Environment Mapping

[omc8db]

Build up a 3-d map of the environment around the rover with rtabmap to avoid obstacles. Depends on #10

[omc8db]

Pull the video logs from the rover and run them through rtabmap on the ground

[omc8db]

Techniques to investigate:

• Stereoscopic Vision
• Structure from Motion
• LIDAR
• Deep Learning
• Some combination of the above

[omc8db]

General Purpose Point Cloud Mapping http://pointclouds.org

[abemorris]

We can successfully run the rtabmap tutorial linked above on a laptop running Ubuntu.
The next step is to get ROS running on the rover (likely on a Pi) and set up as detailed here.

[sdlouhy]

Turns out most rtabmap-related issues have been coming from the fact that the odometry dependency (openni) has been disabled due to bugs on several distributions of ros.

[sdlouhy]

This is also a great forum to look at whenever you have any issues with rtabmap.
Somewhere in the ros wiki, there is a good tutorial for working with two cameras as opposed to one Kinect camera or something of that ilk. Don't think this is the exact tutorial, but I believe it links to it.

[sdlouhy]

Forgot to add the camera calibration link. The one difficult thing about this is making sure the cameras have the same timestamp

[130s]

I'm not following this ticket at all but just FYI a ticket ros-drivers/openni_camera#66 referenced in #15 (comment) is now closed.

[sdlouhy]

Environment mapping plans have changed a little in the past year. The current plan is to use PL_SLAM for environment mapping, and a combination of stereoscopic cameras and a 1-D LiDAR for obstacle avoidance.