Modeling individual differences in risk feeling of autonomous driving behavior with a prediction error

Abstract

Studies on autonomous driving systems are being conducted to realize a safe and secure mobile society. According to these studies, the safety of the system and the driver’s feeling of safety often do not match. To enable automatic driving control according to the driver's comfort, previous researchers introduced a risk feeling index to quantify the driver’s feeling of safety. These studies, however, did not consider the effects of the drivers’ individual differences for the risk feeling index. We hypothesized that the risk perception was formed by the difference between the actual autonomous driving behavior and the driver's prediction for it. The driver's risk feeling is expected to vary depending on his/her prior prediction of autonomous driving behavior. We considered that this difference, which we term as the prediction error, explains an individual difference in the risk perception. The purpose of this study is to reveal the effect of the prediction error on the risk perception. We conducted an experiment to investigate the effect of these factors on the risk feeling in the overtaking scene using a driving simulator. We obtained two types of subjective evaluations to examine the effects of the prediction errors on risk perception. One is perceived risk based on the driver’s original prediction, and the other is perceived risk based on the predictions that we manipulated by conducting a learning session during the experiment. The results suggest that the risk is perceived based on both the experimentally manipulated prediction and the individual driving characteristic of anxiety while driving. These findings will enable the development of a secure automatic driving system that suits each driver by controlling the system appropriately based on the driving characteristic of the driver.

Keywords

• design
• autonomous driving
• risk feeling
• individual differences
• prediction error