Using kalman filters on GPS tracks

Abstract

This paper describes the practical evaluation of the application of the Kalman filters to GPS tracks, gathered by mobile phones or GPS trackers. We try to answer the question whenever the filtering applied on higher layer of the mobile device software may improve the quality of the data provided by the GPS receiver. Two metrics are used for comparison: the average euclidean distance between the points on the GPS tracks and the actual location of the user and the area of the polygon created by intersection of the filtered and real track. We find that the Kalman filtering does not improve those two metrics and the direct use of the data provided by the GPS receiver provides track which is on average more near the real path than result of Kalman filtering. However we observe that this is caused by the errors introduced when the user change the direction and when we evaluate a parts of the path without rapid changes of direction (as e. g. crossing) the filters allow to generate the points which are more near the road taken by the user.