Data-driven Human Modeling based on Temporal Information and Nonlinear Model Predictive Control for Adaptive Cruise Control Reducing Motion Sickeness

This paper proposes a novel nonlinear system modeling technique using a data-driven approach with temporal information for adaptive cruise control (ACC) focused on reducing motion sickness through model predictive control (MPC). We develop an approximated human model from real-world data to enhance motion prediction and integrate it into MPC's cost function and constraints, emphasizing tracking performance, control effort, and motion sickness reduction. Using the ISO 2631-1:1977 standard, motion sickness is evaluated with the motion sickness dose value (MSDV) in the longitudinal axis of human motion. Validated through MATLAB/Simulink simulations, our method improves low-frequency human motion prediction accuracy, reduces RMSE and maximum error by 5.23% and 23.4%, lowers MSDV by 17% in ACC scenarios, and increases car-following performance by 42.3% compared to previous methods.