Shelf Replenishment Based on Object Arrangement Detection and Collapse Prediction for Bimanual Manipulation
, Damien Petit 1, Takao Nishi 1, Kazuyuki Nagata 2, Weiwei Wan 1 and Kensuke Harada 1,2Round 1
Reviewer 1 Report
The paper presents a method for safe replenishment process from a single depth image. The image is processed by collapse map network to predict the existence of collapsing objects. The proposed strategy provides a decisions and actions of safety movement of the shelf content (several objects). The authors provided the experiments with a real bimanual robot which performed three scenarios: shelved, stacked, and random. The objects were randomly placed in each scenario. The experimental results verify the proposed method.
The method is original.
The paper is not well written. It is extremely difficult to read the paper. I recommend the authors to strictly follow the proposed organization of the manuscript (see Robotics | Instructions for Authors (mdpi.com)), where proposed sections should be: 1. Introduction, 2. Material and methods, 3. Results 4. Discussion and 5. Conclusion if section Discussion is too long. This would help a lot that the manuscript would be more readable. I also recommend to follow the strict instructions, how to make citations (numbering of citation is not OK), where to appear the Tables or Figures in the text etc. So, you need to rearrange the manuscript.
By the way, my opinion is that your research is good and present important scientific data, but you have to adapt to the readers, who not need to be exhausted and frustrated by the reading your paper. This would help you to receive more citations for your paper.
The english is good, I found one strange sentence on page 6, line 173, where was written: "... objects to have moved ..." Probably it should be written: "objects to be moved".
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper presents a shelf replenishment system that selects the safest action based on a collapse prediction estimator. The paper is well written and well structured. The work is solid and technically sound. I have one minor comment.
The training time and detection time should be provided, so readers can better understand the overall implementation.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This research presents an approach for generating a replenishment process with a bimanual robot through three step process: firstly, classifying the objects, secondly predicting the safety of different actions, and thirdly selecting and conducting the optimal action plan. The topic is interesting. The manuscript can be accepted for publications after addressing below concerns.
1. Author should add one paragraph in section two showing the research gap and the reasons for conducting this research.
2. Section 5 should be mentioned in line 64 and 65 (page 3).
3. The success for random scenes is quite low. Author should compare the results achieved in this research with other’s and explain the reasons for the differences. Author should propose some solutions to increase the successful rate. How did authors get the number in total column in table 2.
4. Authors currently only consider about the damage of items from collapse, it is important to consider other damage source such as from robot hand’s clamping force, which can directly break the objects.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors have improved the manuscript as was suggested!
