Next Article in Journal
A Multi-Source Domain Adaptation Method for Bearing Fault Diagnosis with Dynamically Similarity Guidance on Incomplete Data
Previous Article in Journal
Event-Triggered Bipartite Formation Control for Switched Nonlinear Multi-Agent Systems with Function Constraints on States
 
 
Article
Peer-Review Record

Robot Closed-Loop Grasping Based on Deep Visual Servoing Feature Network

Actuators 2025, 14(1), 25; https://doi.org/10.3390/act14010025
by Junqi Luo 1, Zhen Zhang 2,*, Yuangan Wang 1,* and Ruiyang Feng 1
Reviewer 1:
Reviewer 2: Anonymous
Actuators 2025, 14(1), 25; https://doi.org/10.3390/act14010025
Submission received: 6 December 2024 / Revised: 28 December 2024 / Accepted: 9 January 2025 / Published: 12 January 2025
(This article belongs to the Section Actuators for Robotics)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

Below are my comments:

1. Performance Testing in Complex Environments

The method proposed in the paper has shown promising results under controlled conditions; however, its performance in scenarios with complete occlusions or rapidly moving targets has not been sufficiently demonstrated. This limitation restricts the applicability of the method in a broader range of real-world applications, where robots often need to operate in complex and variable environments such as outdoors or industrial settings.

2. Validation of Model Generalization

While the model's effectiveness on specific datasets is established, its generalization capabilities when faced with diverse datasets remain untested. This could lead to performance degradation when the model encounters new scenes, especially when there are significant differences in object shape, size, and material from the training data.

 

 

3 . Challenges in Sim2Real Transfer

The paper does not address how to bridge the gap between the model's performance in simulated environments and real-world deployment, which may stem from discrepancies in physical properties and sensor noise between the simulated and real worlds. This could result in suboptimal performance when the model is applied in practical scenarios.

4. Real-time Performance of the Model

In certain application scenarios, such as emergency response or human-robot interaction, robots need to make decisions and respond quickly. The paper lacks an assessment of the model's performance under these strict real-time requirements, which may imply that the model fails to meet the demands of time-sensitive tasks in practical applications.

5. Robustness Testing of the Model

Robustness measures a model's ability to maintain performance in the face of anomalous inputs. The paper does not include tests of the model under adversarial attacks, sensor failures, or unexpected disturbances, which could lead to a decline in performance when the model encounters uncertainties and challenges in the real world.

6. Diversity in Experimental Design

The experimental design should ideally emulate the diversity and complexity of the real world. If the experiments in the paper are too singular, they may not fully assess the model's performance under varying conditions, especially when confronted with novel situations not present in the experiments. This limitation could affect the comprehensive evaluation of the model's practicality and reliability.

 

Comments on the Quality of English Language

It will be better to  The English could be improved to more clearly express the research.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper entitled “Robot Closed-Loop Grasping Based on Deep Visual Servoing Feature Network” presents a novel deep learning-based approach to enhance robotic grasping in complex environments.

 The paper is well structured, the information is delivered in an understandable way, and the contributions of the authors are underlined in an adequate manner.

Nevertheless, there are some aspects that all authors should consider in order to offer a valuable scientific paper.

 

1.      The authors should take into consideration the introduction of a paragraph, at the end of section 1, where a short description of the section could be made.

2.      I suggest providing a detailed explanation for the choice of Darknet-53 as the backbone for the DVSFN architecture. Highlight the trade-offs it offers in terms of computational efficiency, accuracy, and real-time performance compared to other architectures.

3.      I recommend elaborating on how the proposed approach outperforms other state-of-the-art methods, such as Keypoint RCNN, RTMPose, and HRNet. Specifically, describe how the system handles occlusion, low-light conditions, and other challenges better than these alternatives.

4.      I suggest including a discussion about the limitations of the method, such as its reliance on advanced hardware or sensitivity to environmental variations.

5.      I also suggest offering more detailed interpretations of the experimental results, including statistical reliability (e.g., confidence intervals or variability across trials). This will help reinforce the validity and significance of your findings.

6.      Please take into consideration splitting Figure 7 in order to enlarge the resulting figures and clearly present the images from the camera. In the current format, their interpretation is very difficult.

7.      I recommend discussing the extent to which the model can be generalized to other object categories, robotic configurations, or camera setups. Providing insights into its adaptability will enhance the relevance of your work for broader applications in robotic systems.

Best regards

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors, 

congratulations on the improvement of your manuscript.

I have no more comments

Best regards.

Back to TopTop