一种基于预测建模的WSN数据管理方案

A New Approach for WSN Data Management Based on Predictive Modeling

能效、满足用户多样化需求的传感器网络数据管理是当前传感器网络应用的一个重大挑战,本文在基于Proxy/Sensor分层传感器网络数据管理方式基础上,提出了一种新的传感器数据管理方案,利用用户的历史查询建立时间序列预测模型以推断用户的未来数据需求,基于预测结果从Sensor层节点获取感知数据并存储在Proxy层节点,同时Sensor层节点实时监测自身的负载状态,当负载达到阈值后主动上报历史感知数据,以提供对历史数据查询与分析的支持。实验结果表明该方案在满足用户数据需求的基础上,以较低的能耗和较小的查询时延获得了较为全面的性能。

Data management in sensor networks has become a crucial issue in recent years, a new approach using predictive modeling for WSN data management is presented in this paper. Based on a two - tier sensor architecture that comprises proxies at the higher tier, each controlling tens of sensor nodes at the lower tier, the approach constructs time series prediction models on proxies to eliminate energy consumption of extra sensed data transmission from the sensor tier to the proxy tier, monitors the load condition of sensor nodes and accordingly uploads archival data to the proxy tier when the load threshold is reached. Simulation results show that our approach not only provides substantial energy cost benefits but also performs relatively small query latency.