Point Cloud Interpolation by RGB Image to Estimate Road Surface Profile for Preview Suspension Control
摘要
The growing prevalence of autonomous driving is expected to shift passengers’ attention from driving, increasing the demand for enhanced ride comfort. Studies addressing ride comfort have prominently explored active suspension control with recent research on preview suspension control using on-board sensors. The proposed systems often include LiDAR deployment at the front for high-precision road surface profiles. However, these systems often involve costly sensors such as LiDAR, making it impractical for on-board installation. Nonetheless, in recent autonomous vehicles, LiDAR tend to be mounted on the roof. It would be beneficial to leverage this LiDAR for preview control, the point cloud obtained from the roof has insufficient density to accurately perceive the unevenness on the road surface. To overcome the low-density issue in point cloud data obtained from less channels LiDAR, this study applies a supervised machine learning model, developed for autonomous driving, to estimate road surface profiles and enhance the precision of these estimations.