李 宁， 郭江华， 蓝 伟

(中国北方车辆研究所，北京 100072)

基于多线激光雷达的非结构化道路感知技术研究

李 宁， 郭江华， 蓝 伟

(中国北方车辆研究所，北京 100072)

为解决无人平台在越野条件下能够感知识别非结构化道路的问题，采用基于多线激光雷达的测距传感器检测技术，通过对激光点云数据做栅格化处理，并进行距地高度、高度差和梯度差等特征统计及聚类分析、目标跟踪，实现了对道路可通行区域、障碍物、坡道和动态车辆等的检测.结果表明，相比于相机，采用基于多线激光雷达的三维测距传感器检测技术，可以更直接地获取障碍物分布的位置信息，能够为无人平台的自主导航提供局部栅格地图.

非结构化道路；障碍物检测；环境感知；多线激光雷达；栅格地图

Abstract：To solve the accessible area detection under the unstructured road for unmanned ground vehicle, environment perception based on 3D-lidar is proposed. First, point clouds obtained from 3D-lidar are mapped to the grid-based map. Then statistics on features such as absolute height of point clouds, difference of height between adjacent point clouds and gradient difference are performed. After that, cluster analysis and object tracking will be used for road accessible area detection, obstacle detection, slope detection and dynamic vehicle detection and tracking. The results show that compared with the camera based method, the position of obstacles distribution can be directly obtained by point clouds based on 3D-lidar and a local grid-based map will be provided for autonomous navigation of unmanned ground vehicle.

Keywords: unstructured road；obstacle detection；environment perception；3D-lidar；grid-based map

无人平台自主导航技术[1-5]是近年来研究的热点，主要包括环境感知、规划决策、运动控制等关键技术.其中，环境感知[6-7]是无人平台进行自主导航的先决条件.由于非结构化道路环境缺少车道线等明显的道路标识，需要通过提取可通行区域作为无人平台局部路径规划的依据.依靠相机对道路纹理、颜色进行路面提取的算法，受光照影响较大，且不同的道路环境下，特征差异明显，算法通用性不强.激光雷达作为主动三维测距传感器，可以直接给出障碍物的距离信息，且具有精度高，鲁棒性强的特点，在包括Cargo[8]、Boss[9]等在内的多个先进无人平台上取得了成功应用.本研究针对多线激……