Re: RTAB-Map and Unity integration

Posted by viniciusbs on
URL: http://official-rtab-map-forum.206.s1.nabble.com/RTAB-Map-and-Unity-integration-tp4925p5005.html

Hi,

Thanks for the script and the tip on the covariances.

matlabbe wrote

but there is still a TF error between base frame and the camera frame. The laser scans don't match exactly with the cloud created from the camera.

I tried to optimize this, but it seems like there is always some misalignment in one pose or another. For example, when adjusting the x position of the camera, so that the laser readings and the point cloud would be aligned, it seemed that it needed different values depending on whether the obstacle was close or far away. I don't know if this is a problem of the lasers (Hokuyo) or a camera (Asus) distortion, or how to fix it.

matlabbe wrote

Following is another problem, this is only one scan. The bottom shows two layers of scans, how did you merge the scans from the 3 lidars? Maybe some TF tuning to get the scans overlaying each other.

I noticed this problem too (more or less same as above, there is always some misalignment no matter what values are used for tf) but wasn't aware that it could impact the mapping performance. We have 3 laser scanners and also use a custom node to merge the readings together. I was passing the merged reading to hector_mapping. I'll make a test of a run using only one of the lasers, see if it improves something.

In the meantime, I noticed that when loop closure is detected, the 2d map generated by rtabmap gets realigned, but the point clouds seem to don't update as much, remaining misaligned sometimes. I'm wondering if is there some parameter that I should tweak to improve this, or do you think it's just the tf problem?

Cheers,
  Vinicius