Models and Algorithms for Travel Time Reliability Assessment of Urban Road Networks Based-on Moving Source Data
|School||Beijing Jiaotong University|
|Course||Transportation Planning and Management|
|Keywords||Urban Road Network Travel Time Reliability Moving Source Data Log-Normal Distribution Logit Model|
The performance of an urban road network has become one of the most important problems that affect the effective operation and sustainable development of cities. Recently, the road network reliability evaluation technique has been developed rapidly, as a probability index for assessing the performance of urban road networks. Especially, the travel time reliability index has attracted increased interests from travelers, government officials, and traffic engineers, due to its ability to well describe the variability of travel times, which has become a fast-evolving research area. Meanwhile, the development of moving detection techniques has enabled the collection of travel time data in a real-time, dynamic, and accurate fashion.This dissertation is intended to develop travel time reliability evaluation models and algorithms, based on the analysis and mining of moving source data, and thus provides a new approach for assessing the performance of road networks. The research in this dissertation contains the following accomplishments:1. The principles for designing the travel time reliability index are developed as comprehensive, systematic, and practical. A new link travel time reliability evaluation index is defined by utilizing the traffic level-of-services and the corresponding expected travel time, overcoming the shortcoming of existing link travel time reliability evaluation indices, which is that it did not provide an exact definition of the link function.2. The link travel rate distributions at AM peak hours for different days and months and for different classes of roads are modeled respectively by using the Normal distribution, Log-Normal distribution, Gamma distribution, and Weibull distribution. The results indicate that a Log-Normal distribution shows the best curve-fitting performance. Therefore, a Log-Normal distribution Reliability model (LOGNR) is developed to assess the link travel time reliability.3. The expected travel times corresponding to the acceptable level-of-service for different classes of roads in urban road network in China are determined, by analyzing the relationships among the volume, speed, and density, and calculating the percentile of link travel times.4. The minimum sample size model of moving source data for the assessment of the link travel time reliability is developed, by applying the theories of random sampling and interval estimation of normal distribution. A heuristic algorithm is designed for the model, and the minimum sample sizes for typical links are calculated.5. The parameter estimation method of the Expectation Maximization (EM) algorithm-based link travel time reliability evaluation model is developed, by combining the improved random interpolation method and the EM algorithm. Further, the EM algorithm is proved, by a comparison, to be more robust for estimating the parameters of link travel time reliability models, than the Maximum Likelihood (ML) estimates based on the observed data or the randomly interpolated data.6. The Sum of LOG-Normal distribution Reliability (SLOGNR) model is developed for assessing the route travel time reliability, based on the probabilistic characteristics about the sum of lognormal stochastic variables, i.e. "a sum of lognormal random variables also follows the lognormal distribution." Both independent and inter-dependent links are considered in the SLOGNR model.7. The stochastic traveler route choice behaviors are analyzed. The Logit route choice model is improved, which overcomes two shortcomings of the Logit model: (1) the problem that the route choice probability depends on the absolute resistances of routes; and (2) the problem about overlapped links. As a result, the network travel time reliability evaluation model is developed based on the improved Logit route choice model.