Naturalistic Driving Study Data for Machine Learning Model Training

Project Description This collection contains driving scenarios of interest to machine learning developers compiled from the The Second Strategic Highway Research Program (SHRP 2) Naturalistic Driving Study (NDS). The emphasis of this task was to provide data in a format that can reduce pre-processed requirements. The SHRP 2 machine learning training data consists of valid and invalid cases for eight trigger thresholds: Longitudinal deceleration: Longitudinal deceleration ≤ −0.65 g Longitudinal acceleration: Longitudinal acceleration ≥ 0.50 g Lateral acceleration: Lateral acceleration with |absolute value| ≥ 0.75 g ABS: Flag presence Lateral jerk: Lateral jerk with |absolute value| ≥ 1.0 g/s within any 0.8-s window Freeway deceleration: Decelerations on freeways (as determined by map functional class) ≥ 0.3 g, no minimum speed Yaw rate: Yaw rate oscillations larger than ±8 degrees/s and occurring within 0.75 s Swerve: Swerve maneuvers larger than 15 degrees/s^2 occurring within 2 s In each instance of a driving maneuver exceeding a threshold, video-based analysis was used to determine if the maneuver was associated with a safety-relevant event (e.g., crash or near-crash). Maneuvers associated with a safety-relevant event were deemed as valid, while those not associated with safety-relevant events were classified as invalid. Both valid and invalid cases should be included in training and evaluation data sets in order to develop robust methods to detect safety-relevant events. Data Request Scope This data collection includes: 1) Forward video and time series data for invalid triggers. 2) Forward video and time series data for crashes and near-crashes identified by valid triggers. 3) Binary and descriptive data tables of quality assessment performed for each time series data variable on a trip basis. Data Specification See attached data dictionaries for specifics. The data collection elements are as follows: Time Series Data Invalid Triggers Forward Video Invalid Triggers Event Details Invalid Triggers Quality Assessment Binary Output Invalid Trigger Trips Binary indicator of pass (1) or fail (0) for each assessment. Quality Assessment Descriptive Output Invalid Trigger Trips Time Series Data Valid Events Forward Video Valid Events Event Details Valid Events Event Details Valid Triggers Quality Assessment Binary Output Valid Event Trips Binary indicator of pass (1) or fail (0) for each assessment. Quality Assessment Descriptive Output Valid Event Trips Quality Assessment Map Spreadsheet describing all of the quality assessments completed for each time series variable.