Estimating probability of automobile accident from driver's reaction time under different arousal states

The main aim of this study is to quantify the likelihood of automobile accidents by collision. Specifically, we seek a way to compute the probability of collision (POC) from the driver's reaction time (RT) distribution. Since RT distributions change with driver arousal states, we propose [1] classification of RT data by arousal states, which are measured by eye-opening rate (EOR), [2] estimation of two RT distribution functions from the data classified with low and high EORs, and [3] computation of POC separately from these distribution functions. To examine the validity of our proposed method, we conducted an experiment using a driving simulator. We fitted ex-Gaussian functions to RT data to estimate their distributions. The results showed that the fitting was quite successful for all data, whereas there were marked individual differences in the way POC changed with EOR.