Add camera data to obstacle layer or voxel layer in costmap?

Question

Hi. I want to use navigation with multiple sensor inputs. I already have two SICK scanners on my robot, and to avoid the issue with only planar scanning, I want to add a RealSense camera as well.

I see that the obstacle/voxel layers in the costmaps in Navigation2 have support for this, and I am considering two cases:

• Use depthimage_to_laserscan on the camera data and provide the obstacle layer with three laserscan inputs (and no voxel layer).
• Use the pointcloud from the camera as input for the voxel layer, and two scan topics from the lasers for the obstacle layers.

Have anyone tested this and can tell me about how the different alternatives would affect the result, or maybe you can point me to some theory about this?

Thanks!

---

Originally posted by ninamwa on ROS Answers with karma: 81 on 2020-02-20

Post score: 1

Answer 1

Do the second.

If you have rich 3D information, use it. There’s no benefit except for compute savings from projecting the cloud into a scan, but you will lose a bunch of context in the environment. Unless you’re using an RPi or weaker computer, its well worth the CPU.

---

Originally posted by stevemacenski with karma: 8272 on 2020-02-20

This answer was ACCEPTED on the original site

Post score: 2

---

Original comments

Comment by ninamwa on 2020-02-21:
Thank you, I will try that out! :-)