Semantic In-Network Complex Event Processing for an Energy Efficient Wireless Sensor Network

Abstract

Wireless Sensor Networks (WSNs) consist of spatially distributed sensor nodes that perform monitoring tasks in a region and the gateway nodes that provide the acquired sensor data to the end user. With advances in the WSN technology, it has now become possible to have different types of sensor nodes within a region to monitor the environment. This provides the flexibility to monitor the environment in a more extensive manner than before.

Sensor nodes are severely constrained devices with very limited battery sources and their resource scarcity remains a challenge. In traditional WSNs, the sensor nodes are used only for capturing data that is analysed later in more powerful gateway nodes. This continuous communication of data between sensor nodes and gateway nodes wastes energy at the sensor nodes, and consequently, the overall network lifetime is greatly reduced. Existing approaches to reduce energy consumption by processing at the sensor node level only work for homogeneous networks.

This thesis presents a sensor node architecture for heterogeneous WSNs, called SEPSen, where data is processed locally at the sensor node level to reduce energy consumption. We use ontology fragments at the sensor nodes to enable data exchange between heterogeneous sensor nodes within the WSN. We employ a rule engine based on a pattern matching algorithm for filtering events at the sensor node level. The event routing towards the gateway nodes is performed using a context-aware routing scheme that takes both the energy consumption and the heterogeneity of the sensor nodes into account.

As a proof of concept, we present a prototypical implementation of the SEPSen design in a simulation environment. By providing semantic support, in-network data processing capabilities and context-aware routing in SEPSen, the sensor nodes (1) communicate with each other despite their different sensor types, (2) filter events at the their own level to conserve the limited sensor node energy resources and (3) share the nodes' knowledge bases for collaboration between the sensor nodes using node-centric context-awareness in changing conditions. The SEPSen prototype has been evaluated based on a test case for water quality management. The results from the experiments show that the energy saved in SEPSen reaches almost 50% by processing events at the sensor node level and the overall network lifetime is increased by at least a factor of two against the shortest-path-first (Min-Hop) routing approach.