Glance and deceleration based generated baseline and treatment cases

These data was generated through scenario generation using counterfactual simulations. A set of (N=44) rear-end crashes from Volvo Cars crash database was used as “seed crashes”. The evasive maneuver of the following vehicle was removed. A crash causation model using deceleration level (3.3 - 10.3 m/s2) and off-road glance durations (0 -6.6 s), together with a simple driver response model was added instead. Simulations were run to generate all possible combinations (N=44220). This is dataset Baseline.

Dataset Treatment ran exactly the same simulations, but with a simple crash avoidance system applied to the cases. The dataset was used as the ground truth in the illustration of the optimal sub-sampling that selects a subset of the ground truth to represent mean of the safety benefits.

The dataset include the simulation input and output (both the baseline and the treatment) for 44 cases. Each case has a CaseID which starts from 1 and end with 44. Three variables are the input of the simulations including eoff, acc and eoff_acc_prob. The variable eoff is the off-road overshot glance duration with unit of second. The variable acc is the acceleration of the driver for braking and the unit is in m/s2 (The acc is negative as driver brakes). Impact speed was the direct output from the simulations. Variable impact_speed0 is for the baseline scenario and impact_spee1 is for the treatment scenario. The injury risk is also included inside the dataset. The injury risk was calculated based on the injury risk function (Stigson et al., 2012). The variable injury_risk0 is for the baseline scenario and injury_risk1 is for the treatment scenario.