Journal of Transportation Engineering, Part A: Systems

Analysis of driving behaviors and related driver clustering is of great significance for improving driving safety, but traffic operation conditions (especially road types and operating speed) often are neglected in existing clustering studies, and the impact of excluding traffic conditions has not been investigated thoroughly. This research proposes an improved driver clustering framework by accounting for road types and average speed. The clustering results were compared with those without considering traffic conditions. The input data of more than 34 million records of second-by-second vehicle trajectories from 315 vehicles in Beijing were sliced into segments of 30 s, and these seconds were classified by road types (expressway versus non-expressway) and by $10\text{-}\mathrm{km}/\mathrm{h}$ average speed intervals. For each driver, the speed variation coefficients (SVCs), acceleration standard deviations (ASTDs), and average negative accelerations (ANAs) by traffic condition were entered into a Gaussian mixture model for an unsupervised clustering of drivers into types of prudent, normal, and aggressive drivers. The improved clustering framework is capable of capturing the variability of driving behaviors (especially dangerous driving behaviors such as sharp decelerations) across drivers, and the comparison demonstrated significant differences between the improved model and the original model with respect to the proportion of every driver type. The improved clustering framework performs better in both intraclass aggregation and interclass separation, and the results of this research indicate the need to consider traffic conditions in driving behavior–based clustering of drivers.

This work investigated the cooperative control of a highway on-ramp under a connected and automated vehicles (CAVs) environment based on improved variable time headway (IVTH). A mainline section with a connected on-ramp and off-ramp of an intelligent highway were used as research objects. An IVTH model was developed and its stability was proved. The traffic flow model of the section was established and then verified in experiments. A cooperative control strategy of highway on-off-ramps is proposed. The cooperative merging control model (CMCM) based on IVTH (CMCM-IVTH) was constructed in which the influence of off-ramp separation is considered. The model predictive control method based on particle swarm optimization is used to obtain the acceleration and deceleration of platoon leaders with CMCM-IVTH and the time headway of the following vehicles with the traffic flow model based on IVTH. The control effects were determined via simulation experiments. The results indicate that the acceleration and deceleration times of vehicles, fuel consumption, and collision possibility decreased significantly using the proposed method. Under a 20% diverging rate, the maximum number of slow-moving vehicles was reduced by as much as 75.0% and 33.3%, total delay time was reduced by as much as 72.9% and 18.9%, the average traveling velocity was increased by as much as 21.9% and 2.1%, compared with the no-optimization-control method and a method with only cooperative merging control, respectively. The operating efficiency and safety level can be enhanced.

Passengers pass ticket barrier gates without paying in metro stations all over the world. This kind of behavior is called fare evasion, and it is troublesome and costly to prevent. As a typical type of fare evasion, tailgating refers to following a fare-paying passenger through the gate. It can be dangerous because the passenger risks being injured by the barrier gate. To detect tailgating fare evasions automatically, the existing surveillance cameras in stations can be utilized to provide a visual-based method at a low cost and efficiently. However, occlusion by crowds during rush hours can lower the accuracy of regular recognition methods based on convolutional neural networks. Moreover, the behavior of a tailgater can be similar to that of other fare-paying passengers if the positional relationship is not taken into account. Thus, we propose a tailgating recognition method that uses videos as input. First, the estimated human pose data in each frame is obtained, of which incomplete skeletons are retained. Second, the multiple persons appearing in adjacent frames are matched, after which a sequence of skeleton data is generated for each pedestrian. Third, a time series of the positional relationship between passengers and the ticket barrier gate is extracted and the passing interval of passengers is defined as the indicator for detecting tailgating. Our experiments showed that tailgaters could be distinguished effectively from fare-paying passengers, and the time series can cope with missing joints caused by occlusion or misidentification in a few frames.

This paper introduces a methodology to quantify traffic change triggered by the combined effect of weather hazards based on winter weather hazards models. The winter weather hazards models for three vehicle types were developed with weigh-in-motion (WIM) data collected in the commuter highway in the cold Canadian region for 5 years. The developed model was utilized to simulate the variations of the percentage reduction for each vehicle type based on the 239 pairs of weather combinations composed of six cold categories and various amounts of snowfall. The first phase involved measuring the marginal effect of weather factors such as cold category (or temperature) on the percentage reduction in traffic volume. The second phase involved utilizing the same winter weather traffic model to quantify the effect of combined weather factors on the percentage reduction. The percentage reduction of the total traffic and passenger cars increased as temperature deteriorated and snowfall increased. Truck traffic decreased as snowfall increased, but interestingly, as temperature deteriorated, it was estimated that the truck traffic volume increased. This phenomenon assumed that truck traffic moves from low-maintenance to high-maintenance highways as the weather deteriorates. The methodology to quantify traffic volume changes can be adopted by highway agencies to determine the timing of the snowplow operation based on the risk assessed in terms of traffic volume reduction. It can also be used to predict the percentage reduction of traffic and then determine whether to open or close a highway.

This paper presents a novel strategy for giving priority to automated guideway transit vehicles (AGTVs) in mixed traffic flow. From the perspective of road segments, a moving-block operation mode (MBOM) was proposed to help AGTVs eliminate the dependence on dedicated lanes. The car-following and lane-changing behaviors of both AGTVs and general vehicles were modeled using the theory of cellular automata. From the perspective of intersections, an MBOM-based dynamic multirequest signal priority (DMSP) model was developed to support the decision of multiple priority requests at the intersection. Notably, the DMSP model can synchronously deal with early green and green extension requests from the same or different phases. Extensive microsimulation experiments were conducted to examine the proposed strategy at various levels of traffic volume and AGTV headway. The results indicated that the MBOM outperforms the traditional strategy of setting up full-time or intermittent dedicated lanes, particularly in traffic conditions where the headway of AGTVs is longer than 300 s and the traffic volume is lower than $9.0\mathrm{kpcu}/\mathrm{h}$. Furthermore, the MBOM-based DMSP strategy was evaluated in terms of the performance of both traffic and energy efficiency. The per person travel time and coal consumption decreased by 6.93% and 1.61%, respectively, demonstrating the effectiveness of improving operational efficiency and sustainability of public transit.

In the current version (Version 6) of the Highway Capacity Manual (HCM), equal-capacity passenger car equivalencies (EC-PCEs) are used to account for the effect of trucks for capacity analyses. The EC-PCEs for freeway segments were estimated using a microsimulation-based methodology in which the capacity of different traffic scenarios was modeled. However, the current EC-PCEs were estimated under assumptions atypical of past HCM releases, including a new definition of capacity and a new data aggregation level. It is important to assess to what extent these assumptions may affect the passenger car equivalencies (PCE) values published in the HCM-6. Consequently, this paper compares the HCM-6 PCEs, and associated capacity adjustment factors (CAF), with values developed using the HCM-6 EC-PCE methodology with historic HCM assumptions. It was found the EC-PCE values from the proposed approach were, on average, 18.2% greater than those published in the HCM-6. This finding suggests the values published in the HCM-6 could underestimate the effect of trucks in capacity and level of service analyses.

It is common knowledge that rainfall and rain-induced waterlogging have negative impacts on urban road traffic in that they could cause decreased traffic flow capacity and induce traffic congestion from a macroscopic view. What is the ultimate cause behind the phenomenon from a micro perspective, however, is still unclear. This paper proposes a variable cell transmission model (VCTM) considering the influence of rainfall and waterlogging on traffic parameters based on the key findings from a previous experiment-simulation study. A typical local traffic network in a city with seven road sections was chosen as a study area, and a case study was conducted in a scenario of rainfall and waterlogging based on the proposed model. The results comprehensively profile how rainfall and waterlogging induce flow fluctuation and density fluctuation. It is found that road capacity is a dominant factor influencing the flow fluctuation, and the flow fluctuation tends to be much more violent near the ponding areas. It is also found that density increases with increased rainfall intensity and water depth, and the density increase caused by waterlogging seems to be more significant. Another important discovery is made that density of some roads could increase violently due to insufficient capacities of downstream roads. These valuable findings from this study help better understand the influence of rainfall and waterlogging on urban traffic flow. This work can also help guide the emergency response by the corresponding government department when encountering rainfall and waterlogging.

Car following is the most common driving behavior in tunnels. However, current car-following models are not completely suitable for tunnels because they do not take into account the environmental factors affecting vehicles in tunnels. In this paper, we present a comprehensive risk-based car-following model to describe car-following behavior in freeway tunnels. Considering the key factors influencing driving behavior in freeway tunnels, we develop a comprehensive safety field (CSF), which consists of the potential, kinetic, and environment fields to estimate the effect of speed limit signs, leading vehicles, and lighting conditions on driving risks. Then, a car-following model based on comprehensive safety field (CF-CSF) was established to determine a vehicular driving strategy in tunnels. The field force is introduced as a quantitative indicator to assess the current driving risk of vehicles and whose increase causes a greater deceleration of vehicles. Furthermore, considering the effect of low-risk levels on driving behavior is generally insignificant, we develop the risk margin (RM) as a safety indicator to determine whether current driving risk affects the driving behavior, and the driving strategy under a free condition is proposed as well. Finally, the proposed CF-CSF model is validated using a real vehicle test trajectory dataset. The comparison with real driving data and some classic car-following models indicate that our proposed CF-CSF model can more accurately predict actual driving behavior in tunnels. It is expected that the findings in this study could be valuable in modeling, understanding, and replicating features of driving behavior in freeway tunnels.

With the growth of a city’s economy, neighboring cities are gradually integrated to form an urban agglomeration. As travel activities driven by various travel demands frequently take place within an urban agglomeration, it is essential to understand the travel demand between cities and improve the intercity transportation system, which promotes the coordinated development of cities in an urban agglomeration. This paper presents our investigation of the travel demand characteristics of urban agglomeration cities. The Beijing-Tianjin-Hebei area and its passenger flows between transportation hubs in the different cities, which are part of a typical urban agglomeration in China and representative travel demand, are taken as the empirical study objects. First, we introduce a method to extract trip data using mobile phone call detail record (CDR) data, which carry rich geographical information on travelers and has been extensively used in recent transportation research. Based on trip data, directed weighted travel demand networks were constructed, with the nodes representing the transportation hubs and the edges representing passenger flows. The results showed that edge weights can be characterized by power-law distribution, which reveals the phenomenon that most passenger flows are concentrated among a few hubs, implying unbalanced travel demand in the urban agglomeration. Our empirical findings contribute to a method for applying large-scale location-based data to extract human mobility information and to understanding the nature of travel demand on the scale of an urban agglomeration. They also provide guidance to government agencies in developing appropriate transportation policies and enhancing infrastructure in the urban agglomeration.

Dynamic load factors ($\varphi$) relate the magnitude of vertical wheel-to-rail loads in operation (dynamic loads) to static loads resulting from the weight of the rail car and its contents, as a function of train speed. The $\varphi$ equations are often used in the selection of rail steel and cross-sections (weight). Equations for $\varphi$ have been put forth by the American Railway Engineering and Maintenance-of-Way Association (AREMA) and others. A limitation of the existing $\varphi$ equations is that they have been derived from loads measured at instrumented sections of track and observe the many wheel loads but with constant track conditions. For this study, measurements of dynamic loads from two instrumented wheelset (IWS) as it conducted four passes over a 340 km section of track operated by a North American Class 1 freight railway through the Canadian Prairies. These measurements provided dynamic loads from one loaded freight car over various track structures at differing train speeds. This paper presents the IWS data sets and the variation of dynamic loads between multiple passes of the section of track studied, and the statistical distribution of dynamic loads. New $\varphi$ equations are developed for tangent track and nontangent track (inclusive of bridges, grade crossings, curves, and switches).

A crucial component of the proactive traffic safety management system is the real-time crash likelihood prediction model, which takes real-time traffic data as input and predicts the crash likelihood for the next 5+ min. This study aims to investigate the application of trajectory fusion to crash likelihood prediction and improve the predictive accuracy of the deep learning crash likelihood prediction model using the temporal attention mechanism. Two trajectory data were integrated using data fusion techniques. Specifically, trajectory data from Lynx buses and the Lytx fleet were collected using the automatic vehicle locator (AVL) and Lytx DriveCam, respectively. A deep learning model was developed for predicting real-time crash likelihood using features extracted from trajectory data. The proposed model contained a temporal attention–based long short-term memory (TA-LSTM) and a convolutional neural network (CNN). Temporal attention was introduced to capture temporal correlations between time-series data. Experimental results suggested that temporal attention could significantly improve the model’s performance on crash likelihood prediction. The proposed model outperformed other benchmark models in terms of sensitivity and false alarm rate. Moreover, trajectory fusion improved the predictive accuracy of the proposed model, which indicated the importance of having data from different types of vehicles for developing real-time crash likelihood prediction models.

Traffic incidents are a primary cause of traffic delays, which can cause severe economic losses. Effective traffic incident management requires integrating intelligent traffic systems, information dissemination, and the accurate prediction of incident duration. This study develops a clustering-based machine learning model to predict the incident duration. Unlike similar studies that train separate machine learning models for a fixed number of clusters, this study proposes an ensemble learning method based on multiple clustered individual models that can provide good and diverse prediction performance. The K-means clustering method is used in this study as a bootstrapping technique in the ensemble learning approach, with the individual models based on the artificial neural network model and random forest regression model. The models are tested using the incident data from Singapore, and the results show that the ensemble model outperforms both the traditional model with fixed clusters and the classical model without clustering. Additionally, this study attempted to determine the significance of different variables on traffic incident durations using the random forest feature importance function. The prediction of incident duration and the analysis of influence factors can contribute to several aspects of traffic management, such as improving traffic dissemination to mitigate traffic congestion caused by incidents.