What is the best way to model the SLAM problem (offline) for a wheeled robot moving on a flat floor in 3D space (the pose has only 3 degrees of freedom)?

In terms of sensor data, we have the wheel odometry and 3D measurements of landmarks (coming from a Kinect v2).

The bit that I'm confused about is that this is a very standard setting, yet all the examples I've seen for 2D SLAM use 2D features and don't mention anything about what to do with 3D sensor measurements, and vice-versa, I haven't seen a 3D SLAM solution where we could easily incorporate known constraints on the robot pose (has to lie on a plane). It would be great to know what is common practice in this setting. An explanation or a link to an example implemented in some SLAM framework would be of much help.


1 Answer 1


Since you are using Kinect my advise is to use Octomap ROS package to build a 3D map. You can extract the 2D occupancy grid map of the ground floor from the Octomap for path planning and navigation.

One disadvantage of using Octomap is, it doesn't have a pose correction method. It just build the map based on the poses provided by the odometry.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.