ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

How to create 3D map from multiple measurements (LIDAR + IMU)

asked 2021-05-12 05:29:46 -0500

Johann.Haselberger gravatar image

I have several measurement runs (on the same route) with a test vehicle equipped with multiple LIDAR sensors and a high-accuracy GPS and IMU system.

From this data, I would like to create a single 3D map. How can I archive this? Is the precise GPS position useful or is an accurate IMU enough?

Thank you in advance

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2021-05-12 13:47:49 -0500

shonigmann gravatar image

updated 2021-05-12 13:48:35 -0500

are your LIDAR sensors 2D? 3D? [edit - I just noticed the 3DPointCloud2 tag, so presumably 3D] Do you need the map as your robot moves through the environment (SLAM) or do you simply need a resultant map when the robot is done (reconstruction). The former is a more popular research topic and you'll find a larger number of options that may satisfy your needs. The latter takes advantage of being able to "see into the future" when registering points, and can result in a more consistent end map, especially when considering things like loop closure.

There are several ROS packages out there that will take LIDAR and IMU data to produce 3D maps, of varying fidelity. In my experience, IMU and depth information alone are enough for decent mapping, though GPS can certainly help with loop closure over larger maps. Many of them allow you to export and import generated maps to allow for reuse, if that's what you meant by "archive".. Deciding between them will depend on your requirements for computational efficiency, map resolution, map scalability, etc.

The multiple LIDAR sensor fusion might make things a bit more complicated. This question (#261259) suggests cob_scan_unifier to merge 2D LIDAR scans, for use with gmapping, but I haven't used it myself. For multiple 3D scans, presumably you'd first want to apply any transform necessary to get them all in the same frame, then concatenate the pointcloud messages from each sensor into a single combined message.

Once you have your fused LIDAR message, you could see if one of the following meets your needs:

edit flag offensive delete link more

Question Tools



Asked: 2021-05-12 05:29:46 -0500

Seen: 949 times

Last updated: May 12 '21