Journal of Transportation Engineering, Part A: Systems

Accurately identifying traffic crash risk factors is an important way to improve freeway safety. The purpose of this research is to reveal the internal coupling mechanisms of and differences between freeway traffic crashes in various area types, as well as to overcome the defects of compatibility and accuracy in the application of conventional data mining algorithms toward road traffic safety. First, the area types were divided into urban, suburban, and mountainous freeways in this research, based on the UW-DRIVENet (Digital Roadway Interactive Visualization and Evaluation Network, University of Washington) transportation big data platform, where data of more than 30,000 traffic crashes in Washington state in 2016 were extracted. The data set was designed via six dimensions: people, vehicle, road, environment, crash, and time. Furthermore, the weighted orientated multiple dimension interactive Apriori algorithm (WOMDI-Apriori) was proposed. In this improved algorithm, a subjective and objective joint weighting model based on interval analytic hierarchy process (IAHP) and gray relational degree was applied to quantify the weight of data fields. Finally, regarding three different area types of freeways, the improved algorithm was adopted to mine the association rules from the perspective of multidimensional interaction: full mapping crash cause and crash dimension autocorrelation perspectives. The results revealed the differences in traffic crash causes and risk factors of cross-freeways. The results show that the accuracy of the improved WOMDI-Apriori algorithm is 82.7%, 88.5%, and 80.5% higher than that of the conventional Apriori association rule algorithm when applied to urban, suburban, and mountainous area freeways, respectively, which indicates that WOMDI-Apriori algorithm can better reveal the causes of freeway traffic crashes and identify crash precursors more accurately. In conclusion, the WOMDI-Apriori algorithm proposed in this research can be used as an effective approach for risk identification of freeway traffic crashes and can also provide theoretical guidance for future freeway traffic safety improvement.

Public opposition to alternative intersection/interchange designs can be a large barrier to their implementation, even when they have safety and efficiency benefits. This study evaluated the factors influencing perceptions for four alternative designs: the roundabout, the restricted crossing U-turn, the displaced left turn, and the diverging diamond interchange. A survey was designed and distributed to 1,000 adults residing in the State of Indiana. The survey solicited participant acceptance of and driver confidence with alternative designs. Bivariate ordered probit models were estimated to evaluate the impact of socio-demographic and trip-making characteristics on public perception. The findings suggest that respondents who are younger, male, highly educated, who travel more often or for farther distances, and who rate their own driving ability highly, are more likely to accept and be confident using alternative designs. The results of this study can inform future outreach efforts by improving transportation agency understanding of public opinions and primary concerns regarding these designs and provide potential paths for improving public perception.

This paper presents the results of a study that explores the desired monthly cost savings that would motivate older adults to change their regular mobility choices to ridesourcing. Using data from a stated preference survey, various socioeconomic and demographic attributes and personal attitudinal factors were examined to investigate their impacts on older adults’ ridesourcing adoption decisions. Random parameter ordered logit models were developed for older adults (age 65 and older) and young adults (age 64 and younger) separately to investigate whether and how the contributing factors differ. The potential heterogeneity in the impacts of these variables was also examined. Model results showed significant differences between older and younger adults’ motivations for ridesourcing adoption. For older adults, those who were less educated or sensitive to travel costs required lower economic incentives, and were therefore easier to persuade to switch to ridesourcing. In contrast, older adults who had a positive view about private vehicle utilities, had trust issues with traveling with strangers, used private vehicles for regular trips, and shopped online regularly showed higher expectations of monthly cost savings to switching to ridesourcing. While these attitudes played a significant role in older adults’ decision-making, socioeconomic and demographic attributes, such as age and income, were more influential in young adults’ ridesourcing choices. The results provide a better understanding of the older adults’ motivations to switch to ridesourcing for monthly cost savings. The findings could be helpful for planners and policymakers to better plan for older adults to meet their mobility needs.

Abstract

Physical infrastructure systems contribute to human civilization significantly and play a key role in the smooth operation of society. However, such systems may experience loss of functionality because of external shocks, i.e., human-made or natural hazards. To enhance the resiliency of the infrastructure systems, network positions or credentials of infrastructure components (e.g., roads, bridges) based on their topology or connectivity need to be assessed. The empirical literature does not provide specific guidance how such topological credentials may contribute to system resilience, i.e., reducing overall adverse impacts as well as recovering from loss of performance. This study emphasized a coordinated and extensive network of experiments at different geographic scales by applying complex network principles to explore the resiliency of Florida road networks. Geographic modeling was used along with Florida road (with bridges) network data to perform network experiments and prioritize certain bridges based on their network credentials. In particular, the study established a systematic approach to rank the topological credentials of bridges based on the connectivity and attributes (i.e., weights) of road networks. Such credentials change significantly when different weights (i.e., vehicular traffic) are introduced to the network topology at different geographic scales. The practical implications for transportation network resilience as well as the scaling effects are examined by quantifying resilience in terms of average network travel time and developing recovery schemes of bridges for a sample road network. The study developed a credible methodology that would benefit states, municipalities, and other transportation authorities to prioritize risk-based recovery strategies.

Practical Applications

This research presents a methodology that can rank the relative importance of bridges in a road network based on their position in the network. Such topological credentials of bridges may change at different geographic scales (i.e., city or county or state) when different network metrics (i.e., centrality) are considered and/or different attributes are introduced as weights (i.e., traffic volume) to network connectivity or topology. The relative importance of different bridges would allow bridge engineers, maintenance personnel, and other practitioners to decide on which bridge should be inspected, maintained, or constructed first based on the position of the bridges in a network setting. Different transportation agencies engage in solving unprecedented problems observed on local roads or bridges. This study provides novel insights on how to go beyond the local context and incorporate a broader network perspective. Bridges are typically inspected every 2 years for regular maintenance purposes. Because of time and budget constraints, inspection of all the bridges may not be possible often in a timely manner. As a result, bridges that would carry more importance in terms of traffic, construction, or other impacts should be given priority. By having a rank of bridges, activities on bridges can be done more systematically.

Using random parameters in combination with extreme value theory (EVT) models has been shown to capture unobserved heterogeneity and improve crash estimation based on traffic conflicts. However, in existing random-parameter EVT models, the predefined distribution means and variances for random parameters are usually constant, which may not capture unobserved heterogeneity well. Therefore, the present study develops a random-parameter Bayesian hierarchical extreme value model with heterogeneity in means and variances (RPBHEV-HMV) to better capture unobserved heterogeneity. The developed model offers two main advantages: (1) it allows random parameters to be normally distributed with varying means and variances; and (2) it incorporates several factors contributing to a heterogeneous distribution of means and variances of random parameters. Application of the developed model to conflict-based rear-end crash prediction was conducted at four signalized intersections in the city of Surrey, British Columbia, Canada. The modified time to collision was employed to fit the generalized extreme value distribution. Three conflict indicators and three traffic parameters were considered as covariates to capture nonstationarity in conflict extremes as well as heterogeneity in means and variances. The results indicated that the RPBHEV-HMV model outperforms existing RPBHEV models in terms of goodness of fit, explanatory power, and crash estimation accuracy and precision.

In this study, 20 drivers, who maintain a safe distance at all speeds, are monitored to help determine the principal reason for trip delay. The aim was to develop a driver-assisted technology device or devices to enhance driver comfort and safety. The use of vehicle-to-vehicle (V2V), vehicle-to-everything (V2X), or a combination of devices that can quickly collect and analyze data is contemplated. Collecting real-time driver response data is a challenge and extremely expensive. For these reasons, a stochastic car-following model, called a car-following network model (CFNM), was developed. It features 20 simulated passenger-car drivers traveling around a test track with 2 cruising zones and a bottleneck. Each driver strives to minimize the individual trip time. The model accounts for (1) driver decision-making lapses made over time, (2) a driver’s inability to precisely control speed, and (3) a driver’s desire to comfortably accelerate and decelerate. Vehicle acceleration and deceleration rates are constrained for (4) driver comfort, and (5) driver health, i.e., vehicle top and lower speeds are adjusted to limit $g\text{-}\mathrm{forces}$. A linear acceleration model and stochastic differential equation are introduced into the CFNM. The stochastic contribution associated with a driver’s inability to precisely control speed is called traffic noise. A model calibration scheme was developed that ensures that the CFNM forecasts of time, speed, and location are reliable. The relationships between traffic noise and delay and road design and delay are discussed. Analyses show that road design and traffic platooning, which are closely associated with traffic noise and driver concern for safety, are critical factors. No single factor can explain delay. The forecast results suggest that three separate assisted-driver technologies will be needed to deal with start-up, cruise, and arrival delay.

Electric buses (e-buses) have been increasingly introduced to urban transit systems in recent years. E-bus scheduling has become a critical problem to ensure cost-efficient transit service. To support continuous e-bus operation, charging facilities are established at the depots to provide charging service. In real-world scenarios, charging facility capacity measured by the number of chargers is always limited. To reduce charging congestions, it is critical to consider the charging facility capacity in the e-bus scheduling problem. In this study, we addressed a multidepot e-bus scheduling problem considering limited charging facility capacity and vehicle-depot constraint. A mixed-integer programming model based on discrete-event optimization method and a large neighborhood search (LNS) heuristic are proposed to solve the problem. Comprehensive numerical experiments based on the real-world operation cases in Shenzhen were conducted. The results verify the model and show that the LNS heuristic can generate near-optimal solutions for large-scale problem instances. Lack of charging facilities at the depots can lead to more e-buses charging at the peak time regarding the time-of-use tariff, resulting in increased total operational cost.

The unprecedented Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV-2), or COVID-19, pandemic adversely affected all walks of life, causing loss of lives and livelihood. The disruptions caused to the economy, social well-being, and transportation systems are almost unfathomable. The scenario in India was grave during the first wave, where high-density urban conglomerations affected the most. Transmissions of the contagion due to human-to-human interactions forced the government to employ strict lockdown policies as an immediate measure to curb the spread. However, gradual relaxations on lockdowns during the initial stages in India demonstrated similar trends between the rise in mobility and COVID-19 positive cases. This study leverages publicly available activity-based mobility datasets to model and predict the number of virus-positive cases during the first pandemic wave in Indian states. Dynamic regression models, which consider the ripple effects of the response and explanatory variables as feedbacks to the response variable, are utilized to analyze the panel data. In addition to the mobility data, the cumulative number of COVID-19 cases is also related to the regional demographics and other information concerning the infection spread and testing data. The proposed model produces good short-term forecasts for Indian states. Findings from the study concerning mobility point to the positive effects of curtailing travel for the effective control of pandemic diffusion through human interactions. Comprehending the effects of mobility and testing rates on the reported number of cases is essential to devising strategies best suited for a region during such an instance. The methodology and contextual knowledge from the study can aid planners, decision-makers, and researchers to bolster support systems in the future.

Revealing urban community structures of a city is of great importance for investigating urban development and sprawl behind the movement dynamics. However, most studies focus on delineating urban community structure and its variations with a single transit mode without covering hierarchical travel distances. This paper proposes an overarching framework to reveal urban community structures by fusing multisource spatiotemporal transportation data. Network science methods and community detection are applied to construct spatially embedded networks and uncover the urban structure from different perspectives, using 1-week transportation data derived from the metro, taxi, and dockless bike-sharing systems (BSSs) of Shanghai, China, in year 2016. Our finding shows that Shanghai can be clustered into six primary communities and exhibits polycentric patterns with strong monocentric characteristics. Shanghai’s urban structure moves toward an embedded hierarchical pattern: the dispersed monocentric structure and the centralized polycentric structure. It reflects poor functional interdependence and horizontal connectivity between communities. Beneath the complex and coupled travel-flow system, the metro dominants the basic framework of the urban community structure and contributes to form the prototype of the core community, while the taxi and BSS tend to play complementary roles like expanding, enhancing, and refining the structure. This research not only provides a promising bridge from the complex urban transportation networks to urban community structures, but also implies potential urban planning policies from an internal and comprehensive perspective.

Work zones are essential to maintaining and upgrading highways. The constrained driving environment in work zones tends to disturb the normal traffic flow, leading to reductions in speed and road capacity. These conditions have proven to increase crash risk. However, the impact of work zones on secondary crashes is yet to be investigated. This study extends the previous research on secondary crash likelihood models by evaluating the impact of work zones on the occurrence of secondary crashes. This study used data collected between January 2014 and June 2019 on a 77.2-km Homestead Extension of Florida Turnpike (HEFT) corridor and a 45.1-km section on Florida’s Turnpike System Mainline—also known as the mainline south section (MSS) in Miami, Florida. Lane widening activities occurred within HEFT during the study period. The results indicated that HEFT experienced approximately twice as many secondary crashes than MSS, that is, seven secondary $\mathrm{crashes}/\mathrm{km}/\mathrm{year}$ in HEFT and four secondary $\mathrm{crashes}/\mathrm{km}/\mathrm{year}$ in MSS. The higher proportion of secondary crashes on HEFT could be attributed to the presence of construction activities. The model results indicate that the presence of work zones significantly influenced the likelihood of secondary crashes. The study results may assist transportation agencies in identifying strategies to improve the safety of both workers and motorists in work zones.

Lane departure crashes account for more than 50% of traffic fatalities in the US. Among the different roadway treatments, rumble strips was found to be cost-effective. Centerline rumbles strips (CLRS), provided on the centerline for two-lane two-way highways, alert distracted and drowsy drivers from drifting into the opposite driving lane. It is used as a countermeasure for head-on or cross-over crashes. Despite public concerns of the generated noise and perceived decrease in visibility of pavement markings, there is no doubt about the safety benefits of CLRS. During severe snowstorms, compacted snow located on the roadway could negatively influence the performance of CLRS. Higher and more frequent winter maintenance level operation might accelerate the functional restoration of CLRS. The gap in literature was approached by quantifying the effect of CLRS in reducing crashes accounting for different levels of winter maintenance operations. Wyoming highways with distinct winter maintenance levels were assessed in this study, in which safety performance functions (SPFs) were developed. Unobserved heterogeneity in between-route variance was accounted for by utilizing random-intercept negative binomial (NB) model in a Bayesian framework. Crash modification factors (CMFs) were developed using the before–after empirical Bayes (EB) method. Generally, the safety benefits of installing CLRS on the selected routes was between 25% and 68% expected crash reductions at 95% significance level. Higher level of winter maintenance was also found to be associated with higher effectiveness of CLRS. Therefore, the agencies should re-evaluate their winter maintenance plans in order to account for locations with high-risk for head-on collisions.

Battery-assisted trolleybuses (BATs) are among the options to be considered for electrifying bus fleets in cities, with established trolleybus catenary networks. BATs charge on the move using existing trolleybus catenaries and can extend their routes using onboard batteries. As such, in cases where thermal bus lines overlap to some extent with a trolleybus catenary network, these may be replaced by BATs; such a case has the advantage of avoiding either opportunity charging facilities or the use of (heavy) electric buses operating under a depot-charging scheme. Aiming at exploiting these operational advantages, this paper proposes a transit route network electrification problem dealing with the optimal selection of existing bus lines to be operated by BATs. The proposed model was applied to a real-world case study in Athens, Greece, with results showing that design objective prioritization largely affects the degree of electrification attained.

The ferry transit system provides a critical transportation link in coastal areas for both residents and tourists. Like signals in a road network, queuing and waiting are unavoidable at ferry terminals. However, a reliable technology does not exist to measure and communicate waiting times. This research tested the feasibility of applying license plate recognition (LPR) technology to track vehicles and estimate waiting times at ferry terminals. The LPR camera sampling rate, capture rate, read rate, and match rate were adopted as measurements of effectiveness. Based on field data collected over a week at one of the busiest ferry terminals in North Carolina, this research revealed that the tested LPR camera had a sampling rate of 84.2%; the average capture rate and read rate were 84.3% and 87%, respectively. The match rate was found to be 79.4%, which is significantly higher than other commonly used data collection technologies such as Bluetooth devices. For the waiting time distribution, this research found that travelers tended to experience long waiting times during midweek days, particularly during the midday period. Additionally, the demand was found to be the primary factor for wait times during the midday peak period, and travelers’ arrival time in terms of proximity to the scheduled ferry departure time was recognized as the key factor for waiting time during early morning and later evening nonpeak periods.

The transfer between bus and metro is an integral part of the public transportation system, but accurately identifying transfer patterns and real-time transfer demands at each metro station for different transfer modes remains a vital issue. Existing identification methods are based mainly on a one-size-fits-all fixed transfer time threshold, or dynamic transfer identification methods only for the metro-to-bus mode, which is not sufficient for mining the hourly transfer characteristics of different transfer modes at each station. This study combined bus smart card (SC) data and the corresponding bus Global Positioning System (GPS) track data into one data set to infer the bus boarding stops. A new algorithm to identify transfer characteristics based on two rules about dynamic transfer time threshold and distance threshold is proposed for the metro-to-bus mode. A new method to recognize transfer patterns based on dynamic transfer time thresholds is presented for the bus-to-metro mode. The two proposed transfer recognition methods were validated using survey data from Shenzhen, China. The results show that the distribution of transfer time by the proposed methods is consistent with that of the survey data. The $p$-values of Mann–Whitney U tests of transfer data identified by the method and survey data all were greater than 0.05. This indicates that the methods proposed in this study can identify effectively the transfer patterns at each metro station per hour of the day, with good generalizability. Moreover, the proposed methods can explain transfer patterns and hourly transfer demand at each station for different transfer modes better than other methods.

Gap acceptance, the fundamental part of pedestrian crossing behavior, represents a pedestrian’s assessment of how safe an existing gap in traffic flow is to cross the path. This assessment is an essential factor affecting pedestrian safety. However, the complex decision problem of “how pedestrians assess the available gap as safe?” is still ambiguous. The proposed paper addresses this with an integrated analytical hierarchical process-fuzzy logic (AHP-FL) approach in line with a realistic target of vision zero. In this approach, FL stood out in the analysis of the uncertainty of gap acceptance, while AHP supported FL to mitigate the expert judgment effect. Thus, the practical application of an innovative methodology based on the combined use of machine learning and a statistical approach was realized. The data collected at three semicontrolled pedestrian crossings were used. AHP questionnaires were conducted on pedestrians and were divided according to age groups and pedestrian driving habits. The effects of age and driving habits on gap acceptance behavior were examined in-depth with the FL models revised with six different weight sets obtained from AHP. The best model performance was achieved with $0.79R^2$ and 25.19% MAPE. The results revealed that the most critical explanatory factors are vehicle speed and item carriage. The highest prediction error was observed in the elderly pedestrian group. Results revealed that the AHP-FL approach would be useful to better understand the complex decision structure of pedestrian crossing behavior.

Small-range battery electric vehicles (BEVs), with driving ranges less than 200 km, are susceptible to charging defects. This study applied a scenario-based approach by modeling a cost function to recommend an optimal charging strategy for small-range BEV users. The proposed cost function considers the operating and range limitation costs of BEVs. The charging threshold, which reflects the charging preference of a user, is adjusted using the cost function and analyzed by comparing different scenarios using real data collected from Shanghai, China. The simulation results indicate that every 2.5 times increase in the charging threshold results in a decrease of approximately 48.9% in the charging times and a 10% reduction in the electrified proportion of a journey. For electrified proportions below 85%, small-range BEVs are no longer an economical option over internal combustion engine vehicles. Moreover, the optimal charging threshold is 35%–41% for BEVs with a 170 km range, and the corresponding daily ownership costs are more reasonable than those of medium and long-range BEVs. The scenario analysis of the availability of the charging piles and battery price further demonstrates the economic advantages of small-range BEVs. These results provide instructible charging strategies for small-range BEV users to alleviate the shortcomings of the range limitations and frequent charging.

Safety performance functions (SPFs) are the key regression tools in the road safety management process (RSMP) and are used to predict crash frequency given a set of roadway and traffic factors. Although regression-based SPFs have been proven to be reliable tools for road safety predictive analytics, some limitations and constrains have been highlighted in the literature, such as the need to assume a probability distribution, the need to select a predefined functional form, possible correlation between independent variables, and possible transferability issues. An alternative to traditional regression models as predictive tools is the use of machine learning (ML) algorithms. This research compared the prediction performance of three well-known ML algorithms, i.e., support vector machine (SVM), decision tree (DT), and random forest (RF), with that of traditional SPFs, and applied and validated ML algorithms in network screening, which is the first step in the RSMP. To achieve these objectives, traditional SPFs using negative-binomial (NB) generalized linear regression were estimated and compared with ML algorithms using three different goodness-of -fit criteria. A data set of urban signalized and unsignalized intersections from two major municipalities in Saskatchewan (Canada) was considered as a case study. Ranking consistency tests of collision-prone locations identified using ML-based and SPF-based performance measures were conducted. The results showed that the consistency of ML-based measures in identifying hotspots was comparable to that of SPF-based measures, particularly the excess (predicted and expected) average crash frequency. Overall, the results of this research support the use of SVM, DT, and RF as predictive tools in network screening.

Abstract

Three safety concerns have been identified in the conventional design consistency assessment procedure for horizontal curves: approximation of the vehicle by the point-mass model, use of unrealistic speeds, and the assumption that the path radius is equal to the design curve radius. This study presents a novel approach to the consistency assessment of horizontal curves in terms of vehicle stability (represented as the safety margin). The novel approach involves calculating the safety margins using the steady-state bicycle model to account for the nonuniform distribution of friction among individual axles due to grade and speed changes and using more-realistic driver behavior parameters. In this study, the realistic driver behavior parameters refer to the modeled operating speeds and calculated vehicle path radii, which were defined based on naturalistic driving data collected through a real-world experiment on a section of a two-lane state road using a GPS device. The analysis showed that the traditional safety margins for sharp curves are unrealistically large compared with the margins defined by the proposed approach; this indicates that the conventional approach uses a larger friction margin than the optimum friction margin. Consequently, the conventional approach results in the application of curves with radii smaller than those required for safe driving. Moreover, the results show that braking on a sharp curve on a steep downgrade is the most critical situation for the loss of lateral vehicle stability. Overall, this study highlights the significance of incorporating additional factors in the design procedure for horizontal curves that affect the safety of the curve, which are not considered in the point-mass model. The most significant factors considered in this study were the grade, driving maneuver in terms of speed change, operating speed, and vehicle path radius.

Abstract

The primary objective of this research was to establish a reliable road design methodology for reducing single-vehicle accidents on horizontal curves. The proposed approach enables the identification of horizontal curves with insufficient side friction, by calculating the difference between the available side friction factor and side friction demand. In a conventional road design practice, the vehicle is approximated as a point mass, and the effects of grade, speed change, vehicle dimensions, and uneven force distribution on individual axles are ignored. Instead, in the novel approach, the steady-state bicycle model was used, considering the aforementioned factors combined with more-realistic speeds and path radii. The improved approach provides a more comprehensive understanding of the design of safe curves, with an emphasis on previously neglected important influential factors (longitudinal grade and driver behavior in terms of operating speed and path radius). The improved design consistency approach could be used for quantifying the risk of skidding, both in the design stage of new roads and for improvement of existing roads.

The focus of this paper is on developing localized growth factors for estimating annual average daily traffic (AADT) of a local, functionally classified road (referred to herein as a local road). Statewide, regional-level, and county-level median and mean growth factors were computed and compared with localized growth factor estimates from spatial interpolation methods, such as Ordinary Kriging, inverse distance weighted (IDW), and natural neighbor (NN) interpolation. The use of Ordinary Kriging-based localized growth factor is recommended for a local road AADT estimation. If count-based local road AADT (c-AADT) figures for previous years are available, they, along with localized growth factors for those years, should be used to estimate the local road AADT (e-AADT) for the reporting year. The estimated AADT (e-AADT) for the base year and localized growth factors from the base year to the reporting year must be used for estimating e-AADT of a non-covered local road link for the reporting year. The c-AADT or e-AADT of the local road link can be used to compute vehicle miles traveled (VMT) for reporting purposes. The proposed methodology reduces the costs and other resources required for traffic data collection and for e-AADT accounting for spatial variations.