Zed-ros-wrapper: What is the best SLAM algorithm for ZED on ROS

[ArtlyStyles]

Hi, I have built a MIT Racecar platform https://mit-racecar.github.io. Currently, I do not have the Hokuyo lidar. I am wondering if I can run SLAM algorithm with ZED. Which algorithm is the best for ZED?

Thanks,

[rbonghi]

Hi @ArtlyStyles , I used on my robot Panther different type of algorithms:

• gmapping
• rtabmap
• Cartographer
• hector slam

My first approach was use the pointcloud to laserscan converter and use the laserscan message to build my map. More or less your same case with your racecar, however the rtabmap with the information from the ZED can build a 3D map.

You can see one of my reconstructions here https://skfb.ly/6sUS7

[ravikt]

I have used ORBSLAM2 and SVO. Both seems to work well with ZED camera.

[Myzhar]

I close the issue but you can continue to pay comments if you need further help

[belal-ibrahim]

@rbonghi May i ask about the results of gmapping using zed without laser sensor?
Were there any drawbacks.
Can zed along with odom be good in mapping large scale env?

[Myzhar]

@belal-ibrahim gmapping requires a laser scan message to work, so you must convert a 3D point cloud (or the depth map) to a simulated 2D scan message losing a lot of useful information.
I suggest you to take a 3D approach using the 3D mapping feature provided by the ZED SDK or using RTABmap that is the state of the art package for 3D localization and mapping