Design and Field Evaluation of an End Effector for Robotic Strawberry Harvesting

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The work titled Design and Field Evaluation of an End Effector for Robotic Strawberry Harvesting focuses on robotic end effector for strawberry harvesting.

Overall, the paper is interesting and organized but there are some point that should be addressed.

The work premises suggest that the focus is on the platform design and testing but the test is limited to the end effector and manual operation. The system is simulated. The primary focus is instead the optimization of the end effector design.

It is not completely clear how the end effect following the harvest move the berry or manages it. How the time compares with hand picking should be highlighted or how it compares with benchmark competition. Those elements should be highlighted and discussed.

Result section and conclusion seem to be the weakest part of the paper.

 

Some wording should be revised. E.g. page 9 speculated by the kinematics might read better with calculated instead of speculated. Or realization might read better characterization. Please review.

Figure 4 font should be increased, or image broken down.

Overall images should be increased in size and for multiple sub images a letter and specific description should be presented.

The work would benefit from a digital support material.

Some more reference could be added.

Author Response

Reviewer 1

Comments and Suggestions for Authors

The work titled Design and Field Evaluation of an End Effector for Robotic Strawberry Harvesting focuses on robotic end effector for strawberry harvesting.

Overall, the paper is interesting and organized but there are some point that should be addressed.

The work premises suggest that the focus is on the platform design and testing but the test is limited to the end effector and manual operation. The system is simulated. The primary focus is instead the optimization of the end effector design.

It is not completely clear how the end effect following the harvest move the berry or manages it. How the time compares with hand picking should be highlighted or how it compares with benchmark competition. Those elements should be highlighted and discussed.

The general procedures for how the robot will manage/move the strawberry following the harvest can be found in figure 4 wherein the system will generate and follow a path to the collection bin with minimal collisions. However, it is not within the scope of research for WSU-TC to implement and explain those processes, as Abberit was responsible for path planning algorithms/machine vision processes and were not willing to disclose their work until it was finished.

To meet as much as the reviewer’s comments, Table 4 on a comparison of the harvesting times for benchmark competition strawberry harvesters utilizing single robotic arm configurations has been added.

Table 4. Comparison of harvesting times utilizing single robotic arm configurations.

Strawberry harvester	Harvesting time per berry (s)
Yamamoto et al. [11]	31.5
Xiong et al. [15]	6.1
Organifarms “BERRY” [18]	28.2
Octinion “Rubion” [21]	4
Tituaña et al. [22]	7.5
This work	2.8, 5*

* Calculated

 

Result section and conclusion seem to be the weakest part of the paper.

The authors agree with what the reviewer has commented. The result section and conclusion have been reinforced in response to the reviewer’s concern. 

 

Some wording should be revised. E.g. page 9 speculated by the kinematics might read better with calculated instead of speculated. Or realization might read better characterization. Please review.

Thank you very much for the helpful comment. The words have been changed as suggested.

 

Figure 4 font should be increased, or image broken down.

Thank you for the comments. Figure 4 has been redrawn as suggested.

 

Overall images should be increased in size and for multiple sub images a letter and specific description should be presented.

The authors agree with what the reviewer has suggested. Most images have been adjusted in size. Also, letters and specific description have been added as suggested.

The work would benefit from a digital support material.

Thank you for the suggestion. The MATLAB codes for the workspace analysis have been added in a pdf form as suggested.

Some more reference could be added.

Thank you for the comment. More references have been added

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

 

This research designed a modular robotic strawberry harvesting system with a Delta X robot and pneumatic end effector. It optimized the end effector, validated its high success rates (94.74% in simulation, 100% in fields), analyzed the workspace, and demonstrated the hardware's ability to harvest strawberries in 2.8 - 3.8 seconds, advancing the field of agricultural robotics. Here are my comments:

 

The kinematic analysis of the Delta X robot is a solid foundation. However, it might benefit from additional theoretical exploration. For example, incorporate a more comprehensive error analysis in the workspace determination considering factors like manufacturing tolerances, mechanical compliance, and environmental uncertainties (e.g., wind-induced vibrations in the field).

 

The sample size and diversity in the field trials could be improved.

 

The author should calculate confidence intervals for the success rates and performing variance analysis on the harvesting times across different test scenarios (e.g., different growth stages or plant arrangements).

 

This paper lacks a recent related advancements, like Multimodal Strain Sensing System for Shape Recognition of Tensegrity Structures by Combining Traditional Regression and Deep Learning Approaches; and Predicting flow status of a flexible rectifier using cognitive computing.

 

Provide specific examples of how different module combinations were utilized and adjusted during the field trials to handle diverse scenarios such as varying plant heights, stem thicknesses, and foliage densities.

 

The manuscript reports success rates and harvesting times, but how do these metrics account for the quality of the harvested strawberries? Were there any evaluations of post-harvest quality factors such as bruising, shelf-life, or nutrient retention?

Author Response

Reviewer 2

 

Comments and Suggestions for Authors

 This research designed a modular robotic strawberry harvesting system with a Delta X robot and pneumatic end effector. It optimized the end effector, validated its high success rates (94.74% in simulation, 100% in fields), analyzed the workspace, and demonstrated the hardware's ability to harvest strawberries in 2.8 - 3.8 seconds, advancing the field of agricultural robotics. Here are my comments:

 

The kinematic analysis of the Delta X robot is a solid foundation. However, it might benefit from additional theoretical exploration. For example, incorporate a more comprehensive error analysis in the workspace determination considering factors like manufacturing tolerances, mechanical compliance, and environmental uncertainties (e.g., wind-induced vibrations in the field).

 Thank you very much for the helpful comment. The authors did not consider any of these for this paper. Perhaps we can consider that exploration in the next work.  

 

The sample size and diversity in the field trials could be improved.

The authors agree with what the reviewer has suggested. However, due to limited access to strawberry fields local to us, it was difficult to acquire opportunities for field testing.

 

The author should calculate confidence intervals for the success rates and performing variance analysis on the harvesting times across different test scenarios (e.g., different growth stages or plant arrangements).

Thank you for the comments. Since testing opportunities were slim and needed to find the time to work together with Abberit, field testing was limited to July which is known to be the end of the growing season. This means less strawberries were available and more foliage present. As we described in the manuscript, our future work would require that the system be characterized in varying growth stages. 

 

 

This paper lacks a recent related advancements, like Multimodal Strain Sensing System for Shape Recognition of Tensegrity Structures by Combining Traditional Regression and Deep Learning Approaches; and Predicting flow status of a flexible rectifier using cognitive computing.

Thank you very much for the invaluable feedback. However, when Abberit and WSU Tri-Cities proposed an affordable and open-source robotic solution, Abberit has worked on implementing machine vision for identifying the stems of strawberries and path planning for the robot. WSU Tri-Cities have focused on design of an end-effector to incorporate with Delta X1 and field evaluation.

 

Provide specific examples of how different module combinations were utilized and adjusted during the field trials to handle diverse scenarios such as varying plant heights, stem thicknesses, and foliage densities.

Thank you for the helpful comment. Unfortunately, due to testing opportunities being limited to the end of the growing season, the only type of foliage density that was able to be tested was of the most burdensome conditions. This means the robot will handle anything less with ease. The robot did not need accommodation for varying stem thickness, as the pneumatics are quite powerful in cutting/pinching all sizes of stems and need only that the utilized pressure be increased when requiring more power. As briefly mentioned in the workspace validation, the robot's frame was fitted with additional 20/20 extrusions to be used as adjustable "legs" via sliding nuts. We do appreciate this comment, as it gives us advice on how to handle future testing with the variability of different growth stages and a larger sample size. The modularity of the robot is more so convenient with respect to the variable crop row/raised bed dimensioning of other farms, mentioned in the manuscript.

 

The manuscript reports success rates and harvesting times, but how do these metrics account for the quality of the harvested strawberries? Were there any evaluations of post-harvest quality factors such as bruising, shelf-life, or nutrient retention?

Since the available sample size was so small during testing opportunities, post-harvest quality factors were not considered for this work. When testing the robot with Abberit's finished machine vision processes in the future, these metrics will absolutely be considered as they are crucial to any distributor to maintain the same quality as human workers. Knowing that other robotic configurations have used similar methods for their end effector and are nearing a commercial product, there isn't any hesitation that this product is a viable solution with minimal to no bruising. Shelf-life and nutrient retention are likely to be factors that depend heavily on bruising metrics.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Adequate now.