地面无人车辆越野环境多要素合成可通行区域检测

发布时间：2018-10-23 16:16

【摘要】：与结构化城市环境中影响地面无人车辆通过性安全的主要是地面以上的障碍物不同,越野环境中感知系统需要考虑包括地面起伏、凸起正障碍物、负障碍、水体、坡面等多种环境要素。为了使地面无人车辆能够完成越野环境下的安全行驶,感知系统必须具备局部感知这些环境要素的能力,或者具备综合环境信息要素以获得可通行区域的能力。本文提出了一套针对以上这些环境要素的越野环境多要素合成可通行区域检测系统。首先,分析了32线激光雷达几何模型,并基于PCL点云库实现了三维激光点云数据的获取,提出了一种简单32线激光雷达雷达与地面无人车辆安装参数标定方法。其次,采用极坐标栅格地图将车辆周围局部环境激光点云栅格化,对地面凸起正障碍物、负障碍、水体、坡面等多种环境要素进行了分析和提取。(1)通过对极坐标栅格地图扇形块内地面候选点进行直线拟合计算栅格地面参考高度,并计算栅格激光点云与地面参考高度的最大高度差判断栅格状态,实现地面点和正障碍点的分离,同时通过计算地面栅格内激光点高度的方差对地面栅格进行了不平度分析;(2)通过计算相邻层激光线的径向间隔进行负障碍检测;(3)通过对极坐标栅格地图中没有激光点的未知状态栅格进行最大、最小高度填充建立极坐标雷达图,并沿极坐标雷达图圆周方向计算栅格间高度差检测道路边缘障碍;(4)通过回波强度滤波检测水体区域;(5)通过对栅格内激光点采用随机采样一致性方法提取平面特征并结合坡面约束条件实现坡面检测。最后,将极坐标栅格地图正障碍物栅格、负障碍物栅格、道路边缘栅格、水体障碍物栅格以及坡面栅格统一标记为障碍物栅格,基于连通区域标记方法对障碍物栅格进行聚类扩展;通过分析同一角度位置相邻激光扫描线对正障碍物、负障碍物、地面的测量情况,对极坐标栅格地图进行可通行区域扩展,实现可通行区域提取。真实越野环境下的实车实验表明,此系统能够满足地面无人车辆在中等起伏越野环境低速安全行驶的要求。
[Abstract]:Different from the obstacles above the ground that affect the safety of ground-based unmanned vehicles in the structured urban environment, the sensing system in off-road environment needs to be considered including ground undulation, raised positive obstacles, negative obstacles, water bodies, etc. Slope and other environmental factors. In order to make the ground unmanned vehicle complete the safe driving in the off-road environment, the sensing system must have the ability to locally perceive these environmental elements, or the ability of synthesizing the environmental information elements to obtain the passable area. In this paper, we present a set of cross-country environment multi-factor synthesis passable area detection system for these environmental elements. Firstly, the geometry model of 32-line lidar is analyzed, and the 3D laser point cloud data is acquired based on PCL point cloud library. A simple calibration method of installation parameters between 32-line lidar and ground unmanned vehicle is proposed. Secondly, the polar grid map is used to grid the local environment of the vehicle by laser point cloud, which raises positive obstacles, negative obstacles and water bodies on the ground. Many environmental factors such as slope are analyzed and extracted. (1) the reference height of grid ground is calculated by straight-line fitting of candidate points in sector block of polar grid map. The maximum height difference between grid laser point cloud and ground reference height is calculated to judge grid state, and the separation of ground point and positive obstacle point is realized. At the same time, by calculating the variance of the laser point height in the ground grid, the irregularity of the ground grid is analyzed; (2) the negative obstacle detection is carried out by calculating the radial interval of the laser line in the adjacent layer; (3) by the method of no excitation in the polar grid map, the negative barrier is detected by calculating the radial interval of the laser line in the adjacent layer. The unknown state grid of the spot is maximum, Minimum height fill to establish polar radar map, Along the circumferential direction of the polar radar map, the road edge obstacle is detected by calculating the height difference between grids; (4) the water area is detected by echo intensity filter; (5) the plane feature is extracted by random sampling consistency method to the laser points in the grid. And combined with the slope constraints to achieve slope detection. Finally, the polar grid map positive barrier grid, negative barrier grid, road edge grid, water barrier grid and slope grid are labeled as obstacle grid. By analyzing the measurement of positive obstacle, negative obstacle and ground by analyzing the adjacent laser scanning line at the same angle, the accessible area of polar grid map is extended. The passable area can be extracted. The real vehicle experiments in off-road environment show that the system can meet the requirements of low speed and safety driving of ground unmanned vehicles in medium undulating off-road environment.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;U463.6

【相似文献】