Z position drift by running cartographer with KITTI dataset

[guillermoherreraf]

Greetings all!

I am using the configuration files you provided for KITTI datasets in the following dataset: https://s3.eu-central-1.amazonaws.com/avg-kitti/raw_data/2011_09_30_drive_0018/2011_09_30_drive_0018_sync.zip, which I converted to .bag with the kitti2bag tool.

I'm comparing V-SLAM and Cartographer outputs for the same dataset. However I see a drift in the z position of the trajectory for the Cartographer output, as it can be seen in the following figures.

This first one is the 2D trajectory.

But when I plot it in 3D, it can be appreciated that the IMU z coordinate is wrong, as those separated blue lines (start and goal) correspond to the same street.

What do you think may be the issue here? I'm guessing there is something to fix in the .lua configuration files.

Thanks in advance!

[inkyusa]

Hello GuillermoHerreraF, It looks quite off in z-direction, and I agree with your opinion that you can tune the submap size and constraint min score for more loop closures. In addition, these links discussing z-drifit https://github.com/googlecartographer/cartographer_ros/issues/1252 https://github.com/googlecartographer/cartographer_ros/issues/809 Hope this helps, Best,

[mhaboali]

Hi @GuillermoHerreraF

Could you give more information about how did you do this comparison?

Your information would be much appreciated!