Journal of Transportation Engineering, Part A: Systems

This study aimed to identify the segment-specific risk factors for rear-end crashes on an expressway in India. The fatal crash data were analyzed from August 2012 to October 2018 with road geometry details, speed, volume, and roadside data. Negative binomial (NB) models with a random parameter framework accounted for the unobserved heterogeneity across the segments. A NB model with fixed parameters (FPNB) was used as a base model against basic random parameter NB (RPNB) and correlated random parameter NB (CRPNB) models. The results indicated that rear-end crashes comprised 49% of the total fatal crashes. Truck and car occupants constituted the highest share of victims in rear-end crashes. As per the estimate of log-likelihood for the developed models, CRPNB was best. The results of the RPNB model suggested that speed, Annual Average Daily Traffic (AADT), horizontal alignment radius, and vertical curve length were the significant variables. In the case of the CRPNB model, speed, AADT, and vertical curve length were the significant variables. Segments where a village was present along the expressway were comparatively safer. Segments where entry and exit ramps and underpasses were present and segments with hazards had higher chances of being associated with rear-end crashes. The study’s findings provide valuable insights into proposing interventions for reducing rear-end crashes on expressways in heterogeneous traffic conditions. Detailed investigations with additional data are required at segments with entry-exit ramps and horizontal and vertical curves. To sum up, the findings of this study may assist highway agencies and design engineers in designing new or evaluating existing expressways from a safety viewpoint.

Early macroscopic traffic flow models were based on observations of volume, speed, and density. The invention of traffic sensors has supplied a wealth of data for the development of more accurate macroscopic flow models. However, traffic sensors typically collect volume, speed, and occupancy data. Researchers prefer to convert occupancy to density because of the density usage in earlier models; however, for this conversion, the average length of passed vehicles must be determined. This length is frequently estimated by researchers. However, because the explanatory variable (density) is not observed but produced, this estimation weakens the model results. Considering these challenges, this research proposes a novel traffic flow modeling approach based on occupancy. The proposed method was tested in three speed-flow rate relationship regions, one of which is congested and two of which are free flow. Free flow speed, capacity, queue discharge flow, breakpoint flow rate, and optimum speed can all be determined more precisely with this method. Furthermore, the nonlinear relationship between speed and flow rate was clarified. The proposed traffic flow model is extremely useful, especially for dynamic traffic management applications, because it is based on directly gathered data such as volume, speed, and occupancy.

The common practice of buses to different directions dwelling at the same stop on an intersection approach may cause serious traffic bottlenecks, where buses queue to pull in and aggressively change lanes for their target exit. To address this problem without changing the bus route for passenger convenience, the study proposes an extended cell transmission model to explore the optimal bus stop design and lane assignment at an isolated signalized intersection. Specifically, vehicles are categorized as cars or buses dwelling on the intersection approach or the exits, and various interactions between them are carefully captured. The effect on mixed traffic of bus dwelling and aggressive lane changes after pulling out is revealed with lane blockage as well as reduced cell receiving and sending capacities. Considering significant mixed traffic stochasticity, a competitive gradient bandit algorithm is established to compare the feasible solutions of the proposed model, where the weight of better solutions is gradually increased, whereas that of others is correspondingly reduced as proven with mathematical calculation. A case study follows to validate that the proposed algorithm is efficient and accurate in identifying the optimal solution, in comparison with Monte Carlo tests. This research provides insights into improving bus stop design and lane assignment at signalized intersections to promote the efficiency and mitigate the conflicts of mixed traffic.

Traffic incidents represent nonrecurring events that have long been found to deteriorate traffic operations and safety. Studies agree that quick incident clearance would translate into substantial savings for the motoring public. However, many factors influence how quickly an incident can be cleared. This paper investigated the influence of the incident location on incident clearance duration using a hierarchical Bayesian survival model. The study presented a statistical approach to examine disparities in incident clearance duration on different freeway segments (i.e., basic, merge, diverge, weaving, on-ramp, and off-ramp sections). In the analysis, data from 58,167 incidents that occurred on freeways in Jacksonville, Florida, for the years 2014–2017 were analyzed. The Bayesian hypothesis testing revealed credible differences in incident clearance durations among most freeway segment pairs. The model results indicated that basic freeway segments had the longest incident clearance durations, followed by diverge segments and off-ramps. Incidents that were crashes, were severe, resulted in shoulder blockage, occurred on weekends, and occurred on segments with high traffic volumes took significantly longer time to be cleared. The study findings could help practitioners strategically position and allocate appropriate incident response resources along the freeway corridors. Practitioners could consider depots along longer basic freeway segments and diverge segments to enable the quick arrival of incident response teams.

In the United States, motor vehicle crashes are the leading reason of fatalities for young drivers in the age group of 16–19 years. This paper used naturalistic driving study (NDS) data to examine the safety of young drivers. The NDS data consist of comprehensive information related to driving behavior. The statistical modeling method (logistic regression) was used to investigate the relationship between factors of driver attention such as secondary tasks, impairments, hands on wheel, and crash risk to aid in the formulation of teen driving policy. The results indicated that nondriving tasks engagement such as cell phone use, external distraction, food and drink intake, personal hygiene, and reaching and handling objects in the vehicle increase the possibility of involvement in a safety critical event [odds ratio $(\mathrm{OR})=1.67–1.93$]. Risk estimation for the observable impairments indicated that emotional state of the driver ($\mathrm{OR}=6.01$) and impairments (drugs/alcohol, $\mathrm{OR}=5.02$) exerted greater influence on crash likelihood compared to involvement in secondary tasks. The study revealed that not all secondary tasks pose the same risk. Transportation agencies can design behavioral countermeasures targeted at young drivers to help increase driver attention. Countermeasures such as increased enforcement, treatment and monitoring of the offender, and education and outreach programs for teenagers may help alleviate impaired driving.

As one key basis of traffic flow theory, the fundamental diagram (FD) has wide application in traffic management and control. However, the FD of mixed human-driven vehicles (HVs) and cooperative adaptive cruise control (CACC) vehicles remains unclear. The deterministic and stochastic FD of mixed traffic was obtained. First, the longitudinal control model (LCM) and connected LCM (CLCM) were selected as the car-following model of HVs and CACCs. Next, in deterministic simulation, the mathematical expression of mixed-traffic FD was conducted considering the CACC penetration rate and platooning intensity. Then, the stochastic distribution of perception and response time calibrated by the vehicle trajectory data set NGSIM was introduced to derive the stochastic FD. After that, in stochastic simulation, taking the variance and computation effort into account, the optimal number of simulated vehicles to obtain steady-state FD was obtained. The impact of CACC penetration rates and platooning intensity was also explored. It revealed that the optimal number of simulated vehicles is 400. CACC can improve the capacity and critical density of highways. In mixed traffic, due to the large variance of HV driver response time, scattering remains unchanged regardless of CACC penetration when there are enough vehicles. Pure CACC traffic has minimum scattering. More importantly, the proposed framework is applicable to other CACC, adaptive cruise control (ACC), and HV car-following models, including further experimental models.

High-occupancy toll (HOT) lanes are emerging as a solution to resolve capacity waste from high-occupancy vehicle (HOV) lanes. This paper proposes a dynamic toll pricing strategy considering traffic efficiency and user experience to balance system and user optimum. First, a long short-term memory-convolutional neural network (LSTM-CNN) is developed to predict the lane-choice behavior of low-occupancy vehicles (LOVs). Real-world data from a 5.57 km segment of Interstate 405 in Bellevue, Washington, in the US are used for validation. Compared with the other four lane-changing models, the proposed model shows the lowest mean absolute error (MAE) at 1.15 for predicting the lane-changing ratio. Next, a toll pricing optimization model was developed. Total travel time and price-performance ratio obtained by traffic status estimation approaches (Greenberg’s traffic flow model and cumulative arrival-departure diagram) represent the two optimums in the objective function, respectively. Then, a dynamic weight function that varies with demand is constructed to realize the tradeoff. Based on the previously identified lane-choice behavior, an optimal toll can be calculated dynamically according to real-time transaction and traffic data. After adopting the new strategy, the total travel time reaches an average drop of 18.12%, and the performance-price ratio reaches an average increase of 28.12% during peak hours under deterministic conditions. The results show that the proposed dynamic pricing strategy helps to bring a significant reduction in the total travel time and at the same time provides a satisfactory toll rate for users.

This paper proposes a roundabout controller that operates in a connected and automated vehicle (CAV) environment. The proposed controller bears the following features: (1) improves the mobility and capacity of roundabouts, (2) has the potential for real-world implementation, and (3) is easy to upgrade for partially connected and automated traffic. The proposed controller is evaluated against a conventional roundabout controller to quantify its performance. The results show that the proposed controller is able to improve throughput by approximately 14% and reduce delays ranging from 66% to 94%. The computational time of the proposed controller was demonstrated to be constantly less than 0.02 s across all traffic demands. This confirms the potential of the proposed controller for real-world application.

During the last train period, passengers will not successfully reach their destination if the timetables of the various transfer station lines are not coordinated. This paper proposes a general model to optimize the last train timetables of urban rail networks and maximize origin-destination (OD) accessibility for metro networks during the last train period. By using a generalized cost function, all transportation activities involved in an urban rail network are effectively described, and a generalized cost function is established for the last train as a criterion for its accessibility. A mixed-integer programming model that pays more attention to passengers’ effective travel routes and accessibility of the entire OD route is proposed. A genetic algorithm combined with a $K$-shortest path algorithm is developed to calculate a rescheduling solution within a satisfactory running time (e.g., 35 min). Finally, comparative experiments are conducted using the Xi’an urban rail network as a case study. The results show that the objective function of OD accessibility is more favorable to last train passengers than to the number of successful transfer passengers at transfer stations. This increases the number of passengers successfully arriving at their destination by 12.03%. The proposed methodology assist in cost-effective coordination of the last train timetables and make accurate recommendations to improve passenger OD accessibility using information guidance.

Using trip data from the SHRP-2 Naturalistic Driving Study (NDS) database collected at 12 sites in three states across the United States, this paper investigates driver behavior at freeway exit ramp terminals. First, the study qualitatively assesses driver speed behavior as they navigate the speed change lane (SCL) and the ramp. Starting at the beginning of the SCL and continuing after diverging onto the ramp controlling curve, a trend of continuous vehicle deceleration was evident, which continued throughout the SCL and ramp. It was also evident that a portion of drivers have a tendency to diverge onto the SCL on the taper and before the SCL has begun, where this behavior is dominant on the taper-type SCL. In general, statistical analysis revealed that the speed measures of driver behavior follow a normal distribution. The speed and deceleration measures at the study sites were statistically and significantly different, with the differences likely related to the geometric characteristics of each site. The data were then used to develop prediction models for the speed and deceleration measures. To account for the repeated measures induced by the same drivers in the dataset, linear-mixed models were developed for the speed and deceleration behavior measures.

Parking has become a major problem in urban traffic management, stemming from insufficient understanding of vehicle parking demand. The rapidly growing intelligent transportation system enables traffic managers to detect vehicle movement continuously on a large scale and supports individual parking behavior analysis. Basic-overnight parking demand, as the subject of this study, is the important part of parking demand. This study proposed a method to calculate overnight parking behavior based on automatic vehicle identification (AVI) data and took Shanghai as a case study to analyze overnight parking demand characteristics. After delineation of traffic analysis zones (TAZs) and reconstruction of travel paths, a vehicle state estimation model based on spatiotemporal information and the parking location model based on density-based spatial clustering of applications with noise (DBSCAN) clustering was developed to calculate overnight parking behavior. This was the first time that individual overnight parking behavior at the metropolitan level was directly calculated, instead of estimated, to master the city’s total overnight parking demand. The accuracy of overnight parking position calculation reached 85.39% in the validation experiment. This study systematically analyzed both macro urban characteristics and micro individual characteristics of overnight parking demand. Most vehicles have the characteristics of commuters in parking for 14 h overnight and being more elastic on weekends, which supports refined parking management in the temporal dimension. The spatial distribution of overnight parking demand significantly correlates with residential districts but shows severe imbalance with parking supply. The micro analysis indicates that vehicles show an obvious polarization in the tendency of overnight parking in any area. Generally, vehicles have multiple overnight parking positions but only park in one or two positions on most nights. The quantitative results are helpful to further understand urban overnight parking demand and individual overnight parking behavior, supporting accurate formulation of parking policies.

After reports of self-driving car accidents, public acceptance of autonomous vehicles (AVs) has shown a downward trend. To explore factors influencing this decline of acceptance and to bring acceptance back to the normal range, this study proposed a new factor, human-manipulated risk perception (HMRP), in the context of human–machine collaborative driving. On this basis, we explored the relationship between HMRP and acceptance of AVs and the mechanism of influence. Furthermore, we proposed a behavioral intervention approach to mediate and moderate the relationship between the two. First, an improved acceptance model was developed based on HMRP. Subsequently, a survey was executed, and 300 valid questionnaires were returned. Test results for the improved model showed that HMRP had a significant negative effect on acceptance of AVs. A hybrid model of mediation and moderation was designed, and user experience was introduced as a moderating variable. The results showed that the proposed behavioral intervention strategy was effective in attenuating the negative effect of HMRP on acceptance of AVs. This study provides theoretical support and methodological guidance for developing behavioral interventions to improve user acceptance.

Queue length is an important parameter to evaluate the congestion level of urban intersections. Abundant methods based on different data sources, such as loop detector data and mobile sensor data, for estimating queue length have received wide attention. A license plate recognition (LPR) system recording vehicle information can be used as a data source to estimate lane-level queue length. In this study, an improved method is proposed for cycle-based queue length estimation at an isolated signalized intersection using LPR data, considering the residual queued vehicles under oversaturated conditions. First, a modified interpolation method is developed to infer the travel time of unmatched vehicles. Then, the complete arrival and departure information are processed as three characteristic parameters of vehicles, which are the input to the maximum probability function for queue length estimation at each lane. Finally, the proposed model is validated using actual cycle maximum queue length collected from a video camera and LPR data in Changsha city, China. The numerical results showed that the proposed model can achieve accurate estimation for lane-level queue length.

In this paper, a multiagent-based control method is proposed to design a transit signal priority (TSP) scheme at urban traffic networks. One agent controls an intersection. Coordination among different intersection agents is deployed to guarantee the benefits from TSP at upstream intersections. At one intersection there are usually many bus routes, and thus multiple TSP requests and coordination requests can occur at one phase simultaneously. Therefore, the proposed method also aims at resolving conflicting TSP or coordination requests. A multilevel fuzzy controller consisting of a transit signal priority controller, Green Time Adjustment Controller 1, fuzzy negotiation controller, and Green Time Adjustment Controller 2 is then introduced to realize these objectives. Following that, fuzzy inference decisions and algorithms are provided for describing the control process of the proposed multiagent method. An urban traffic network with 25 intersections is selected to conduct a case study to evaluate the proposed method by comparison and sensitivity analysis. The proposed method performed better than the other three methods under different scenarios in accordance with the simulation results.

Forced merge behavior on freeway merging areas is a major cause of traffic conflicts. Intelligent transportation systems can provide the possibility of information interaction and collaborative control, and optimized trajectory planning methods for merging vehicles in a vehicle-to-infrastructure (V2I) environment can improve the safety and efficiency of freeway merging areas. However, it is difficult for vehicles to determine suitable gaps for merging from the ramp to the mainline. Although some researchers have developed models to determine the appropriate insertable gap, they usually assume that the vehicles are driving at the same speed and acceleration, and there is no corresponding selection strategy for different traffic density scenarios. This study develops a trajectory planning method for merging vehicles according to characteristics of different mainline densities in a V2I environment, which can be applied in for the development of dynamic merging assistance systems to distribute control signals to drivers. The trajectory planning problem is transformed into a nonlinear optimal control problem, and the goal is to optimize the insertion gap, energy consumption, and passenger comfort. The decision variables include the time-varying longitudinal acceleration of a group of vehicles and the optimal insertable gap; the optimization problem is then solved by a heuristic algorithm. To verify the applicability and safety of the proposed method, numerical simulation experiments are carried out in different density traffic flow scenarios, and the safety of the experimental results was evaluated from the individual vehicle level and traffic flow level. Results show that the proposed method is suitable for different mainline densities and can effectively improve merge safety. Simulation examples are included to compare the performance of the proposed model to a baseline model with results showing significant improvements in both travel time and safety.

Efficient operations at intersections are associated with smooth, safe, and sustainable travel at the network level. It is often challenging to prevent congestion at these locations, especially during rush hours, owing to high traffic demand and restraints of road geometry. The primary cause of intersection traffic congestion is large variations and fluctuations in traffic demand during the day. Dynamic lane assignment (DLA) is an intelligent transportation system (ITS) technology that can improve traffic operations at signalized intersections, making better use of available space by optimizing lane allocation based on real-time fluctuating traffic demands. This study used a brute-force optimization model to produce a vast synthetic dataset of turning movements at a four-legged signalized intersection. The model then found the optimal lane assignment and cycle length for each case. The produced data were used to develop a deep learning model to predict the optimal lane assignment and cycle length simultaneously at any four-legged signalized intersection without applying optimization software per se. The deep learning model results developed were compared with the results of the $k$-nearest neighbors (KNN), random forest, and decision tree algorithms. It was found that the deep learning model was superior, with accuracy and F1 score of 95.26% and 93.56%, respectively. The proposed solution is expected to require insignificant additional resources and services for successful implementation, particularly in developing countries.

Engineers use an array of yardsticks, ranging from visual inspection to physical measurements, occasionally combined with users’ costs assessed from average long-term traffic data, to characterize a transportation facility’s condition and prioritize maintenance needs. This approach is inconsistent with basic economics principles that recognize that transportation facility users optimize their well-being by making rational decisions regarding travel choices. This paper outlines a joint engineering-economic modeling framework that converts engineering indicators of facility condition to a measure of consumer welfare. The feasibility of this framework is demonstrated by applying it to a case study that assesses the welfare of users of a network composed of two substitutable bridges with deteriorating riding surfaces. The analysis accounts for increased users’ costs as well as the reduction in traffic volume on a deteriorating facility due to partial diversion of its traffic to the alternate facility. Such joint engineering-economic analyses would give better appreciation of the economic impact of facility deterioration and maintenance deferrals, eventually leading to more informed maintenance decisions compared with those that rely solely on traditional engineering criteria.

As urban road networks experience increasing demand, it is not always feasible to expand them physically. To efficiently use the available space, we need to find solutions that do not require physical expansion. Modifying design factors such as driveway access points, medians, and right turn on red control types are among the solutions that are comparatively easy and quick to implement, do not require more land, and can have a significant impact on traffic flow. This study investigates the impact of several design factors on the number of lane changes on urban arterials near freeway interchanges. While lane changing has a significant impact on traffic flow and safety, it has not received adequate research attention since lane changing maneuvers are difficult to measure. Moreover, existing lane changing research is mainly focused on freeways and highways. This study is focused on urban arterials just downstream and upstream of freeway exits and entry points near diamond interchanges, but the methodology and findings can be applied to arterials near other intersection types. The results show that driveway options, median options, and the location of the right turning bays significantly impact lane changes on urban roads near freeway interchanges. The purpose of this study is to prove that low-cost design modifications can impact lane changes, which ultimately impact the operational performance of roadways.

With the large amounts of available traffic data, it becomes necessary to develop tools that can perform several tasks related to the collected data. These tasks include storing the data in a standard format, filtering the data/flagging suspicious records, processing the data and calculating useful quantitative traffic indices, and finally, visualizing the outcomes. In this paper, a data-driven, yet novel, data-filtering approach was proposed to flag outliers in daily cycling counts at automatic traffic counters (ATCs). The approach was motivated by the spatiotemporal relationship of cycling counts collected at permanent count stations. The proposed approach is flexible because it assumes no prior knowledge about which locations may experience sensor malfunction (i.e., outliers). The approach was tested using a large data set of more than 111,000 daily bicycle volumes collected in 4 years (2016–2019) at more than 60 different permanent count stations in the City of Vancouver, Canada. The approach was validated using complete annual sets of data at four count stations in 2016. Scenarios of undercounting and overcounting were simulated using different percentages of the actual counts. The results showed that the proposed approach has a strong ability in detecting and removing most outliers, especially for cases of substantial undercounting.

The use of predictive models to investigate the performance of automated network management (ANM) under multiple control strategies and given traffic demands will help assess control options in a connected and automated vehicular environment without testing them in a simulation environment. The current prediction models used to evaluate the traditional networks are not applicable in an automated environment. The goal of this study is to develop delay and surrogate conflict predictive models to examine the performance of a reservation-based intersection control strategy in a flexible automated network. Efforts have been made to identify the correlation between the delay and the total number of conflicting requests with types and characteristics of the conflict points and conflicting flows. To generate an extensive origin-destination level dataset, three combined flexible lane assignment and reservation-based intersection control (CFALRIC) models were used in which the through traffic is shared in six various proportions with one or both turning lanes and tested under two different traffic demands, leading to a total of 36 scenarios. The selection of appropriate input variables was a crucial step in this study. Linear regression and multigene genetic programming (MGGP) approaches are utilized to derive new prediction models for the delay and the number of conflicting requests. Results reveal that the regression model is a reasonably appropriate fit for both models; however, the accuracy of the MGGP approach is higher for the delay prediction model. These models have great potential to be applied for planning purposes such as traffic assignment in an automated environment.

Weigh-in-motion (WIM) stations are effective but expensive tools for the monitoring and enforcement of truck weight regulations. Faced with the increasing challenges of maintaining highways with ever shrinking financial resources, traffic authorities may need to couple WIM stations with other cost-effective measures for regulating truck weights. This study explores the idea of integrating WIM stations and truck-prohibited roads to reduce the impact of overloaded trucks on highway pavements and bridges. The problem of optimally locating WIM stations and truck-prohibited roads is formulated as a bilevel programming model. The upper-level model mimics the decision-making behavior of the traffic authority, whose aim is to determine the optimal locations for the deployment of WIM stations and the selection of truck-prohibited roads so as to minimize the total cost while considering route choices of overweight trucks. The lower-level models are used to determine the least cost paths for the overweight trucks of different origin–destination (OD) pairs in response to the authority’s decisions in the upper-level model. A heuristic that iteratively solves the upper-level and the lower-level models is developed to efficiently solve the bilevel model, which is characterized as NP-hard. The proposed model and heuristic were evaluated using test instances generated based on the Nevada road network. The results demonstrate the advantage of using hybrid or multiple enforcement measures to prevent damage to the road network by overweight trucks in comparison with the traditional WIM-only approach.

Recently, the traffic control infrastructure and public transport fare records in Tehran have produced massive data. Such data can be analyzed to monitor the city’s transportation condition and also to improve its transportation network. Understanding the fundamental features of traffic, including flow, density, and velocity, on important roadways is vital for a reliable traffic control mechanism. Obtaining this data from traffic cameras scattered across the city was the goal of this research. The average speed of cars, as recorded by automated license plate readers (ALPRs), was also obtained, and a relationship for estimating the flow density is proposed that simulates cars’ behaviors. The curve plotted by the proposed relationship has about an 25% average speed deviancy from that determined by the probe vehicles for all cameras. This error seems acceptable, considering the limitations of ALPRs and/or diversity of roads.

Crashes involving pedestrians are most likely to occur when the pedestrian is crossing the road. In the particular context of road traffic systems in China, with frequent on-crosswalk cycling and pedestrian green signals overlapping with vehicle right- and left-turn green signals, walking on a signalized intersection crosswalk seems much more challenging. As such, pedestrians need to be more cautious and perform properly. The present study designed tasks for safe crossing at this form of intersection and, using video cameras, explored the performance of pedestrians. Taking smartphones as source of distraction, comparisons were made between distracted and nondistracted pedestrians. Of the phone use rate of 20.3% while crossing, 13.4% were visual-manual interactions, 3.5% were hand-free phones, and 3.4% were hand-held phones. For both genders, pedestrians who crossed while visually-manually interacting with a mobile or talking/listening on a mobile were less likely to glance before initiating crossing, less likely to look at traffic while crossing, less likely to avoid on-crosswalk opposite pedestrians and bicycles, and less likely to make eye contact with right/left-turning drivers. Surprisingly, female pedestrians who were texting took longer to react to the pedestrian green signal than males who were texting. Findings of this study suggest that distraction while crossing roads exposes pedestrians to a high risk of accident. Hence, to improve pedestrian safety, the authors advise relevant actions such as unsafe-crossing warning technologies; pedestrian, bicyclist, and driver education; and law enforcement. Detailed advice is found in the summary section at the end of the paper.