A New SLAM Algorithm with Key-Frame Estimation and Local Map Upgrade Scheme

带关键帧和可靠平面表示的激光定位算法

School of Computer Science, Sichuan University, Chengdu 610000, China

Received: 23 April 2020

Abstract

Simultaneous Localization and Mapping(SLAM) is the most important tool in creating map and auto-navigation, which is an indispensable link for pilotless automobile. The current SLAM algorithms suffer from unreliable feature matching and large registration error. To reduce those deficiencies, we proposed a key-frame estimation and local map upgrade scheme, which include the following 3 parts:1) Local map matching strategy; 2) Local map updating scheme; 3) Key-frame selecting scheme. Experimental results proved that our scheme improved the performance of current localization methods.

摘要

激光实时定位与地图构建（simultaneous localization and mapping，SLAM）是创建地图和实时导航的重要手段之一，也是无人驾驶不可缺少的一环。针对目前激光SLAM算法对特征匹配可靠性不足、配准误差较大等问题，基于平面拟合算法，提出一种局部地图改进和关键帧估计的方案。具体包括以下3个方面：①局部地图匹配规则和平面表示方法；②局部地图更新方案；③关键帧筛选机制。该方法解决了目前激光定位方案中缺乏关键帧估计，以及局部地图中平面多向性问题，实验表明该方法使得局部地图能够保留更具多样性的激光雷达帧，同时平面表示和匹配也更为可靠。