Learning to Predict 3D Rotational Dynamics from Images of a Rigid Body with Unknown Mass Distribution

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This article proposes a novel dynamic model estimation and prediction method for non-uniform or irregular objects based on Reinforcement Learning, and creative exploration was conducted on the application of image processing in dynamic model prediction, making the article highly innovative and research valuable. After reading this article, my suggestions are as follows：

1. Some of the basic knowledge used in the article requires some explanation and background introduction. For example, in the article, the SO(3) rotation group is used to describe the rotational kinematics of a rigid body, but the definition and basic principles of SO(3) are not introduced, which makes it difficult for those with insufficient understanding of the theory of rotation group to understand.

2. The article provides a detailed introduction to the modeling methods and datasets used, but only provides a brief introduction to the reinforcement learning methods used, and then provides the final results. This makes the coherence of the article slightly lacking, making it difficult to explain the effectiveness of the method at the algorithm level.

3. If a comparison chart of the results between this method and other estimation and prediction methods can be provided, it will better reflect the effectiveness and superiority of the method adopted in the article.

Author Response

Please see the attachment. Comments to this reviewer are under "Reply to Reviewer 1".

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper concerns an application of current concern for space research, it is well-structured and presented, and the results interesting. 

Some points that should be addressed in the paper are:

- some information about computational training time of the approach / expected inference time 

- the approach is formulated based on dynamics of rigid bodies, however, cases with satellite with flexible appendages - such as solar panels - which influence the attitude of the S/C (and induce a change in time of the mass distribution) are not considered (from table 1, you have higher errors in case of satellites and not prism/cubes). The authors should elaborate on this aspect.

- there is a typo in the affiliation #2 "Aerosapce". Please revise the paper to correct more typos

Comments on the Quality of English Language

The quality of English is good (paper written by native speakers)

Author Response

Please see the attachment. Response to this reviewer are under "Reply to Reviewer 2".

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The comments and suggestions are included in the attached file.

Comments for author File: Comments.pdf

Author Response

Please see the attachment. Response to this reviewer are under "Reply to Reviewer 3".

Author Response File: Author Response.pdf