Efficient power management for Wireless Sensor Networks: A data-driven approach

Abstract

Providing energy-efficient continuous data collection services is of paramount importance to Wireless Sensor Network (WSN) applications. This paper proposes a new power management framework called Data-Driven Power Management (DDPM) as the infrastructure for integrating various energy efficient techniques, such as the approximate querying and the sleep scheduling. By utilizing the beneficial properties of these techniques, we can achieve better energy efficiency while still meeting the application specific criteria, such as data accuracy and communication latency. The distinguishing feature of DDPM is that it starts by exploiting the natural tradeoff between the quality of the sensor data and the energy consumption, and then it generates a precision-guaranteed estimation for each sensor node as its maximum sleep time. Eventually deterministic schedules can be made by the DDPM based on these estimations. We further propose two decentralized algorithms so that the undesirable communication delays caused by staggered local sleep schedules can be avoided. The experimental results show that the nodespsila sleep times can be significantly increased while incurring only a minor rise in latency.