Mobility Management in WSNs using Fuzzy Logic: An Industrial Application Scenario Zinon Zinonos * 1 , Chrysostomos Chrysostomou * 2 , and Vasos Vassiliou † 3 * Department of Computer Science, University of Cyprus, Nicosia, CYPRUS 1 zinonos@cs.ucy.ac.cy 3 vasosv@cs.ucy.ac.cy † Department of Computer Science and Engineering, Frederick University, Nicosia, CYPRUS 2 ch.chrysostomou@frederick.ac.cy Abstract—Mobility management in Wireless Sensor Networks is considered to be of an upmost importance for today’s critical applications. In this paper, we present a soft mobility manage- ment solution where the mobility procedures are supported by fuzzy logic techniques. Our solution was designed and imple- mented to support the movement of a mobile worker inside an oil refinery area. The results show that the proposed system provides high reliability and control over the handover actions. I. I NTRODUCTION In recent years, sensor networks characteristics have led to their incremental utilization in different types of applications such as environmental monitoring and control, healthcare, military, and industrial automation. The majority of the ap- plications expect the presence of a large number of static sensor nodes, which is an assumption that is also generally made in WSN research. However, this assumption is not valid in applications where the existence of mobile sensor nodes is required. In order to efficiently monitor or control a mobile entity, the node associated with it must be able to handoff between different networks while performing its usual functions. In addition, packet losses and disconnection periods, provoked by the handoff, should be minimized and controlled. In critical applications, like personnel safety in an industrial environment, the real-time monitoring system must be available requiring the existence of a proper mobility protocol to control the handoff procedure and to handle the movement of the mobile node efficiently. In this paper, we present a mobility management protocol that has been used to support mobile worker operations in a real industrial environment where performance requirements are critical. This work was implemented within the context of an FP7 European project named, GINSENG [1]. The end user of the GINSENG project is the Petrogal oil refinery at Sines, Portugal. II. PROBLEM STATEMENT An industrial or (petro)chemical plant, like an oil refinery, often has many hazardous areas that need regular maintenance. The cleaning and condition assessment of storage tanks is one of those activities that may be considered hazardous, since they typically contain a toxic atmosphere and residues of their previous contents. When employees enter such hazardous areas of the refinery there is a possibility to loose conscious- ness or become dizzy and fall. Using orientation and heart or pressure monitoring sensors attached to employees, their condition can be monitored and alarms can be signalled when an employee is lying on the floor. Although packet losses should be minimized, this application can be tolerant to a small amount of loss, however minimal delay is of the essence. The assumptions and requirements for the Personnel Safety Monitoring scenario are the following: • Data Delivery: The data sent by the mobile node must arrive to sink node within one second. • Packet loss: Few packets can be lost. The targeted packet loss is less than 1%. • Network: The network is consisted of 1 sink node, 12 fixed nodes and 1 mobile node. • Topology: The nodes construct a 3-2-1 tree topology meaning that at each time there are 2 free available positions for the mobile node to handoff. • The mobile node sends periodic upstream data to the sink. • The mobile node data frequency is equal to one packet per second. III. MOBILITY SOLUTION The GINSENG system can provide to the system operator a variety of network performance metrics. Mostly, these metrics are centralized available at the end-system (sink) even though some useful metrics such as the link-to-link packet loss and the Received Signal Strength Indicator (RSSI) are available at each node. Since our target is to provide a reliable solution, the main metric that we will consider is the End-to-End (E2E) packet loss, which is only available at the sink. Using the refinery testbed, we run a small number of experiments to identify any relationship between the E2E packet loss and the RSSI. The selection of the RSSI was based on previously published research [5] where it was shown that the RSSI can be used as a reliable metric. Our tests shown that after a specific RSSI value (we call this a threshold) the E2E packet loss increases, therefore, we consider this threshold as a trigger to start seeking a new network attachment point.