Efficient Information Dissemination in

Wireless sensor networks constitute an emerging technology that has received recently significant attention both from industry and academia. On the one hand, there is an ever-widening range of attractive applications (e.g., disaster and environmental monitoring, wildlife habitat monitoring, target tracking, intrusion detection, security surveillance) sensor networks can be used for. On the other hand, sensor networks are self-organizing ad-hoc systems where optimized energy consumption is of paramount importance; therefore, relaying information between sensors and a sink node, possibly over multiple wireless hops, in an energy-efficient manner is a challenging task that preoccupies the research community for some time now. Optimizing energy consumption in wireless sensor networks is of paramount importance. Sensors are tiny devices with sensing, processing, and transmitting capabilities; they are of low cost, but have a consequently low storage and computational capacity, and a limited energy supply. It is usually considered impossible or impractical (from a technical or economical point of view) to recharge their batteries; thus, they should be managed in such a way to ensure the unattended operation of the network for a long enough time period (e.g., several months). There is a recent trend to deal with this problem by introducing mobile elements (sensors or sink nodes) in the network. The majority of these approaches assume time-driven scenarios. However, there are several real-life applications for which an event-based is more appropriate. In this paper we propose to adaptively move the sink node inside the covered region, according to the evolution of current events, so as to minimize the energy consumption of the dissemination of the event-related data. Both analytical and simulation results are given. 1.0 INTRODUCTION Wireless sensor networks constitute an emerging technology that has received recently significant attention both from industry and academia. On the one hand, there is an ever-widening range of attractive applications (e.g., disaster and environmental monitoring, wildlife habitat monitoring, intrusion detection, security surveillance) sensor networks can be used for. On the other hand, sensor networks are selforganizing ad-hoc systems where optimized energy consumption is of paramount importance; therefore, relaying information between sensors and a sink node, possibly over multiple wireless hops, in an energyefficient manner is a challenging task that preoccupies the research community for some time now. Vincze, Z.; Vidács, A.; Vida, R. (2006) Efficient Information Dissemination in Wireless Sensor Networks using Mobile Sinks. In Dynamic Communications Management (pp. 4-1 – 4-24). Meeting Proceedings RTO-MP-IST-062, Paper 4. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp. Efficient Information Dissemination in Wireless Sensor Networks using Mobile Sinks 4 2 RTO-MP-IST-062 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Sensors are tiny devices with sensing, processing, and transmitting capabilities; they are of low cost, but have a consequently low storage and computational capacity, and a limited energy supply. It is usually considered impossible or impractical (from a technical or economical point of view) to recharge their batteries; thus, they should be managed in such a way to ensure the unattended operation of the network for a long enough time period (e.g., several months). Information gathering in sensor networks can follow different patterns, depending mostly on the specific needs of the applications. In a time-driven scenario all sensors send data periodically to the sink. As opposed to this, in the event-driven case sensors start communicating with the sink only if sensing an event, i.e., a situation that is worth reporting. Finally, in a query-driven scenario a sensor transmits its data only if the sink asks for it. Most of the research papers in the area address the time-driven scenario, and provide energy-efficient solutions for homogeneous networks, with sensors having constant and equal amounts of data to send in all parts of the covered region. However, there are a large number of applications (e.g., intrusion detection, seismic activity monitoring, animal movement tracking) where an event-driven approach is more appropriate. Hence, in our paper we address only this scenario. As we noted before, energy efficiency is the main requirement for the operation of a sensor network. Sensors consume energy for sensing the field, for digitizing and processing the data, but the most penalizing task is by far the transmission of the information [1]. In the most commonly accepted power attenuation model [2], signal power falls as d, where d is the distance from the transmitter antenna and α is a constant dependent on the wireless transmission environment, typically between 2 and 4. Therefore, assuming that all receivers have the same power threshold for signal detection, typically normalized to one, the energy required to support communication between the two nodes is d. In such conditions it is straightforward to assert that by minimizing the distance between a sensor and a sink node we can efficiently reduce power consumption, both for singleand multi-hop communications (reducing the length of the multi-hop path results in fewer and/or shorter hops, i.e., less energy is needed to relay data to the sink). Besides analyzing the general case of an event-driven scenario, we intend also to have a closer look on a specific example where events move inside the observed region following a correlated random walk model. There are several concrete use cases this example can be relevant for. In [3] authors show that animal movements can be described as a correlated random walk. A similar result is obtained in [4] for the specific case of caribous. Moreover, the model should fit intrusion detection and target tracking applications as well. We propose sink moving strategies in this paper for both the singleand multi-hop event-driven WSNs. At first we examine how to replace the sink periodically when the sensor network is using single-hop communication. We evaluate the proposed sink relocating strategies through simulations. After that we analyze, both analytically and through simulations, the efficiency of adaptively moving the sink node so as to react to dynamic events that follow a correlated random walk mobility model in multi-hop WSNs. The rest of this paper is organized as follows. In section 2.0 we present related work in the area of energy optimization and sink mobility. In section 3.0 we propose the sink moving strategies for the case of singlehop WSNs. In section 4.0 we describe the assumed network model, and calculate the overall energy requirement an event poses on the network, as well as the maximum energy consumption of a specific sensor in case of multi-hop WSNs. According to these analytical results, in we show how to find the optimal position of the sink inside the network so as to minimize overall or maximum energy consumption. We also present simulation results evaluating the performance of the proposed strategies, while section 5.0 concludes the paper. Efficient Information Dissemination in Wireless Sensor Networks using Mobile Sinks RTO-MP-IST-062 4 3 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED 2.0 RELATED WORK There were many proposals recently targeting the energy efficiency of wireless sensor networks. Some approaches focused on energy conserving routing techniques, i.e., finding optimal routes in terms of consumed power, and balancing the energy consumption among all nodes [6], [7], [8], [9]. Others were based on topology control schemes, i.e, deploying sensor and sink nodes in an efficient way or reshaping the topology through dynamic power control of the participating sensors [10], [11], [12], [13], [14]. Clustering techniques are also widely employed; the network is divided into small clusters, a cluster head being responsible for aggregating and relaying towards the sink the information gathered from the sensors of its cluster [15],[16]. In all the above approaches the elements of the network are all considered static. However, there is a recent trend to explore mobility as a way of enhancing energy efficiency. In [17] sensors dynamically react to the environmental changes and move towards areas where events occur frequently. In [18] sensor mobility is exploited at the deployment phase, to eliminate coverage holes that are discovered through the use of Voronoi diagrams. Mobile sensors are also considered in [19] to provide an extension of a stationary sensor network. Moving the sink node is also a widely explored solution. The approaches can be classified into three categories: random, predictable, and controlled mobility of the sink. In [20] the authors propose an architecture that builds on the random mobility of mobile agents, called data MULEs (Mobile Ubiquitous LAN Extensions), to collect sensor data in sparsely deployed networks. A similar approach, but for dense networks, is used by SENMA (SEnsor Networks with Mobile Agents) [21]; data is sent directly to the mobile agent that is flying above the sensor field, the transmission being triggered based on the estimated fading state of each sensor in its communication with the agent. [22] uses a random walk model for a mobile relay to theoretically derive parameters such as delay and data delivery ratio. A serendipitous movement of the sink nodes is also assumed in [23]. However, the authors propose an inversed scenario, where there is a single sensor that transmits data to a large number of mobile sinks. They describe the SEAD (Scalable Energy-efficient Asynchronous Dissemination) protocol to build and maintain an energyefficient dissemination tree that covers all the sink nodes. A predictable mobility solution is presented in [24]. The sink (called observer) moves along a predefined path, and pulls data from sensors in single-hop communication when arriving near to them. A predefined path of the sink is used in [5] as well. The authors show that moving the sink at the periphery of the covered circular region ensures energy-efficient operation; the approach is considerably different from other mobile sink solutions in that it assumes multi-hop communication between all the sensors and the sink. There are also several solutions that propose a controlled mobility of the sink nodes. In the AIMMS (Autonomous Intelligent Mobile Micro-server) system [25], [26] a mobile micro-server moves across the network, along a specific trail, to route data from the deeply embedded nodes. Its mobility is controlled in order to spend extra time (e.g., stop or slow down) in regions where there is a large amount of data to send or the communication channel is constrained. The idea of using mobile nodes for message ferrying is also considered in [27]; these nodes provide non-random proactive routes in highly-partitioned wireless ad-hoc networks. An attempt to determine specific sink movements for energy optimization is presented in [28]. The authors argue that multi-hop communication results in the sensors neighboring the sink being depleted at a fast pace. Therefore, they propose to employ multiple sink nodes that periodically change their locations, and present an ILP (Integer Linear Programming) model to obtain the optimal positions of these sinks. A linear programming solution to determine the movement of the sink and its sojourn time in different points of the network is given in [29] as well. Both the sensors and the sink are placed on a bi-dimensional grid. Efficient Information Dissemination in Wireless Sensor Networks using Mobile Sinks 4 4 RTO-MP-IST-062 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED The sink moves along the grid, sojourns times in the specific grid points being calculated so as to maximize the network lifetime. Finding the optimal position of the sink is addressed in [30] as well, even if mobility is not involved. The authors assume a time-driven scenario, where all sensors send data at a constant rate; the problem is how to deploy n sink nodes so as to ensure an energy-efficient operation of the network. [31] addresses the static deployment problem as well, as it proposes to find the optimal locations of multiple sinks in sparse networks of aggregator points that send data directly to these sinks. In this paper we propose a solution that is significantly different from all the above approaches. We assume an event-driven scenario, where sensors that detect an event send data to the sink node in a singleor multi-hop manner. Our goal is to control the mobility of the sink so as to ensure an energy-efficient operation of the network. The sink node is alerted about the current events, estimates the evolution of these events, and decides about the optimal neighboring place where it should move so as to maximize network lifetime. 3.0 ADAPTIVE SINK MOBILITY IN EVENT-DRIVEN SINGLE-HOP WSNS In this section we define adaptive mobility strategies for the sink node in case that the sensors use singlehop communication, so as to reduce the distance nodes have to send reports to. The solution perfectly fits the event-driven scenario, as the mobile sink can adapt and get closer to the nodes that are currently active, significantly reducing their energy consumption. It is also different from the existing controlled mobility solutions, as sensors do not wait to report until the sink gets close to them; thus, it can support delaysensitive real-time applications. constraints, and is able to move anywhere inside the network. We also assume that nodes are able to adjust their radio power depending on their distance d from the sink. Although sensing also requires energy, this is far less than the energy used for communication; thus, we neglect it. 3.1 Minimizing average energy consumption Let A be the set of active cluster heads (CHs) that have data to send to the sink. Let (x0,y0) denote the coordinates of the sink, and (xi,yi) the coordinates of the ith CH. Let di denote the distance between the sink and the ith CH. In the most commonly accepted attenuation model, signal power falls as di , where α is the attenuation exponent, with values ranging from 2 to 5. Thus, the energy needed for the ith CH to transmit data is