An integrated algorithm for fusing travel times, local speed and flow

Traffic information plays an important role in traffic management and control. In order to better estimate traffic states from varieties of traffic sensors, traffic data fusion techniques are employed. This paper proposes an algorithm which can fuse travel time information and local loop data. Travel times from in-car GPS or cameras can provide average journey speeds over a few road segments but fails to provide the traffic details on each segment. Loop detectors can provide local traffic information, but there are biased errors in speed measures and the error in vehicle counts is accumulated over the time. The proposed algorithm exploits the strength of each type of data and avoids their weakness. In contrast to the previously often-used traffic fusion techniques, this algorithm needs very few assumptions on traffic behavior but can fuse more types of data. The validation shows this fusion algorithm can improve the estimation accuracy by up to 10 times.