Zhen (Sean) Qian

Investigating driver injury severity patterns in rollover crashes using support vector machine models

Rollover crash is one of the major types of traffic crashes that induce fatal injuries. It is important to investigate the factors that affect rollover crashes and their influence on driver injury severity outcomes. This study employs support vector machine (SVM) models to investigate driver injury severity patterns in rollover crashes based on two-year crash data gathered in New Mexico. The impacts of various explanatory variables are examined in terms of crash and environmental information, vehicle features, and driver demographics and behavior characteristics. A classification and regression tree (CART) model is utilized to identify significant variables and SVM models with polynomial and Gaussian radius basis function (RBF) kernels are used for model performance evaluation. It is shown that the SVM models produce reasonable prediction performance and the polynomial kernel outperforms the Gaussian RBF kernel. Variable impact analysis reveals that factors including comfortable driving environment conditions, driver alcohol or drug involvement, seatbelt use, number of travel lanes, driver demographic features, maximum vehicle damages in crashes, crash time, and crash location are significantly associated with driver incapacitating injuries and fatalities. These findings provide insights for better understanding rollover crash causes and the impacts of various explanatory factors on driver injury severity patterns.

The morning commute problem with heterogeneous travellers: the case of continuously distributed parameters

We study the morning commute problem with a heterogeneous travelling population whose early/late arrival penalty parameters are continuously distributed. Following Arnott et al.s (1988) [Schedule delay and departure time decisions with heterogeneous commuters, Transportation Research Record, 1197, 56–67] model, where the ratio of the value of early schedule delay (VESD) over the value of late schedule delay (VLSD) is assumed to be constant across the population, we first derive the user-optimal travel profiles and the corresponding total travel time (TTT) for a one-to-one network with a single route. We show that although every commuter is better off if the bottleneck capacity is enlarged, commuters with high values of early arrival penalty (EAP) benefit more than those with low values. In addition, the homogeneity assumption overestimates the queuing delay and thus the TTT. Then we extend our analysis to a network with two routes, a freeway and an arterial road (AR) connecting a single origin–destination (O–D) pair. In this case, we find that the travellers who weigh schedule delay more will first shift to the AR, prompted by an increase in total demand. However, the critical travel demand at which travellers start to use the AR remains the same if the common EAP for homogeneous users is equal to the expected value of EAP distribution for heterogeneous users. Interestingly, there exists a critical value of EAP parameter, such that all the travellers whose EAP is less than it choose the freeway, and those whose EAP is larger than it choose the AR. Sensitivity analysis on TTT is performed with respect to freeway capacity, AR capacity and free-flow travel time on the AR. The results indicate that enlarging freeway (AR) capacity will always reduce the TTT of the whole network and the TTT of the AR (freeway), and increase the demand share of the freeway (AR). We also show that every commuter is better off if either the freeway capacity or the AR capacity is enlarged. Finally, we provide numerical examples to demonstrate the considerable differences in flow patterns and network performances between a homogeneous and a heterogeneous population.

Optimal dynamic pricing for morning commute parking

Morning commute parking is a challenging issue for both commuters and transportation planners. This paper proposes a procedure to maximise the benefits of parking management by optimal dynamic parking pricing. A generic parking model is presented for a set of sequential parking areas. Travellers make parking location choices and departure time choices to minimise their generalised travel cost, when provided with either information regarding time-varying parking occupancies and prices, or after sufficient day-to-day parking experience. We consider expected parking cruising time to be a function of parking occupancy. When parking demand is inflexible in departing schedules, we demonstrate that the system optimal (SO) pricing solution is not unique. The non-uniqueness offers flexibility to achieve additional parking management goals by dynamic prices, such as constant arrival rates (CAR). For the case where travellers have full flexibility to change their departure times, the SO solution is not practically achievable. We propose the CAR solution where the total cost is minimised constrained by that the arrival rates to each area exactly match a desired rate (e.g. flow capacity). We show that each parking area should be priced in such a way that a maximum occupancy cap should be imposed. Convenient parking spots should be high priced at the beginning of the morning commute and saved for the peak traffic. In addition, we derive the optimal parking prices with respect to average occupancies, which holds potentials to guide parking price adjustment for traffic management.

Computing Individual Path Marginal Cost in Networks with Queue Spillbacks

“Individual path marginal cost” (IPMC) is defined as the change in travel cost of one unit of flow on a time-dependent path caused by one unit of flow on another time-dependent path. Knowledge of IPMC is central to dynamic transportation modeling, for instance, to compute system-optimal network performance, to solve a dynamic origin–destination (O-D) estimation problem, and to analyze equity issues for travelers with different origins and destinations. This paper proposes a method of approximating IPMC for general networks, in which a cell transmission model–based kinematic wave model is used to model traffic dynamics. By tracing the changes in the cumulative flow curves of the bottleneck links on which queues form during dynamic network loading, an approximation method is developed to obtain the IPMC for the cases of merge junctions, diverge junctions, and general junctions. This method was applied to compute the total path marginal cost in a network. The results showed that vehicles at the beginning of the congestion duration had significantly larger marginal travel costs than other vehicles. The method was then applied to solve a dynamic O-D estimation problem with partial link-flow counts and historical O-D trip tables. With the incorporation of IPMC into the estimation procedure, both the O-D demands and the observed path travel times were successfully reproduced.

A novel work zone short-term vehicle-type specific traffic speed prediction model through the hybrid EMD–ARIMA framework

This paper presents a hybrid short-term traffic speed prediction framework through empirical mode decomposition (EMD) and autoregressive integrated moving average (ARIMA). The goals of this paper are to investigate (1) does the hybrid model provide better short-term traffic conditions (i.e. traffic speeds) than the traditional models? (2) how the performance of the hybrid model varies for varying scenarios such as mixed traffic flow and vehicle-type specific traffic prediction in a work zone, on-ramp, and off-ramp; and (3) why hybrid models provide better prediction than other single-staged models. Using empirical data from a work zone on interstate I91 in Springfield, MA and the on/off-ramp data from the Georgia State Route 400, the proposed hybrid EMD–ARIMA modelling framework is tested in the four distinct scenarios aforementioned. The prediction results of the hybrid EMD–ARIMA model are evaluated against the experimental data and also compared with the results from the traditional ARIMA, the Holt–Winters, the artificial neural network models, and a naive model. The evaluation results showed that the hybrid EMD–ARIMA model outperforms the traditional forecasting models in different scenarios.

Estimating Time-Dependent Freeway Origin-Destination Demands with Different Data Coverage: Sensitivity Analysis

To obtain more credible estimates of time-dependent travel demand, various data sources should be exploited jointly to improve the observability of origin–destination (O-D) trip tables. A comprehensive case study uses a real freeway network to reveal how different data coverage affects the quality of estimated O-D tables. The dynamic O-D estimation problem is formulated as a variational inequality (VI) problem that provides a flexible framework to incorporate different data sources and to encapsulate realistic traffic flow dynamics. Traffic surveillance data considered include traffic counts from vehicle detectors, historical O-D tables (static or dynamic), and travel time measurements on subpaths. A novel method is employed to evaluate marginal path travel times, which is a key procedure to properly incorporate travel time measurements into the VI formulation. Numerical experiments generate a number of guidelines to the proper selection of data coverage for obtaining improved O-D estimates.

Empirical Mode Decomposition–Autoregressive Integrated Moving Average: Hybrid Short-Term Traffic Speed Prediction Model

Short-term freeway traffic speed prediction is essential to improving mobility and roadway safety. It has been a challenging and unresolved issue. Traffic speed prediction can be applied to enhance the intelligent freeway traffic management and control for applications such as operational and regulation planning. For example, with more reliable traffic speed prediction, the advanced traveler information system can provide travelers with predictive travel time information and optimal routing, which allows them to arrange their schedules accordingly. Moreover, traffic managers can use the predicted information to deploy various traffic management strategies to increase system efficiency. In this paper, a hybrid empirical mode decomposition (EMD) and autoregressive integrated moving average (ARIMA) (or EMD-ARIMA) approach was developed to predict the short-term traffic speed on freeways. In general, there were three stages in the hybrid EMD-ARIMA forecasting framework. The first was the EMD stage, which decomposed the freeway traffic speed time series data into a number of intrinsic mode function (IMF) components and a residue. The second stage was to find the appropriate ARIMA model for each IMF and residue and then make predictions on the basis of the appropriate ARIMA model. The third stage was to combine the prediction results of each IMF and residue to make the predictions. The experimental results indicated that the proposed hybrid EMD-ARIMA framework was capable of predicting short-term freeway traffic speed with high accuracy.

Full Closure or Partial Closure? Evaluation of Construction Plans for the I-5 Closure in Downtown Sacramento

Three alternative construction plans with full highway closures and limited capacity closures (partial closures) are evaluated for a large-scale freeway reconstruction project in the Sacramento metropolitan area. By using a dynamic network analysis tool, the potential impact resulting from different construction plans is estimated and analyzed measured by changes in total demand, travel delays, vehicle miles/hours traveled (VMT/VHT), fuel consumptions, and emissions. The results show that, for this Interstate 5 closure, if only user delay costs are considered, the actual full closure plan is the best. It produces considerably less total system travel delay than the two partial closure plans although it closed more lanes. This plan, however, may induce more VMT than the other two. In fact, the other two partial closure plans produce slightly fewer emissions and consume slightly less fuel. The outcome of which plan comes out on top crucially depends on two other factors, besides the chosen objective: the amount of demand reduction a plan induces and the availability of alternative routes to divert traffic away from the work zone area. Developing efficient demand management measures will be the key to improving the network performance and to reduce the emissions. Moreover, when data are available, construction costs and emissions produced by construction activities should also be included in the evaluations, in addition to the user costs considered in this study.

Statistical inference of probabilistic origin-destination demand using day-to-day traffic data

Abstract Recent transportation network studies on uncertainty and reliability call for modeling the probabilistic O-D demand and probabilistic network flow. Making the best use of day-to-day traffic data collected over many years, this paper develops a novel theoretical framework for estimating the mean and variance/covariance matrix of O-D demand considering the day-to-day variation induced by travelers’ independent route choices. It also estimates the probability distributions of link/path flow and their travel cost where the variance stems from three sources, O-D demand, route choice and unknown errors. The framework estimates O-D demand mean and variance/covariance matrix iteratively, also known as iterative generalized least squares (IGLS) in statistics. Lasso regularization is employed to obtain sparse covariance matrix for better interpretation and computational efficiency. Though the probabilistic O-D estimation (ODE) works with a much larger solution space than the deterministic ODE, we show that its estimator for O-D demand mean is no worse than the best possible estimator by an error that reduces with the increase in sample size. The probabilistic ODE is examined on two small networks and two real-world large-scale networks. The solution converges quickly under the IGLS framework. In all those experiments, the results of the probabilistic ODE are compelling, satisfactory and computationally plausible. Lasso regularization on the covariance matrix estimation leans to underestimate most of variance/covariance entries. A proper Lasso penalty ensures a good trade-off between bias and variance of the estimation.

Parking Sensing and Information System: Sensors, Deployment, and Evaluation

This paper describes a smart parking sensing and information system that disseminates parking availability information to public users in a cost-effective and efficient manner. The hardware framework of the system is built on advanced wireless sensor networks and cloud service over the Internet, and the system is highly scalable. The parking information provided to the users is set in the form of occupancy rates and expected cruising time. Both are obtained from an analytical algorithm that processes historical and real-time data and are then visualized in a color theme. The entire parking system is deployed and extensively evaluated at Stanford University, California, Parking Structure 1.