployment at Tungurahua volcano, Ecuador, in which Lance was used to drive data collection for an eight-node network collecting seismic and acoustic signals from the active volcano.