Improved area estimates for localization in wireless sensor networks

To accurately determine the location of every sensor in large wireless sensor networks is normally very computation intensive and hence, is not cost-effective. On the other hand, a coarse estimate of the sensors' location is usually sufficient for many applications. The Area Localization Scheme (ALS) has been proposed as a scheme that estimates the position of every sensor within a certain region rather than identifying its exact location. Experimental results have shown that ALS is a promising technique with an accuracy of over 80% of sensors successfully localized. However, the accuracy of the scheme is dependent of the location of the sensors, as the regions into which the sensors are localized are not of uniform sizes. If a sensor is predicted to lie in a small region, then there is a higher probability that the prediction is inaccurate, and vice versa. In this paper, we aim to correct such discrepancy in defining positioning error. We propose to aggregate smaller regions to form more uniform regions within which nodes are localized. This effort improves the overall scheme's accuracy according to some predefined threshold.