可移动坐标框架的智能车辆同时定位与建图  

Movable Coordinate Frame Based Simultaneous Localization and Mapping of Intelligent Vehicle

作　　者：钱钧[1] 杨汝清[1] 杨明[2] 伍舜喜[2] 王春香[1] 

机构地区：[1]上海交通大学机器人研究所,上海200240 [2]上海交通大学自动化系,上海200240

出　　处：《上海交通大学学报》2009年第6期857-861,共5页Journal of Shanghai Jiaotong University

基　　金：欧盟第六框架CyberCars-2资助项目(EC-FP6-IST-028062);上海市科委登山计划资助项目(062107035)

摘　　要：针对智能车辆在进行未知环境探测时,为实现完全自主需要解决同时定位与建图问题.基于扩展卡尔曼滤波器方法中较大的车辆方向角方差导致明显的不一致性,提出一种可移动坐标框架方法.在检测到新的特征时,将参考框架移动到车辆位姿处,使特征的初始方差与车辆位姿估计误差无关,同时车辆方向角方差由于仅受局部环境影响而始终保持为较小值,从而获得较好的一致性估计性能.人工环境和自然环境中的实验结果表明,该方法可以获得精确的车辆轨迹估计和环境特征地图.When an intelligent vehicle enters an unknown environment, it must solve the problem of simultaneous localization and mapping in order to pursue high autonomy. Extended Kalman filter based method is subject to inconsistency when the variance of vehicle＇s heading angle is big. Based on it, this paper described a movable coordinate frame based method, which transforms the reference frame to where the vehicle is once a new feature is discovered. This transformation makes sure that each feature has small variance at the initialization. In addition, the heading angle keeps small uncertainty since it is only affected by local environment. So the consistency of estimation could be improved. The experimental results of artificial and natural environments show that high quality map can be built as well as accurate vehicle trajectory.

关 键 词：智能车辆 激光雷达 同时定位与建图 扩展卡尔曼滤波器 一致性 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]