Ballari, Daniela and Wachowicz, Monica
The design of a Bayesian Network for mobility management in Wireless Sensor Networks.
In: "GIScience 2010: 6th international conference on Geographic Information Science", 14/09/2010 - 17/09/2010, Zurich, Suiza. ISBN 978-3-642-15299-3.
Mobility in Wireless Sensor Networks (WSNs) is achieved by attaching sensors to mobile objects such as animals (Juang et al. 2002), people (Campbell et al. 2008), and robots (Dantu et al. 2005). Currently, the research about WSN management is mainly focused on energy management functions to control how sensors should use their power; fault management functions to solve sensor problems; quality of services (QoS) management functions to quantify and control the performance; and mobility management functions to detect the sensor movement so that the network wireless connectivity is always maintained (Wang et al. 2010; Ruiz et al. 2003). However, the sensor mobility has not only an impact on the network connectivity, but also on the network spatial coverage. In mobile WSNs, the extension of the spatial coverage is often changing, and as a result, the region of interest might be inaccurately sensed by the mobile sensors. Therefore, the representation of a movement context is important to avoid making interpretations and decisions outside of the situation in which the WSN is capturing information; and make possible to decide where, when and how the sensing is performed in order to obtain the most suitable spatial coverage of a region of interest. This paper proposes a Bayesian network (BN) approach for making explicit the structural and parametric components of a movement context using WSN metadata. The aim is to infer mobility management requirements when a spatial coverage is incorrectly covering a Region of Interest (ROI), regardless the network connectivity. The BN approach provides several advantages regarding to the probabilistic representation of a movement context, the inference of mobility management requirements based on such a context, and the dynamic updating of the movement context every time new metadata are retrieved from the WSN. Previous research works in WSNs have used a similar approach focusing on energy management (Elnahrawy and Nath 2004) and prediction of sensor movement directions (Coles et al. 2009). The main contribution of our work is the analysis of how well a ROI is being covered by mobile sensors, and what are the requirements to improve that coverage given a movement context. A controlled experiment was carried out and the results show that, when the ROI is not being sufficiently covered by a WSN, the BN can probabilistically infer different mobility management requirements, based on a given movement context. Two movement contexts have been used to illustrate this approach. They are related to whether the sensing is being carried out in an emergency situation or not.