As of October 1, 2016, the SWUTC concluded its 28 years of operation and is no longer an active center of the Texas A&M Transportation Institute. The archived SWUTC website remains available here.


SWUTC Ph.D. Candidate Assistantship Project Description

New Methodologies for Analyzing Freeway Traffic Flow Characteristics

University: Texas A&M University

Principal Investigator:
Yajie Zou
Zachry Department of Civil Engineering
(979) 845-5674

Faculty Supervisor:
Yunlong Zhang
Zachry Department of Civil Engineering
(979) 845-9902

Funding Source: USDOT Funds

Total Project Cost: $32,079

Project Number: 600451-00020

Date Started: 9/1/12

Estimated Completion Date: 8/31/13

Project Summary

Project Abstract:
The knowledge of speed, headway and travel time is necessary because these variables are fundamental measures of traffic performance of a highway system. Therefore, developing reliable and innovative analytical techniques for analyzing these variables is very important. The primary goal of this research is to develop some new methodologies for analyzing and describing the characteristics of speed, headway and travel time.

Project Objectives:
The primary goal of this research is to develop new methodologies for analyzing the characteristics of speed, headway and travel time. To accomplish this goal, following objectives are planned to be addressed in this research. The corresponding contributions of each part are also provided.

The first part concerns the heterogeneity problem in freeway vehicle speed data. Previously, normal, log-normal, gamma and other forms of distribution have been used to fit speed data. If the characteristics of speed data are homogeneous (or unimodal distribution), the distribution of speed can be generally modeled by normal, log-normal and gamma distributions. However, if the speed data exhibit excess skewness and bimodality (or heterogeneity), unimodal distribution function does not give a satisfactory fit. Thus, the mixture model has been proposed by May (1990) for traffic stream that consists of two classes of vehicles or drivers. So far, the mixture models used in previous studies to fit bimodal distribution of speed data considered normal density as the specified component; therefore, other types of component density were not fully investigated. To capture excess skewness, kurtosis and bimodality present in speed distribution, we propose skew-normal and skew-t mixture models to fit freeway speed data. Skew-normal and skew-t distributions are known for their flexibility, allowing for heavy tails, high degree of kurtosis and asymmetry. To investigate the applicability of mixture models with skew-normal and skew-t component density, we fit a 24-hour speed data collected on IH-35 in Texas using skew-normal and skew-t mixture models with the Expectation Maximization type algorithm.

This part shows that finite mixture of skew distributions can significantly improve the goodness of fit of speed data. Moreover, the proposed skew-normal and skew-t mixture models can better account for heterogeneity problem in freeway vehicle speed data. Thus, the methodology developed in this part can be used in analyzing the characteristics of freeway speed data. The findings can also be used in development and validation of microscopic simulation of freeway traffic. For example, considering that many traffic analytical and simulation models use speed as an input for travel time, delay, and level of service determination, the developed models can generate more accurate speed value as the input and help improving the reliability of the analysis output.

The second part concerns the correlation between freeway speed and headway data. Traditionally, speed and headway are assumed to be independent in the microscopic simulation models. As a result, the same headway distribution may be assumed for different speed levels and this assumption neglects the possible variability of headway distribution across speed values. Moreover, a number of developed microscopic simulation models generate vehicle speeds and vehicle arrival times as independent inputs to the simulation process.  Up to date, few studies have been directed at exploring the dependence between speed and headway. Considering the potential dependence between speed and headway, it is useful to construct bivariate distribution models to describe the characteristics of speed and headway. Compared with one dimensional statistical models representing speed or headway separately, bivariate distributions have the advantage that the possible correlation between speed and headway is taken into consideration. Given this advantage, bivariate distribution models can be used to improve the accuracy or validity of microscopic simulation models. The objective of second part is to develop a bivariate distribution for capturing the dependence, and describing the characteristics of speed and headway simultaneously. To examine the applicability of the proposed methodology, the developed model is applied to a 24-hour speed and headway dataset collected on IH-35 in Austin, Texas.

The second part shows that the bivariate distribution for speed and headway has been successfully constructed. The proposed methodology can overcome the correlation problem associated with the traditional approach. The developed bivariate distribution can be used to determine the headway distribution for given speed values, and vice versa. That is, given a specific value of speed, the headway distribution corresponding to that speed value can be derived. Thus, the proposed method can provide a better framework for developing accurate microscopic simulation models to generate vehicle speeds and vehicle arrival times simultaneously by considering the dependence between speed and headway.

The third part concerns the freeway link travel time prediction. Because of the highly dynamic and nonlinear evolvement of traffic condition over time and space, travel time prediction remains a difficult yet important challenge for transportation engineers. So far, there are many techniques and models developed for short-term travel time prediction. Previously, van Lint et al. (2005) categorized the explored techniques into three major strands: model-based approaches, instantaneous approaches and data-driven approaches. Despite a large number of short-term travel time prediction approaches have been developed in the past decade, few studies take into account spatial and temporal travel time information simultaneously in the prediction model. Such techniques are particularly useful in predicting the freeway link travel times since the traffic condition on the neighboring links can help identifying the traffic condition on the target link. In this part, our objective is to develop a space-time diurnal method, which merges the spatial and temporal travel time information to obtain accurate short-term travel time prediction of freeway corridors under different traffic conditions. In this part, the space-time diurnal method is examined using the travel time data collected on a test bed along the US-290 in the Houston area.

In this part, we develop a space-time diurnal method to obtain accurate short-term travel time prediction of freeway corridors under different traffic conditions. Contrary to the most existing methods that yield the point prediction of short-term travel time, the developed methodology can overcome the drawbacks of the point prediction and describe the uncertainty of the future travel times using the prediction intervals. Unlike some other prediction methods which lack a good interpretation of the model, the proposed technique can make use of geographically dispersed travel time observations as predictors to obtain short-term prediction and can yield theoretically interpretable prediction models. Finally, the computational requirements of the developed method are modest and this technique can be implemented in online travel time prediction.

Task Descriptions:

Task 1– Literature review

Task 2– Mixture modeling of freeway speed data

Task 3– Bivariate modeling of freeway speed and headway data

Task 4– Predicting freeway link travel time with a space-time diurnal method

Task 5 & 6– Documentation and Conclusions