Map-based Design for Autonomic Wireless Sensor Networks

A prominent functionality of a Wireless Sensor Network (WSN) is environmental monitoring. For this purpose theWSN creates a model for the real world by using abstractions to parse the collected data. Being cross-layer and application-oriented, most ofWSN research does not allow for a widely accepted abstraction. A few approaches such as database-oriented and publish/subscribe provide acceptable abstractions by reducing application dependency and hiding communication details. Unfortunately, these approaches ignore the spatial correlation of sensor readings and still address single sensor nodes. In this work we present a novel approach based on a “world model” that exploits the spatial correlation of sensor readings and represents them as a collection of regions called maps. Maps are a natural way for the presentation of the physical world and its physical phenomena over space and time. Our Map-based World Model (MWM) abstracts from low-level communication issues and supports general applications by allowing for efficient event detection, prediction and queries. In addition our MWM unifies the monitoring of physical phenomena with network monitoring which maximizes its generality. From our approach we deduce a general modeling and design methodology for WSNs. Using a case study we highlight the simplicity of the proposed methodology. We provide the necessary tools to use our architecture and to acquire valuable WSN insights in the established OMNeT++ simulator.