Adaptive Filtering Algorithm for Estimating Signalized Arterial Travel Times

Abstract

Based on advanced cross-sectional detection technology, such as automatic number-plate recognition, travel-time data can be obtained directly by matching vehicle information. However, because of drivers’ behaviors such as en route stopping, detouring, or driving at an abnormal speed, the corresponding travel times may not be representative of the road section. Filtering out such data can improve the estimation accuracy of travel time and enhance its applicability to subsequent research. Considering that the control delay occupies a large proportion of the travel time on signalized arterials, this paper divides the travel time into non-stopped travel time, control delay, and activity delay. Then, we determine the stops threshold in each cycle based on the stops progression pattern and modify the threshold using a series of fault-tolerance approaches. Finally, the modified thresholds are used to implement the cycle-based travel-time filtering for signalized arterials. The method has been comprehensively validated using both empirical and simulation data. The results indicate that the proposed method can provide accurate and reliable threshold windows, effectively track the travel-time variations, and maintain robust performance under various conditions.