Computing and Mathematical Sciences School, Liverpool John Moores University,. Liverpool, UK. E-mail :{ cmpfbouh, M.Merabti, H.M.Mokhtar}@livjm.ac.uk.
A Coordination Protocol for Wireless Sensor and Actor Networks F.Bouhafs, M. Merabti, and H. Mokhtar Computing and Mathematical Sciences School, Liverpool John Moores University, Liverpool, UK E-mail :{ cmpfbouh, M.Merabti, H.M.Mokhtar}@livjm.ac.uk
Abstract-Unlike Wireless sensor networks (WSANs) are composed of static, low-resources, densely distributed, sensors and, mobile, resource-rich, sparsely distributed actors. In this kind of networks, sensors are responsible for monitoring and gathering information about physical phenomenon, while actors are responsible for performing appropriate actions according to the information provided by sensors. In this paper we propose a sensor-sensor coordination protocol for WSAN based on clustering. In this protocol each actor builds a Voronoi region containing its nearest sensors. Furthermore, this protocol groups nodes detecting the same event in a cluster in order to ensure that only the nearest actor to the event area is informed about the event. I. INTRODUCTION
Sensor networks technologies have developed very quickly in the last few years. This kind of networks shows a big potential in the future security and monitoring applications. Although these networks were designed in the beginning to just monitor study fields and collect information, they could be used actively by deploying active nodes called actors. Actors are nodes that could perform actions in the study filed according to information collected by sensor nodes. For instance, in fire detection application actors could be mobile engines equipped with necessary material to put out the fire. These actors are, generally, mobile, energy-rich, equipped with better processing and transmission capabilities, and sparsely deployed comparing to sensor nodes which are densely deployed. The main idea behind sensor and actor networks (WSAN) [1] is to perform the adequate action correspondent to the detected event with higher precision. Upon a detection of an event a sensor node must signal this event to an actor to deal with these phenomena. In the example of fire detection application, a sensor node that detects a fire must inform the actor in order to put out this fire. Thus, in addition to sensor-sensor communication there is a real need to design a sensor-actor communication scheme. However, since sensor nodes are densely deployed, there is a high possibility that many nodes
ISBN: 1-9025-6013-9 © 2006 PGNet
detect the same event and try to inform an actor independently from each other which could lead to an overlapping between many actors. Thus, the most important challenge is to design a communication architecture that offers coordination mechanisms between sensors and actors. In this paper we address this challenge by developing a new coordination protocol for WSAN that organizes the WSAN in Voronoi regions such as each region contains only one actor and its nearest sensors. To ensure that only one actor is informed when a physical event occurs, we propose to use semantic clustering [2, 3] to group nodes detecting the same physical phenomena in the same cluster. The remainder of the paper is structured as follows. Section 2 will review the major issues in WSANs and the related works found in the literature. Section 3 will introduce some concepts and definitions related to Voronoi diagrams and regions. Section 4 will give an overview on our coordination protocol and describe its different phases. Section 5, will draw our conclusions and outline open research issues and future works. II. BACKGROUND
In wireless sensor and actor networks (WSANs), sensors and actors are deployed to collect data from the environment and perform appropriate actions based on this collected information. WSAN’s could be used in two modes : Automated Mode: In these mode sensors detecting a phenomenon send their collected data to the actor nodes which process all incoming data and initiate appropriate actions. Semi-Automated Mode: In contrast to the automated mode, in this mode all sensors send their data to the sink which will coordinate all the acting process. The advantage of the semi-automated mode is that it is has an architecture similar to architecture used in wireless sensor networks, thus actual works on routing and communication
schemes could fulfil the requirement of such networks. However this mode has two major drawbacks: Latency If an phenomenon is detected by some sensors, performing an action towards this phenomenon will take time since each node has to send its data to the sink and each actor have to wait until it receive orders from the sink. In automated-mode such latency will be less important since sensors send their data to actors directly. Network Lifetime As all sensors have to send their data to the sink wherever the phenomenon happened, all the collected data will pass through the sensors situated at one hop from the sink. Thus, these sensors will have excessive burden of relying. Such burden could lead to a total failure of all the networks. Similarly, in automated-mode, the nodes within one hop from the actors may have a higher load of relaying packets. However, this load will not be constantly the same since such situation depends on the event area. Thus, in the automated-mode the WSAN will have longer lifetime than the semi-automated. As discussed before, the main goal of deploying sensor and actors in a study field is to trigger actors in order to deal with a specific event upon detected by one or many sensors. Thus, as soon as a sensor detects this event it must inform the appropriate actor about it. However, this deployment confronts two main challenges: •
As many sensors could detect the same event in the same time, each sensor could inform an actor that it considers as the appropriate one to deal with this event, which could lead to invoking many unnecessary actors, and overlapping between them, a specially when the area where the event is detected is too small.
•
Since an actor is generally triggered by the reception of information from a sensor node, it could perform the same action each time that it receives information from a sensor node. In the case of hundreds of sensors detecting the same event in the same time and sending its related information to the same actor, this one will perform the appropriate action hundreds of time depleting its energy unnecessarily while eventual other sensors may need this node to deal with another event.
Form these two main points we can see clearly that coordination and synchronization between the different nodes is the key element in the design of an efficient WSNA. The
sensors-sensors coordination may refers to the clustering concepts already well know in wireless sensor networks; however, unlike traditional clustering schemes where clusters are built generally at the deployment of the sensor network, here clusters are more dependants to the detected event and becomes more semantic. In other words, the clustering criteria must take in consideration the nature of sensed data. Many works have been proposed around clustering in sensor networks. The aim of clustering in these works is to maintain the energy consumption of sensor nodes by involving them in single hop or multi-hop communication within a particular cluster [4-7]. The cluster building operation is generally started by the cluster-head which is, generally, the rich-power node among a set of sensors. However, this type of clustering does not fill with the requirements of wireless sensor and actor networks (WSANs). Indeed, while these works focus only on energy conservation, WSAN needs more sophisticated clustering algorithms where nodes grouped in the same cluster must share almost the same information or detecting the same event. Thus, there is a need of new type of clustering approaches that take in consideration the nature of the sensed data in the cluster building operation. III. CONCEPTS AND DEFINITIONS
As mentioned previously, in WSAN actors are sparsely deployed comparing to the sensors which are densely deployed. Let s a sensor node in the WSAN network, and i and j two actors deployed in the same network. Let di and dj the distance that separate the sensor node s from respectively the actors i and j. If di>dj, the sensor is closer to the actor i than to the actor j and thus lies in the half plane that contains the actor i. By applying this rule to multiple pairs of sensor-actor distances, we can define the actor’s region of influence as the region where all sensors are much closer to this actor than any other actor in the WSAN. As the sensors are static, each actor can determine, initially, its nearest sensors. These nearest sensors results in a Voronoi region [8].
{
Let S = p1 , p 2, ..., p N
} be a finite set in the n-dimensional
space ℜ , and let d(p,q) the distance that separates the two points p and q. For p, q ∈ S let B(p, q)={x| d(p,x)=d(q,x)} be the bisector of p and q. B is the perpendicular line through the centre of the n
segment pq . It separates the half-plane:
D(p,q)={x| d(p,x)
