PickingDK: A Framework for Industrial Bin-Picking Applications

Round 1

Reviewer 1 Report

1- Abstract

The motive of the research is briefly highlighted. Method chosen and findings from the research are also included. However, the implications of the analyses carried out and the novelty of the research in the area should be highlighted.

 

2- Introduction

Line 17-19 - Does not cite consistently. Who claimed that “in the last years, the appearance of new sensors combined with new developments in Artificial Intelligence is expanding the use of robots to more complex tasks”?

Line 64 - Research objectives are given and is clear

Use more recent and relevant references throughout the introduction section.

 

3- Related work

Revise every citation style. For instance:

Line 91 - [4] presents a 3D object localization method from point…

Change to Drost [4] presents a 3D object localization method from point…

 

Others:

[5] uses a structured light sensor mounted

[6] employs neural networks that

[22] exploits a ROS based architecture to adapt different bin picking

[25] presents a versatile interface-based framework for the planning,

 

Line 78 – “Various works focusing on specific modules in a bin picking process are presented in the recent years”. Please include a table outlining the past research.

 

4- Overview of PickingDK

What kind of research framework did you suggest?

Any comparisons to the prior framework are possible?

 

5- Conclusions

“Finally, Section VI draws the conclusions and discusses future work”. Future work is not being discussed in Section VI. Please revised the conclusions and add the limitations of the research.

 

6- References

Most of the references, though, are outdated. References should contain more recent ones, particularly those from the previous three years.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

In the paper, the authors propose a bin-picking framework called PickingDK. The presented framework  architecture is based on plugins  that integrate different types of sensors, robots, tools.

The solution seems interesting, although it does not exhaust the topic.

The weak point of the work is:

- comparing the bin-picking framework only with RSO,

- bin-picking framework comparison made only on the basis of information presented on their websites,

- the paper contains typos and mistakes which should be improved.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript entitled "Picking DK: A Framework for Industrial Bin-Picking

Applications." is well written and is scientifically correct. I recommend that this manuscript can be accepted for publication after the following minor revisions:

1. Remove full stop ( . ) from the title.

2. Presentation of the manuscript lacks and is not up to the good standards and it so should be improved.

3. Write all references on the same pattern. Some references are incomplete, these should be completed.

4. Comparative analysis needs improvement.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

Good article in bin-picking for robotic applications. For readability of the paper,  step by step presentation of PickingDK from concept to implementation described all of necessary tools including hybridization. Stresses on hybridization and optimization.  .  addresses higher safety standard.  Good collection of references.      

Describes the architecture of pickingDK – what does this acronym stand for?

What is 7 DOF in frames or links??

Why only 13 pieces are used in validation? What are they, real ones or synthetic.

Magnetic gripper does not need AI to detect shape? Can magnetic gripper pick multiple objects when one is desired?

 Why weight is not considered as data parameter for object recognition: the shape and size, not weight.

AI metrics Percision, Recall are used, how about  F1, why was it excluded? Three go together.

Does 4.1s perframe mean  4.1 fps?

“it is critical to have zero FP to ensure efficacy in the 338 picking process.”  It is a good expectation, but not realistic. May use the word virtually zero.

What is the nature of two different end-effectors? Describe them, also  why not 4?

What is virtual plug-in elaborate, define? Synthetic data generation can be subjective to demonstrate, the system amy break don on real data.

Too many typos, grammar error to enumerate here. Large sentences are ambiguous and sidetrack the intent.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Authors have addressed most of the issues well. I recommend this paper for further approval procedures and publishing in the Applied Sciences journal.