This paper presents a novel approach to understanding the factors influencing autonomous vehicle (AV) acceleration in mixed traffic environments, crucial for the smart transformation of urban mobility systems. The study introduces a pioneering Regularized Stacked Long Short-Term Memory (RS-LSTM) model for predicting AV acceleration. Employing explainable AI techniques, including SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDPs), the study interprets factors shaping AV behavior. OpenACC dataset is utilized for model training, testing, and factor exploration. Results reveal that leader acceleration is the most influential factor in determining follower AV acceleration. Additionally, space headway and follower speed exhibit critical thresholds (32 meters and 28 m/s, respectively), beyond which the relationship with AV acceleration predictions undergoes a change. These findings contribute to a deeper understanding of AV behavior in mixed traffic scenarios, with implications for optimizing AV performance and integration in realworld traffic conditions.

This study investigates the response times of autonomous vehicles (AVs) equipped with adaptive cruise control (ACC) and traditional human-driven vehicles (TVs) in mixed traffic scenarios. The primary objective is to assess how these response times impact the stability and safety of mixed traffic flow, considering the growing prevalence of ACC technology in vehicles worldwide. Utilising a trajectory dataset from OpenACC totalling 3389.70 s, this research introduces a response time estimation framework that combines cross-correlation and partial autocorrelation techniques. The study calibrates Gazis, Herman, and Rothery's (GHR) car-following model to evaluate mixed traffic flow stability and employs a modified time-to-collision (MTTC) surrogate for safety analysis. The study also delves into the influence of vehicle manufacturer diversity on study outcomes. Key findings reveal that the AVs exhibit significantly longer response times, ranging from 1.10–3.20 s, compared to the 0.30–1.90-second range of traditional vehicles (p value < 0.005). These extended response times in AVs contribute to prolonged traffic flow instability and increased traffic conflicts. Moreover, the type of lead vehicle does not significantly affect the response times of either AVs or TVs (p value > 0.005). The study also highlights that vehicle manufacturer diversity does not substantially affect these response times. Additionally, the examination of fitted GHR parameters underscores AVs' heightened sensitivity to spacing and relative speed, providing insights into AV dynamics in the presence of mixed traffic.

This study aims at evaluating the performance of three commonly used car-following models in representing the longitudinal movement of autonomous vehicles (AVs) following traditional human-driven vehicles (TVs), based on both real-world data and simulation. Our ultimate goal is to assess the extent of discrepancies, if any, between the simulated and real-world AV response. We utilize real-world vehicle trajectory data collected from adaptive cruise control (ACC) equipped vehicles to compare the speed, acceleration profiles and vehicle position over time of modeled AVs and real-world observed AVs. Our observations suggest that the modeled AV longitudinal responses differ significantly from the real-world AVs. Overall, our study highlights the necessity of using the appropriate car-following models to ensure that all conclusions and decisions drawn about the operations and safety of AVs in platoons are well founded.

This study examines if the Gini coefficient, which is a widely used measure of income/wealth inequality in a country/society, could be used as a measure of travel time reliability. To be considered a valid measure of travel time reliability, the Gini coefficient should at least agree with existing and widely used reliability measures. We study if the Gini coefficient is in alignment with existing reliability measures by examining the degree to which it correlates with the established measures. Since travel times generally follow right-skewed distributions, we simulate the parameters of four right-skewed distributions: i.log-normal, ii. burr, iii. gamma, and iv. Weibull distributions following which we compute the existing reliability measures as well as the Gini coefficient for each of the parameters. We then compute the correlation coefficients between the Gini coefficient and other established reliability measures. We find that the Gini coefficient is strongly correlated with travel time reliability measures like Buffer-Index, Coefficient of Variation, Width and Skewness for Log-Normal and Gamma distribution. Finally, we test our hypothesis on a travel time data collected using Bluetooth detectors. We perform K-means clustering to segregate the travel rates from the data into two clusters. After determining the suitable distribution for travel rates in each cluster, we compute the Gini coefficient and other reliability measures based on the parameters of the fitted distributions. This study shows that not only does the Gini coefficient agree with existing reliability measures for some distributions, it also helps in intuitively discerning and communicating travel time reliability.

Surrogate safety measures (SSMs) are an effective alternative to assess traffic safety in the absence of crash count data. Time-based and distance-based safety surrogate measures are extensively used for traffic conflict reporting. However, the threshold selection for time-based SSMs and stopping distance calculation for distance-based SSMs are mostly made based on response time assumptions. There is little research that have used calibrated response time from real world traffic to measure time-based and distance-based SSMs. This research analyzed the shortcomings of applying time and distance-based SSMs in mixed traffic environment, using real-world vehicular trajectory data. In this context, mixed traffic is defined when traditional human-driven vehicles (TVs) share the road with autonomous vehicles (AVs). We find that the surrogate safety measures' threshold selection for mixed traffic is an unexplored area that needs to be addressed. Furthermore, our findings point to the need for the use of calibrated response time when using time-based and distance-based SSMs.

This study tests the hypothesis that an analytically estimated driver reaction time required for asymptotic stability, based on the macroscopic Gazis-Herman-Rothery (GHR) model, serves as an indicator of the impact of traffic oscillations on rear-end crashes. If separate GHR models are fit discontinuously for different traffic regimes, the local drop in required reaction time between these regimes can also be estimated. This study evaluates the relationship between rear-end crash rates and that drop in required reaction time. Traffic data from 28 sensors were used to fit the GHR model. Rear-end crash rates, estimated from four years of crash data, exhibited a positive correlation with the drop in required reaction time at the congested regime's density-breakpoint. A linear relationship provided the best fit. These results motivate follow-on research to incorporate macroscopically derived reaction time in road-safety planning. More generally, the study demonstrates a useful application of a discontinuous macroscopic traffic model.

Autonomous and connected vehicle technologies have the potential to bring profound changes in travel behavior and transportation network performance with moderate to significant market penetration rates (MPRs) within the next few decades. To better understand the long-term impacts of these technologies, this study predicts the network-level effects of privately owned autonomous vehicles (AVs) and connected and autonomous vehicles (CAVs) for the Triangle Region, North Carolina, in the year 2045. Market penetration scenarios of personal AVs and CAVs along with results from microscopic mixed-traffic simulations and travel behavior assumptions are incorporated into a regional travel demand model. Results indicate that a 75% MPR of personal AVs deteriorates the performance of the network, leading to a 5.4% increase in vehicle-hours traveled, and a 17.2% increase in hours of delay. The opposite holds for private CAV adoption, which is found to result in higher peak-period link speed and less congestion. The results of this research help planners and engineers to make informed transportation planning decisions and work toward harnessing the benefits of these technologies while minimizing any negative impacts.

Effective management of highway networks requires a thorough understanding of the conditions under which vehicular crashes occur. Such an understanding can and should inform related operational and resource allocation decisions. This paper presents an easily implementable methodology that can classify all reported crashes in terms of the operational conditions under which each crash occurred. The classification methodology uses link-based speed data. Unlike previous secondary collision identification schemes, it neither requires an a priori identification of the precipitating incident nor definition of the precipitating incident’s impact area. To accomplish this objective, the methodology makes use of a novel scheme for distinguishing between recurrent and non-recurrent congestion. A 500-crash case study was performed using a 274 km section of the I-40 in North Carolina. Twelve percent of the case study crashes were classified as occurring in non-recurrent congestion. Thirty-seven percent of the crashes in non-recurrent congestion classified were identified within unreported primary incidents or crashes influence area. The remainder was classified as primary crashes occurring in either uncongested conditions (84%) or recurrent congestion (4%). The methodology can be implemented in any advanced traffic management system for which crash time and link location are available along with corresponding archived link speed data are available.

This paper presents a monitoring system that was developed to assess travel time reliability for observed operating conditions by utilizing traffic stream and non-transportation related data. A prototype was created for an interstate highway route in the Research Triangle region of North Carolina. It describes how the input datasets were obtained, the required data fusion procedures, how the data were analyzed to create the monitoring system outputs, and relevant insights obtained from the reliability monitoring system prototype.

Traffic congestion costs drivers an average of $1,200 a year in wasted fuel and time, with most travelers becoming less tolerant of unexpected delays. Substantial efforts have been made to account for the impact of non-recurring sources of congestion on travel time reliability. The 6th edition of the Highway Capacity Manual (HCM) provides a structured guidance on a step-by-step analysis to estimate reliability performance measures on freeway facilities. However, practical implementation of these methods poses its own challenges. Performing these analyses requires assimilation of data scattered in different platforms, and this assimilation is complicated further by the fact that data and data platforms differ from state to state. This paper focuses on practical calibration and validation methods of the core and reliability analyses described in the HCM. The main objective is to provide HCM users with guidance on collecting data for freeway reliability analysis as well as validating the reliability performance measures predictions of the HCM methodology. A real-world case study on three routes on Interstate 40 in the Raleigh-Durham area in North Carolina is used to describe the steps required for conducting this analysis. The travel time index (TTI) distribution, reported by the HCM models, was found to match those from probe-based travel time data closely up to the 80th percentile values. However, because of a mismatch between the actual and HCM estimated incident allocation patterns both spatially and temporally, and the fact that traffic demands in the HCM methods are by default insensitive to the occurrence of major incidents, the HCM approach tended to generate larger travel time values in the upper regions of the travel time distribution.

In this study, a two-regime, steady-state, traffic stream model is developed by applying the macroscopic Gazis–Herman–Rothery model to fixed sensor data on freeways. The uncongested and congested regimes are modeled discontinuously with an overlap range defined in terms of density. The overlap is important as various phenomena related to the change in traffic state can be modeled by introducing this overlap. Two empirical tools for removing non-stationary, mixed-state, and erroneous observations are applied at different stages of the model development process. Three constraints justified by the Highway Capacity Manual (HCM) were applied to fit the model so that the fitted parameters have reasonable and physically interpretable values. The proposed model is applied to one year of data (2013) obtained from fixed sensors located at five freeway basic segments near Raleigh, North Carolina. The resulting fundamental diagrams show that the fitted models reasonably represent the steady-state observations. Two forms of the freeway flow model described in the HCM were applied to the same observations to provide a continuous model comparison. Two statistical performance measures, mean squared error of flow rate and Bayesian Information Criterion, verify that the proposed model is preferable to the HCM models both in terms of fit alone and when considering the tradeoff between fit and model complexity. It is expected that the proposed discontinuous steady-state model will be useful to researchers and practitioners to study various site-specific freeway traffic stream characteristics.

Over the past decade, traffic heteroscedasticity has been investigated with the primary purpose of generating prediction intervals around point forecasts constructed usually by short-term traffic condition level forecasting models. However, despite considerable advancements, complete traffic patterns, in particular the seasonal effect, have not been adequately handled. Recently, an offline seasonal adjustment factor plus GARCH model was proposed in Shi et al. 2014 to model the seasonal heteroscedasticity in traffic flow series. However, this offline model cannot meet the real-time processing requirement proposed by real-world transportation management and control applications. Therefore, an online seasonal adjustment factors plus adaptive Kalman filter (OSAF+AKF) approach is proposed in this paper to predict in real time the seasonal heteroscedasticity in traffic flow series. In this approach, OSAF and AKF are combined within a cascading framework, and four types of online seasonal adjustment factors are developed considering the seasonal patterns in traffic flow series. Empirical results using real-world station-by-station traffic flow series showed that the proposed approach can generate workable prediction intervals in real time, indicating the acceptability of the proposed approach. In addition, compared with the offline model, the proposed online approach showed improved adaptability when traffic is highly volatile. These findings are important for developing real-time intelligent transportation system applications.

Identification of recurrent bottlenecks is an effective way to hone an appropriate investment in current facilities to relieve congestion. Furthermore, it would enable the ranking or prioritisation of bottlenecks since bottleneck removal and its associated impact alleviation are hampered by limited sources. It is imperative that transportation jurisdiction understand and identify the basis for ranking bottlenecks by exploring: how often they are active; how long it takes the congestion to disappear; and how many miles of road are affected. Previous bottleneck identification schemes have focused on identifying congestion with little attention to distinguishing the recurrent level at the same 'bottleneck' location. In contrast to traditional schemes, a data-driven approach for identifying recurrent bottlenecks is introduced, using probe vehicle speed reports. The historical spatiotemporal characteristics of bottlenecks are investigated through a comprehensive analysis of 2253 miles of all state-wide interstates in North Carolina. Using the characteristics determined the recurrent bottleneck locations with a historical time span of bottleneck activation are revealed and tested. The findings of the proposed identification schemes generate critical information in order to quantify and diagnose a bottleneck and its associated impact area.

Weigh stations are necessary for safeguarding highway infrastructure by enforcing truck weight limits. However, mandating all trucks to stop at all weigh stations decreases travel time reliability. This decrease in travel time reliability adversely impacts the productivity of the trucking industry and to a lesser degree impacts personal travel reliability as well. This study, conducted at the Lumberton weigh station on Interstate 95 in North Carolina, quantifies the impact of weigh stations on truck travel time reliability. Truck travel times were observed over periods of weigh station operation and weigh station closure. Comparison of these two states sheds light on the variability in travel time caused by weigh station operation. Results show that when the weigh station is operational, truck travel time reliability degrades significantly. VISSIM microsimulation software was used to quantify the expected impact of weigh in motion (WIM) on truck travel time reliability assuming different scenarios of WIM truck bypass. The model results indicate that WIM technology does increase travel time reliability and provides benefits to both trucking companies and enforcement agencies. It was also found that increases in the proportion of WIM bypass result in increases in route travel time reliability. However, this simulation model improvement was not uniform, with the highest marginal improvement occurring in the 30%–40% WIM bypass range.

This paper aims to examine the effect of age and various characteristics of non-driving related activities during highly automated driving on subsequent performance in notified takeovers among younger and older drivers. The paper presents new analyses of data collected in our earlier study (Clark & Feng, 2016). Non-driving-related activities that participants voluntarily chose to engage in during automated driving were categorized according to their cognitive dimensions in information processing. Using hierarchical multiple regressions, we analyzed the effect of driver age, total duration and number of engagement in non-driving-related activities, the duration and cognitive dimensions of the last activity prior to takeover on average speed during takeover and the response time to a takeover notification. We found that older drivers speed was negatively predicted by age while their response time to a notification was not predicted by any factor. In contrast, younger drivers showed a trend of positive relationship between age and average speed and the characteristics of the last task engagement explained a significant portion of the variance of response time to a notification.

This study investigated the main factors affecting the severity of injury to pedestrians in taxi–pedestrian crashes on urban arterial roads. Video data recorded by an in-car black box were used. Because the video data provided direct crash observation, they were more reliable than the crash data, and video images and speed profiles retrieved from the black box were advantageous for safety studies. For analysis of the black box data, this study defined new explanatory variables that affected injury severity; these variables could not have been identified by the conventional method, which was based on crash reports. A multiple-indicator and multiple-cause model was used to investigate the relationship between the explanatory variables and injury severity. A total of 484 taxi–pedestrian crash scenes over 2 years was used for the multivariate analysis in the city of Incheon, South Korea. The crash characteristics most strongly associated with increased crash severity were failure by the pedestrian to watch for approaching vehicles, jaywalking by the pedestrian, the pedestrian being elderly, excessive vehicle speed, failure by the driver to immediately stop, limited driver vision, and nighttime. This study emphasized the potential of individualized black box video recording data for crash severity analysis and investigation of the causal factors of crashes.

Travel time reliability has become a subject of increasingly intense interest for researchers and practitioners. Carriers and shippers have focused on reliability in terms of operational control ev...