Localization in Wireless Sensor Networks

Abstract

The technique of nding physical co-ordinates of a node is known as localization. Importance of localization arises from the need to tag the sensed data and associate events with their location of occurrence. Location information of a sensor node can be obtained by using GPS. But, installing GPS in every node is not a feasible solution. This is because: (i) sensor nodes are deployed in a very large number. Installing GPS at every node will increase the cost
as well as size, (ii) GPS consume power, which will effect the network lifetime. Moreover, location cannot be pre-programmed as it is un-known where nodes will be deployed during their operational phase.

In this thesis, we have made an attempt to address localization in static as well as
mobile sensor networks. For static network we have proposed two distributed range based localization techniques called (i ) Localization using a single anchor node (LUSA), (ii ) Dis- tributed binary node localization estimation (DBNLE). Both the techniques are proposed for grid environment. In LUSA, we have identied three types of node: anchor, special and unknown node. For every anchor node there exists two special node and they are placed perpendicular to the anchor node. Localization in LUSA is achieved by a single anchor node and two special nodes. Localization occurs in two steps. First special nodes are localized and then the unknown nodes. We have compared LUSA with a closely related localization technique called Multi-duolateration (MDL). It is observed that the localization error and
localization time is lesser in LUSA. In DBNLE a node is localized with only two location aware nodes instead of three nodes in most localization techniques. This not only reduces the localization time but also the dependency.

For mobile WSNs, we have proposed a distributed localization technique called dead reckoning localization in mobile sensor networks (DRLMSN). In DRLMSN, localization is done at discrete time intervals called checkpoint. Unknown nodes are localized for the rst time using three anchor nodes. In their subsequent localization, only two anchor nodesare used. Using Bezouts theorem, we estimate two possible locations of a node. A dead reckoning approach is used to select one among the two estimated locations. We have used Castalia simulator to evaluate the performance of the schemes.