Removing Systematic Error in Node Localization Using Scalable Data Fusion

摘要

Methods for node localisation in sensor networks usually rely upon the measurement of received strength, time-of-arrival, and/or angle-of-arrival of an incoming signal. In this paper, we propose a method for achieving higher accuracy by combining redundant measurements taken by different nodes. This method is aimed at compensating for the systematic errors which are dependent on the specific nodes used, as well as their spatial configuration. Utilising a technique for data fusion on the physical layer, the time complexity of the method is constant and independent of the number of participating nodes. Thus, adding more nodes generally increases accuracy but does not require additional time to report measurement results. Our data analysis and simulation models are based on extensive experiments with real ultrasound positioning hardware. The simulations show that the ninety-fifth percentile positioning error can be improved by a factor of three for a network of fifty nodes.