Efficient Lane Detection and Keeping for Autonomous Vehicles in Real-World Scenarios

In this study, we developed an auto-lane-keeping assist system to contribute to autonomous driving. The system will keep the vehicle in its designated lane to reduce road accidents. The main challenge of this research is to make the system suitable for real-world road conditions. To achieve an efficient lane-keeping system, we detected the lane lines automatically using canny edge detection and a sliding window algorithm. The proposed lane-keeping system has another stage: lane-keeping. The lane-keeping task includes the calculation of the offset value, meaning how much the vehicle deviates from its lane or if the vehicle is in the center line of the lane, and the steering angle correction of the vehicle to be in the lane. We use the Sliding Window Algorithm for the offset value calculation, which makes the detection of curved lane lines more accurate than the previous works and also performs well in various weather and lighting conditions. The proposed system’s performance is evaluated using the IROADS dataset in this work. We implemented lane keeping in both software and hardware models, introducing the AURIX TC-375 microcontroller for hardware integration, which enhances overall lane-keeping system output through the integration of detection and keeping processes.