Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Deutsches Museum - TSDF Tuning #1058

Closed
alexmillane opened this issue Oct 11, 2018 · 2 comments
Closed

Deutsches Museum - TSDF Tuning #1058

alexmillane opened this issue Oct 11, 2018 · 2 comments

Comments

@alexmillane
Copy link

Thanks @kdaun for the addition of TSDF mapping; this is great work and I'm excited to use it!

I was just wondering if someone had played around and found some good tunings for the Deutsches Mueseum dataset? With the default params in the scans appear to "slide around" quite a bit relative to the (sub)map.

@kdaun
Copy link
Contributor

kdaun commented Oct 12, 2018

The TSDF is quite sensitive to the parameter configuration, here a configuration for the Deutsches Museum dataset and the resulting map:
deutsches_museum_tsdf
In my experience, in comparison to the Occupancy Grid pipeline, TSDF can improve the accuracy, the robustness to poor initial guesses and the convergence rate of the scan matching at the cost of a more expensive map updated and worse upper bounds for the branch-and-bound constraint search.
The airplane hall in the Deutsche Museum yields a reduced accuracy, as the TSDF is more sensitive to the strongly moving point observations when scanning the wings.

I'm still working on the TSDF approach and looking forward to hear about your results.

@alexmillane
Copy link
Author

Cheers for this! Works great! Looking forward to seeing how things progress. I'll let you know if I manage to come up with anything interesting.

Again, great work and thanks for the help.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants