This study proposes an autonomous driving system that integrates LiDAR-based drivable-area perception with GPS reliability assessment to enable stable driving even when GPS reliability deteriorates or is lost due to environmental factors. The system e...
This study proposes an autonomous driving system that integrates LiDAR-based drivable-area perception with GPS reliability assessment to enable stable driving even when GPS reliability deteriorates or is lost due to environmental factors. The system evaluates GPS reception status in real time using GNSS fix type, the number of satellites, and HDOP, and defines three driving modes accordingly: NORMAL, DEGRADED, and DISABLE. In the NORMAL mode, polynomial candidate paths are generated in a Frenet frame with respect to a pre-built global reference path, and a Linear Quadratic Regulator (LQR) controller tracks the selected reference path. In the DEGRADED mode, road–grass boundaries and structural boundaries are extracted using LiDAR intensity and height information; these boundaries are transformed into boundary obstacles in the Frenet frame and incorporated into a collision-aware cost evaluation so that the selected path reduces the risk of boundary collisions. In the DISABLE mode, GPS information is not used; instead, a centerline is generated from the left and right boundaries of the LiDAR-perceived drivable area and used as the reference path. Path tracking is performed in a LiDAR-based relative coordinate frame by computing the lateral offset to the generated centerline and applying the same LQR structure without relying on GPS. The proposed system is validated on a real vehicle platform equipped with a MID-360 LiDAR and a GNSS/INS sensor suite in a non-structured outdoor environment containing road–grass and structural boundaries, and experimental results confirm stable path tracking and collision-free driving across GPS reliability variations and mode transitions.