Intelligent Reflective Surface-Assisted Visible Light Communication with Angle Diversity Receivers and RNN: Optimizing Non-Line-of-Sight Indoor Environments
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper explores enhancing indoor VLC systems through the integration of Intelligent Reflecting Surfaces (IRS) and Angular Diversity Receivers (ADRs), utilizing an RNN-based optimization framework. It dynamically optimizes IRS positioning to maximize received power and SNR, even in challenging NLoS environments. Comprehensive simulations validate the system's performance improvements, demonstrating robustness and scalability for practical applications.
1.The experiments on received power and SNR optimization in VLC scenarios based on RNN can be compared with other models, such as CNN and BP networks, to demonstrate better performance.
2.It appears that only one LED is used as the signal transmitter in the study. However, in typical indoor environments, there are usually multiple LEDs. What would be the results if multiple LED transmitters were present?
3.For a multivariable function optimization problem, such as the angle optimization issue of ADR, could a more suitable network, like KAN network, be employed?
4.In the VLC scenario, increasing the number of epochs improves both the received power and SNR. This makes the experiment seem redundant. It is suggested to replace it with other metrics.
Author Response
Thank you to the reviewer for your valuable comments and suggestions in improving the paper, which were fully addressed and responded to in the attached document and the modified paper.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper presents a novel strategy combining IRS and ADR in VLC systems to improve transmission efficiency under shadowing conditions. The idea is innovative, but some issues need to be addressed:
1. Section 1: Why was RNN chosen as the optimization method? Many existing studies, including traditional and other learning-based approaches, offer better performance with lower costs. Please explain why RNN is necessary over these alternatives.
2. Section 7: Add a figure showing the RNN structure used, including the network layers and their components. This will help evaluate the algorithm.
3. Section 8.1, Figure 4(c): The description states the maximum received power is 12.11 dB, but the colorbar doesn't exceed 10 dB. Please check and correct this inconsistency, as well as similar figures.
4. Figure 4 Description: If the figure is correct, explain why MRC Hemispherical (12.11 dB) has a lower maximum received power than MRC Pyramidal (15.92 dB), even though MRC Hemispherical is stated to perform better? How does this align with the conclusions?
5. Section 8.5, Figure 7: Why does the received power stay the same as training epochs increase? Shouldn't performance improve? Also, if MRC Pyramidal reaches 20 dB even when training epoch=0, then what is the point of using RNN? Please add sufficient explanations for this issue.
6. Abstract: The abstract is detailed but could be more concise. Shorter sentences would improve readability and clarity.
Comments on the Quality of English LanguageLanguage could be more concise.
Author Response
Thank you to the reviewer for your valuable comments and suggestions in improving the paper, which were fully addressed and responded to in the attached document and the modified paper.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe author has addressed all the comments raised in the previous review, implementing necessary revisions that significantly improve the manuscript. Consequently, I recommend accepting the manuscript for publication.
Author Response
Thank you for all the comments, which helped to greatly improve the article.