Analyzing relationships between latent topics in autonomous vehicle crash narratives and crash severity using natural language processing techniques and explainable XGBoost.

Safety is one of the most essential considerations when evaluating the performance of autonomous vehicles (AVs). Real-world AV data, including trajectory, detection, and crash data, are becoming increasingly popular as they provide possibilities for a realistic evaluation of AVs' performance. While substantial research was conducted to estimate general crash patterns utilizing structured AV crash data, a comprehensive exploration of AV crash narratives remains limited. These narratives contain latent information about AV crashes that can further the understanding of AV safety. Therefore, this study utilizes the Structural Topic Model (STM), a natural language processing technique, to extract latent topics from unstructured AV crash narratives while incorporating crash metadata (i.e., the severity and year of crashes). In total, 15 topics are identified and are further divided into behavior-related, party-related, location-related, and general topics. Using these topics, AV crashes can be systematically described and clustered. Results from the STM suggest that AVs' abilities to interact with vulnerable road users (VRUs) and react to lane-change behavior need to be further improved. Moreover, an XGBoost model is developed to investigate the relationships between the topics and crash severity. The model significantly outperforms existing studies in terms of accuracy, suggesting that the extracted topics are closely related to crash severity. Results from interpreting the model indicate that topics containing information about crash severity and VRUs have significant impacts on the model's output, which are suggested to be included in future AV crash reporting.

Impact of event encoding and dissimilarity measures on traffic crash characterization based on sequence of events.

Crash sequence analysis has been shown in prior studies to be useful for characterizing crashes and identifying safety countermeasures. Sequence analysis is highly domain-specific, but its various techniques have not been evaluated for adaptation to crash sequences. This paper evaluates the impact of encoding and dissimilarity measures on crash sequence analysis and clustering. Sequence data of interstate highway, single-vehicle crashes in the United States, from 2016 to 2018, were studied. Two encoding schemes and five optimal matching based dissimilarity measures were compared by evaluating the sequence clustering results. The five dissimilarity measures were categorized into two groups based on correlations between dissimilarity matrices. The optimal dissimilarity measure and encoding scheme were identified based on the agreements with a benchmark crash categorization. The transition-rate-based, localized optimal matching dissimilarity and consolidated encoding scheme had the highest agreement with the benchmark. Evaluation results indicate that the selection of dissimilarity measure and encoding scheme determines the results of sequence clustering and crash characterization. A dissimilarity measure that considers the relationships between events and domain context tends to perform well in crash sequence clustering. An encoding scheme that consolidates similar events naturally takes domain context into consideration.

Intersection two-vehicle crash scenario specification for automated vehicle safety evaluation using sequence analysis and Bayesian networks.

This paper introduces a test scenario specification procedure using crash sequence analysis and Bayesian network modeling. Intersection two-vehicle crash data was obtained from the 2016-2018 National Highway Traffic Safety Administration (NHTSA) Crash Report Sampling System (CRSS) database. Vehicles involved in the crashes are specifically renumbered based on their initial positions and trajectories. Crash sequences are encoded to include detailed pre-crash events and concise collision events. Based on sequence patterns, the crashes are characterized as 55 types. A Bayesian network model is developed to depict the interrelationships among crash sequence types, crash outcomes, human factors, and environmental conditions. Scenarios are specified by querying the Bayesian network's conditional probability table. Distributions of operational design domain (ODD) attributes (e.g., driver behavior, weather, lighting condition, intersection geometry, traffic control device) are specified based on conditions of sequence types. Also, distribution of sequence types is specified on specific crash outcomes or combinations of ODD attributes.

Using empirical traffic trajectory data for crash risk evaluation under three-phase traffic theory framework.

This study employed surrogate safety measures to evaluate the crash risks in different traffic phases and phase transitions according to the three-phase theory. The analysis was conducted from a microscopic perspective based on empirical vehicle trajectory data collected from the Interstate 80 in California, USA, and the Yingtian Expressway in Nanjing, China. Traffic phases were identified based on traffic flow variables and their correlations. Two advanced crash risk indexes from vehicle trajectories were conducted to evaluate the safety performance in each traffic state. The effects of various traffic flow variables (i.e. flow rate, density, average speed) on crash risks were explored based on speed-density plane, speed-flow plane and flow-density plane. Three regression models were developed to quantify the effects of traffic flow variables and traffic states on crash risks. The results show significant disparities of safety performance among different traffic states. Synchronized flow and wide moving jam are found to be the most dangerous phases. High density and low speed are associated with high crash risk. The best crash risk prediction performance is achieved when integrating both traffic phases and traffic parameters.

Automated vehicle crash sequences: Patterns and potential uses in safety testing.

With safety being one of the primary motivations for developing automated vehicles (AVs), extensive field and simulation tests are being carried out to ensure AVs can operate safely on roadways. Since 2014, the California Department of Motor Vehicles (DMV) has been collecting AV collision and disengagement reports, which are valuable data sources for studying AV crash patterns. A crash sequence of events describes the AV's interactions with other road users before a collision in a temporal manner. In this study, sequence of events data extracted from California AV collision reports were used to investigate patterns and how they may be used to develop AV test scenarios. Employing sequence analysis methods and clustering, this study evaluated 168 AV crashes (with AV in automatic driving mode before disengagement or collision) reported to the California DMV from 2015 to 2019. Analysis of subsequences showed that the most representative pattern in AV crashes was "collision following AV stop". Analysis of event transition showed that disengagement, as an event in 24% of all studied AV crash sequences, had a transition probability of 68% to an immediate collision. Cluster analysis characterized AV crash sequences into seven groups with distinctive crash dynamic features. Cross-tabulation analysis showed that sequence groups were significantly associated with variables measuring crash outcomes and describing environmental conditions. Crash sequences are useful for developing AV test scenarios. Based on the findings, a scenario-based AV safety testing framework was proposed with sequence of events embedded as a core component.

Understanding driver distractions in fatal crashes: An exploratory empirical analysis.

INTRODUCTION: Driver distraction has become a significant problem in transportation safety. As more portable wireless devices and driver assistance and entertainment systems become available to drivers, the sources of distraction are increasing. METHOD: Based on the results of different studies in the literature review, this paper categorizes different distraction enablers into six subcategories according to their fundamental characteristics and how they would affect a driver's likelihood of engaging in non-driving related activities. The review also discusses the characteristics and influence of external and internal distractions. The objective of this study is to examine the effect of different distraction sources in fatal crashes with the consideration of a driver's age and sex. Tukey test, chi-square test of independence, Nemenyi post-hoc test, and Marascuilo procedure have been used to investigate the top distraction sources, the trend of distraction-affected fatal crashes, the effect of different distractions on drives in different age groups, and their influence on female and male drivers. RESULTS: It was found that inner cognitive inferences accounted for the greatest proportion of driver engagement in distractions. Young drivers show a larger probability of being distracted by in-vehicle technology-related devices/objects. Within the group of young drivers, female drivers showed a higher probability than their male counterparts of engaging in distracted driving caused by in-vehicle technology-related devices. Among six subcategories of distractions, drivers older than 80 years old were found to be most likely affected by inner cognitive interferences.

A conceptual framework for dynamic extension of the red clearance interval as a countermeasure for red-light-running.

This manuscript describes the development and evaluation of a conceptual framework for real-time operation of dynamic on-demand extension of the red clearance interval as a countermeasure for red-light-running. The framework includes a decision process for determining, based on the real-time status of vehicles arriving at the intersection, when extension of the red clearance interval should occur and the duration of each extension. A zonal classification scheme was devised to assess whether an approaching vehicle requires additional time to safely clear the intersection based on the remaining phase time, type of vehicle, current speed, and current distance from the intersection. Expected performance of the conceptual framework was evaluated through modeling of replicated field operations using vehicular event data collected as part of this research. The results showed highly accurate classification of red-light-running vehicles needing additional clearance time and relatively few false extension calls from stopping vehicles, thereby minimizing the expected impacts to signal and traffic operations. Based on the recommended parameters, extension calls were predicted to occur once every 26.5cycles. Assuming a 90scycle, 1.5 extensions per hour were expected per approach, with an estimated extension time of 2.30s/h. Although field implementation was not performed, it is anticipated that long-term reductions in targeted red-light-running conflicts and crashes will likely occur if red clearance interval extension systems are implemented at locations where start-up delay on the conflicting approach is generally minimal, such as intersections with lag left-turn phasing.

Injury outcomes and costs for cross-median and median barrier crashes.

INTRODUCTION: The objective of this research was to quantify the injury outcomes and develop reliable and comprehensive injury costs for cross-median crashes (CMC) and median barrier crashes (MBC). METHOD: A three-step methodology was developed to quantify the crash costs for each crash severity and type. All CMC and MBC between 2001 and 2007 in Wisconsin were identified and used in this analysis. The Wisconsin CODES database provided comprehensive injury costs based on the injury types and severities suffered by participants in study crashes. RESULTS: As expected, multi-vehicle CMC result in more total injuries and more severe injuries than single-vehicle CMC. Injury costs for the same injury level on KABCO scale are different for different crash types. Injury costs for concrete MBC are 33% to 50% less than those of multi-vehicle CMC, while the injury costs of concrete MBC for lower severities (B and C) are similar to those of single-vehicle CMC for the same severities; but for incapacitating injuries the costs are 30% less. As expected, concrete MBC result in lower severities than CMC. The costs, by crash severity, vary significantly between different crash types. Concrete median barrier injury crashes are roughly 20% of multi-vehicle CMC costs and 50% of single-vehicle CMC costs. CONCLUSIONS: Results indicate that using one set of crash costs for all crash types biases any evaluation. Therefore, it is recommended that crash-type-specific costs be used in applications such as development of median barrier warrant where specific types of crashes are considered (CMC and MBC). IMPACT ON INDUSTRY: Using crash specific costs can lead to a more realistic benefit-cost analysis and enable better decision-making.

Rainfall effect on single-vehicle crash severities using polychotomous response models.

As part of the Wisconsin road weather safety initiative, the objective of this study is to assess the effects of rainfall on the severity of single-vehicle crashes on Wisconsin interstate highways utilizing polychotomous response models. Weather-related factors considered in this study include estimated rainfall intensity for 15 min prior to a crash occurrence, water film depth, temperature, wind speed/direction, stopping sight distance and deficiency of car-following distance at the crash moment. For locations with unknown weather information, data were interpolated using the inverse squared distance method. Non-weather factors such as road geometrics, traffic conditions, collision types, vehicle types, and driver and temporal attributes were also considered. Two types of polychotomous response models were compared: ordinal logistic and sequential logistic regressions. The sequential logistic regression was tested with forward and backward formats. Comparative models were also developed for single vehicle crash severity during clear weather. In conclusion, the backward sequential logistic regression model produced the best results for predicting crash severities in rainy weather where rainfall intensity, wind speed, roadway terrain, driver's gender, and safety belt were found to be statistically significant. Our study also found that the seasonal factor was significant in clear weather. The seasonal factor is a predictor suggesting that inclement weather may affect crash severity. These findings can be used to determine the probabilities of single vehicle crash severity in rainy weather and provide quantitative support on improving road weather safety via weather warning systems, highway facility improvements, and speed limit management.

Using eye movements to evaluate effects of driver age on risk perception in a driving simulator.

Novice drivers (16-year-olds with < or = 6 months' driving experience) have the highest crash involvement rates per 100 million vehicle miles (161 million vehicle km). In the past, this was attributed to greater risk taking or poorly developed psychomotor skills. More recently, however, their high crash involvement rate has been hypothesized to be attributable largely to their relative inability to acquire and assess information in inherently risky situations. The current study seeks to evaluate this hypothesis by recording eye movements while 72 participants (24 novice drivers, 24 younger drivers, and 24 older drivers) drove through 16 risky scenarios in an advanced driving simulator. There were significant age-related differences in driver scanning behavior, consistent with the hypothesis that novice drivers' scanning patterns reflect their failure to acquire information about potential risks and their consequent failure to deal with these risks. Actual or potential applications of this research include modification of these scenarios for display on a PC as a basis for a training module that would enable novice drivers to recognize risky scenarios before they encounter them on the road, in the hope of reducing their high fatality rate.