The environmental perception of unmanned vehicles is a vital technology for automatic driving. The usage of three-dimensional (3D) LiDAR for obstacle detection becomes a popular research topic. In this paper, we first introduce the classification of obstacle detection methods for an unmanned vehicle according to different sensors. The basic principle of obstacle detection based on 3D LiDAR is then introduced in detail along with an analysis of the traditional method of obstacle detection using 3D LiDAR. Deep learning is an important method for two-dimensional object detection and classification. We analyze the characteristics of the 3D LiDAR point clouds and the challenges of deep learning for point clouds. Finally, we analyze the research status and development trend of deep learning in the point cloud obstacle detection application and introduce relevant datasets in the field of automatic driving, such as KITTI and ApolloScape.