SUMMARY
Probe-based data collection system, which deploys movable sensors in traffic streams to collect traffic records, is commonly used for area-wide measurement of travel times and conditions of traffics on road networks. However, due to sparse distribution of polling positions from a low probe sampling frequency in practice, the sampled points from probe vehicles do not necessarily match with the end points of each link, and the travel time between two consecutive polling points needs to be positioned into an individual link for link travel time calculation. This paper presents the new technique that addresses the allocation of travel times based on data gathered from sparse GPS probe vehicles. The travel time allocation procedure by applying the relationship among instantaneous speeds, sampled positions and tracked times is proposed. The proposed technique is discussed and compared with a ground truth benchmark of high resolution field data recorded from urban roadway corridor in both congested and uncongested traffic conditions. Results suggest that addressing travel time allocation problems by the proposed method significantly outperforms the benchmark method in both complete link and local levels which are demonstrated by an increasing estimation accuracy up to 60%. It is also observed from the local level analysis that the accuracy figures are lower at the intersections comprising high stopped delays, which normally contain highly fluctuated speed profiles. Nonetheless, the proposed technique outperforms the baseline approach in both intersections with low or high stopped delays. The proposed technique offers superior results to the baseline technique in all scenarios, particularly on the segments with highly fluctuated speed profiles.