关键词: 无线传感器网络, 数据精简, 鲁棒性, 轻量级

Abstract: Wireless Sensor Networks applications are frequently deployed in severe and restricted communication scenarios because of limited wireless bandwidth, computing, batter power and unexpected noises. To efficiently and accurately transfer collected data from sensors to the sink, data reduction algorithms are proposed to decrease the transmission by sending those readings that deviate from the prediction by an error budget. However, few such algorithms consider the effect of packet lose in practical scenarios. In this paper, LRPH（Lightweight Robust algorithm with Plain Heartbeat） is proposed, which works on a lightweight prediction model and a heartbeat mechanism to resist the influence caused by packet loss and to be aware of the state of sensors in time. Furthermore, LRSH（Lightweight Robust algorithm with Smart Heartbeat） as an optimization is proposed to reduce the heartbeat cost in LRPH. Experiments with real traces show that the LRPH approach only requires 4.15% transmission to recover all data by a prediction model within an error threshold, and furthermore, the LRSH approach requires 3.63%. Both two algorithms can significantly resist packet loss in severe communication scenarios comparing with previous approaches.

Key words: Wireless Sensor Network（WSN）, data reduction, robustness, lightweight