ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A

Accuracy of SLAM in outdoor environments using 2D versus 3D Lidar?


Hello, I am working on a startup, and we are thinking through LIDAR sensors we might use. Our platform will be used outdoors, especially in settings like orchards and berries on farms, where you have a lot of leaves, trunks, etc.

A function of our platform will be the ability to use SLAM to build a map, which the platform can then retrace back through to a set point. We’ve been debating what sensors we could use to do this, and are considering both 2D and 3D lidars. For 2D lidars, we are experimenting with the Sweep 360, and the SICK TIM 561. For 3D lidars, we have been trying to get some early low cost units from Quanergy, which has a 120 degree horizontal/10 degree vertical FOV Lidar.

Can anyone comment on the accuracy of SLAM in this type of environment using 3D Lidars versus a 2D lidar (likely mounted about 18 inches from the ground)? In the real world, does the added data generated by 3D Lidars tend to make SLAM mapping more accurate, or no?


You can solicit sensor placement and unit advice from the community of course, but it is clear that you are going to run into much more fundamental issues later, since you have to solicit this kind of advice from the community to begin with.

I would look for a perception/SLAM engineer to join your startup.


The discussion about 2D or 3D environment perception is usually more related to robustness rather than to accuracy. If your environment is highly 3D, meaning ramps, holes, steps, furniture, … then a 2D SLAM might fail due to a poor representation, not due to inaccuracies. Otherwise, if your environment is flat enough, and features have a strong vertical expansion (walls , posts) , then 2D could be enough robust and very accurate, and generally cheaper in both, dollars and computation.
My advice would be that think about your environment, and try to imagine if a 2D slice can represent it in a robust way to navigate. Then take a decision (and then accept that you may fail in your decision …)