Closed bigbigpark closed 2 years ago
Hi bigbigpark,
This is the right place for these questions, thanks for your interest in the project !
So in the paper, we implemented the LoopClosure in pyLidar-SLAM.
Historically, pyLiDAR-SLAM was the main project where I could implement / benchmark SLAM algorithms, We made a wrapper for CT-ICP in python and integrated it into pyLidar-SLAM.
Since then, (and really for time management of my PhD thesis), my focus has been on C++ code and ct_icp improvements.
So we are currently working on an implementation of the loop closure in C++ (with ROS support), as we found that it was more relevant in the short-term for most use-case of SLAM. It will probably appear in another gitlab project, but there will be pointers to it in this project. It should appear in the coming weeks (there is a working version which still needs some polishing). In the mean time, you will find the LoopClosure implementation we used in our paper in pyLidarSLAM.
However our aim is (when I will have some time) to update the python wrapping to integrate ct-icp with pyLiDAR-SLAM, so we can use the neat parameter grid-search that comes with hydra.
So right now our SLAM (in this project) is really an odometry : ie correct locally, but no compensation for large trajectory errors.
Hope it answers your question,
Best Regards,
I totally understood what I didn't know. Thank you for your kind reply.
Thank you for sharing your wonderful project and giving good inspiration to many people.
Good luck.
I don't know whether this Github issue pages are right place for question...
I have read your paper
CT-ICP: Real-time Elastic LiDAR Odometry with Loop Closure
.I successfully have built your ct_icp code using KITTI dataset with viz3d
Here is my question
What is exactly the role of pyLiDAR-SLAM ?
The paper says that ct_icp is lidar-only odometry and has loop closure itself using g2o
Is it enough for SLAM to use ct_icp ?
I would appreciate it if you could answer my question...!
Best regards.