Lane-Keeping Assistance System Based on Fast Lane Line Detection Network and Model Predictive Control

The lane-keeping assistance system is one of the essential active safety technologies for autonomous vehicles. This paper introduces a lane-keeping system based on a fast lane line detection network and model predictive control. A camera is utilized to capture the visual data of the road ahead and analyze it to extract lane lines. Based on this, a model predictive lateral controller is developed to obtain the appropriate steering angle required for lane keeping. A joint simulation using CarSim and Simulink was conducted to verify the feasibility of the lane-keeping system. At low speeds, the lateral deviation of the vehicle when steering is a maximum of 0.2 m. The steering deviation at high speeds does not exceed 0.45 m to ensure the vehicle's stability. The results indicate that the vehicle deploying the lane-keeping assistance system in the paper can drive along the lane center trajectory smoothly and accurately.