SM based localization using statistical approach
This thesis presents a localization technique for wireless sensor networks that precisely provides the location of the client, whose position is to be estimated, with acceptable error range. The basic principle involved is with increase in transmission distance between transmitter and receiver, the number of messages received at the receiver decreases. The building block of this technique is "Secure localization (SecLoc)" technique for wireless sensor networks.
The hardware requirements for SecLoc include seven reference points (RPs), three secure messages transmitters and a localization controller. SecLoc works under ten different power levels. The first step of SecLoc is to determine two RPs that are closest to the client. This is done by finding the disparity/difference between the number of secure messages, transmitted at different power levels by the secure message transmitters, and received by each of the reference points and the client whose position is to be found out. The two reference points that have the least disparity in the number of secure messages from that of the client are selected and the client's location is estimated using these two reference point co-ordinates and their corresponding disparities from the client.
Our proposed technique, "SM based localization using statistical approach" incorporates the SecLoc system and deals with it's drawbacks, claimed by us, in a more sophisticated manner using statistical approaches discussed later.
The results show that "SM based localization using statistical approach" outperformed SecLoc in the aspect of locating the client under contention. The error obtained for two particular locations of the client using SecLoc was in the range from 5.3 feet to 14.9 feet* while our SM based localization using statistical approach yielded an error ranging between 3.8 feet and 4.65 feet (77% of the time) for confidence limits of one sigma and 3.19 feet to 4.9 feet (77% of the time) for confidence limits of two sigma when the client is at position 73. For the client at position "63" our method produced an error between 3.2 feet and 4.3 feet (75% of the time) for the confidence interval of one sigma and 2.7 feet to 4.7 feet (75% of the time) for the confidence limits of two sigma. The hardware requirements are same as that of SecLoc. No additional hardware is required.