上海交通大学学报

上海交通大学学报（自然版） ›› 2018, Vol. 52 ›› Issue (9): 1120-1124.doi: 10.16183/j.cnki.jsjtu.2018.09.017

• 学报（中文） • 上一篇    下一篇

基于直线特征的机器人自主定位方法

杨晶东，彭坤，顾浩楠，师艳伟   

  1. 上海理工大学 光电信息与计算机工程学院, 上海 200093
  • 通讯作者: 杨晶东（1973-），男，黑龙江省齐齐哈尔市人，博士, 副教授，研究方向为智能机器人与计算机视觉. 电话(Tel.): 021-55271064; E-mail:eerfriend@yeah.net.
  • 基金资助:
    国家自然科学基金(61374039),上海市自然科学基金(15ZR1429100)，沪江基金(C14002)资助项目

Autonomous Localisation Method Based on Linear Feature for Robots

YANG Jingdong,PENG Kun,GU Haonan,SHI Yanwei   

  1. School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China

摘要： 针对迭代最近点(ICP)算法容易造成局部最优和正态分布变换(NDT)算法在长距离导航中会出现较大转角偏差的问题，提出一种基于直线特征的ICP自主定位(CICP)方法.利用激光扫描点集与其拟合直线的协方差函数关系，以快速匹配扫描点间的直线特征，通过扫描匹配点集与拟合直线的闭环误差获得当前时刻全局环境中相邻匹配扫描点间的位姿，并通过室内环境下的自主定位实验验证其可靠性.结果表明，与传统的ICP算法和NDT算法相比，CICP算法具有较好实时性和较高的定位精度，能够降低ICP算法的矩阵运算量，提高机器人的自定位精度.

关键词: 自主机器人, 全局定位, 迭代最近点, 扫描匹配, 滤波

Abstract: The convergence for iterative closest point (ICP) algorithm depends much on the inputs, which often results in local optimum. And normal distribution transform (NDT) algorithm usually acquires more scans in scan matching, and the large angular deviation occurs during long-term autonomous navigation for robots. An improved ICP self-localization algorithm based on line feature is proposed in the paper. We derive functional relations of covariance matrix between matching scans and their fitting lines to scan and match line features rapidly. And the robot updates global poses in terms of closed form correlations. We test the reliability of CICP algorithm in indoor locating experiments. The experiments show that CICP algorithm, compared with ICP or NDT, owns higher real-time and location accuracy, and can reduce matrix computation and improve self-localization precision efficiently.

Key words: autonomous robots, global localization, iterative closest point (ICP), scan matching, filtering

中图分类号: