A Stochastic Model for Environment Sensing Correction

Abstract

Wireless Sensor Networks (WSN) are growing in popularity and penetrating newer fields of applications more than ever. Gradually, we are relying on WSNs to perform more complex tasks with increasing cognitive abilities. At the core of the WSN research transpires, the need is to achieve accurate sensing at real-time. Unfortunately, confidence in sensors’ readings decreases in harsh environments and as a result of normal reading errors, message loss, or even low battery operations. The complex problem of dealing with corrections and in some cases shredding the outcome of entire deployments leads to loss of effort, time, and money. Classic approaches for correcting laboratory experiments like curve fitting and least square are well known and have been established for decades. But little research attempts have been made to correct and recalibrate sensors observations in real-time. Furthermore, classic approaches for correcting sensor observations require higher interaction between sensors to a level we cannot afford in deployments where battery, network, and memory represent scarce resource. In addition, classical corrections lack the ability to contemplate the physical properties of the underlying sensor environment. In this article, we present a sensor correction model that relies on clustered WSN. Our approach employs autonomous selection mechanism to elect cluster heads by applying a stochastic competition between cluster members while maintaining the underlying physical properties as the bases to locate competition winner. Then, we perform the stochastic competitive correction at real-time by referencing the underlying physical properties of the environment represented by selected relation as suggested by the physics of the environment. Finally, sensors adapt the minor changes and maintain a relation to their surroundings by continuously monitoring and assimilating information received from surrounding sensors. We show that this approach has smaller footprint in terms of processing, communicating, and storage. We present our approach and apply it on an environment of known physical property.

Keywords:

• Sensors
• Stochastic learning
• Sensor fusion
• Sensor localization

Additional information

Notes on contributors

Morgan Yasser

Morgan Yasser (M–05) received his Bsc. and Msc. from Cairo University, Cairo, Egypt in 1986 and 1997 in the field of computer science. Dr. Morgan received his PhD. from Carleton University, Ottawa, Canada in 2005 in the field of computer science. Dr. Morgan led many industrial research teams in 3M-Innovation, Nortel Networks, and Siemens. Dr. Morgan (Ph.D, P. Eng.,) is an Associate Professor in University of Regina and has three patents filed and the latest is uniquely assigned to him. Dr. Morgan’s research spans sensor networks, vehicular communications, and intelligent transportation systems. He has a close interest in the IEEE standards development and has contributed to the IEEE workgroup for 802.11p and P1609.1/.2/.3/and. 4. He is also a part of the Dedicated Short Range Consortium (DSRC). Beside his conference and journal publications, he authored and co-authored standards publications with the IEEE, IEEE-SA, IETF, and 3GPP. He has published several refereed research papers in conferences and journals relevant to his research areas. E-mail: [email protected]

Bais Abdul

Bais Abdul received the M.Sc. degree in Electrical Engineering from N.W.F.P University of Engineering and Technology, Peshawar Pakistan, in 2003, and the Ph.D. degree from Vienna University of Technology, Vienna, Austria, in 2007. He is currently a Postdoctoral Fellow at the Faculty of Engineering and Applied Science, University of Regina, Regina, SK, Canada. His research interests include vehicular communication, sensor networks, multimedia content distribution, image processing, and computer vision. E-mail: [email protected]

Moustafa El-Gindy

Moustafa El-Gindy (M’87) is an associate professor at the University of Ontario Institute of Technology (UOIT) and former senior scientist and director of the Vehicle Dynamics and Simulation Research Center at Pennsylvania State University. He was the PI of industrial funded projects related to design and development of rear and front under-ride guards of heavy trucks. A patent application of a newly designed front under-ride guard was submitted jointly with the industrial partner. Dr El-Gindy has extensive knowledge and experience of the non-linear finite element analysis codes, such as LSDYNA and Pam-Crash, which will be used in this project. He also has several journal and conference publications in this field. Dr El-Gindy is also the founder and executive editor of International Journal of Heavy Vehicle Systems. He serves as fellow of the American Society of Mechanical Engineers (ASME) and former Chair of its Vehicle Design Committee. Dr El-Gindy has over 150 technical publications in the field of Heavy Vehicle Systems, Crash Simulation, and Highway Safety. E-mail: [email protected]