A Machine Learning Approach to Improve the Accuracy of GPS-Based Map-Matching Algorithms (Invited Paper)

Advanced map-matching algorithms use location and heading of GPS points along with geometrical and topological features of digital road networks to find the road segment on which the vehicle is moving. However, GPS errors sometimes impede map-matching algorithms in finding the correct segment, especially in dense and complicated parts of the network, such as near intersections with acute angles or on close parallel roads. In this paper an artificial neural network (ANN) approach is explored to improve the segment identification accuracy of map-matching algorithms. The proposed ANN is continuously trained by using the horizontal shift imposed on GPS points and once it is trained, it will be used to correct raw GPS points before inputting them into the map-matching algorithm. Integrating the proposed ANN enabled an existing map-matching algorithm to find the correct segments for some of the GPS points where the original map-matching algorithm had failed to do so.