Comments to author (Associate Editor)
=====================================

The authors proposed a new Imitation Learning method to
lower the workload of human users in robot training. The
problem is important and the formulation is complete. The
paper is acceptable. Please see the comments of the
reviewers and address those minor issues. 


----------------------------------------
Comments on Video:

The video is clearly presented. More details of the
methodology overview should be provided if possible. 
Comments to author (Editor)
================================

Please follow the AE and reviewers' comments to prepare the
revision.


----------------------------------------
Comments on Video:

Accept












The authors proposed a LazyDAgger method to improve the
efficiency of interactive imitation learning by reducing
context switches between supervisor and autonomous control.
I consider this research is very important in the field,
and it is a good fit for this conference. 

I have several suggestions for this manuscript:

1. How the LazyDAgger compared with the reinforcement
learning framework? I think there are many similarities
between these two. It will be good if the authors can
discuss it.

2. What are the assumptions for LazyDAgger? 

3. For the supervisor controls, are they always be the
same? Or they also have some variances that can influence
the control process?

4. In Fig.3, what is the meaning of shadows? Are they
represent the uncertainty? 

----------------------------------------
Comments on the Video:

The video is helpful to better understand the research and
case study demonstrated in this paper.














This is a well-written paper for an important topic of
interactive imitation learning with minimum supervisory
needed during system switch. The paper is solid in the
methodologies and demonstrated with sufficient simulation
and real case studies.

----------------------------------------
Comments on the Video:

The video and annotations are very clear.