Cooperative control for lane departure prevention based on model predictive control and gated recurrent unit model

Abstract

Human-machine conflict has a significant impact on driving safety, posing a vital challenge in the design of lane departure prevention (LDP) systems. To address the issue, this paper proposes a driver-intelligent vehicle cooperative steering torque assistance control strategy. The lane departure decision-making module based on the gated recurrent unit (GRU) is used to predict the lateral deviation of the vehicle and to make real-time decisions regarding the switching of the model predictive control (MPC) based assistance controller. Next, the conflict performance between the MPC lane keeping and conflict reduction (MPC-LKCR) controller’s torque and the driver’s torque is added to the optimization objective of the MPC lane keeping (MPC-LK) controller, while the lane keeping performance is continually retained. That is because a shared factor based on the fuzzy model is designed with the ability to adjust the assistance torque within the MPC-LKCR controller according to the driver’s intention. Finally, after the overall optimization of the MPC-LKCR controller, the final torque after the superposition of driver and assistance torque acts on the steering column to realize the human-machine cooperative steering control. The driving data from 52 drivers were collected to train the GRU model offline. The proposed strategy was simulated and analyzed under different driving scenarios, and hardware-in-the-loop experiments were completed on a driving simulator to validate it. Hardware-in-the-loop results show that the average conflict intensity and conflict time ratio are reduced by 23.8% and 34.4% under the MPC-LKCR controller compared to the MPC-LK controller. The strategy not only accomplishes the task of vehicle lane departure but also effectively reduces the time and intensity of human-machine conflicts.