FAST_LIO_SAM 融入后端优化的FASTLIO SLAM 系统前端：FAST_LIO2 后端：LIO_SAM

FAST_LIO_SAM

Front_end : fastlio2 Back_end : lio_sam

indoor

Videos : FAST-LIO-SAM Bilibili_link

Source code : FAST_LIO_SAM

Related worked

1.FAST-LIO2为紧耦合的lio slam系统，因其缺乏前端，所以缺少全局一致性，参考lio_sam的后端部分，接入GTSAM进行后端优化。

2.FAST_LIO_SLAM的作者kim在FAST-LIO2的基础上，添加SC-PGO模块，通过加入ScanContext全局描述子，进行回环修正,SC-PGO模块与FAST-LIO2解耦，非常方便，很优秀的工作。

3.FAST_LIO_LC的作者yanliang-wang,在FAST_LIO_SLAM的基础上添加了：1.基于Radius Search 基于欧式距离的回环检测搜索，增加回环搜索的鲁棒性；2.回环检测的优化结果，更新到FAST-LIO2的当前帧位姿中，幷进行ikdtree的重构，进而更新submap。

Contributions

FAST_LIO_SAM的主要贡献：

1.对比FAST_LIO_SLAM 与 FAST_LIO_LC 使用外部接入的PGO回环检测模块进行后端优化，FAST_LIO_SAM 将LIO-SAM的后端GTSAM优化部分移植到FAST-LIO2的代码中，数据传输处理环节更加清晰。

2.增加关键帧的保存，可通过rosservice的指令对地图和轨迹进行保存。

3.FAST_LIO_SLAM中的后端优化，只使用了GPS的高层进行约束，GPS的高层一般噪声比较大，所以添加GPS的XYZ三维的postion进行GPS先验因子约束。

Prerequisites

Ubuntu 18.04 and ROS Melodic
PCL >= 1.8 (default for Ubuntu 18.04)
Eigen >= 3.3.4 (default for Ubuntu 18.04)
GTSAM >= 4.0.0(tested on 4.0.0-alpha2)

Build

cd YOUR_WORKSPACE/src
git clone https://github.com/kahowang/FAST_LIO_SAM.git
cd..
catkin_make

Quick test

Loop clousre：

1 .For indoor dataset

dataset is from yanliang-wang 's FAST_LIO_LC ,dataset which includes

/velodyne_points

(10Hz) and

/imu/data

(400Hz).

roslaunch fast_lio_sam mapping_velodyne16.launch
rosbag play  T3F2-2021-08-02-15-00-12.bag

indoor

2 .For outdoor dataset

dataset is from LIO-SAM Walking dataset: [Google Drive]

roslaunch fast_lio_sam mapping_velodyne16_lio_sam_dataset.launch
rosbag  play  walking_dataset.bag

outdoor_1

outdoor_2

3.save_map

输入如下指令到terminal中，地图文件将会保存在应文件夹中

rosservice call /save_map "resolution: 0.0
destination: ''" 
success: True

4.save_poes

输入如下指令到terminal中，poes文件将会保存在相应文件夹中

rosservice call /save_pose "resolution: 0.0
destination: ''" 
success: False

evo 绘制轨迹

evo_traj kitti optimized_pose.txt without_optimized_pose.txt -p

evo1 evo2

5.some config

# Loop closure
loopClosureEnableFlag: true# use loopclousre or not 
loopClosureFrequency: 4.0# Hz, regulate loop closure constraint add frequency
surroundingKeyframeSize: 50# submap size (when loop closure enabled)
historyKeyframeSearchRadius: 1.5# meters, key frame that is within n meters from current pose will be considerd for loop closure
historyKeyframeSearchTimeDiff: 30.0# seconds, key frame that is n seconds older will be considered for loop closure
historyKeyframeSearchNum: 20# number of hostory key frames will be fused into a submap for loop closure
historyKeyframeFitnessScore: 0.3# icp threshold, the smaller the better alignment# visual iktree_map  
visulize_IkdtreeMap: true# visual iktree_map  
recontructKdTree: true

savePCDDirectory: "/fast_lio_sam_ws/src/FAST_LIO_SAM/PCD/"# in your home folder, starts and ends with "/". Warning: the code deletes "LOAM" folder then recreates it. See "mapOptimization" for implementation

Use GPS：

1.dataset

dataset is from LIO-SAM Park dataset: [Google Drive]

roslaunch fast_lio_sam mapping_velodyne16_lio_sam_dataset.launch
rosbag  play  parking_dataset.bag

Line Color define: path_no_optimized(blue)、path_updated(red)、path_gnss(green)

gps_optimized_path

gps_optimized_with_map

2.save_map

输入如下指令到terminal中，地图文件将会保存在应文件夹中

rosservice call /save_map "resolution: 0.0
destination: ''" 
success: True

FAST-LIO Map (no gnss prior factor) Red ; FAST-LIO-SAM (with gnss prior factor) Blue

gps_map

3.save_poes

输入如下指令到terminal中，poes文件将会保存在相应文件夹中

rosservice call /save_pose "resolution: 0.0
destination: ''" 
success: False

evo 绘制轨迹

evo_traj kitti gnss_pose.txt optimized_pose.txt  -p

FAST-LIO (no gnss prior factor)FAST-LIO-SAM (with gnss prior factor) evo_no_optimized evo_optimized

4.some config

# GPS Settings
useImuHeadingInitialization: false# if using GPS data, set to "true"
useGpsElevation: false# if GPS elevation is bad, set to "false"
gpsCovThreshold: 2.0# m^2, threshold for using GPS data
poseCovThreshold: 0#25.0                      # m^2, threshold for using GPS data  位姿协方差阈值 from isam2

5.some fun

when you want to see the path in the Map [satellite map](http://dict.youdao.com/w/satellite map/#keyfrom=E2Ctranslation)，you can also use Mapvizp plugin . You can refer to my blog on CSDN.

mapviz_1

mapviz_2

Attention:

1.FAST-LIO2中对pose姿态是使用so3表示，而gtsam中，输入的relative_pose姿态是Euler RPY形式表示，需要使用罗德里格斯的公式进行转换更新。

2.参考yanliang-wang FAST-LIO-LC中的iktree reconstruct

3.在walking数据集中，因为有个别数据是在同一个地方不断手持旋转激光雷达，旋转激光雷达的角度达到了保存关键帧的阈值，在短时间内，保存了多帧相似的关键帧，导致ISAM2出现特征退化，进而里程计跑飞，可以根据数据集的情况适当调整关键帧选取的阈值参数。

4.添加GPS prior 先验因子的部分diamante，参考lio_sam的先验因子部分，对比于kim的FAST-LIO-SLAM，FAST-LIO-SLAM中只是用了GPS的高层约束，并没有使用xy方向的约束，而GPS在高层(Z轴)的误差比较大，优化过程中容易引入误差。

5.GPS先验因子中，“useGpsElevation”是否选择GPS的高层约束，默认不使用，因为GPS的高层噪声比较大。

6.LIO-SAM 中使用ekf_localization_node这个ROS Package 把GPS的WGS84 坐标系转到 World系下，FAST-LIO-SAM考虑到尽量与外部的ROS package 解耦，调用 GeographicLib进行坐标转换。

some problems:

1.GNSS的经纬高噪声协方差没有转换到World系下，暂时使用latitude longtitude 的cov noise 作为x y 向的cov nosie

2.应该使用的是ENU坐标系，但是使用GeographicLib转换后的结果得到的坐标系是NED坐标系下的，原因暂时没捋清楚，待解决。（X: E Y: N Z: -D ）

3.在跑较大的数据集(600s)时，偶尔出现程序崩的现象，暂时没定位问题所在，待解决。

Acknowledgements

In this project, the LIO module refers to FAST-LIO and the pose graph optimization refers to FAST_LIO_SLAM and LIO_SAM.The mainly idea is for FAST_LIO_LC.Thanks there great work .

Also thanks yanliang-wang、minzhao-zhu、peili-ma 's great help .

edited by kaho 2022.6.20

标签：前端算法机器学习

本文转载自: https://blog.csdn.net/weixin_41281151/article/details/125371285
版权归原作者 KaHoWong 所有，如有侵权，请联系我们删除。