Development of Continuum Robot Arm and Gripper for Harvesting Cherry Tomatoes

Round 1

Reviewer 1 Report

In this paper, a Cherry Tomato Harvesting System Based on line driven discrete manipulator and neural network technology is developed. The mechanical structure, gripper design and the development of cherry tomato recognition system are introduced. Thank you for your efforts in developing this work and the results shared in your article. If you like, I would like to share some suggestions:

1. The last sentence of the abstract is missing punctuation, and the text needs careful proofreading. Some keywords are too broad, such as "gripper, design";

2. "202. robot design" should be "2. robot design", so the text content and language of the full text need to be carefully checked;

3. Figure 3 (b) is too small to be seen clearly

4. When calculating distance, what is image plane (r) and object plane (R)?

5. In "4. kinematic and kinetic formulas", the title sequence number of each section is incorrect.

6. How is the data set used for training collected and what is its structure?

7. There are few references in recent 3 years. Please add the research status in recent years.

8. In Chapter 4, the figures and most formulas are the same as those in reference [24][25]. Please clarify the contribution or improvement of this paper.

9. How to prove that YOLOv5 has the best performance in Cherry Tomato detection scenario? Proved by comparing other algorithms.

Thank you very much.

Author Response

Response to the Reviewer

Dear Reviewer thank you so much for your professional feedback and suggestions. (point by point responses would be in the bracket with a bold text )

1. The last sentence of the abstract is missing punctuation, and the text needs careful proofreading. Some keywords are too broad, such as "gripper, design"; (The Abstract was rewritten completely based on suggested comments, also the robot gripper specified as gripper tool. )
2. "202. robot design" should be "2. robot design", so the text content and language of the full text need to be carefully checked; (The typos of the whole text had been refined )
3. Figure 3 (b) is too small to be seen clearly (The figure split by the position and now the figure fonts are clearly seen)
4. When calculating distance, what is image plane (r) and object plane (R)?(This is great question, r is an object image and R is an object image projection, same as a human eye, Human eye perception works based on objects with predefined sizes and based on object size in a horizon , human can identify approximate distance.)
5. In "4. kinematic and kinetic formulas", the title sequence number of each section is incorrect. (The title sequence numbers are refined )
6. How is the data set used for training collected and what is its structure?( the dataset had been collected in the greenhouse by web camera and that web camera had been utilized in the experiment. )
7. There are few references in recent 3 years. Please add the research status in recent years. ( The introduction part of the paper had been rewritten completely and the reference list of the paper are updated )
8. In Chapter 4, the figures and most formulas are the same as those in reference [24][25]. Please clarify the contribution or improvement of this paper.

(Dear Reviewer this is for readers convenience, because in control part we applied the forward and inverse kinematics, after I have received strong recommendations to add kinematic/kinetic formulations)

9. How to prove that YOLOv5 has the best performance in Cherry Tomato detection scenario? Proved by comparing other algorithms.

The YOLOv5 is compatible with microprocessor based boards and it works easily, I have referred a one great paper on it (ref 30) shows nice evidence on it)

Thank you very much for your constructive feedback after your comments the paper became much better than before.

This is the link for the experimental video: https://www.youtube.com/watch?v=wauiqdEP_ak

Author Response File: Author Response.docx

Reviewer 2 Report

R: Request of the Reviewer

This paper proposes present a new tomato harvesting robot system with an arm and a tomato detection system. However, the topic of study is very interesting and experimentally investigated, the reviewer has some issues indicated below;

R: The whole abstract should be rewritten because it should be revised by the following order; background, purpose, method, results and conclusion.

R: In entire manuscript it is better to use “In this paper, it is done…” besides using first person plural like “we”. For example, in conclusion authors

R: The originality of the work are not obviously mentioned in the abstract and the content.

R: Introduction part and literature review is not satisfactory. More specific statements about the aspects of the problem already studied by other researchers. Author should add more up-to-date literature into the introduction part.

R: All figures and their text should be at the same format. And all figures and their text should be at the same size and format. Figures should be replaced with more descriptive ones. In some Figures grids could be added.

R: A proof reading by a native English speaker should be conducted to improve both language and organization quality. Entire article should be checked for typos.

R: In the conclusion the main results and contributions of the study should be clearly summarized by additional explanations. The results should be shown by data. Conclusion should be rewritten to clarify the solutions.

Author Response

Respond to the Reviewer

Dear Reviewer thank you so much for your professional review and suggestions. I have replied to your comments next to the comment in the brackets.

R: The whole abstract should be rewritten because it should be revised by the following order; background, purpose, method, results and conclusion.

(The abstract had been rewritten completely based on provided comments)

R: In entire manuscript it is better to use “In this paper, it is done…” besides using first person plural like “we”. For example, in conclusion authors

( I have replaced personal pronouns from the paper as much as possible)

R: The originality of the work are not obviously mentioned in the abstract and the content.

(The abstract rewritten and I have included the paper novelty ,thank you for such useful suggestion)

R: Introduction part and literature review is not satisfactory. More specific statements about the aspects of the problem already studied by other researchers. Author should add more up-to-date literature into the introduction part.

(The paper introduction rewritten by 90 percent and the reference list are updated as well)

R: All figures and their text should be at the same format. And all figures and their text should be at the same size and format. Figures should be replaced with more descriptive ones. In some Figures grids could be added.

(Dear Reviewer I would like apologize beforehand in this paper figures due to the background colors the fonts and sizes are should be different )

R: A proof reading by a native English speaker should be conducted to improve both language and organization quality. Entire article should be checked for typos.

 ( The paper was passed through proofreading service)

R: In the conclusion the main results and contributions of the study should be clearly summarized by additional explanations. The results should be shown by data. Conclusion should be rewritten to clarify the solutions.

(the conclusion part is rewritten and additional figure included to enhance the paper results.)

 

Dear Reviewer

Thank you for your constructive comments and suggestions, after refining of the paper based on above mentioned comments the paper definitely looks better.

Here is the link for the experimental video:

https://www.youtube.com/watch?v=wauiqdEP_ak

Author Response File: Author Response.docx

Reviewer 3 Report

Comments to Authors

The manuscript entitled “Development of Continuum Robot Arm and Gripper for Harvesting Cherry Tomatoes” is very interesting. This paper designs a complete set of tomato picking device and new robot arm and gripper are designed to pick tomatoes. The structure of TakoBot is introduced in detail and motion analysis of the TakoBot is shown through kinematic and kinetic formulations. Functional implementation in the experiment presented in the manuscript which is suitable for publication. However, some major revision needed before the final acceptance. There are so many shorting coming from abstract to conclusion (needs improvement) that needs to be fixed before it processed for publication sector.

Specific comments

Title: The title is good.

Abstract:

1. There are punctuation errors in the abstract.

2. Add some more brief description of results. The abstract lack of such information.

Introduction:

1. Overall reading of the manuscript is recommended for minor spell checking. (The second line of the introduction, “a typical lavour intensive work?”, a typical labour intensive work?)

2. Check the serial number. (“202. Robot design”)

Design section:

1. Please check the title format carefully, such as bold, serial number. (“4. Kinematic and Kinetic formulations”)

2. Why choose YOLO for the recognition of different types of tomatoes, has it been compared with other algorithms?

3. How can the recognition system effectively identify tomatoes that are obscured by leaves?

4. Only the precision of tomato recognition is mentioned, but the time required for tomato recognition is not mentioned.And after recognizing the tomato, the time for the movement of TakoBot is also lacking.

Conclusion: Should be improved and need some more specific information. Tomato harvesting process took average 65 seconds, but the time of identification, extraction and other processes is not mentioned. And no future specific plan mentioned.

Reference: It's best to update the references and replace them with new.

General Comments

Please double check the spelling, punctuation, ordinal numbers in the manuscript.

Double check that all references are cited within the text, and that all citations within the text have a corresponding reference. Double check the spelling of the author names and its affiliation.

Author Response

Respond to the reviewer

Dear Reviewer thank you so much for your professional review and suggestions. I have replied to your comments next to the comment in the brackets.

Specific comments

Title: The title is good.

Abstract:

1. There are punctuation errors in the abstract. (The abstract rewritten completely and checked for typos as well.)

 

2. Add some more brief description of results. The abstract lack of such information. ( In a new abstract I have included brief information about results as well).

Introduction:

1. Overall reading of the manuscript is recommended for minor spell checking. (The second line of the introduction, “a typical lavour intensive work?”, a typical labour intensive work?) (The introduction part had been rewritten completely as well and th references are also updated)
2. Check the serial number. (“202. Robot design”) (Already fixed this typo, thank you for check )

Design section:

1. Please check the title format carefully, such as bold, serial number. (“4. Kinematic and Kinetic formulations”) (The serial number of the chapters are corrected, thank you for check)
2. Why choose YOLO for the recognition of different types of tomatoes, has it been compared with other algorithms? ( That’s a great question, actually YOLO model is convenient in case of application by microprocessor based boards like Rasberry PI. Therefore, in the experiment exactly YOLOv4 and YOLOv5 demonstrated more stable work. )
3. How can the recognition system effectively identify tomatoes that are obscured by leaves? ( For our model detecting of ¼ of the tomato is enough to conduct the harvesting operation. For such case I build an obstacle to generate real environment)
4. Only the precision of tomato recognition is mentioned, but the time required for tomato recognition is not mentioned. And after recognizing the tomato, the time for the movement of TakoBot is also lacking. (This is the terrific question, I have included one more figure regards robot tomato harvesting trajectory and time. In total, the tomato recognition and picking of tomato takes 56 seconds, this is achievement in tomato harvesting, because we failed thousand times and process took more than 2 minutes.)

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thank you for sharing your work in this article and your careful revision. There are still some problems to be pointed out.

1. the serial number of Chapter 2 is incorrect

2. in Figure 25, it may be more intuitive to use a positive time axis

3. yolov5 needs more experimental results (such as the result graph of MSE). It is better to have a comparative test with other algorithms to illustrate the superiority of yolov5 performance

Thank you for your work.

Author Response

Response to the Reviewer

Dear Reviewer thank you so much for your professional feedback and suggestions. (point by point responses would be in the bracket with a bold text )

Thank you for sharing your work in this article and your careful revision. There are still some problems to be pointed out.

1. the serial number of Chapter 2 is incorrect (I have corrected the chapter 2 serial numbers, thank you so much )
2. in Figure 25, it may be more intuitive to use a positive time axis (The figure 25 is refined )
3. yolov5 needs more experimental results (such as the result graph of MSE). It is better to have a comparative test with other algorithms to illustrate the superiority of yolov5 performance (I have added a paragraph and comparison table with another algorithm Mask R-CNN)

Thank you so much for your spared time and feedback comments. Now the paper looks even better.

Reviewer 2 Report

The authors revised the paper and authors answers and revisions to reviewers’ requests are well explained and satisfactory.

Author Response

Dear Reviewer 

Thank you so much for your professional revision and spared time 

Provided feedback comments and suggestions definitely improved the paper. 

In the paper, I have included a paragraph and table of comparison of YOLOv5 with Mask R-CNN. Also some minor corrections