摘要
路口是典型的城市交通场景之一,为了能够比较准确的刻画路口交通场景,为无人车运动规划提供充分可靠的场景信息,提出了创建基于栅格图拼接的多层次路口地图的方法。上述方法首先以栅格图为输入信息,依据ICP(iterative closest point)算法,离线建立路口静态地图,再以此作为先验信息,使用无人车行驶过程中的实时感知数据对路口静态地图进行在线扩充与更新,并且分别依据激光雷达和相机的感知结果实时更新动态障碍物列表和交通标志等信息,解决了路牙、树木等非结构化障碍物信息难以刻画的问题。所提方法实现简单,高效,仿真结果表明了方法的有效性。
Intersection is a common urban traffic scene. In order to obtain an accurate description of the intersection for motion planner of the autonomous vehicle,this paper proposed an algorithm of mapping based on grids and ICP algorithm. Firstly,a grids based off-line map was built via pose conversion and ICP based registration,and this can serve as priori. Secondly,we updated the map based on on-line grids as the off-line map building way,and maintained the dynamic obstacle list and traffic lights information based on perception results. The proposed algorithm can solve the trouble that curbstones and trees are difficult to describe. And the simulations show the effectiveness of the proposed algorithm.
作者
屈盼让
薛建儒
於二军
孙少华
QU Pan-rang;XUE Jian-ru;YU Er-jun;SUN Shao-hua(Aeronautics Computing Technique Research Institute,Xi'an Shanxi 710065,China;Institute of Artificial Intelligence and Robotics,Xi'an Jiaotong University,Xi'an Shanxi 710049,China)
出处
《计算机仿真》
北大核心
2019年第1期342-347,共6页
Computer Simulation
基金
国家自然科学基金(91320301)
航空科学基金(2016ZC31004)
关键词
路口
无人车
栅格
Intersection
Autonomous vehicles
Grid
