Abstract
Recently, link travel time distribution (LTTD) estimation has gained a lot of interest since the probabilistic model not only captures the dynamic features of link travel time but also provides abundant knowledge like the mean and variance which can be used as indicators to analyze link travel time reliability. However, existing methods still suffer from a number of problems: 1) most studies employ parametric models, e.g., Gaussian, which is only suitable in the limited traffic conditions like free flow or congestion. 2) many techniques heavily rely on the measurements detected on the roads. They cannot be applied to the whole road network since there is absence of observations in some roads due to the limited number of traffic detectors installed in the road network. In lieu of the aforementioned challenges, in the paper, we employ kernel density estimator (KDE) to model LTTD which is validated to be effective in any state of traffic condition. Further, motivated by the network tomography techniques, we propose an expectation maximization (EM) based algorithm to estimate model parameters only with end-to-end (E2E) measurements detected by traffic detectors at or near some road intersections. With 3.0e+07 GPS trajectories collected by the taxicabs in Xi'an, China, the experimental results show that the LTTD estimated by our proposed method are in excellent agreement with the empirical distributions, and better than its counterparts adopting Gaussian and log-normal models.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have