A Dual-Level Model Predictive Control based Driver Model for Curve Driving

To achieve autonomous driving, many advanced control methods have been conducted. This paper proposes a driver model coupling lateral and longitudinal control for curve driving based on dual-level model predictive control (MPC). During the curve driving, actual expert drivers can conduct vehicle’s lateral motion control in coordination with longitudinal control, so a safer and smoother driving performance can be achieved. In order to better reflect this behavior in the control model for autonomous driving, a dual-level MPC is applied to improve trajectory following ability on curve road. Firstly, a lateral following module based on two-wheel bicycle model and MPC is established. Then, the lateral decision is used as the input of another longitudinal control module based on MPC. In this way, longitudinal control module will ensure the vehicle keeps at a stable motion state during the whole curve. Finally, the effectiveness of the proposed method is verified by a simulated experiment. The simulation result corresponds with the skillful manner of the expert drivers.