Research Article
Multiple abstraction levels in performance analysis of WSN monitoring systems
@INPROCEEDINGS{10.4108/ICST.VALUETOOLS2009.7736, author={Marco Beccuti and Daniele Codetta-Raiteri and Giuliana Franceschinis}, title={Multiple abstraction levels in performance analysis of WSN monitoring systems}, proceedings={1st International ICST Workshop on Performance Methodologies and Tools for Wireless Sensor Networks}, publisher={ACM}, proceedings_a={WSNPERF}, year={2010}, month={5}, keywords={Wireless Sensor Network optimization simulation Markov Decision Well-formed Net Stochastic Activity Network.}, doi={10.4108/ICST.VALUETOOLS2009.7736} }- Marco Beccuti
Daniele Codetta-Raiteri
Giuliana Franceschinis
Year: 2010
Multiple abstraction levels in performance analysis of WSN monitoring systems
WSNPERF
ICST
DOI: 10.4108/ICST.VALUETOOLS2009.7736
Abstract
In this paper, we illustrate the use of different methods to support the design of a Wireless Sensor Network (WSN), by using as a case study a monitoring system that must track a moving object within a given area. The goal of the study is to find a good trade off between the power consumption and the object tracking reliability. Power saving can be achieved by periodically powering off some of the nodes for a given time interval. Of course nodes can detect the moving object only when they are on, so that the power management strategy can affect the ability to accurately track the object movements. We propose two models and the corresponding analysis and simulation tools, that can be used in a synergistic way: the first model is based on the Markov Decision Well-formed Net (MDWN) formalism while the second one is based on the Stochastic Activity Network (SAN) formalism. The MDWN model is more abstract and is used to compute an optimal power management strategy by solving a Markov Decision Process (MDP); the SAN model is more detailed and is used to perform extensive simulation (using the M¨obius tool) in order to analyze different performance indices, both when applying the power management policy derived from the first model and when using different policies.
