Offloading Strategy Based on Graph Neural Reinforcement Learning in Mobile Edge Computing

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper under review introduces an approach to enhance decision-making in Mobile Edge Computing (MEC) by integrating dynamic graph with deep reinforcement learning.

Here are my comments:

• A major concern is the perceived lack of novelty in the methodology. The use of Graph Neural Networks (GNN) and Deep Reinforcement Learning (DRL) is well-established in the field, with many studies already applying similar techniques.

• Given the numerous equations and variables introduced, it would be beneficial to include a notation table for clarity.

• Citations are needed for some of the referenced equations.

• An analysis of the algorithms' complexity and scalability would be advantageous.

• The scope of the data and parameter settings should be clarified, explained, or supported with references to enhance understanding and validation of the experiment result.

Comments on the Quality of English Language

This paper is well-written, making it easily understandable for readers, which is crucial for conveying complex technical concepts.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Comments for author File: Comments.pdf

Comments on the Quality of English Language

There are some minor typos and English language mistakes throughout the manuscript; please correct them.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper introduces an innovative offloading strategy for MEC by integrating GNN with DRL. The proposed method, M-GNRL, addresses the limitations of traditional DRL in handling dynamic, graph-structured data in MEC environments. By aggregating node features with GNN and integrating them into a hierarchical DRL framework, M-GNRL improves the accuracy and efficiency of task offloading decisions. Overall this paper is well written and technically correct. Here are some revision suggestions:

1) The Introduction Section would benefit from a more explicit statement of the research gap and how this work addresses it.

2) The system model description is good, however, including a simple illustrative figure early in the section might help readers understand the setup better. In addition, some parameters and their relevance are not immediately clear, such as the environmental influence coefficient θ.

3) The explanation of the GraphSAGE model and its integration could be more concise.

4)The notation used is sometimes inconsistent or unclear, making it hard to follow the equations and their purposes.

5)The experimental setup lacks a discussion on the choice of parameters and their justification.

6) The performance metrics used (convergence speed, task arrival rate, system cost) are appropriate, but the comparative analysis could be more in-depth, with a better discussion on the limitations and potential improvements.

7) There are minor grammatical errors and inconsistencies that affect readability.

Comments on the Quality of English Language

Overall, the paper is well written. Here are some minor comments:

1. There are minor grammatical errors throughout the paper, including issues with verb tense consistency.

2. Some sentences are too complex and lengthy, making them difficult to read and understand. Breaking these down into shorter, more concise sentences would improve clarity.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have well addressed my concerns. The language could be further improved.

Comments on the Quality of English Language

Minor editing of English language is needed.

Author Response

再次感谢审稿人的提醒，我们已经仔细检查了整篇文章，纠正了一些语法错误和单词，现在我们已提交修改后的文章。