Steering or braking avoidance response in SHRP2 rear-end crashes and near-crashes: A decision tree approach.

OBJECTIVE: The paper presents a systematic analysis of drivers' crash avoidance response during crashes and near-crashes and developed a machine learning-based predictive model that can determine driver maneuver using pre-incident driver behavior and driving context. METHODS: We analyzed 286 naturalistic rear-end crashes and near-crashes from the SHRP2 naturalistic driving study. All the events were manually reduced using face video (face and forward) and kinematic responses. In this paper, we developed new reduction variables that enhanced the understanding of drivers' gaze behavior and roadway attention behavior during these events. These features reflected how the event criticality, measured using time to collision, related to drivers' pre-incident behavior (secondary behavior, gaze behavior), and drivers' perception of the event (physical reaction and maneuver). The imperative understanding of such relations was validated using a random forest- (RF) based classifier, which efficiently predicted if a driver was going to brake or change the lane as an avoidance maneuver. RESULTS: The RF presented in this paper effectively explored the nonlinear patterns in the data and was highly accurate (∼96 %) in its prediction. A further analysis of the RF model showed that six features played a pivotal role in the decision logic. These included the drivers' last glance duration before the event, last glance eccentricity, duration of 'eyes on road' immediately before the event, the time instance and criticality when the driver perceives the threat as well as acknowledge the threat, and possibility of an escape path in the adjacent lane. Using partial dependency plots, we also showed how different thresholds of these feature variables determined the drivers' maneuver intention. CONCLUSIONS: In this paper we analyzed driving context, drivers' behavior, event criticality, and drivers' response in a unified structure to predict their avoidance response. To the best of our knowledge, this is the first such effort where large-scale naturalistic data (crashes and near crashes) was analyzed for prediction of drivers' maneuver and determined key behavioral and contextual factors that contribute to this avoidance maneuver.

Do the benefits outweigh the costs? Societal benefit-cost analysis of three large truck safety technologies.

Although research has found advanced safety technologies to be effective at preventing large truck crashes, limited empirical data exists regarding their cost effectiveness to the U.S. society. Without these data, carriers are hesitant to adopt costly technologies and government agencies are hesitant to create regulation mandating their use. The objective of this study was to provide scientifically-based estimates of the societal benefits and costs of large truck automatic emergency braking (AEB), lane departure warning (LDW), and video-based onboard safety monitoring (OSM). For each technology, benefit-cost analyses were performed for installing the technology on all large trucks (including retrofitting existing trucks) and for equipping new large trucks only. Sensitivity analyses examined three cost estimates (low, average, high; values technology-specific), two estimates of system efficacy (low and high; values technology-specific), and three discount rates (0%, 3%, 7%) for each technology. Equipping trucks with LDW and video-based OSM systems were found to be cost effective for all combinations of costs, efficacy, and discount rates examined, for both new and existing trucks. Results for AEB and were mixed. Only a $500 AEB system was cost effective when equipping new trucks and retrofitting existing trucks. However, all cost estimates were cost effective with a 28% efficacy rate when only equipping new large trucks. Overall, these data suggested all three technologies can be cost-effective for new large trucks provided the current costs and efficacy rates can be maintained or improved upon.

A synthetic approach to compare the large truck crash causation study and naturalistic driving data.

Truck crashes represent a significant problem on our nation's highways. There is a great opportunity to learn about crash causation by analyzing and comparing the Large Truck Crash Causation Study (LTCCS) and naturalistic driving (ND) data. These data sets provide in-depth information, but have contrasting strengths and weaknesses. The LTCCS contains information on high-severity crashes (crashes and fatal crashes), but relied on data collected during crash investigations. The LTCCS identified principal driver errors in the crash, such as the Critical Reason, but not detailed behaviors or scenario sequences. The ND data sets relate primarily to non-crashes that are detectable from dynamic vehicle events, such as hard braking, swerve, etc., provide direct video observations of the driver and the surrounding driving scene and precise information on driver inputs (kinematics) and captured events, and provide certain types of exposure data that cannot easily be obtained using crash reconstruction data. The ND data are collected continuously, thereby capturing both safety-critical events and normative driving (i.e., baseline). The current project evaluated large-truck crash data from the LTCCS and two large-truck ND data sets, the Naturalistic Truck Driving Study and the Drowsy Driver Warning System Field Operational Test. A synthetic risk ratio analysis on the associated factor, Following Too Closely, indicated that truck drivers in the LTCCS were 1.34 times more likely to be involved in a crash, than an ND crash-relevant conflict, if they were following too closely (i.e., tailgating). Given several caveats noted in the paper, this study suggests it's possible to use the ND data set to calculate the exposure of a given behavior and use the LTCCS data set to calculate the crash exposure to the same behavior.

Benefit-cost analysis of lane departure warning and roll stability control in commercial vehicles.

INTRODUCTION: This paper presents the cost benefits of two different onboard safety systems (OSS) installed on trucks as they operated during normal revenue deliveries. Using a formal economic analysis approach, the study quantified the costs and benefits associated with lane departure warning (LDW) systems and roll stability control (RSC) systems. METHODS: The study used data collected from participating carriers (many of these crashes were not reported to state or Federal agencies), and the research team also reviewed each crash file to determine if the specific OSS would have mitigated or prevented the crash. The deployment of each OSS was anticipated to increase the safety of all road users, but impact different sectors of society in different ways. Benefits that were inherent in each group (e.g., industry, society) were considered, and different benefit-cost analyses (BCAs) were performed. RESULTS: This paper presents two BCAs: a BCA focused on the costs and benefits in the carrier industry by implementing each OSS, and a BCA that measured the societal benefits of each OSS. In addition, a BCA for a theoretical mandatory deployment option for each OSS is presented. CONCLUSIONS: BCA results for LDW and RSC clearly showed their benefits outweighed their costs for the carrier and society. Practical applications: Cost information is a crucial factor in purchasing decisions in carriers; similarly, regulators must consider the cost burden prior to mandating technologies. The results in this study provide carrier decision makers and regulators with information necessary to make an informed decision regarding RSC and LDW.

An assessment of commercial motor vehicle driver distraction using naturalistic driving data.

OBJECTIVE: This study analyzed naturalistic driving data from commercial trucks (3-axle and tractor-trailer/tanker) and buses (transit and motorcoach) during a 3-month period. METHODS: The data set contained 183 commercial truck and bus fleets comprising 13,306 vehicles and included 1085 crashes, 8375 near crashes, 30,661 crash-relevant conflicts, and 211,171 baseline events. Study results documented the prevalence of tertiary tasks and the risks associated with performing these tasks while driving. RESULTS: Results indicated the odds of involvement in a safety-critical event differed as a function of performing different cell phone-related subtasks while driving. Although the odds ratio for talking/listening on a cell phone while driving was found to not significantly increase the likelihood of involvement in a safety-critical event, other cell phone subtasks (e.g., texting, dialing, reaching) were found to significantly increase the odds of involvement in a safety-critical event. CONCLUSIONS: The results suggest that cell phone use while driving should not be considered a simple dichotomous task (yes/no). Consideration should instead be made for a set of discrete cell phone subtasks that are each associated with varying levels of risk. Several hypotheses are presented to explain why cell phone use while driving was found to not increase the likelihood of involvement in a safety-critical event.

Evaluating the 2003 revised hours-of-service regulations for truck drivers: the impact of time-on-task on critical incident risk.

On January 4th, 2004, the Federal Motor Carrier Safety Administration implemented a revised set of regulations concerning the hours-of-service (HOS) of commercial motor vehicle (CMV) drivers. One central component of the revised HOS regulations was a one-hour increase in allowable driving time, from 10 to 11h. The current study evaluated the impact of the additional driving-hour on critical incident risk. Data from a naturalistic truck driving study, which resulted in over 2 million driving miles of continuously collected data, were analyzed. Driving hour bins (hours 1 through 11) were created and the frequency of critical incidents for each hour, and trips (opportunities) per each hour, were identified. A relative frequency was then calculated (critical incidents divided by opportunities) for each hour and odds ratios were determined. Analyses found an elevated risk in the 1st driving-hour, but no consistent significant difference between hours 2 through 11. Analyses on time-of-day, where incident rates were calculated for each of the 24h in the day, were also conducted. The results found a strong positive correlation to national traffic density data. As an impact on U.S. national transportation policy, the results of this study do not support the hypothesis that there is an increased risk resulting from CMV drivers driving in the 11th driving-hour as compared to the 10th driving-hour, or any hour.