Citation

Yu Z, Shi S, Wang S, Mu Y, Li W, Xu C, Zhang M. Transp. Res. D Trans. Environ. 2017; 54: 114-141.

Copyright

(Copyright © 2017, Elsevier Publishing)

DOI

10.1016/j.trd.2017.04.029

PMID

unavailable

Abstract

A driving cycle is a speed-time profile for a vehicle driving under a specified condition. It is usually developed from vehicle driving data collected by experiments to represent the real-life diving patterns. Because of lacking corresponding sampling theory, it is difficult for engineers to determine when vehicle driving cycle experiments should be stopped. In order to obtain sufficient experimental data, engineers normally choose to prolong the time of experiments wasting time and money. How to build a synthetical sampling subset of data representing a larger one becomes a main problem of sampling experiments. This paper, based on statistical inference theory, proposed a method to solve this problem at the city zone scale. First, the information entropy of road intersections was applied to determine the reasonable zone size. Then, according to one-month driving data of Changchun taxis and one-week driving data of Beijing taxis, it was found that the traffic distribution in city zone were able to be described by Nakagami distribution. It can pass the K-S test under the 0.05 significance level. In the order to fully use driving data, the bootstrap method was employed to conduct three resampling experiments in Changchun and five in Beijing. After analyzing the confidence intervals of distribution parameters, this paper discovered that the quality of the sampling data could be indicated by the accuracy of each zone's per car per day per square kilometers travel times. The linear relationship between the expectation of zone travel times variable coefficient and the expectation of a ab which was used to evaluate the similarity between sampling distribution and population distribution was discovered. This relationship was also proved in this paper theoretically. Since the expectation of variable coefficient can be computed by sampling data, engineers are able to estimate the quality of these data in real time. If the α_αb reaches the preset threshold, experiments can be stopped.

Language: en

Keywords

Speed; Time; Data mining; Taxicabs; Beijing (China); Changchun (China); Driving cycle; Statistical inference