• Traditional metrics like maximum acceleration and jerk are insufficient to accurately evaluate passenger comfort in autonomous vehicles.
• Motion complexity features significantly improve the explanation of passenger comfort compared to simple motion characteristics.
• A real-time passenger comfort measurement using EMG and stepwise regression enables seamless data collection in naturalistic driving studies.
• A machine learning-based method using only vehicle motion information can accurately estimate passenger comfort in real-time, highlighting the importance of motion complexity for future AV design.