Ehsan Mazloumi

Using GPS Data to Gain Insight into Public Transport Travel Time Variability

Transit service reliability is an important determinant of service quality, which has been mainly studied from the perspective of passengers waiting at stops. Day-to-day variability of travel time also deteriorates service reliability, but is not a well-researched area in the literature partly due to the lack of comprehensive data sets on bus travel times. While this problem is now being addressed through the uptake of global positioning system (GPS)-based tracking systems, methodologies to analyze these data sets are limited. This paper addresses this issue by investigating day-to-day variability in public transport travel time using a GPS data set for a bus route in Melbourne, Australia. It explores the nature and shape of travel time distributions for different departure time windows at different times of the day. Factors causing travel time variability of public transport are also explored using a linear regression analysis. The results show that in narrower departure time windows, travel time distributions are best characterized by normal distributions. For wider departure time windows, peak-hour travel times follow normal distributions, while off-peak travel times follow lognormal distributions. The factors contributing to the variability of travel times are found to be land use, route length, number of traffic signals, number of bus stops, and departure delay relative to the scheduled departure time. Travel time variability is higher in the AM peak and lower in the off-peak. The impact of rainfall on travel time variability is only found significant in the AM peak. While the paper presents new methods for analyzing GPS-based data, there is much scope for expanding knowledge through wider applications to new data sets and using a wider range of explanatory variables.

Prediction Intervals to Account for Uncertainties in Travel Time Prediction

The accurate prediction of travel times is desirable but frequently prone to error. This is mainly attributable to both the underlying traffic processes and the data that are used to infer travel time. A more meaningful and pragmatic approach is to view travel time prediction as a probabilistic inference and to construct prediction intervals (PIs), which cover the range of probable travel times travelers may encounter. This paper introduces the delta and Bayesian techniques for the construction of PIs. Quantitative measures are developed and applied for a comprehensive assessment of the constructed PIs. These measures simultaneously address two important aspects of PIs: 1) coverage probability and 2) length. The Bayesian and delta methods are used to construct PIs for the neural network (NN) point forecasts of bus and freeway travel time data sets. The obtained results indicate that the delta technique outperforms the Bayesian technique in terms of narrowness of PIs with satisfactory coverage probability. In contrast, PIs constructed using the Bayesian technique are more robust against the NN structure and exhibit excellent coverage probability.

Prediction intervals to account for uncertainties in neural network predictions: Methodology and application in bus travel time prediction

Neural networks have been employed in a multitude of transportation engineering applications because of their powerful capabilities to replicate patterns in field data. Predictions are always subject to uncertainty arising from two sources: model structure and training data. For each prediction point, the former can be quantified by a confidence interval, whereas total prediction uncertainty can be represented by constructing a prediction interval. While confidence intervals are well known in the transportation engineering context, very little attention has been paid to construction of prediction intervals for neural networks. The proposed methodology in this paper provides a foundation for constructing prediction intervals for neural networks and quantifying the extent that each source of uncertainty contributes to total prediction uncertainty. The application of the proposed methodology to predict bus travel time over four bus route sections in Melbourne, Australia, leads to quantitative decomposition of total prediction uncertainty into the component sources. Overall, the results demonstrate the capability of the proposed method to provide robust prediction intervals.

An Integrated Framework to Predict Bus Travel Time and Its Variability Using Traffic Flow Data

Information about bus travel time and its variability is a key indicator of service performance, and it is valued by passengers and operators. Despite the important effect of traffic flow on bus travel time, previous predictive approaches have not fully considered a traffic measure making their predictions unresponsive to the dynamic changes in traffic congestion. In addition, existing methodologies have primarily concerned predicting average travel time given a certain set of input values. However, predicting travel-time variability has not received sufficient attention in previous research. This article proposes an integrated framework to predict bus average travel time and its variability on the basis of a range of input variables including traffic flow data. The framework is applied using GPS-based travel-time data for a bus route in Melbourne, Australia, in conjunction with dynamic traffic flow data collected by the Sydney Coordinated Adaptive Traffic Systems loop detectors and a measure of schedule adherence. Central to the framework are two artificial neural networks that are used to predict the average and variance of travel times for a certain set of input values. The outcomes are then used to construct a prediction interval corresponding to each input value set. The article demonstrates the ability of the proposed framework to provide robust prediction intervals. The article also explores the value that traffic flow data can provide to the accuracy of travel-time predictions compared with when either temporal variables or scheduled travel times are the base for prediction. While the use of scheduled travel times results in the poorest prediction performance, incorporating traffic flow data yields minor improvements in prediction accuracy compared with when temporal variables are used.

Lane-Changing Decision Model for Heavy Vehicle Drivers

Lane-changing maneuvers of heavy vehicles have significant influence on surrounding traffic characteristics because of the physical effects that the heavy vehicles impose on surrounding vehicles. These effects are the result of heavy vehicles’ length, size, weight, and limitations in their maneuverability. Because of the importance of heavy vehicle lane-changing maneuvers, this article presents an exclusive fuzzy logic lane-changing decision model with 2 and 3 fuzzy sets for heavy vehicle drivers on freeways. To examine the accuracy of the heavy vehicle drivers’ lane-changing decision model, the authors compared the number of estimated lane-changing maneuvers of heavy vehicles with the observed number of heavy vehicle lane-changing maneuvers. In addition, the authors compared estimated traffic flow measurements (e.g., traffic flow, average speed) with the observed traffic flow measurements through microscopic traffic simulations. Results show that the fuzzy logic heavy vehicle lane-changing model can better replicate the observed lane-changing decisions of heavy vehicle drivers than can the current lane-changing decision models. Furthermore, using an exclusive heavy vehicle lane-changing decision model can increase the accuracy of the microscopic traffic simulation software in estimating the macroscopic traffic flow measurements.

Exploring the Value of Traffic Flow Data in Bus Travel Time Prediction

The accurate prediction of transit travel times has a range of applications to benefit operators and passengers. Transit travel time is affected by several factors such as traffic flow and passenger demand, which have to be considered to make accurate predictions. However, previous studies have not considered real world traffic flow variables in their prediction models. This paper develops artificial neural network (ANN) models to predict bus travel time on the basis of a range of input variables including traffic flow data collected from a bus route in Melbourne, Australia. To overcome the drawback of ANNs in determining the effect of each input variable on the independent variable, the paper adopts a regression analysis to determine the important input variables for prediction. The paper examines the value that traffic flow data would make to the prediction accuracy. To this end, two alternative models are developed and the results are compared with those obtained from the traffic flow data–based models. A historical data–based ANN in which temporal variables are substituted with the traffic flow variable and a timetable-based model that traditionally utilizes scheduled travel times are developed. Although the use of scheduled travel times results in the poorest prediction performance, incorporating traffic flow data yields minor improvements in prediction accuracy compared with when temporal variables are used.

Enhanced Evaluation of Heavy Vehicle Lane Restriction Strategies in Microscopic Traffic Simulation

Heavy vehicles can considerably affect traffic flow particularly during heavy-traffic conditions. Large numbers of heavy-vehicle lane-changing maneuvers can contribute to increase the number of traffic accidents and hence to reduce the freeway safety. The increase in the number of heavy vehicles on freeways has been the motivation to establish strategies to reduce the interaction between heavy vehicles and passenger cars. Previous studies have examined different lane-restriction strategies for heavy vehicles using microscopic traffic-simulation packages. Those packages mostly use a general lane-changing model to estimate the lane-changing behavior of heavy-vehicle and passenger-car drivers. The general lane-changing models ignore the fundamental differences in the lane-changing behavior of passenger cars and heavy vehicles. However, an exclusive lane-changing model for heavy vehicles can increase the accuracy of simulation models. The application of such a model can result in a more reliable evaluation of lane restriction strategies. In this paper, different lane restriction strategies are defined for heavy vehicles. For each strategy, the macroscopic and microscopic traffic measurements of two freeway sections in California are analyzed, using the VISSIM default lane-changing model and an exclusive heavy- vehicle lane-changing model. The results show that the VISSIM default model unrealistically overestimates the observed number of heavy-vehicle lane-changing maneuvers and potentially overestimates the number of traffic accidents. Using the exclusive lane-changing model for heavy vehicles enhances the accuracy of the VISSIM traffic simulation model in microscopically estimating the lane-changing maneuvers of heavy vehicles.