Navigation with uncertain map data for automated vehicles

Safe navigation is an indispensable component of automated vehicles. Due to limited computational resources and sensor range, the planning problem is often divided into several sub-problems. A global planner first plans a route based on map information without considering local obstacles. The local planner then plans a trajectory based on the given route and the local environment model. HD maps are often used for this purpose. The generation of such maps is costly and the approach is poorly scalable.

In our work, we developed a concept for automated navigation in unstructured environments based on freely available map information. First, a route from the starting point to the destination is planned based on an OpenStreetMap map using the A* algorithm. Subsequently, a local sampling-based trajectory planning is performed considering the inaccurate route information. A local occupancy grid is used as the environment model, which is created from the 3D lidar point cloud of our test vehicle.