A LiDAR-Based Obstacle-Detection Framework for Autonomous Driving

This paper presents a LiDAR-based obstacle detection framework for autonomous driving. The real-time processing of LiDAR data is achieved by using robust grid-based ground extraction and fast rule-based points clustering. The framework is evaluated by LiDAR data from different scenarios, and compared to classical methods. The results demonstrate that the proposed ground extraction method can remove flat, non-flat roads and slope-changing terrains, and the proposed detection framework requires less computational time while achieving comparable detection performance.