Assessing the applicability of impact speed injury risk curves based on US data to defining safe speeds in the US and Sweden.

Vision Zero is an approach to road safety that aims to eliminate all traffic-induced fatalities and lifelong injuries. To reach this goal, a multi-faceted safe system approach must be implemented to anticipate and minimize the risk associated with human mistakes. One aspect of a safe system is choosing speed limits that keep occupants within human biomechanical limits in a crash scenario. The objective of this study was to relate impact speed and maximum delta-v to risk of passenger vehicle (passenger cars and light trucks and vans) occupants sustaining a moderate to fatal injury (MAIS2+F) in three crash modes: head-on vehicle-vehicle, frontal vehicle-barrier, and front-to-side vehicle-vehicle crashes. Data was extracted from the Crash Investigation Sampling System, and logistic regression was used to construct the injury prediction models. Impact speed was a statistically significant predictor in head-on crashes, but was not a statistically significant predictor in vehicle-barrier or front-to-side crashes. Maximum delta-v was a statistically significant predictor in all three crash modes. A head-on impact speed of 62 km/h yielded 50% (±27%) risk of moderate to fatal injury for occupants at least 65 years old. A head-on impact speed of 82 km/h yielded 50% (±31%) risk of moderate to fatal injury for occupants younger than 65 years. Compared to the impact speeds, the maximum delta-v values yielding the same level of risk were lower within the head-on crash population. A head-on delta-v of 40 km/h yielded 50% (±21%) risk of moderate to fatal injury for occupants at least 65 years old. A head-on delta-v of 65 km/h yielded 50% (±33%) risk of moderate to fatal injury for occupants younger than 65 years. A maximum delta-v value of approximately 30 km/h yielded 50% (±42%) risk of MAIS2+F injury for passenger car occupants in vehicle-vehicle front-to-side crashes. A maximum delta-v value of approximately 44 km/h yielded 50% (±24%) risk of MAIS2+F injury for light truck and van occupants, respectively, in vehicle-vehicle front-to-side crashes.

Finding and understanding pedal misapplication crashes using a deep learning natural language model.

OBJECTIVE: The objective of this study was to develop a system which used the BERT natural language understanding model to identify pedal misapplication (PM) crashes from their crash narratives and validate the accuracy of the system. METHODS: The training dataset used for this study was 11 cases from the NMVCCS study and 952 cases from the North Carolina state crash database. Cases for this study were selected from their respective full datasets using a keyword search algorithm containing terms indicative of a pedal-related mistake. A BERT language model was used to classify each case narrative as either no pedal misapplication, PM by vehicle 1, PM by vehicle 2, or PM by vehicle 3. After training, the language model was used to determine the incidence of pedal misapplication in a test dataset of 8,668 North Carolina and NMVCCS cases and these results were compared to a manual review of the dataset. After manual review, 2,969 cases were pedal misapplications. RESULTS: The model's AUC ROC performance at detecting PM was quantified on the entire testing dataset to evaluate the power of the system to generalize to case narratives unseen at training time. The AUC ROC value was 0.9835, indicating strong generalization to all crash narratives. By choosing the optimal threshold using the ROC curve, the system correctly identified PM in 95.7% of crash narratives. When pedal misapplication was correctly identified, the correct vehicle was identified in 95.9% of cases. A total of 3,062 pedal misapplications were identified. The model labeled cases 353 times faster than a researcher. CONCLUSIONS: The strong performance of the model suggests that the automated interpretation of case narratives can be used for future research studies without any manual review. This would save time and enable the use of datasets where manual review would be infeasible. The automated extraction of information from crash narratives using deep learning natural language models has not been demonstrated previously in the literature, to the best of the authors' knowledge. This technique can be applied to large, infrequently used datasets of crash narratives and extended to extract useful vehicle, occupant, or environment information to make these datasets amenable to traditional statistical analyses.

Differential benefit of sensor system field-of-view and range in pedestrian automated emergency braking systems.

OBJECTIVE: Current Pedestrian Automatic Emergency Braking (P-AEB) systems often use a combination of radar and cameras to detect pedestrians and automatically apply braking to prevent or mitigate an impending collision. However, these current sensor systems might have a restricted field-of-view (FOV) which may not detect all pedestrians. Advanced sensors like LiDAR can have a wider FOV that may substantially help improve detection. The objective of this study was to determine the influence of FOV and range on the effectiveness of P-AEB systems to determine the potential benefit of advanced sensors. METHODS: This study utilized vehicle-pedestrian crashes from the Pedestrian Crash Data Study (PCDS) to calculate pre-crash pedestrian and vehicle trajectories. A computational model was then applied to simulate the crash with a hypothetical P-AEB system. The model was designed to be able to vary the system's field-of-view (FOV), range, time-to-collision of activation, and system latency. In this study we estimated how the FOV and range of advanced sensors could affect P-AEB system effectiveness at avoiding crashes and reducing impact speed. Sensor range was varied from 25 - 100 m and sensor FOV was varied from ±10° to ±90°. RESULTS: Sensors simulated with a range of 50 m or greater performed only approximately 1% better than with a 25 m range. Field-of-view had a larger effect on estimated system avoidance capabilities with a ± 10° FOV sensor estimated to avoid 46-47% of collisions compared to 91-92% for a ± 90° FOV sensor. The system was able to avoid a greater percentage of cases in which the vehicle was traveling straight at sensor FOVs of ±30° and below. Among the unavoided crashes with a sensor FOV of ±90°, the average impact velocity using a 100 m range sensor was 7.4 m/s which was 3.1 m/s lower than a 25 m range sensor. CONCLUSIONS: Sensor ranges above 25 m were not found to significantly affect estimated crash avoidance potential, but had a small effect on impact mitigation. Sensor FOV had a larger effect on crash avoidance up to a FOV of ±60° with little additional benefit at larger FOVs.

Considering real-world sightline obstructions in crash and injury prevention estimates for left turn across path/opposite direction intersection active safety systems.

OBJECTIVE: Previous analyses of active safety systems in left turn across path/opposite direction (LTAP/OD) crashes have shown that sensor sightline obstructions might affect the performance of these systems. National retrospective crash databases provide little information about the proportion of cases which have sightline obstructions. One promising alternative are naturalistic driving studies (NDS). The objective of this study was to estimate the proportion of LTAP/OD crashes and near-crashes which have sightline obstructions using a large-scale NDS and update previous estimates of intersection active safety system effectiveness using this information. METHODS: LTAP/OD crash and near-crash cases were identified from the Second Strategic Highway Research Program (SHRP 2) dataset. Each case was reviewed for the presence of obstructing vehicles when the left turning vehicle began turning. This study considered 241 crash and near-crash LTAP/OD events selected from SHRP 2. SHRP 2 was an NDS which collected 80 million kilometers of driving from approximately 2,500 participants over a 2.5 year period. The sightline obstruction ratio was defined as the proportion of cases which had sightline obstructions when the turning vehicle began turning. A logistic regression model was used to determine the statistical significance of factors which affected the sightline obstruction ratio, which included event severity, traffic control device, subject vehicle crash configuration, and turning lane presence. LTAP/OD active safety system effectiveness was quantified in a prior study for cases with and without sightline obstructions separately. System effectiveness was re-computed by weighting the results according to the worst-case sightline ratio computed in this study. RESULTS: Traffic control device, subject vehicle crash type (turning or traveling through), and turning lane presence were not found to affect sightline obstruction ratio. In crash cases, the sightline obstruction ratio was 40%. In near-crash cases, the sightline obstruction ratio was 18%. Finally, the effectiveness of an intersection active safety system was evaluated using this sightline obstruction ratio. CONCLUSIONS: This study quantified the sightline obstruction ratio, an important parameter needed to evaluate intersection active safety systems. This study also establishes a baseline for evaluating the presence of sightline obstructions in a future naturalistic driving study when road infrastructure has changed.

Expansion of NASS/CDS for characterizing run-off-road crashes.

OBJECTIVE: Run-off-road (ROR) crashes account for one-third of all annual crash fatalities in the US. The National Automotive Sampling System Crashworthiness Data System (NASS/CDS) is a dataset which may be used to understand the nature of ROR crashes. Despite the wealth of coded data available in NASS/CDS, this dataset lacks coded information about the roadside environment and the off-road trajectory of the vehicle. This information would be useful for determining lane departure warning (LDW) benefits, residual safety problems, performance of current safety hardware, lane marking inventory, LDW test procedure development, radius of curvature characterization, and effectiveness of ESC. The purpose of this paper is to demonstrate a methodology for expanding the data available in NASS/CDS to form and validate a specialized road departure database. METHODS: Observed, measured, and reconstructed data elements were extracted from NASS/CDS and compiled into the National Cooperative Highway Research Program (NCHRP) 17-43 database. Observed variables were primarily coded from the scene photographs and included information such as the lane markings, and geometry of the roadside cross-section. Additional variables were measured from the scaled scene diagrams including the path of the vehicle, road dimensions, and roadside object positions. The vehicle impact speed and departure speed were reconstructed using the WinSMASH delta-v, roadside object characteristics, and vehicle path. Two studies were conducted to demonstrate the usefulness of the NCHRP 17-43 database in evaluating both vehicle-based and infrastructure-based ROR countermeasures. RESULTS: The resulting NCHRP 17-43 database includes 1,581 NASS/CDS cases representing 510,154 ROR crashes. Analysis of the database found that drivers which crashed following an overcorrection were younger than drivers which did not overcorrect. This may indicate that inexperienced drivers are more likely to overcorrect when departing the roadway. The 85th percentile impact severity of ROR crashes, which occur on roads with a speed limit greater than 65 mph, is higher than the practical worst-case test conditions for roadside barriers. CONCLUSIONS: The NCHRP 17-43 database contains information extracted from NASS/CDS cases to better understand the nature of ROR crashes, driver behavior in these crashes, and the potential benefits of both vehicle-based and infrastructure-based ROR countermeasures.

Estimating near side crash injury risk in best performing passenger vehicles in the United States.

The objective of this paper was to develop an injury risk model relating real world injury outcomes in near-side crashes with U.S. New Car Assessment Program (NCAP) test performance, crash, and occupant properties. The study was motivated by the longer-term goal of predicting injury outcomes in a future fleet in which all vehicles are expected to have passive safety performance equivalent to a 5-star NCAP rating level (the highest star rating and lowest risk of injury). The dataset used to evaluate injury risk was the National Automotive Sampling System / Crashworthiness Data System (NASS/CDS). Case years 2010-2015 were used. An injured occupant was defined as a vehicle occupant who experienced an injury of maximum Abbreviated Injury Scale (AIS) of 2 or greater, or who were fatally injured. Injury severity was scored using AIS-2005 (2008 update). Cases were selected in which front-row occupants of late-model vehicles were exposed to a near-side crash. Logistic regression was used to develop an injury model with delta-v, belt status, age, and gender as predictor variables. The side crash performance of each vehicle was identified and added to the model by matching each case with the associated performance in the NCAP moving deformable barrier side impact crash test. NCAP MDB test performance, delta-v, and occupant age, sex, and BMI were found to be significant predictors of injury risk. The effect of a 5 % higher risk in the MDB test (approximately one star rating worse) was comparable to a 2.84 km/h increase in delta-v. This model informs the development of active safety systems in a future fleet where vehicle passive safety performance is higher, quantified by the NCAP MDB test.

Residual road departure crashes after full deployment of LDW and LDP systems.

Objective: Road departures are one of the most severe crash modes in the United States. To help reduce this risk, vehicles are being introduced in the United States with lane departure warning (LDW) systems, which warn the driver of a departure, and lane departure prevention (LDP) systems, which assist the driver in steering back to the roadway. Previous studies have estimated that LDW/LDP systems may prevent one third of drift-out-of-lane road departure crashes. This study investigates the crashes that were not prevented, to potentially set research priorities for next-generation road departure prevention systems. Methods: The event data recorder (EDR) data from 128 road departure crashes in the National Automotive Sampling System Crashworthiness Data System (NASS-CDS) from 2011 to 2015 were mapped onto the vehicle trajectory and simulated with LDW/LDP to assess the potential for crash avoidance. The model predicted that 63-83% of single-vehicle road departure crashes may not be prevented by an LDW system and 49% may not be prevented by an LDP system. Results and Conclusions: For LDP systems, which were assumed to have zero latency, no crashes were avoided if the time-to-collision (TTC) from lane crossing to impact was less than 0.55 s. Obstacles such as guardrails and traffic barriers, which tend to be very close to the road, were more common among the remaining crashes. The study shows that LDW/LDP systems are limited by two factors, driver reaction time and TTC to the roadside object. Thus, earlier driver response and longer TTC may help in these situations.

Estimated benefit of automated emergency braking systems for vehicle-pedestrian crashes in the United States.

Objective: The objective of this research study is to estimate the benefit to pedestrians if all U.S. cars, light trucks, and vans were equipped with an automated braking system that had pedestrian detection capabilities. Methods: A theoretical automatic emergency braking (AEB) model was applied to real-world vehicle-pedestrian collisions from the Pedestrian Crash Data Study (PCDS). A series of potential AEB systems were modeled across the spectrum of expected system designs. Both road surface conditions and pedestrian visibility were accounted for in the model. The impact speeds of a vehicle without AEB were compared to the estimated impact speeds of vehicles with a modeled pedestrian detecting AEB system. These impacts speeds were used in conjunction with an injury and fatality model to determine risk of Maximum Abbreviated Injury Scale of 3 or higher (MAIS 3+) injury and fatality. Results: AEB systems with pedestrian detection capability, across the spectrum of expected design parameters, reduced fatality risk when compared to human drivers. The most beneficial system (time-to-collision [TTC] = 1.5 s, latency = 0 s) decreased fatality risk in the target population between 84 and 87% and injury risk (MAIS score 3+) between 83 and 87%. Conclusions: Though not all crashes could be avoided, AEB significantly mitigated risk to pedestrians. The longer the TTC of braking and the shorter the latency value, the higher benefits showed by the AEB system. All AEB models used in this study were estimated to reduce fatalities and injuries and were more effective when combined with driver braking.

Crash and injury prevention estimates for intersection driver assistance systems in left turn across path/opposite direction crashes in the United States.

Objective: The objective of this research study was to estimate the number of left turn across path/opposite direction (LTAP/OD) crashes and injuries that could be prevented in the United States if vehicles were equipped with an intersection advanced driver assistance system (I-ADAS). Methods: This study reconstructed 501 vehicle-to-vehicle LTAP/OD crashes in the United States that were investigated in the NHTSA National Motor Vehicle Crash Causation Survey (NMVCCS). The performance of 30 different I-ADAS system variations was evaluated for each crash. These variations were the combinations of 5 time-to-collision (TTC) activation thresholds, 3 latency times, and 2 different response types (automated braking and driver warning). In addition, 2 sightline assumptions were modeled for each crash: One where the turning vehicle was visible long before the intersection and one where the turning vehicle was only visible within the intersection. For resimulated crashes that were not avoided by I-ADAS, a new crash delta-V was computed for each vehicle. The probability of Abbreviated Injury Scale 2 or higher injury in any body region (Maximum Abbreviated Injury Scale [MAIS] 2+F) to each front-row occupant was computed. Results: Depending on the system design, sightline assumption, I-ADAS variation, and fleet penetration, an I-ADAS system that automatically applies emergency braking could avoid 18-84% of all LTAP/OD crashes. Only 0-32% of all LTAP/OD crashes could have been avoided using an I-ADAS system that only warns the driver. An I-ADAS system that applies emergency braking could prevent 47-93% of front-row occupants from receiving MAIS 2 + F injuries. A system that warns the driver in LTAP/OD crashes was able to prevent 0-37% of front-row occupants from receiving MAIS 2 + F injuries. The effectiveness of I-ADAS in reducing crashes and number of injured persons was higher when both vehicles were equipped with I-ADAS. Conclusions: This study presents the simulated effectiveness of a hypothetical intersection active safety system on real crashes that occurred in the United States. This work shows that there is a strong potential to reduce crashes and injuries in the United States.

Evaluation of alternative intersection treatments at rural crossroads using simulation software.

OBJECTIVE: Rural roads are characterized by hazardous roadsides and suboptimal geometry yet allow for high travel speeds and unfavorable impact angles. In Victoria, 25% of persons seriously injured and 52% of fatalities occur on rural roads, with 30% occurring at intersections. In the United States, almost twice the number of traffic fatalities occur in rural areas than in urban areas, while accounting for less than half of all vehicle miles traveled and 21% of the population. The choice of safety countermeasure is therefore paramount. Simulation software provides a cost-effective means of analyzing alternative intersection treatments with a view to identifying their effectiveness in mitigating crashes. The aim of this research was to assess the safety benefits of 4 alternative intersection treatments using in-depth crash data with an advanced crash reconstruction process. METHOD: Using a single serious injury real-world crash from the Monash University Accident Research Centre Enhanced Crash Investigation Study and crash reconstruction software, an exemplar rural crash was reconstructed and validated against real-world data. The crash involved a passenger vehicle (European New Car Assessment Programme 5-star) approaching from a minor road and failing to yield at a give-way sign; the posted speed limit was 80 km/h. The vehicle was struck on the right/driver side by a rigid truck (B-vehicle; 1990) traveling on the major approach (100 km/h). The driver of the case vehicle was seriously injured. Four alternative intersection treatments appropriate for the crash site were constructed in computer-aided design software (Rhinoceros Ver. 5): roundabout; rumble strips; a reduced speed limit; and the combination of lower speed limit and rumbles to determine the reduction in crash forces in the presence of the countermeasures. RESULTS: The hypothetical scenarios demonstrate substantial reductions in impact force and different points of impact, resulting in a significantly lower injury severity for the struck driver. Speed limit reduction to 80 km/h on the main approach (from 100 km/h) in combination with rumble strips on both intersection approaches had the most favorable outcome with the crash avoided entirely, assuming speed compliance. DISCUSSION: The findings have implications for understanding the role of speed in crashes and hence the design of effective countermeasures. Simulation software, validated using real-world data, provides a cost-effective means of evaluating alternative intersection treatments for rural intersections. Scaled up, implementing these treatments would have significant safety benefits and reduce the road trauma currently associated with rural roads.

Injury mitigation estimates for an intersection driver assistance system in straight crossing path crashes in the United States.

OBJECTIVE: Accounting for one fifth of all crashes and one sixth of all fatal crashes in the United States, intersection crashes are among the most frequent and fatal crash modes. Intersection advanced driver assistance systems (I-ADAS) are emerging vehicle-based active safety systems that aim to help drivers safely navigate intersections. The objective of this study was to estimate the number of crashes and number of vehicles with a seriously injured driver (Maximum Abbreviated Injury Scale [MAIS] 3+) that could be prevented or reduced if, for every straight crossing path (SCP) intersection crash, one of the vehicles had been equipped with an I-ADAS. METHODS: This study retrospectively simulated 448 U.S. SCP crashes as if one of the vehicles had been equipped with I-ADAS. Crashes were reconstructed to determine the path and speeds traveled by the vehicles. Cases were then simulated with I-ADAS. A total of 30 variations of I-ADAS were considered in this study. These variations consisted of 5 separate activation timing thresholds, 3 separate computational latency times, and 2 different I-ADAS response modalities (i.e., a warning or autonomous braking). The likelihood of a serious driver injury was computed for every vehicle in every crash using impact delta-V. The results were then compiled across all crashes in order to estimate system effectiveness. RESULTS: The model predicted that an I-ADAS that delivers an alert to the driver has the potential to prevent 0-23% of SCP crashes and 0-25% of vehicles with a seriously injured driver. Conversely, an I-ADAS that autonomously brakes was found to have the potential to prevent 25-59% of crashes and 38-79% of vehicles with a seriously injured driver. I-ADAS effectiveness is a strong function of design. Increasing computational latency time from 0 to 0.5 s was found to reduce crash and injury prevention estimates by approximately one third. For an I-ADAS that delivers an alert, crash/injury prevention effectiveness was found to be very sensitive to changes in activation timing (warning delivered 1.0 to 3.0 s prior to impact). If autonomous braking was used, system effectiveness was found to largely plateau for activation timings greater than 1.5 s prior to impact. In general, the results of this study suggest that I-ADAS will be 2-3 times more effective if an autonomous braking system is utilized over a warning-based system. CONCLUSIONS: This study highlights the potential effectiveness of I-ADAS in the U.S. vehicle fleet, while also indicating the sensitivity of system effectiveness to design specifications. The results of this study should be considered by designers of I-ADAS and evaluators of this technology considering a future I-ADAS safety test.

Predicting crash-relevant violations at stop sign-controlled intersections for the development of an intersection driver assistance system.

OBJECTIVE: Intersection crashes resulted in over 5,000 fatalities in the United States in 2014. Intersection Advanced Driver Assistance Systems (I-ADAS) are active safety systems that seek to help drivers safely traverse intersections. I-ADAS uses onboard sensors to detect oncoming vehicles and, in the event of an imminent crash, can either alert the driver or take autonomous evasive action. The objective of this study was to develop and evaluate a predictive model for detecting whether a stop sign violation was imminent. METHODS: Passenger vehicle intersection approaches were extracted from a data set of typical driver behavior (100-Car Naturalistic Driving Study) and violations (event data recorders downloaded from real-world crashes) and were assigned weighting factors based on real-world frequency. A k-fold cross-validation procedure was then used to develop and evaluate 3 hypothetical stop sign warning algorithms (i.e., early, intermediate, and delayed) for detecting an impending violation during the intersection approach. Violation detection models were developed using logistic regression models that evaluate likelihood of a violation at various locations along the intersection approach. Two potential indicators of driver intent to stop-that is, required deceleration parameter (RDP) and brake application-were used to develop the predictive models. The earliest violation detection opportunity was then evaluated for each detection algorithm in order to (1) evaluate the violation detection accuracy and (2) compare braking demand versus maximum braking capabilities. RESULTS: A total of 38 violating and 658 nonviolating approaches were used in the analysis. All 3 algorithms were able to detect a violation at some point during the intersection approach. The early detection algorithm, as designed, was able to detect violations earlier than all other algorithms during the intersection approach but gave false alarms for 22.3% of approaches. In contrast, the delayed detection algorithm sacrificed some time for detecting violations but was able to substantially reduce false alarms to only 3.3% of all nonviolating approaches. Given good surface conditions (maximum braking capabilities = 0.8 g) and maximum effort, most drivers (55.3-71.1%) would be able to stop the vehicle regardless of the detection algorithm. However, given poor surface conditions (maximum braking capabilities = 0.4 g), few drivers (10.5-26.3%) would be able to stop the vehicle. Automatic emergency braking (AEB) would allow for early braking prior to driver reaction. If equipped with an AEB system, the results suggest that, even for the poor surface conditions scenario, over one half (55.3-65.8%) of the vehicles could have been stopped. CONCLUSIONS: This study demonstrates the potential of I-ADAS to incorporate a stop sign violation detection algorithm. Repeating the analysis on a larger, more extensive data set will allow for the development of a more comprehensive algorithm to further validate the findings.

Injury Outcome in Crashes with Guardrail End Terminals.

OBJECTIVE: The goal of this study is to evaluate the crash performance of guardrail end terminals in real-world crashes. Guardrail end terminals are installed at the ends of guardrail systems to prevent the rail from spearing through the car in an end-on collision. Recently, there has been a great deal of controversy as to the safety of certain widely used end terminal designs, partly because there is surprisingly little real-world crash data for end terminals. Most existing studies of end terminal crashes used data from prior to the mid-1990s. Since then, there have been large improvements to vehicle crashworthiness and seat belt usage rates, as well as new roadside safety hardware compliant with National Cooperative Highway Research Program (NCHRP) Report 350, "Recommended Procedures for the Safety Performance Evaluation of Highway Features." Additionally, most existing studies of injury in end terminal crashes do not account for factors such as the occurrence of rollover. This analysis uses more recent crash data that represent post-1990s vehicle fleet changes and account for a number of factors that may affect driver injury outcome and rollover occurrence. METHODS: Passenger vehicle crashes coded as involving guardrail end terminals were identified in the set of police-reported crashes in Michigan in 2011 and 2012. End terminal performance was expected to be a function of end terminal system design. State crash databases generally do not identify specific end terminal systems. In this study, the coded crash location was used to obtain photographs of the crash site prior to the crash from Google Street View. These site photographs were manually inspected to identify the particular end terminal system involved in the crash. Multiple logistic regression was used to test for significant differences in the odds of driver injury and rollover between different terminal types while accounting for other factors. RESULTS: A total of 1,001 end terminal crashes from the 2011-2012 Michigan State crash data were manually inspected to identify the terminal that had been struck. Four hundred fifty-one crashes were found to be suitable for analysis. Serious to fatal driver injury occurred in 3.8% of end terminal crashes, moderate to fatal driver injury occurred in 11.8%, and 72.3% involved property damage only. No significant difference in moderate to fatal driver injury odds was observed between NCHRP 350 compliant end terminals and noncompliant terminals. Car drivers showed odds of moderate to fatal injury 3.6 times greater than LTV drivers in end terminal crashes. Rollover occurrence was not significantly associated with end terminal type. CONCLUSIONS: Car drivers have greater potential for injury in end terminal crashes than light truck/van/sport utility vehicle drivers. End terminal designs compliant with NCHRP 350 did not appear to carry different odds of moderate driver injury than noncompliant end terminals. The findings account for driver seat belt use, rollover occurrence, terminal orientation (leading/trailing), control loss, and the number of impact events. Rollover and nonuse of seat belts carried much larger increases in injury potential than end terminal type. Rollover did not appear to be associated with NCHRP 350 compliance.

Comparison of Expected Crash and Injury Reduction from Production Forward Collision and Lane Departure Warning Systems.

OBJECTIVES: The U.S. New Car Assessment Program (NCAP) now tests for forward collision warning (FCW) and lane departure warning (LDW). The design of these warnings differs greatly between vehicles and can result in different real-world field performance in preventing or mitigating the effects of collisions. The objective of this study was to compare the expected number of crashes and injured drivers that could be prevented if all vehicles in the fleet were equipped with the FCW and LDW systems tested under the U.S. NCAP. METHODS: To predict the potential crashes and serious injury that could be prevented, our approach was to computationally model the U.S. crash population. The models simulated all rear-end and single-vehicle road departure collisions that occurred in a nationally representative crash database (NASS-CDS). A sample of 478 single-vehicle crashes from NASS-CDS 2012 was the basis for 24,822 simulations for LDW. A sample of 1,042 rear-end collisions from NASS-CDS years 1997-2013 was the basis for 7,616 simulations for FCW. For each crash, 2 simulations were performed: (1) without the system present and (2) with the system present. Models of each production safety system were based on 54 model year 2010-2014 vehicles that were evaluated under the NCAP confirmation procedure for LDW and/or FCW. NCAP performed 40 LDW and 45 FCW tests of these vehicles. RESULTS: The design of the FCW systems had a dramatic impact on their potential to prevent crashes and injuries. Between 0 and 67% of crashes and 2 and 69% of moderately to fatally injured drivers in rear-end impacts could have been prevented if all vehicles were equipped with the FCW systems. Earlier warning times resulted in increased benefits. The largest effect on benefits, however, was the lower operating speed threshold of the systems. Systems that only operated at speeds above 20 mph were less than half as effective as those that operated above 5 mph with similar warning times. The production LDW systems could have prevented between 11 and 23% of drift-out-of-lane crashes and 13 and 22% of seriously to fatally injured drivers. A majority of the tested LDW systems delivered warnings near the point when the vehicle first touched the lane line, leading to similar benefits. Minimum operating speed also greatly affected LDW effectiveness. CONCLUSIONS: The results of this study show that the expected field performance of FCW and LDW systems are highly dependent on the design and system limitations. Systems that delivered warnings earlier and operated at lower speeds may prevent far more crashes and injuries than systems that warn late and operate only at high speeds. These results suggest that future FCW and LDW evaluation should prioritize early warnings and full-speed range operation. A limitation of this study is that additional crash avoidance features that may also mitigate collisions-for example, brake assist, automated braking, or lane-keeping assistance-were not evaluated during the NCAP tests or in our benefits models. The potential additional mitigating effects of these systems were not quantified in this study.

Evaluation of Vehicle-Based Crash Severity Metrics.

OBJECTIVE: Vehicle change in velocity (delta-v) is a widely used crash severity metric used to estimate occupant injury risk. Despite its widespread use, delta-v has several limitations. Of most concern, delta-v is a vehicle-based metric which does not consider the crash pulse or the performance of occupant restraints, e.g. seatbelts and airbags. Such criticisms have prompted the search for alternative impact severity metrics based upon vehicle kinematics. The purpose of this study was to assess the ability of the occupant impact velocity (OIV), acceleration severity index (ASI), vehicle pulse index (VPI), and maximum delta-v (delta-v) to predict serious injury in real world crashes. METHODS: The study was based on the analysis of event data recorders (EDRs) downloaded from the National Automotive Sampling System / Crashworthiness Data System (NASS-CDS) 2000-2013 cases. All vehicles in the sample were GM passenger cars and light trucks involved in a frontal collision. Rollover crashes were excluded. Vehicles were restricted to single-event crashes that caused an airbag deployment. All EDR data were checked for a successful, completed recording of the event and that the crash pulse was complete. The maximum abbreviated injury scale (MAIS) was used to describe occupant injury outcome. Drivers were categorized into either non-seriously injured group (MAIS2-) or seriously injured group (MAIS3+), based on the severity of any injuries to the thorax, abdomen, and spine. ASI and OIV were calculated according to the Manual for Assessing Safety Hardware. VPI was calculated according to ISO/TR 12353-3, with vehicle-specific parameters determined from U.S. New Car Assessment Program crash tests. Using binary logistic regression, the cumulative probability of injury risk was determined for each metric and assessed for statistical significance, goodness-of-fit, and prediction accuracy. RESULTS: The dataset included 102,744 vehicles. A Wald chi-square test showed each vehicle-based crash severity metric estimate to be a significant predictor in the model (p < 0.05). For the belted drivers, both OIV and VPI were significantly better predictors of serious injury than delta-v (p < 0.05). For the unbelted drivers, there was no statistically significant difference between delta-v, OIV, VPI, and ASI. CONCLUSIONS: The broad findings of this study suggest it is feasible to improve injury prediction if we consider adding restraint performance to classic measures, e.g. delta-v. Applications, such as advanced automatic crash notification, should consider the use of different metrics for belted versus unbelted occupants.

Driver Behavior During Overtaking Maneuvers from the 100-Car Naturalistic Driving Study.

OBJECTIVE: Lane changes with the intention to overtake the vehicle in front are especially challenging scenarios for forward collision warning (FCW) designs. These overtaking maneuvers can occur at high relative vehicle speeds and often involve no brake and/or turn signal application. Therefore, overtaking presents the potential of erroneously triggering the FCW. A better understanding of driver behavior during lane change events can improve designs of this human-machine interface and increase driver acceptance of FCW. The objective of this study was to aid FCW design by characterizing driver behavior during lane change events using naturalistic driving study data. METHODS: The analysis was based on data from the 100-Car Naturalistic Driving Study, collected by the Virginia Tech Transportation Institute. The 100-Car study contains approximately 1.2 million vehicle miles of driving and 43,000 h of data collected from 108 primary drivers. In order to identify overtaking maneuvers from a large sample of driving data, an algorithm to automatically identify overtaking events was developed. The lead vehicle and minimum time to collision (TTC) at the start of lane change events was identified using radar processing techniques developed in a previous study. The lane change identification algorithm was validated against video analysis, which manually identified 1,425 lane change events from approximately 126 full trips. RESULTS: Forty-five drivers with valid time series data were selected from the 100-Car study. From the sample of drivers, our algorithm identified 326,238 lane change events. A total of 90,639 lane change events were found to involve a closing lead vehicle. Lane change events were evenly distributed between left side and right side lane changes. The characterization of lane change frequency and minimum TTC was divided into 10 mph speed bins for vehicle travel speeds between 10 and 90 mph. For all lane change events with a closing lead vehicle, the results showed that drivers change lanes most frequently in the 40-50 mph speed range. Minimum TTC was found to increase with travel speed. The variability in minimum TTC between drivers also increased with travel speed. CONCLUSIONS: This study developed and validated an algorithm to detect lane change events in the 100-Car Naturalistic Driving Study and characterized lane change events in the database. The characterization of driver behavior in lane change events showed that driver lane change frequency and minimum TTC vary with travel speed. The characterization of overtaking maneuvers from this study will aid in improving the overall effectiveness of FCW systems by providing active safety system designers with further understanding of driver action in overtaking maneuvers, thereby increasing system warning accuracy, reducing erroneous warnings, and improving driver acceptance.

Analysis of Driver Evasive Maneuvering Prior to Intersection Crashes Using Event Data Recorders.

OBJECTIVE: Intersection crashes account for over 4,500 fatalities in the United States each year. Intersection Advanced Driver Assistance Systems (I-ADAS) are emerging vehicle-based active safety systems that have the potential to help drivers safely navigate across intersections and prevent intersection crashes and injuries. The performance of an I-ADAS is expected to be highly dependent upon driver evasive maneuvering prior to an intersection crash. Little has been published, however, on the detailed evasive kinematics followed by drivers prior to real-world intersection crashes. The objective of this study was to characterize the frequency, timing, and kinematics of driver evasive maneuvers prior to intersection crashes. METHODS: Event data recorders (EDRs) downloaded from vehicles involved in intersection crashes were investigated as part of NASS-CDS years 2001 to 2013. A total of 135 EDRs with precrash vehicle speed and braking application were downloaded to investigate evasive braking. A smaller subset of 59 EDRs that collected vehicle yaw rate was additionally analyzed to investigate evasive steering. Each vehicle was assigned to one of 3 precrash movement classifiers (traveling through the intersection, completely stopped, or rolling stop) based on the vehicle's calculated acceleration and observed velocity profile. To ensure that any significant steering input observed was an attempted evasive maneuver, the analysis excluded vehicles at intersections that were turning, driving on a curved road, or performing a lane change. Braking application at the last EDR-recorded time point was assumed to indicate evasive braking. A vehicle yaw rate greater than 4° per second was assumed to indicate an evasive steering maneuver. RESULTS: Drivers executed crash avoidance maneuvers in four-fifths of intersection crashes. A more detailed analysis of evasive braking frequency by precrash maneuver revealed that drivers performing complete or rolling stops (61.3%) braked less often than drivers traveling through the intersection without yielding (79.0%). After accounting for uncertainty in the timing of braking and steering data, the median evasive braking time was found to be between 0.5 to 1.5 s prior to impact, and the median initial evasive steering time was found to occur between 0.5 and 0.9 s prior to impact. The median average evasive braking deceleration for all cases was found to be 0.58 g. The median of the maximum evasive vehicle yaw rates was found to be 8.2° per second. Evasive steering direction was found to be most frequently in the direction of travel of the approaching vehicle. CONCLUSIONS: The majority of drivers involved in intersection crashes were alert enough to perform an evasive action. Most drivers used a combination of steering and braking to avoid a crash. The average driver attempted to steer and brake at approximately the same time prior to the crash.

Population distributions of time to collision at brake application during car following from naturalistic driving data.

PROBLEM: Forward collision warning (FCW) systems are designed to mitigate the effects of rear-end collisions. Driver acceptance of these systems is crucial to their success, as perceived "nuisance" alarms may cause drivers to disable the systems. In order to make customizable FCW thresholds, system designers need to quantify the variation in braking behavior in the driving population. The objective of this study was to quantify the time to collision (TTC) that drivers applied the brakes during car following scenarios from a large scale naturalistic driving study (NDS). METHODS: Because of the large amount of data generated by NDS, an automated algorithm was developed to identify lead vehicles using radar data recorded as part of the study. Using the search algorithm, all trips from 64 drivers from the 100-Car NDS were analyzed. A comparison of the algorithm to 7135 brake applications where the presence of a lead vehicle was manually identified found that the algorithm agreed with the human review 90.6% of the time. RESULTS: This study examined 72,123 trips that resulted in 2.6 million brake applications. Population distributions of the minimum, 1st, and 10th percentiles were computed for each driver in speed ranges between 3 and 60 mph in 10 mph increments. As speed increased, so did the minimum TTC experience by drivers as well as variance in TTC. Younger drivers (18-30) had lower TTC at brake application compared to older drivers (30-51+), especially at speeds between 40 mph and 60 mph. DISCUSSION: This is one of the first studies to use large scale NDS data to quantify braking behavior during car following. The results of this study can be used to design and evaluate FCW systems and calibrate traffic simulation models.

Improved method for roadside barrier length of need modeling using real-world trajectories.

The 2011 AASHTO Roadside Design Guide (RDG) contains perhaps the most widely used procedure for choosing an appropriate length of need (LON) for roadside barriers. However, this procedure has several limitations. The procedure uses a highly simplified model of vehicle departure, and the procedure does not allow designers to specify an explicit level of protection. A new procedure for choosing LON that addresses these limitations is presented in this paper. This new procedure is based on recent, real-world road departure trajectories and uses this departure data in a more realistic way. The new procedure also allows LON to be specified for a precisely known level of protection - a level which can be based on number of crashes, injury outcomes or even estimated crash cost - while still remaining straightforward and quick to use like the 2011 RDG procedure. In this analysis, the improved procedure was used to explore the effects of the RDG procedure's assumptions. LON recommendations given by the 2011 RDG procedure were compared with recommendations given by this improved procedure. For 55 mph roads, the 2011 RDG procedure appears to lead to a LON sufficient to intercept between 80% and 90% of right-side departures that would otherwise strike a hazard located 10 m from the roadway. For hazards closer than 10 m, the 2011 RDG procedure intercepts progressively higher percentages of real-world departures. This suggests the protection level provided by the 2011 RDG procedure varies with the hazard offset, becoming more conservative as the hazard moves closer to the roadway. The improved procedure, by comparison, gives a consistent protection level regardless of hazard location.

Comparison and validation of injury risk classifiers for advanced automated crash notification systems.

OBJECTIVE: The odds of death for a seriously injured crash victim are drastically reduced if he or she received care at a trauma center. Advanced automated crash notification (AACN) algorithms are postcrash safety systems that use data measured by the vehicles during the crash to predict the likelihood of occupants being seriously injured. The accuracy of these models are crucial to the success of an AACN. The objective of this study was to compare the predictive performance of competing injury risk models and algorithms: logistic regression, random forest, AdaBoost, naïve Bayes, support vector machine, and classification k-nearest neighbors. METHODS: This study compared machine learning algorithms to the widely adopted logistic regression modeling approach. Machine learning algorithms have not been commonly studied in the motor vehicle injury literature. Machine learning algorithms may have higher predictive power than logistic regression, despite the drawback of lacking the ability to perform statistical inference. To evaluate the performance of these algorithms, data on 16,398 vehicles involved in non-rollover collisions were extracted from the NASS-CDS. Vehicles with any occupants having an Injury Severity Score (ISS) of 15 or greater were defined as those requiring victims to be treated at a trauma center. The performance of each model was evaluated using cross-validation. Cross-validation assesses how a model will perform in the future given new data not used for model training. The crash ΔV (change in velocity during the crash), damage side (struck side of the vehicle), seat belt use, vehicle body type, number of events, occupant age, and occupant sex were used as predictors in each model. RESULTS AND CONCLUSIONS: Logistic regression slightly outperformed the machine learning algorithms based on sensitivity and specificity of the models. Previous studies on AACN risk curves used the same data to train and test the power of the models and as a result had higher sensitivity compared to the cross-validated results from this study. Future studies should account for future data; for example, by using cross-validation or risk presenting optimistic predictions of field performance. Past algorithms have been criticized for relying on age and sex, being difficult to measure by vehicle sensors, and inaccuracies in classifying damage side. The models with accurate damage side and including age/sex did outperform models with less accurate damage side and without age/sex, but the differences were small, suggesting that the success of AACN is not reliant on these predictors.