An emerging class of sensor network applications focus on reliable collection of high-resolution signals from across the network. The need to capture raw signals at high data rates makes them unable to take advantage of existing approaches to in-network data aggregation. Instead, these systems must attempt to maximize the amount of value collected for the application in face of bandwidth and energy limitations.
We have developed Lance, a general approach to bandwidth and energy management in wireless sensor networks. Combining an applicationdriven notion of value with a system-driven notion of cost, we frame the data collection challenge as a multi-dimensional knapsack problem. Using energy vectors to represent the magnitude and distribution of energy costs associated with collecting specific pieces of data we compare several online heuristic algorithms approximating the optimal knapsack solution. We evaluate our system through simulation, testbed experiments, and a field deployment at an active volcano.