Hi @ziv-lin, when we detect there is a potential loop, before we use different voxel resolution to filter the curr anchor keyframe and history anchor history keyframe, we set the initial trans use keyframe's center, link is here. I guess we calculate transform is: the TF from curr to history in world frame. Here the transform direction is reffering to frame, not vectors in frame. Why we don't set the trans initial is history minus curr,

Eigen::Matrix< double, 3, 1 > transform_T =
( keyframe_b->get_center() - keyframe_a->get_center() ).template cast< double >();

It is the vector from curr to history. Thanks for your help very much!

First, thanks for your code and bag.But when I try to transform the pointcloud to the pcd.Here is a question:

/opt/ros/kinetic/lib/pcl_ros/pointcloud_to_pcd: error while loading shared libraries: libpcl_io.so.1.7: cannot open shared object file: No such file or directory

But how could I find the libpcl_io.so.1.7 when I use the pcl1.9?

I would be very happy for yous any reply!

Hi,

I saw the Livox Horizon device and thought this would be a good match for this kind of software.
Before I buy such a device I want to make sure it will be compatible with this Software or at least I want to help making it compatible.

Thanks for this awesome project!

CU
KirK

Hi,Thanks for your open source.I test the new code on Ubuntu16.04 && ROS Kinetic.But I cannot reproduce experimental results using the example data——loop-simple.bag.
Here are some screen shots.
a)loop close error

b)big delay

c)ceres-solver optimization error

Any suggestions?

Hi, NICE WORK! but what camera did you choose to fit the 3D printing toos?

I try to roslaunch the program with the downloaded file and try to input like this:
roslaunch loam_livox rosbag.launch rosbag play /home/Downloads/CYT_02.bag
get the roslaunch error: The following input files do not exist: /home/Downloads/CYT_02.bag, I don't know why?

Jiarong, Thanks for sharing your great work!
In the current version from Github, does it use edges from reflectivity and filter out points with e_pt_reflectivity_high? I do see it filters out points with e_pt_reflectivity_low.

thanks

template <typename T>
static T vector_project_on_vector( const T &vec_a, const T &vec_b )
{
    return vec_a.dot( vec_b ) * vec_b / ( vec_b.norm() * vec_b.norm() );
}

template <typename T>
static T vector_project_on_unit_vector( const T &vec_a, const T &vec_b )
{
    return vec_a.dot( vec_b ) * vec_b;
}

您好，感谢您的工作！我在研究代码时，对上面这段代码有点疑惑（在eigen_math.hpp中）。这两个函数的名称本意和实现方式是否搞反了呢？下面那个函数本意是想投影到单位向量，但是实现方式却错了。
在ceres_icp.hpp中调用的是vector_project_on_unit_vector，看了代码原理，确实应该是vec_a投影在vec_b的单位向量上。
期待您的回复，谢谢！

Hi @ziv-lin, after transformed curr's points into map, when insert these celles into m_keyframe_of_updating_list called in add_cells into each keyframe of updating_list, we only insert the first cell into each keyframe, right? why we don't insert all of cells into each of them?
Thanks for your help!

    FUNC_T void add_cells( const std::set< Mapping_cell_ptr > &cells_vec )
    {
	unsigned int last_cell_numbers = 0;
	for ( typename std::set< Mapping_cell_ptr >::iterator it = cells_vec.begin(); it != cells_vec.end(); it++ )
	{
	    m_set_cell.insert( *( it ) );
	    m_map_pt_cell.insert( std::make_pair((*(it))->get_center(), *it) );
	    if(last_cell_numbers!= m_set_cell.size()) // New different cell
	    {
	        if(last_cell_numbers == 0)
	        {
	            // Init octree
	            m_pcl_cells_center->push_back( pcl::PointXYZ( (*it)->m_center( 0 ), (*it)->m_center( 1 ), (*it)->m_center( 2 ) ) );
	        }
	        last_cell_numbers = m_set_cell.size();
	    }
	};
	m_accumulate_frames++;
    }

Hi,

I know that you are not using IMU/GPS sensor for this loam_livox, can the result of loam_livox converted in las/laz file for RS/GIS analysis? Can the loam_livox rosbag be georeferenced 3D Pointcloud?

However, if you have IMU/GPS sensor attached to livox mid-40, what would be a possible scenario regarding using this loam_lidar for georeferencing 3D pointcloud?

Thanks.

Noli

While driving the robot around if I yaw too quickly ScanRegistration crashes with exit code -11. Sometimes it happens while just driving straight ahead. Is there a way to turn off loop closure ?

Hello,

I am using Ubuntu 16.04 LTS and installed kinetic version of ROS (which has PCL 1.7 installed). I encountered these error when I catkin_make under my catkin workspace:

/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/libpcl_features.so: undefined reference to `_ZNK3pcl6search6SearchINS_11PointXYZRGBEE12radiusSearchERKNS_10PointCloudIS2_EEidRSp6vectorIiSaIiEERS8_IfSaIfEEj'
collect2: error: ld returned 1 exit status
loam_livox/CMakeFiles/livox_scanRegistration.dir/build.make:462: recipe for target '/home/derek/workspace/ros/devel/lib/loam_livox/livox_scanRegistration' failed
make[2]: *** [/home/derek/workspace/ros/devel/lib/loam_livox/livox_scanRegistration] Error 1
CMakeFiles/Makefile2:7030: recipe for target 'loam_livox/CMakeFiles/livox_scanRegistration.dir/all' failed
make[1]: *** [loam_livox/CMakeFiles/livox_scanRegistration.dir/all] Error 2
Makefile:138: recipe for target 'all' failed
make: *** [all] Error 2

Thanks in advance!

Hi, I wonder how can I export the result to *.pcd files. Now I have recorded a rosbag file, and it looks good in rviz, but I can't get the coordinates of the points. How can I export them to *.pcd files? Then I can process them in pcl. Thank you!

Hello, I am trying to use the given rosbag to recreate the result, I followed the readme, used the command: roslaunch loam_livox rosbag.launch + rosbag play HKU_ZYM.bag, but the process scanRegistration was crashed right after I played the rosbag, and I got the message as followed:

I am not sure if it is the same issue as the issue #34, but I have updated to the latest version but still have this problem, please take a look, thanks a lot!

Thanks for sharing code.
I tried to use this code to run KITTI Data. I changed three places of rosbag.launch:

1. scan_line -> 64
2. lidar_type -> velodyne
3. remap from="/laser_points" to = "/livox/lidar" -> remap from="/laser_points" to = "/kitti/velo/pointcloud"
   but nothing shows up. I don't know why. Please help me.

Hi,
thanks for your project. when I run "roslaunch loam_livox livox.launch", I get the warning like this:
[ WARN] [1575016816.066039723]: Drop lidar frame in mapping for real time performance !!!
Is that something wrong? Can you give me some advices?

Hi,

It seems that the pdf for A fast, complete, point cloud based loop closure for LiDAR odometry and mapping publication is wrong.

I downloaded the right one, found using google search. 1909.11811v1.pdf

Noli

Hi @ziv-lin , i use rosbag_loop.launch and don't modify any param in performance_precision.yaml , play with loop_hku_zym.bag. I run the demo 5 times, and i found sometimes it doesn't trigger loop, but sometimes it can. Did you meet this ? Here is This is the result of a experiment not loop.

Hi @ziv-lin , similar to Running large_scale demo died. #32, after pose graph, the trajectory is more sharp than before loop closure trajectory, just like the following picture, my PCL is 1.9.1, built from source with default params. The trajectory size before and afte pose graph is equal to each other. What do you think of this?

Hi，recently i get a livox sensor, due to their unique Non-repetitive Scanning Technology, the adjacent frame differs a lot, it's hard to do frame-frame match work.
do you have any papers relative to this project, or do you have any introductions to their unique Non-repetitive Scanning Technology. Thanks a lot

Hi,
when I run "roslaunch loam_livox livox.launch
"
the terminal print

[ WARN] [1571727835.757896886]: time Map corner and surf num are not enough ****** min max timestamp = [29.464680, 29.512922] ****** map corner num 0 surf num 108 [ WARN] [1571727835.807914851]: time Map corner and surf num are not enough ****** min max timestamp = [29.512922, 29.558619] ****** map corner num 0 surf num 108 [ WARN] [1571727835.858688485]: time Map corner and surf num are not enough ****** min max timestamp = [29.558619, 29.609156] ****** map corner num 0 surf num 108 [ WARN] [1571727835.908180349]: time Map corner and surf num are not enough ^C****** min max timestamp = [29.609156, 29.669964] ****** map corner num 0 surf num 108 [ WARN] [1571727835.958653270]: time Map corner and surf num are not enough

Is that something wrong?

First of all, thank you very much for the exciting work of open source. When I read the source code, I found that the Velodyne option is provided in laser_feature_extractor.hpp. I tried to modify and compile, but rviz showed nothing when playing the rosbag of the Velodyne lidar.
Is it possible to use Velodyne radar by changing the parameters? Thank you very much!

Trying to start up ROS navigation stack with output from loam-livox. Not having much luck. Is there a document that explains the topics published by loam-livox? Can you point me to resources explaining how to do this?

Hi！I use rosbag_mid100.launch to processing data from livox-mid100. The result is better than livox-mid40. But, why there are some points on the trajectory in /velodyne_cloud_registered and /laser_cloud_surround? I checked the raw data and there are no these points. How can I delete these points?

Thank you for your excellent work but I'm unable to match results of demo material . Testing inside of an interior space 5x5x5 meters and the result looks more like a ball of yarn , Is this a limitation of Loam or something I'm doing wrong

the setup is on a tripod yet the mapping location is jittering all over the place

您好,看您這邊建圖的demo,包括提供的rosbag中
https://github.com/hku-mars/loam_livox/blob/master/pics/demo_cyt.gif
livox單幀的點雲很稠密, 但是我們這邊在拿點雲的原始數據時,只能依次獲取點數量很少很稀疏的小數據包, 請問您這邊是有進行什麼預處理嘛,比如直接將鄰近的幾幀數據包拼接起來作爲一幀點雲, 或者是使用了硬件同步?
非常感謝

Hi,I found that laser mapping will crash after loop closure.It happens both on rosbag_loop_simple and rosbag_loop.Here are some screen shots.
loop_simple

loop

Any suggestions?

I would like to point out that identifiers like “__EIGEN_MATH_HPP__” and “_T” do not fit to the expected naming convention of the C++ language standard.
Would you like to adjust your selection for unique names?

您好，感谢您的付出。我想请问你们有没有尝试过在windows上运行算法。我想迁移到windows上，请问您有什么建议吗？例如ros的版本要求，pcl的版本要求。

I'm trying to integrate loam-livox with the ROS nav stack on my rover. The stack is working with a ZED camera and Rtabmap now. When I try and switch it over to loam-livox I can see the map if I run velodyne_cloud_registered into octomap_server. So that appears to be working. However when the robot moves I see the aft_mapped odometry moving but its not moving the robot model in rviz like its supposed too. Noticed that loam-livox doesn't publish odom or map like the other mapping software I've used does. So thinking that incorrect setup of the tf is what is causing my issue. Have you used it with ROS navigation?

Hello,
The livox_laserMapping node is crashing with some sort of memory issue:

 Error in `/home/finostro/Code/ros_ws/devel/lib/loam_livox/livox_laserMapping': double free or corruption (out): 0x00007fdcaddc1bc0 

Maybe you could run your code with address sanitizer to check for memory problems?

Cheers

Hi, great contributors,

Today i'm testing your awesome project. I downloaded the large-scale HKUST_01.bag . But it somehow corrupted after a few seconds.

This is my terminal and problem:

Looking forward to your reply.
THANKS

How did you build this under ROS? When I try it fails looking for a function that isn't in the opencv libs that ROS installs. I built ROS from source and got it to compile but no livox/scan or velodyne_cloud_registered messages when I start it that way. I built ROS with opencv3.4.3 and that takes care of the missing function which was introduced in 3.3.2 from what I've read. Nvidia Xavier installs 3.3.1 as part of the flash. If any 3.3.1 or 3.2 opencv libs exist it fails. The error is undefined reference to cv::Mat::updateContinuityFlag().

How change your code so that can make only one of the mid-100 work( Mid-100 is made up of three mid-40s )

The CMakeLists.txt file in the src directory refers to kinetic only. I've built it and it runs on melodic with out changing the soft link but I'm wondering if this needs to point to the correct toplevel.cmake file for the version of ROS you are running? Also with other mapping packages I've been using the package provides an odom and map frame. Are there any plans to add this?

Hello, thank you for your amazing work, but I am not quite clear about what to do after connecting lidar, could you please provide a detailed operation instruction

Managed to get loam-livox working with the ROS navigation stack. Using octomap_server to create a 2-d map from velodyne_cloud_registered for the costmaps. I did run into an issue though. When I first set it up there were constant errors from ros_tf2 complaining about / in front of the camera_init frame. Tf2 doesn't allow / in front of frame names. Couldn't find any other source of the error so I edited laser_mapping.hpp and laser_feature_extractor.hpp changing all the instances of /camera_init to camera_init. The errors went away and the costmaps started working. Also using velodyne_cloud_registered as input for the costmaps instead of using pointcloud_to_laserscan to create a LaserScan topic. Doesn't seem to work with amcl however it wants a Pointcloud. I'm using robot_localization instead with a ukf filter using aft_mapped_to_init and the imu as inputs. In the robot model I made camera_init the parent of base_link. Robot_localization adds the map topic to complete everything. V2 performance seems to be better. I'll get it outside in a day or so and see how it works. Good Job !!!

If you leave loam-livox running for about an hour it will eventually take up all the memory and swap and crash. This is on a nvidia Xavier which has 16Gb of ram. Ubuntu 18.04. Easy to see with jtop. It takes a long enough period of time to crash that it isn't really going to effect my testing that much but I thought you might like to know.

I used a usb camera as my web cam with usb_camera package installed. I executed the same process as given in the readme.md file on the repo. The below screenshot is taken with the camera and lidar attached. Any help is appreciated.

Hi!
Thanks a lot for your excellent work! Will the next version of the code support IMU?

Currently unable to run in raspbian nspawn 64bit , is this not supported?

I found that the output frequency of the topic /aft_mapped_path is almost 6hz，Whether the output frequency of pose can reach 10 Hz？

Hi:
Thank you for your outstanding work!
As the README instructions, I configured my environment. But when I Updated my PCL to version 1.9.1, something is wrong.
My motified CMakeLists.txt as follow:

Then "1.rosbag play CYT_02.bag 2. roslaunch loam_livox livox.launch", an error is throwed out, as follow pictures show.

However, when I used PCL1.7, it running well. But the results did not have the effect of the show.

Hi,

It seems that CloudCompare can open and view various XYZ data files.

https://en.wikipedia.org/wiki/CloudCompare

I think it would good if possible that converted bag files can output format that can be opened in CloudCompare.

It seem that PCL is used in this app.
http://pointclouds.org/documentation/tutorials/pcd_file_format.php

Thanks.

Noli

Hi @ziv-lin , thanks for sharing your work to community. when calculating cell's center cell_map_keyframe.hpp func find_cell_center() and covariance func get_covmat() in code is as follows, different from the paper equ1. and equ3. repectively, does this matter?

cell_center(0) = (std::round((pt(0) - half_of_box_size) / box_size)) * box_size + half_of_box_size;
cell_center(1) = (std::round((pt(1) - half_of_box_size) / box_size)) * box_size + half_of_box_size;
cell_center(2) = (std::round((pt(2) - half_of_box_size) / box_size)) * box_size + half_of_box_size;

for (size_t i = 0; i < pt_size; i++) {
m_cov_mat = m_cov_mat + (m_points_vec[i] * m_points_vec[i].transpose()).template cast<COMP_TYPE>();
}
m_cov_mat -= pt_size * (m_mean * m_mean.transpose());
m_cov_mat /= (pt_size - 1);

When calculating feature direction in func feature_direction(), in paper V-C(2D histogram of keyframe), " we choose the direction with positive X components", in code we transform points whose x part is negative into positive, could you please explain it? Looking forward to your reply, thanks for your help!

if (vec_3d[0] < 0) {
vec_3d *= (-1.0);
}

Hi，I cann't download the bags，can you upload them to baidu cloud？

Hi!

Thanks a lot for this package. I've run some tests already with Mid-40 and really like it so far! Would you be able to give me some pointers on how to implement Mid-100 support?

If you take pull request I'd be more than happy to expose this feature as a parameter and make a PR for it.