LIO-SAM框架下的智能车辆SLAM算法优化与实现  

Optimization and Implementation of SLAM Algorithm for Intelligent Vehicles under LIO-SAM Framework

作　　者：张家鑫 田国富[1] 常天根 张森 Zhang Jiaxin;Tian Guofu;Chang Tiangen;Zhang Sen(Shenyang University of Technology,Shenyang 110870)

机构地区：[1]沈阳工业大学,沈阳110870

出　　处：《汽车技术》2024年第12期23-30,共8页Automobile Technology

摘　　要：为解决激光SLAM在局部地图构建时重复提取关键帧及回环检测中的无效回环问题,基于LIO-SAM框架,采用vector容器、Kd-tree最近邻搜索与VoxelGrid滤波器,避免当前帧附近关键帧的重复提取。在回环检测方面,引入基于扫描上下文与基于距离的回环检测算法,通过设置帧序列差异阈值筛选回环帧,减少在红绿灯等待或礼让行人场景中的回环检测次数。试验结果表明,与LIO-SAM相比,所提出算法在不影响建图精度的前提下,平均缩短当前帧附近关键帧提取时间31.39%,回环检测次数减少32.5%,显著提升计算效率和鲁棒性,为优化资源利用和算法性能提供了有效方法。To address the issues of redundant keyframe extraction during local map construction and invalid loop closures in loop detection for laser-based SLAM,this study adopts a method based on the LIO-SAM framework.It utilizes a vector container,Kd-tree nearest neighbor search,and VoxelGrid filter to avoid redundant extraction of keyframes near the current frame.For loop detection,algorithms based on scan context and distance-based methods are introduced.By setting a threshold for frame sequence differences,loop frames are screened to reduce the number of loop detections in scenarios such as waiting at traffic lights or yielding to pedestrians.Experimental results show that,compared to LIO-SAM,the proposed algorithm reduces the average keyframe extraction time near the current frame by 31.39%and the number of loop detections by 32.5%,without compromising mapping accuracy.This significantly enhances computational efficiency and robustness,providing an effective method for optimizing resource utilization and algorithm performance.

关 键 词：激光SLAM LIO-SAM 局部地图 关键帧提取 回环检测 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]