A Map-matching Algorithm to Improve Vehicle Tracking Systems Accuracy

The satellite-based vehicle tracking systems accuracy can be improved by augmenting the
positional information using road network data, in a process known as map-matching.
Map-matching algorithms attempt to estimate vehicle route and location in a particular
road map (or any restricting track such as rails, etc), in spite of the digital map errors and
GPS inaccuracies. Point-to-curve map-matching is not fully suitable to the problem since
it ignores any historical data and often gives inaccurate, unstable, jumping results. The
better curve-to-curve matching approach consider the road connectivity and measure the
curve similarity between the track and the possible road path (hypotheses), but mostly
does not have any way to manage multiple route hypotheses which have varying degree of
similarity over time. The thesis presents a new distance metric for curve-to-curve mapmatching
technique, integrated with a framework algorithm which is able to maintain
many possible route hypotheses and pick the most likely hypothesis at a time, enabling
future corrections if necessary, therefore providing intelligent guesses with considerable
accuracy. A simulator is developed as a test bed for the proposed algorithm for various
scenarios, including the field experiment using Garmin e-Trex GPS Receiver. The results
showed that the proposed algorithm is able to improve the map-matching accuracy as
compared to the point-to-curve algorithm.
Keywords: map-matching, vehicle tracking systems, Multiple Hypotheses Technique,
Global Positioning System.