LiDARs for MRS SLAMs (A-LOAM, LIO-SAM, Octomap): What are MRS-compatible and budget-friendly options?

[MJavadZallaghi]

I need to use MRS packages for SLAM which are LiDAR-based for indoor and outdoor settings, like Octomap mapper and planner, aloam and lio-sam. Regarding the LiDAR sensors, Ouster LiDARs like OS0-128 or OS1 are used in the simulation examples I found in the MRS packages with provided links. These are expensive 3D LiDARs, make them hard to be used with a limited budget.

Considering your experience and expertise, what are the other options as 3D LiDARs with reasonable price for running SLAM of Octomap mapper and planner, aloam and lio-sam? Even with a limited horizontal FoV and shorter range of another 3D LiDAR, we need to use these MRS SLAM packages for our research. So it would be great if you can tell me which alternatives I could have as 3D LiDAR for this purpose. Compatibility of any other 3D LiDAR with MRS System is a major requirement as I want to use MRS SLAM packages mentioned here.

Also as a cheap solution, noticed about hector, which uses a 2D LiDAR (Slamtec RPLIDAR A3 360°). From the simulation example I saw, hector SLAM of MRS generates a 2D map, not decent for path planning in 3D indoor settings. Is it possible to generate 3D map with MRS hector SLAM? Because conceptually if altitude be measurable with garmin range finder — that is the case — the implemented SLAM should be able to create map in different layers, meaning having a 3D map with minimum details, but decent for path planning and obstacle avoidance.

Generally, I really appreciate any comment and insight on SLAMs from MRS, the LiDARs I can use, the previous experiences you have had, any lessons, as my aerial robot will depend on this MRS autonomy stack. It would really a nice capability if I be able to have proper LiDAR setup to use Octomap mapper and planner, aloam and lio-sam, and hector.

Thanks in advance for your valuable answers!

[migsdigs]

Hi @MJavadZallaghi
It seems we are more or less looking for the same information in the MRS stack. I am looking to use it for indoor drone flight, and the state estimation is particularly important to me at this point. Like you, I am looking for solutions with cheaper 3D LiDARs like the Livox MID-360, or some RGB-D/other VIO. I have not been able to make much progress yet, and I am also waiting for some answers on these topics, but I have come across Point-LIO, which is listed as one of the localization plugins in the documentation, but is not included in the MRS stack by default (at least not in the apptainer). I think that might be of interest to you. Still, I think I would like to get the A-LOAM state estimation working with a cheaper LiDAR if possible, so I would appreciate any feedback from you side. If you come across anything valuable please share it in your discussions, I am also interested in any answers you receive in your other posted discussions.

[MJavadZallaghi]

Hi @migsdigs. Thanks for making this reply! Yeah while I will keep posting my questions and also update the discussion I've opened up, it's good we share what we learn to use all capabilities of MRS stack with each other! :)

[matemat13]

To add to this reply, you can also consider e.g. the Ouster OS1-16 or -32 (which are the 16 row and 32 row variants), that are cheaper than the "full" 128 row variants and are probably sufficient for most use-cases. Another option are LiDARs from Unitree, but AFAIK we don't have much experience with these. Lastly, you can simply use an RGB-D camera for the mapping, such as the Intel Realsense (our driver is here) or similar.

[migsdigs]

Hi @matemat13, thank you very much for your reply!

If you don't mind me asking, what has been your experience using VIO, such as open-vins (with some RGB-D or other cameras) methods for state estimation in indoor environments? Would you say it is generally sufficient, or it is not reliable/robust enough and we should rather stick to LIO-based methods?

Thanks again!

[snakehaihai]

I need to use MRS packages for SLAM which are LiDAR-based for indoor and outdoor settings, like Octomap mapper and planner, aloam and lio-sam. Regarding the LiDAR sensors, Ouster LiDARs like OS0-128 or OS1 are used in the simulation examples I found in the MRS packages with provided links. These are expensive 3D LiDARs, make them hard to be used with a limited budget.

Considering your experience and expertise, what are the other options as 3D LiDARs with reasonable price for running SLAM of Octomap mapper and planner, aloam and lio-sam? Even with a limited horizontal FoV and shorter range of another 3D LiDAR, we need to use these MRS SLAM packages for our research. So it would be great if you can tell me which alternatives I could have as 3D LiDAR for this purpose. Compatibility of any other 3D LiDAR with MRS System is a major requirement as I want to use MRS SLAM packages mentioned here.

Also as a cheap solution, noticed about hector, which uses a 2D LiDAR (Slamtec RPLIDAR A3 360°). From the simulation example I saw, hector SLAM of MRS generates a 2D map, not decent for path planning in 3D indoor settings. Is it possible to generate 3D map with MRS hector SLAM? Because conceptually if altitude be measurable with garmin range finder — that is the case — the implemented SLAM should be able to create map in different layers, meaning having a 3D map with minimum details, but decent for path planning and obstacle avoidance.

Generally, I really appreciate any comment and insight on SLAMs from MRS, the LiDARs I can use, the previous experiences you have had, any lessons, as my aerial robot will depend on this MRS autonomy stack. It would really a nice capability if I be able to have proper LiDAR setup to use Octomap mapper and planner, aloam and lio-sam, and hector.

Thanks in advance for your valuable answers!

depends on the task you are dealing with.

For general lightweight, you may try "Airslam: An efficient and illumination-robust point-line visual slam system" which can run on jetson orin in real time.

For dealing with robust depth perception, fast lio2 with basic Mid360 would work.

However, if you area also dealing with smooth trajectory planning with continuous time representation, then you may try the SLICT2 or GPTR series of work.

For a degenerative environment or a smoke environment, it is a bit harder to cope, MM-LINS could do

For 3D map with 2D LIDAR, I also did sth b4 by using IMU + 1D LIDAR + 2D lidar and make it explore within a ship, it is achievable but i don't think it is needed as one can just buy a cheap unitree lidar for complete 360 degree x180 degree coverage. https://www.unitree.com/LiDAR