Real-World Road Experiment-Based Design of Human-Aligned Autonomous Vehicle Motion from Psychophysiological Responses

The transition from driver to passenger in autonomous vehicles (AVs) introduces novel human factors challenges, particularly concerning motion comfort. Traditional comfort standards, derived from whole-body vibration models, are inadequate for characterizing motion comfort in AVs. This study introduces a data-driven framework that establishes quantitative AV motion planning thresholds derived from passenger psychophysiological responses. Through a rigorously designed real-road experiment involving 30 young, healthy participants and 765 valid maneuvers, we synchronized vehicle kinematics data with surface electromyography (sEMG) signals from passengers. We quantified the level of motion-induced discomfort via sEMG and semantically mapped it to three discrete levels. Key findings reveal that during predictable turns, passengers tolerated lateral accelerations up to 4.20 m/s 2 , far exceeding traditional standards. Crucially, jerk was identified as a critical disturbance factor; for instance, within a moderate acceleration range (2.37–2.73 m/s 2 ), a jerk exceeding 5.14 m/s 3 elevated discomfort levels. A classification and regression tree model, utilizing these parameters, achieved a prediction accuracy of 82%–85%. These data-driven thresholds provide ergonomic guidelines for the design of passenger-centric AV motion planning algorithms, prioritizing human psychological and physiological wellbeing.