Abstract—In monitoring applications the accuracy of data is paramount. When considering wireless sensor networks the quality of readings taken from the environment may be hampered by outliers in raw data collected from transmitters attached to nodes' analogue-to-digital converter ports. To improve the data quality sent to the base-station, a real-time data analysis should be implemented at nodes' level, while taking into account their computing power and storage limitations. This paper deals with the problem of outliers detection and accommodation in raw data. The proposed approach relies on univariate statistics within an hierarchical multi-agent framework. Results from experiments on a real monitoring scenario, at a major oil refinery plant, show the relevance of the proposed approach.

Index Terms—Oil refinery plant, real-time monitoring, wireless sensor networks, outliers detection and accommodation, multi-agent systems.

Paulo Gil and Alberto Cardoso are with the CISUC-Informatics Engineering Department, University of Coimbra, 3030-290, Coimbra, Portugal.
Amâncio Santos is with the Polytechnic Institute of Coimbra, ISEC, DEIS, 3030-199, Coimbra, Portugal.
Luís Palma and Paulo Gil are with the UNINOVA-CTS, Department of Electrical Engineering, Faculty of Science and Technology, New University of Lisbon, 2829-516, Caparica, Portugal.

Cite:Paulo Gil, Amâncio Santos, Luís Palma, Alberto Cardoso, "A Multi-Agent System for Outliers Accommodation in Wireless Sensor Networks," International Journal of Computer and Electrical Engineering vol. 6, no. 6, pp. 457-468, 2014.