Evaluation of intersection crashes using naturalistic driving data through the lens of future I-ADAS.

OBJECTIVE: Intersection advanced driver assistance systems (I-ADAS) with the capability to detect possible collisions and perform evasive braking have the potential to reduce the number of intersection crashes. However, these systems will encounter many challenges caused by the complexity of real-world driving conditions. The purpose of this study is to use real-world naturalistic driving data to conduct an initial exploration of the potential challenges for future I-ADAS in straight crossing path (SCP), left turn across path/lateral direction (LTAP/LD), and left turn across path/opposite direction (LTAP/OD) crash configurations. METHODS: Intersection crashes were selected from the Second Strategic Highway Research Program (SHRP 2) Naturalistic Driving Study. The SHRP 2 dataset includes front-facing, driver-facing, rear-facing, and a hands/feet-facing video and vehicle speed, steering, accelerator, and brake time-series data. This data was reviewed to understand driver sightline obstructions, driver distractions, and initiation of driver responses. The estimated time to collision (TTC) from the precipitating event, defined as when either vehicle entered the intersection without the right-of-way, was computed based on the distance to the impact point divided by the current velocity of the subject vehicle. RESULTS: The median impact speed was 18.0 km/h for SCP and LTAP/LD crashes and 16.1 km/h for LTAP/OD crashes. The median TTC from the precipitating event was 1.35 s for SCP and LTAP/LD crashes and 1.44 s for LTAP/OD crashes. For SCP crashes, the three main sightline obstruction scenarios were slower vehicles traveling in the same direction waiting to turn right, vehicles in the closer crossing lane, and a parked truck. For LTAP/OD crashes, the sightline obstruction was often oncoming vehicles in a closer lane blocking the view of another vehicle. CONCLUSION: Sightline obstructions could present a challenge for future I-ADAS to activate in SCP, LTAP/LD, and LTAP/OD crashes. This study utilized naturalistic driving data to complete a comprehensive analysis of intersection crashes, including driver distractions, evasive maneuvers, and sightline obstructions that can assist in the development of I-ADAS. This analysis is not possible with police-reported crash data only, which does not contain necessary details on the driver and surrounding environment.

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.

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.

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.

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.