Rough Road Ahead Off-road Terrain Detection Through Enhanced Sensor Fusion

Abstract

Abstract This thesis examines the development of the Automatic Terrain Detection and Adaptation (ATDA) feature for off-road vehicles. The ATDA system is designed to improve vehicle safety, efficiency, and comfort in demanding off-road environments characterized by uneven terrain, diverse obstacles, and unpredictable road types and conditions. Traditional methods for detecting off-road terrain and obstacles, which rely on either image-based or point cloud-based models, have proven insufficient for accurate predictions. To overcome these limitations, this research integrates camera-based and point LiDAR-based methodologies through a sensor fusion approach. This integrated method is anticipated to yield more accurate results, significantly improving the vehicle’s navigation capabilities in complex off-road terrains. Aligned with the latest trends in sensor fusion for environmental perception in autonomous vehicles, this project prioritizes evaluating path traversability using static vehicle features rather than the traditional focus on obstacle density. This innovative approach enhances the vehicle’s ability to detect and assess off-road terrain and obstacles, thereby improving navigation safety and efficiency. With an emphasis on the integration of camera and LiDAR technologies for enhanced terrain and obstacle detection and classification.