Lidar-based Traversable Region Detection in Off-road Environment

Traversable region detection is a fundamental problem in the field of autonomous driving. This paper proposes a fast method to detect obstacles and obtain the traversable region in the off-road environment. Our method takes advantage of both radial features and transverse features based on the high definition of 3D Lidar points. First, we manage Lidar points by scanning lines and sectors in the polar system at the same time. Then the most obstacles can be quickly detected by using radial features in the polar system. For the false detection, transverse features are applied to verify the results. Finally, the constrained region within the nearest obstacle points in each sector defines the traversable region around the vehicle. Our method can detect positive obstacles, negative obstacles, and hanging obstacles in real-time. The experimental results show the robustness and accuracy of the proposed method in different kinds of off-road environments.