Thanks very much for posting these lectures! I will definitely be trying out the adaptive rho function approach, it looks fantastic.

Will there be lectures on Direct methods?

Do you have a link to the EM-based adaptive robust kernel estimation paper? Haven’t been able to find it online

@CyrillStachniss

4 жыл бұрын

Still under submission but hopefully out in the not to far future

@SpatialAIKR

3 жыл бұрын

I think this is the paper. arxiv.org/abs/2004.14938

Is the paper of the ICP example under submission?

@CyrillStachniss

4 жыл бұрын

Yes. IROS 2020.

@SpatialAIKR

3 жыл бұрын

I think this is the paper. arxiv.org/abs/2004.14938

how to applay it in a case of mixed outliers/outliers-free measurements ?

@CyrillStachniss

Ай бұрын

You always need also outlier-free observations. If you have 100% outliers, no state estimator can do its job…

@kaptorkin

Ай бұрын

@@CyrillStachniss write it a little bit clearer. For example we have odometry measurements (they don't have outliers, and we decided their errors distribution as Gaussian one) and in the same time we have landmarks visual measurement (that contains outliers) . So we have 2 groups of measurements, without outliers and with outliers. In case we apply robust kernel for landmarks visual measurements and for odometry measurements we apply usual L2 kernel ? Or we just mix all kinds of measurements and apply one robust kernel for each kinds of measurements?