Adaptive Autonomous Driving Style Considering Human Trust

The lack of driver's trust in autonomous driving is likely to lead to more unnecessary takeovers, thereby increase the driver's workload and diminishing effectiveness of driving automation. This study introduces a Partially Observable Markov Decision Process (POMDP) model to tailor the autonomous driving behavior, aiming to enhance driver trust and forestall an increase in workload. Data retrieved from a real-car-based driving simulator were employed to accomplish the parameter estimation required for the model, followed by the computation of the optimal policy, which was meticulously analyzed. Subsequent validation experiments proves that the conclusion policy can control the driver state to high trust and low workload. By adjusting the driving style to improve the driver's trust, this study contributes to improving the practicality of autonomous driving.