Advances in microelectronics technology have made it possible to build inexpensive, low-power, miniature sensing devices. Equipped with a microprocessor, memory, radio, and battery, such devices can now combine the functions of sensing, computing, and wireless communication into miniature smart sensor nodes , also called motes . Since smart sensors need not be tethered to any infrastructure because of on-board radio and battery, their main utility lies in being ad hoc, in the sense that they can be rapidly deployed by randomly strewing them over a region of interest. Several applications of such wireless sensor networks have been proposed, and there have also been several experimental deployments. Example applications are:
● Ecological Monitoring: wild-life in conservation areas, remote lakes, forest fi res.
● Monitoring of Large Structures: bridges, buildings, ships, and large machinery, such as turbines.
● Industrial Measurement and Control: measurement of various environment and process parameters in very large factories, such as continuous process chemical plants.
● Navigation Assistance: guidance through the geographical area where the sensor network is deployed.
● Defense Applications: monitoring of intrusion into remote border areas; detection, identification, and tracking of intruding personnel or vehicles.
The ad hoc nature of these wireless sensor networks means that the devices and the wireless links will not be laid out to achieve a planned topology. During the operation, sensors would be difficult or even impossible to access and hence their network needs to operate autonomously. Moreover, with time it is possible that sensors fail (one reason being battery drain) and cannot be replaced. It is, therefore, essential that sensors learn about each other and organize into a network on their own. Another crucial requirement is that since sensors may often be deployed randomly (e.g., simply strewn from an aircraft), in order to be useful, the devices need to determine their locations. In the absence of a centralized control, this whole process of self-organization needs to be carried out in a distributed fashion. In a sensor network, there is usually a single, global objective to be achieved. For example, in a surveillance application, a sensor network may be deployed to detect intruders. The global objective here is intrusion detection. This can be contrasted with multihop wireless mesh networks , where we have a collection of source – destination pairs, and each pair is interested in optimizing its individual performance metric. Another characteristic feature of sensor networks appears in the packet scheduling algorithms used. Sensor nodes are battery-powered and the batteries cannot be replaced. Hence, energy-aware packet scheduling is of crucial importance.
A smart sensor may have only modest computing power, but the ability to communicate allows a group of sensors to collaborate to execute tasks more complex than just sensing and forwarding the information, as in traditional sensor arrays. Hence, they may be involved in online processing of sensed data in a distributed fashion so as to yield partial or even complete results to an observer, thereby facilitating control applications, interactive computing, and querying. A distributed computing approach will also be energy-efficient as compared to mere data dissemination since it will avoid energy consumption in long haul transport of the measurements to the observer; this is of particular importance since sensors could be used in large numbers due to their low cost, yielding very high resolutions and large volumes of sensed data. Further, by arranging computations among only the neighboring sensors the number of transmissions is reduced, thereby saving transmission energy. A simple class of distributed computing algorithms would require each sensor to periodically exchange the results of local computation with the neighboring sensors. Thus the design of distributed signal processing and computation algorithms, and the mapping of these algorithms onto a network, is an important aspect of sensor network design.
Design and analysis of sensor networks must take into account the native capabilities of the nodes, as well as architectural features of the network. We assume that the sensor nodes are not mobile . Further, nodes are not equipped with position-sensing technology , like the Global Positioning System (GPS). However, each node can set its transmit power at an appropriate level — each node can exercise power control . Further, each node has an associated sensing radius ; events occurring within a circle of this radius centered at the sensor can be detected.
In general, a sensor network can have multiple sinks, where the traffic generated by the sensor sources leaves the network. We consider networks in which only a single sink is present. Further, we will be concerned with situations in which sensors are randomly deployed . In many scenarios of practical interest, preplanned placing of sensors is infeasible, leaving random deployment as the only practical alternative; e.g., consider a large terrain that is to be populated with sensors for surveillance purposes. In addition, random deployment is a convenient assumption for analytical tractability in models. Our study will also assume a simple path loss model , with no shadowing and no fading in the environment.
Source of Information : Elsevier Wireless Networking Complete 2010