Ask Your Question

Revision history [back]

There are two well developed ROS systems for doing exactly what you want. RTAB-Map is the one we prefer in our lab, but the alternative rgbslam is also very good and depending on your exact use case may be a better solution.

Both of these systems are designed specifically for RGB-D sensors such as the Kinect, and have fairly good setup guides.

Hope this helps.

There are two well developed ROS systems for doing exactly what you want. RTAB-Map is the one we prefer in our lab, but the alternative rgbslam is also very good and depending on your exact use case may be a better solution.

Both of these systems are designed specifically for RGB-D sensors such as the Kinect, and have fairly good setup guides.

Hope this helps.

Update: You're trying to do 2D mapping with a 3D sensor, how is the robot moving? If it's driving over a flat floor and will not tilt, drive up ramps etc then you can extract a 2D slice from the kinect point cloud and convert it to a (mock) laser scan. If you robot will be driving in a 3D fasion then 2D mapping is not appropriate and you'll have to build a 3D map.

Out of interest why is a 3D map not suitable to your needs?

There are two well developed ROS systems for doing exactly what you want. RTAB-Map is the one we prefer in our lab, but the alternative rgbslam is also very good and depending on your exact use case may be a better solution.

Both of these systems are designed specifically for RGB-D sensors such as the Kinect, and have fairly good setup guides.

Hope this helps.

Update: You're trying to do 2D mapping with a 3D sensor, how is the robot moving? If it's driving over a flat floor and will not tilt, drive up ramps etc then you can extract a 2D slice from the kinect point cloud and convert it to a (mock) laser scan. If you robot will be driving in a 3D fasion then 2D mapping is not appropriate and you'll have to build a 3D map.

I think the documentation of rgbdslam is not up-to date. We've made it work on kinetic without any problems.

Out of interest why is a 3D map not suitable to your needs?

There are two well developed ROS systems for doing exactly what you want. RTAB-Map is the one we prefer in our lab, but the alternative rgbslam is also very good and depending on your exact use case may be a better solution.

Both of these systems are designed specifically for RGB-D sensors such as the Kinect, and have fairly good setup guides.

Hope this helps.

Update: Update: You're trying to do 2D mapping with a 3D sensor, how is the robot moving? If it's driving over a flat floor and will not tilt, drive up ramps etc then you can extract a 2D slice from the kinect point cloud and convert it to a (mock) laser scan. If you robot will be driving in a 3D fasion then 2D mapping is not appropriate and you'll have to build a 3D map.

I think the documentation of rgbdslam is not up-to date. We've made it work on kinetic without any problems.

Out of interest why is a 3D map not suitable to your needs?

Update: Theoretically 3D mapping is more complicated, but since the software has already been made for you, and you're using a 3D sensor it will be easier for you to implement 3D than converting your sensor to 2D.

You can find the ROS installation instructions to RTAB map here and there are good set of tutorials, including this one here which describes how to do mapping with several sensors including the kinect you're working with.

Hope this gets you started.