A Novel Scheduling Algorithm for Improved Performance of Multi-Objective Safety-Critical WSN Using Spatial Self-Organizing Feature Map

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper examines IoT advancements, emphasizing Wireless Sensor Networks (WSN). It highlights the importance of dependability in applications like fire detection. The study proposes a Self-Organizing Feature Map (SOFM) to improve node scheduling in WSNs, demonstrating that SOFM outperforms the Bat node scheduling algorithm in coverage metrics. Overall, the organization is well and the writing is clear. However, the following issues should be further considered and clearly clarified before the final publication:

1. How does the SOFM algorithm ensure the Quality of Service (QoS) for critical applications, and how is it quantified?

2. The Multi-Objective Optimization (MOO) problem was briefly mentioned. Could you elaborate on how the trade-offs between coverage, connectivity, and network lifetime are addressed using SOFM?

3. While the simulation results are promising, have there been any real-world deployments or tests of the proposed SOFM node scheduling algorithm to validate its effectiveness in actual scenarios?

4. The SOFM technique, as described, emphasizes the iterative update based on the BMU. However, the mathematical robustness behind the selection of the BMU and the subsequent update mechanism is not clearly presented. Could you provide a more detailed mathematical analysis or proof that validates the BMU selection method, ensuring its optimality for the given application in WSN?

5. The paper mentions the use of the Euclidean distance as a method to determine the closeness of nodes. How did you arrive at the decision to use the Euclidean distance over other potential distance measures, especially given the multi-dimensional nature of data in WSN? A mathematical justification or comparative analysis would strengthen this choice.

6. The optimization mechanism seeks to simplify the problem domain by reducing its dimensions. While the qualitative benefits are mentioned, there's a lack of quantitative analysis or proof showcasing the extent of data preservation post-dimensionality reduction using SOFM. Could you delve deeper into the mathematical proof that guarantees no significant loss of information when moving from a higher-dimensional space to a lower one using this technique?

7. Since the authors considered a real-time system based on IoT devices, it is suggested that the authors refer to some papers about the metric of data timeliness in your revised introduction part, such as ‘Age of information in energy harvesting aided massive multiple access networks’ and ‘On the role of age of information in the Internet of Things’. Just to name a few.

Author Response

Dear Reviewer 1,

Thanks for your valuable feedback and comments that have been addressed thoroughly. It is also worth mentioning that all typos, grammar, and language issues have been addressed. The entire paper has been proofread by Native Academic English. In addition, the following are the replies to all your comments:

1. To address the following question “How does the SOFM algorithm ensure the Quality of Service (QoS) for critical applications, and how is it quantified?” The QoS for critical applications is often quantified using specific metrics and benchmarks relevant to the application domain. These might include response time, throughput, error rates, and reliability. Integration with QoS Metrics: in our study, the SOFM node scheduling algorithm itself directly quantifies QoS, in its results providing the optimal solution. The SOFM node scheduling algorithm features can be integrated into a system that uses QoS metrics. For example, the SOFM node scheduling algorithm considers coverage, connectivity, and network lifetime metrics as part of the whole system for the WSN network, the reduction in false positives or negatives can be a QoS metric

 

2. The Multi-Objective Optimization (MOO) problem was briefly mentioned. Could you elaborate on how the trade-offs between coverage, connectivity, and network lifetime are addressed using SOFM? In this work the SOFM node scheduling algorithm is being used to address the MOO problem through the visualisation of the spatial space in the WSN where we considered network configuration based on node residual energy, Cluster Head, child nodes, coverage and connectivity configuration.
3. While the simulation results are promising, have there been any real-world deployments or tests of the proposed SOFM node scheduling algorithm to validate its effectiveness in actual scenarios? We have used simulation test bed along with statistical analysis; however, we considered the use of a hybrid approach as a future work.
4. The SOFM technique, as described, emphasizes the iterative update based on the BMU. However, the mathematical robustness behind the selection of the BMU and the subsequent update mechanism is not clearly presented. Could you provide a more detailed mathematical analysis or proof that validates the BMU selection method, ensuring its optimality for the given application in WSN? The entire mathematical formulations in the problem formulation section have been revised and we have referenced the use of the general structure of the SOFM equations. The Euclidean distance is used to calculate the BMU for the winner victor and its subsequent neighbour. We have provided three examples to further illustration.
5. The paper mentions the use of the Euclidean distance as a method to determine the closeness of nodes. How did you arrive at the decision to use the Euclidean distance over other potential distance measures, especially given the multi-dimensional nature of data in WSN? A mathematical justification or comparative analysis would strengthen this choice. This has been answered in question 4. However, the Euclidean distance is based on the geometric distance between points in a multi-dimensional space. It corresponds to the length of the straight line connecting two points. This geometric interpretation aligns well with the intuitive notion of distance in physical space, making it a natural choice for spatial data in WSN. In addition, the Euclidean distance is computationally straightforward to calculate. In a multi-dimensional space, it involves taking the square root of the sum of squared differences along each dimension. This simplicity makes it efficient to compute, especially in real-time or resource-constrained WSN environments.
6. The optimization mechanism seeks to simplify the problem domain by reducing its dimensions. While the qualitative benefits are mentioned, there's a lack of quantitative analysis or proof showcasing the extent of data preservation post-dimensionality reduction using SOFM. Could you delve deeper into the mathematical proof that guarantees no significant loss of information when moving from a higher-dimensional space to a lower one using this technique? Although, there are multiple techniques to proof no significant loss of information when moving from a higher-dimensional space to a lower one, to name a few such as the Principal Component Analysis (PCA) which is the standard one, Incremental, or stream techniques etc., our study (SOFM node scheduling algorithm) is based on the statistical analysis where we obtained the P value as illustrated in Table 1.
7. Since the authors considered a real-time system based on IoT devices, it is suggested that the authors refer to some papers about the metric of data timeliness in your revised introduction part, such as ‘Age of information in energy harvesting aided massive multiple access networks’ and ‘On the role of age of information in the Internet of Things’. Just to name a few. In reply to this comment the research overview at the first level to address the metrics that are involved in the MOO problem that are coverage, connectivity and network lifetime. In the second level we will look thoroughly on the metrics recommended, hence we consider it as future work.

 

Reviewer 2 Report

Comments and Suggestions for Authors

In this paper, the author proposed SOFM node scheduling algorithm aims to spatially explore the state space of the solution domain and obtain an optimal solution. The description of the SOFM proposed in this paper is unclear and confusing. And I also paid attention to other issues, such as：

In the abstract, there is too much description of the background of the problem, so a concise abstract is recommended.

The figure in the article are not clear, such as Figures 1, 2, etc. 

What does w in Figure 2 stand for?

What does CH mean? Please give the full name the first time an abbreviation appears.

At the end of the introduction, it is recommended to give the contribution of this article so that readers can understand the content of this article.

In related work, please review the coverage, connectivity, and network lifetime of WSN, and discuss the shortcomings and challenges of current research.

I don't understand Eq. 6, what does it mean?

In the paragraph starting on page 7, it is not recommended to use below and it is recommended to change it to as shown in Algorithm 1.

 

It is recommended that the solution proposed in this article be described in detail.

The experiment recommends adding 2 latest research solutions as baseline.

Figure 4 Why are the x-axis time (rounds) of the 2 algorithms inconsistent?

The conclusion is redundant and it is recommended to simplify it.

There are too few references. It is recommended to cite relevant literature from MDPI journals in the past 3 years.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Dear Reviewer 2,

Thanks for your valuable feedback and comments that have been addressed thoroughly. It is also worth mentioning that all typos, grammar, and language issues have been addressed. The entire paper has been proofread by Native Academic English. In addition, the following are the replies to all your comments:

The entire abstract has been fully revised in accordance to your comments. This includes: rewriting the abstract so it has a problem definition, contribution, methodology, and results.

Figure 1 and 2 have been fully redesigned and they are much clearer now. Please note that the legend is fully addressed and written within the figure. In figure 2 w means weight.

CH means Cluster Head and all abbreviations have been defined at their first appearance in the article.

At the end of the introduction the contribution of the article has been presented in line with your comments. Please refer to line 51.

In related work, please review the coverage, connectivity, and network lifetime of WSN, and discuss the shortcomings and challenges of current research. That has been addressed

In relation to the following comment “I don't understand Eq. 6, what does it mean?”, the entire problem formulation section has been revised. Every equation has been explained with all defined values and made clearer for the reader to understand including equation 6.

According to your comment “In the paragraph starting on page 7, it is not recommended to use below and it is recommended to change it to as shown in Algorithm 1”, this has been addressed.

 It is recommended that the solution proposed in this article be described in detail. As stated earlier, the entire paper has been revised and structured coherently. All sections were rewritten and elaborated. The solution was discussed in different sections. However, it is mostly discussed in more detail and mathematically in section 3 the problem formulation section. Examples have been included in that section along with defining all used variables and algorithm. Whereas the simulated solution was discussed and evaluated against the Bat node scheduling algorithm in section 4, results and discussion section.

“The experiment recommends adding 2 latest research solutions as baseline”. It is worth noting that this is a continues research work where we have addressed the better solution in chronological order. From the best improved to the latest. In this work the best improved is the bat and the oldest is the HMM where we have referenced this in the paper.

“Figure 4 Why are the x-axis time (rounds) of the 2 algorithms inconsistent?”. This is a different metrics which requires the use of more time, hence the time (Rounds) ranging from 0 to 1800 so the base line can fully appear.  

The conclusion has been fully and clearly revised and crafted.

The referencing list has been fully revised and therefore, enriched from various journal related papers specifically from the MDPI. 

Reviewer 3 Report

Comments and Suggestions for Authors

The solution proposed in the paper can  contribute to scientific knowledge, but the paper did not  properly process the proposed solution.

-The abstract has 212 words while the maximum is 200 words because of that i strongly recommend that the authors rewrite the abstract including the problems, methods and results in concisely. Also, introducing abbreviations in the abstract is not recommended.

-The introduction is too short and lot from  the content provided by the instructions for authors is missing: The introduction should briefly plhouldace the study in a broad context and highlight why it is important. It should define the purpose of the work and its significance, including specific hypotheses being tested. The current stateof the research field should be reviewed carefully and key publications cited.  Finally, briefly mention the main aim of the work and highlight the main conclusions.

-The subject matter is not presented in a comprehensive manner and organization of the paper is not good:

-Subsections 1.1. and 1.2. which deal with existing methods should  be in the section 2.Materials and Methods  which must be after now Section 2. Related work which shouldd be a part of very short first section 1.Introduction.

- The paper lacks with suitable limitations of the proposed framework.

-Section conclusion is too short without part about scientiffic contribution of this manuscript and without the part about future work.

- The number  from total 14 references in the paper  is too small for the publication of the paper  in such an eminent journal as Electronics .

-Figures 1 and 3  are not clear.

Author Response

Dear Reviewer3,

Thanks for your valuable feedback and comments that have been addressed thoroughly. It is also worth mentioning that all typos, grammar, and language issues have been addressed. The entire paper has been proofread by Native Academic English. In addition, the following are the replies to all your comments:

 

The solution now has been fully structured where segments of each contribution is flowing coherently for a better presentation. 

The entire abstract has been fully revised in accordance to your comments. This includes: rewriting the abstract so it has a problem definition, contribution, methodology, and results. Although it has been recommended not to introduce the abbreviations within the body of the abstract, we chose to go along with a first reviewer recommendation which indicates clearly to introduce the abbreviations at first appearances even if in the abstract.

 

The introduction of the paper has been fully revised and rewritten accordingly to you feedback. The new introduction covers all importance of the study including the motivation and contribution.

 

The structure of the paper has been fully revised and reorganised where it is presented in a comprehensive manner. Therefore, the subsections 1.1. and 1.2 mentioned in your previous comments have been emerged in the new structure. Subsequently in reply to your comment we would like to draw your attention to that this paper is presented in a coherence way to serve the purpose of this research. 

The limitation of the proposed framework has been addressed in a concise way in the conclusion. 

The entire conclusion is revised and now it includes the contribution of this manuscript and the future work of the study.

The referencing list has been fully revised and therefore, enriched from various journal related papers specifically from the MDPI. 

Figure 1 and 3 have been fully redesigned and they are much clearer now.

Reviewer 4 Report

Comments and Suggestions for Authors

In this manuscript, the authors present a method based on Self-Organizing Feature Map (SOFM) to improve the scheduling among nodes in Wireless Sensor Networks (WSN). The algorithm is described and simulation results are provided, comparing the SOFM scheduling method to the BAT technique in terms of energy efficiency, connectivity, coverage, lifetime, etc. The results highlight the ability of the proposed SOFM algorithm to outperform the baseline. The paper is interesting and relevant to the topics of Electronics. However, some minor comments need to be addressed:

Figure captions should be more descriptive instead of simply mentioning the variable depicted (e.g., connectivity or coverage). Please explain to the reader in one or two sentences what is shown in the figure.

Moreover, in Figure 7 the two lines are interchangeable. Please use some other means (e.g., dotted line) to illustrate the connectivity.

It would be very interesting to include either in the “Introduction” or in the “Related work” sections relevance of your work with PPDR (Public Protection and Disaster Relief) applications and MCS (Mission Critical Services) technical solutions that use state-of-the-art ML methods. Relevant references are:

 

1.       Petrov, Vitaly, et al. "Achieving end-to-end reliability of mission-critical traffic in softwarized 5G networks." IEEE Journal on Selected Areas in Communications 36.3 (2018): 485-501.

2.       Spantideas, Sotirios, et al. "Intelligent Mission Critical Services over Beyond 5G Networks: Control Loop and Proactive Overload Detection." 2023 International Conference on Smart Applications, Communications and Networking (SmartNets). IEEE, 2023.

3.       Lu, Sally, et al. "Applications of artificial intelligence and machine learning in disasters and public health emergencies." Disaster medicine and public health preparedness 16.4 (2022): 1674-1681.

Moreover, you may also want to check ML-assisted techniques that are used for feature knowledge distillation:

 

1.       Chung, Inseop, et al. "Feature-map-level online adversarial knowledge distillation." International Conference on Machine Learning. PMLR, 2020.

2.       Yang, Zhendong, et al. "Vitkd: Practical guidelines for vit feature knowledge distillation." arXiv preprint arXiv:2209.02432 (2022).

 

 

Comments on the Quality of English Language

Detailed proof-reading is required for minor errors, e.g., line 165 time-series analysis instead of time serious analysis.

Please explain the abbreviations the first time they appear in the manuscript (e.g., HVAC, HMM, etc.).

Author Response

Dear Reviewer 4,

Thanks for your valuable feedback and comments that have been addressed thoroughly. It is also worth mentioning that all typos, grammar, and language issues have been addressed. The entire paper has been proofread by Native Academic English. In addition, the following are the replies to all your comments:

“Figure captions should be more descriptive instead of simply mentioning the variable depicted (e.g., connectivity or coverage). Please explain to the reader in one or two sentences what is shown in the figure”. That has been addressed

Moreover, in Figure 7 the two lines are interchangeable. Please use some other means (e.g., dotted line) to illustrate the connectivity. This has been addressed

It would be very interesting to include either in the “Introduction” or in the “Related work” sections relevance of your work with PPDR (Public Protection and Disaster Relief) applications and MCS (Mission Critical Services) technical solutions that use state-of-the-art ML methods. Relevant references are:

1. Petrov, Vitaly, et al. "Achieving end-to-end reliability of mission-critical traffic in softwarized 5G networks." IEEE Journal on Selected Areas in Communications 36.3 (2018): 485-501.
2. Spantideas, Sotirios, et al. "Intelligent Mission Critical Services over Beyond 5G Networks: Control Loop and Proactive Overload Detection." 2023 International Conference on Smart Applications, Communications and Networking (SmartNets). IEEE, 2023.
3. Lu, Sally, et al. "Applications of artificial intelligence and machine learning in disasters and public health emergencies." Disaster medicine and public health preparedness16.4 (2022): 1674-1681.

Moreover, you may also want to check ML-assisted techniques that are used for feature knowledge distillation:

1. Chung, Inseop, et al. "Feature-map-level online adversarial knowledge distillation." International Conference on Machine Learning. PMLR, 2020.
2. Yang, Zhendong, et al. "Vitkd: Practical guidelines for vit feature knowledge distillation." arXiv preprint arXiv:2209.02432(2022).

 Thank you for spending your time to find relevant papers to our study. We have incorporated most of the above references in our work.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors' response to Reviewer 1's comments is comprehensive, yet there are a few issues that could be improved or clarified:

 

1. Mentioning future work in real-world testing is good, but you might also want to discuss any potential challenges or limitations that could arise in practical deployments. This would provide a more balanced view of your approach.

2. The authors acknowledged the need to explore metrics of data timeliness but haven't provided a clear direction on how they plan to incorporate these into your future study. It would be helpful to outline a more detailed plan for integrating these metrics into their future work, possibly by setting specific objectives or hypotheses to test.

 

Comments on the Quality of English Language

none

 

Author Response

Dear Reviewer 1, 

Thank you for your further comments that we have addressed as follows:

1. In order to improve the dependability of safety-critical systems in WSN, we coined a definition in which we state "A WSN is dependable for a given application if its maximum lifetime is equal or greater than its expected service time. In addition, a WSN must be rigorously tested to state its dependability". Therefore, in our testing environment, we considered the following as important:
   1. Realistic assumptions. For example, assumptions that are realistic to be applied for real-life WSN deployment. In the literature, there are one-for-all solutions. Such solutions are designed, but placed on the shelve collecting dust, as it is very difficult to be applied for specific scenarios or it does not meet the modern safety-critical system requirements.
   2. The difference in the very fabric of the sensor might have an impact on the performance whether small or big. For example, any factory that produces sensor components might have impurities within the materials of the sensor itself. This might hinder the sensor node's performance. 
   3. The choice of defining new metrics that can measure and quantify the result obtained from the real-life testing deployment.

2. In regrade to the reviewer’s second comment, in order to improve the dependability of safety-critical systems for such applications we need to consider the following solution which has already been mentioned in the future work of this research paper as follows:

In future work, this limitation will be addressed by introducing the LSTM model whose approach is entirely on a time-dependent framework. Although the metrics involved in this research are concerned with the MOO problem, in future work we will look thoroughly at metrics that are concerned with real-time system applications to robustly assess the proposed SOFM node scheduling algorithm. For example, our hypothesis will extend to propose an approach that approximates the end-to-end delay distribution of the links between immediate nods. This shall enable certain sequences of messages to be sent through a chain of selected paths within time intervals. Thereby, enabling the application of WSN to exploit the quality of service trade-off based on timeliness requirements. This shall guarantee the service availability and reliability, and in turn, improve the dependability of safety-critical WSN systems.

 

Kind Regards,

Issam Al-Nader

Reviewer 2 Report

Comments and Suggestions for Authors

The author has done a good work in revising the article and the quality of the article has been improved. However, some of the latest relevant work has been misssing by the author, for example, doi: 10.1109/ACCESS.2020.3008373

https://doi.org/10.1016/j.dcan.2022.09.005

doi: 10.1109/ACCESS.2018.2889717

Author Response

Dear Reviewer 2, 

Thank you for your further comments and input. We have included all three recommended relevant work in our paper:

1. doi: 10.1109/ACCESS.2020.3008373,
2. https://doi.org/10.1016/j.dcan.2022.09.005,
3. doi: 10.1109/ACCESS.2018.2889717.

Kind Regards,

Issam Al-Nader

Reviewer 3 Report

Comments and Suggestions for Authors

The authors of the paper correctly corrected the paper, and they took into account almost all my remarks and included them in the correction of the paper.

The limitation of the proposed framework has been addressed in a concise way in the conclusion, but it must have been addressed earlier in the discussion of the obtained results and must be described in more detail.

Author Response

Dear Reviewer 3, 

Thank you for your further comments. The limitation of the proposed framework has been addressed in the discussion of the obtained results in more detail as recommended at the end of each metric in sections 4.1, 4.2, and 4.5. Further, we have expanded in explaining the limitation in the conclusion.

Kind Regards,

Issam Al-Nader

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed all my concerns.