On Efficient Data Reduction for Network Partition Forecasting in WSNs

Abstract

WSNs (Wireless Sensor Networks) are generally deployed for long-lived missions. However, they rely on finite energy resources which lead to network partitioning. Network partitioning limits the dependability of WSN by making relevant spatial regions disconnected thus requiring the maintenance of the network. The network maintenance necessitates early warning and consequently forecasting of the network partitioning such that some early action can be taken to mitigate the problem. There exist approaches allowing for detection of network partitioning but none for its forecasting. We present an efficient approach for a proactive network ParFor (Partition Forecasting) based on energy maps. ParFor implements spatial and temporal suppression mechanisms such that from energy weak regions only a few nodes report short alarms to the sink. Using these alarms the forecasting is done centrally at the sink. Using simulations we highlight the efficiency and accuracy of ParFor.

Keywords

• WSNs
• Partition Forecasting
• Sensor Map
• Energy Regions.