wh200720041 / floam

Fast LOAM: Fast and Optimized Lidar Odometry And Mapping for indoor/outdoor localization IROS 2021

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

major rewrite, would you take a PR?

flynneva opened this issue · comments

first off - amazing work!

I wanted to ask before I got too far down the path - I've done a pretty major rewrite of the code over on my forked repository and wanted to see if you would accept a PR from me to bring these changes back into this original repo?

if not, then I'll probably just go ahead and release the floam package myself to ROS noetic since I think it could be really useful for other people.

let me know what you think! and looking forward to seeing if we can combine our efforts!

Hi @flynneva

Thanks for your extension work based on FLOAM. It is a pretty good rewrite.

I noted that there are several major improvements in your code:

  1. Use nodelet to improve the efficiency of data transmission between nodes.
  2. The feature extraction method has been changed. You use the normal estimation from pcl library to extract edge and planar features.
  3. add data synchronisation for callback function.

I couldn't read through every details of your code due to the time limit, please correct me if I am missing anything. I think these improvements are very important to make the performance more reliable.

Unfortunately I am unable to access my PC recently. I am not able to evaluate the improved codes (I am really interested in the performance of your improved version since there are really significant changes). I will consider merge your code into this repo when I have some test on my PC.

You are welcomed to release the floam package to ROS, I think it would be pretty useful to the robotic society and I really appreciate your extension work.

Great thanks again and you are also welcomed contact me via my email if you would like to discuss or share your thoughts about floam

please correct me if I am missing anything

no you pretty much nailed it! first was nodelets, then was separating out the feature extraction stuff for scanning lidars vs imaging lidars...I havent implemented the scanning one yet but it should be able to just use the same logic from your original code I think. next was rewriting the odom and mapping nodes to use the generic sync features from ROS to sync up the different topics.

You are welcomed to release the floam package to ROS
this is awesome to hear, and thank you again for your work with it so far! i'll be sure to keep you as author, all your copyrights as well as link to your original repo here as well when I do eventually publish it after testing.

I'll keep you updated as soon as I'm able to test it but if you or anyone else end up testing it before I do let me know how it goes!!! it'd be great to get floam developed into a real powerhouse for real-time lidar localization and mapping.

so I just was able to add in the scanning lidar logic as you had before over the weekend. it seems to be working OK (the edge and surface detection) however i'd love to get some feedback how this might be able to be improved or adjusted. i dont know what parameters to tune in order to make it perform better so any experience from your end or recommendations what should be parameters or adjustable and what shouldnt would be extremely helpful.

@wh200720041 i've got the rewrite fully functional now for the kitti dataset if you want to check out the link above with the talked about changes.

let me know what you think!

@wh200720041 just to keep you in the loop, I just opened a PR for the initial release of floam! should be exciting getting it published and easier for people to install.

would love to collaborate and see if we can make it even more efficient!

@wh200720041 fyi i think floam should be available to download now via apt install ros-noetic-floam. you can see it on ros index now.

would love to collaborate more with you on enhancing floam or on another project!

good job

@flynneva
great thanks for your help, I am glad to see it is available on ROS. let's work together to improve floam in the future:)

@flynneva
Hi, I wonder how much using the normal estimation from pcl library will increase the speed of extracting edge and planar features?

@FishInWave no idea but it would be great to test! would be great to get more help maintaining/adding features to flynneva/floam if you'd like to give it a shot!

also would be great to migrate it to ROS 2 sometime 😄