Differences between rtabmap's VIO aprroches and zed's

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Differences between rtabmap's VIO aprroches and zed's

Williamc17
Dear people,

I am a graduate student in China, and also a beginner in ROS and rtabmap. Now I am using rtabmap_ros based on zed, trying some VIO approaches (if there really are some), to help my biomimetic bipep climbot in school's lab to do 3D mapping in the featureless environment.

I have seen many topics in the forum and Github, but I am still so confused.

What I want to get IS A good indoor 3D mapping, which's odommetry can be showed and compared to the ground truth.

These are my questions:

1. When I using rtabmap with zed, are there really different VIO approaches I can try? How I can know which is the best one?

2. As the tutorial in RGB-D Handheld Mapping of rtabmap_ros said, if I choose B) With zed odometry, does it means that the published odometry is from zed, and it's nothing to dot with rtabmap? After all, this is the zed's VIO approach. If I don't have zed's source code, I can't tune parameters of the odometry. Am I right?

3. If I want to use rtabmap's VIO approaches, HOW should I do? May I ask for some tutorials?

4. In on topic in the forum, I found no matter using what VIO approach, tranform between sensors should be calibrated. So finishing camera/IMU calibration with kalibr tool, what to do next with those XXX.yaml files? How I can link them with rtabmap? Maybe I should publish them in tf tree or are there are other approaches?

I hope my words can be understood except for some syntax error and typo. And it's very kind of you to let me know how you help me correct them.

Looking forward to your reply.

Yours,
William
Cheers,
William
Reply | Threaded
Open this post in threaded view
|

Re: Differences between rtabmap's VIO aprroches and zed's

matlabbe
Administrator
Hi William,

1. For VIO, if you have a ZED-MINI, I think the zed sdk does VIO. Otherwise, you would have to look for open source VIO approaches, like Okvis, MSCKF_VIO, rovio or VINS.

2. Yes, if you choose the zed odometry commands, odometry from zed node will be sent directly to rtabmap node. Yes, if they don't expose parameters of their odometry approach, you won't be able to tune them.

3. On ROS, I recommend that you use the ROS packages of the VIO approaches you will find.

4. OKVIS has some documentation on how to start it with config from Kalibr.

cheers,
Mathieu