Review Reports - Universal Learning Approach of an Intelligent Algorithm for Non-GNSS Assisted Beamsteering in V2I Systems

Round 1

Reviewer 1 Report

The interesting subject, but somewhat abstract, is the treatment of comparative notions. Title or abstract.

The template used does not seem to be downloaded from the journal page or possibly latex was used and there are things that do not resemble the original.

References are missing and in their place are question marks.

Explain the figures in more detail.

Try to present the characteristics or parameters of the simulations.

The article has no specific purpose, try to highlight the scientific contribution and the motivation as much as possible.

Highlight the new elements and compare the approach with other works in the literature.

There are aspects that need to be improved, try to present the results in detail and anchor the obtained data extremely well in relation to the current level.

The article has a good structure, but it needs to benefit from some changes in order to be published.

Author Response

Dear Editor,

We are grateful to you for the opportunity to refine our manuscript and to the reviewers for the helpful and insightful comments and suggestions. In addition to the revision based on the reviewer's comments, we've improved the manuscript in the following ways: 1) The paper's title was corrected to describe better its novelty, main idea, and methodology. The new title is "Universal learning approach of an intelligent algorithm for a non-GNSS assisted beamsteering in V2I systems"; 2) To the same end, we changed the structure of the manuscript, which now includes the following sections: Introduction, Beamsteering system concept, PAA's simulation model, The learning approach of the intelligent algorithm, Simulation and Results, and Conclusions; 3) general proof-reading of the text has been done to improve the grammar and readability.

Along with the clear version of the revised manuscript, please find attached the version with the highlighted changes that have been made to address the reviewers' comments.

Please find below the detailed answers to the reviewer's comments.

Thank you very much for your valuable suggestions and comments. Below we present the changes that we have made according to them.

1.The interesting subject, but somewhat abstract, is the treatment of comparative notions. Title or abstract.

We changed the manuscript’s title and abstract (lines 3-7) to fix the comment and make the novelty more precise.

2.The template used does not seem to be downloaded from the journal page or possibly latex was used and there are things that do not resemble the original.

We checked the latex template we used and fixed some design mistakes.

3.References are missing and in their place are question marks.

We checked the reference list and the text and fixed the missed references.

4.Explain the figures in more detail.

We have redesigned Figure 2 to be more illustrative and to show different system scenarios. We also detailed the description of Figure 2 and Figure 3 (lines 167-170 and 189-194, respectively) and adjusted the Figure 4 caption.

5.Try to present the characteristics or parameters of the simulations.

We have added the training hyperparameters for the XGBoost algorithm on lines 306 to 312. Lines 316 to 326 now contain the signal propagation and SUMO parameters for traffic simulation.

6.The article has no specific purpose, try to highlight the scientific contribution and the motivation as much as possible.

We have changed the structure of the article, introduction, and conclusion for a more accurate presentation of the work's novelty and the study's essence.

7.Highlight the new elements and compare the approach with other works in the literature.

We described the new elements in our approach more precisely (introduction and Section 3) and have added the results of comparing the operation of the simulated system with [27] in the lines in lines 364-368.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper presents a beam-steering technique of a phased antenna array based on gradient boosting. However, the beamsteering system for a traffic system should be described more realistically.

Figure 2 shows the directed radiation beams to the vehicle in a less dense intersection scenario. In real applications, V2I may need to cover up to 300 m or longer, covering a larger number of vehicles on the same road. It is necessary to provide a strong signal to all such vehicles. Provide a such illustration of beamsteering considering both longer road coverage and also four directions coverage of an intersection.

Why are there 10 sub-sectors? How are they selected?

What kinds of data are considered to exchange between V and I? How is such an exchange initiated? Describe the cause of the delay to understand the role of ML better.

The details of ML, i.e., inputs and outputs, should be described clearly.

What is the meaning of “dynamic” vehicle dimension L?

How “the time delay between the start of request message transmission” is obtained from SUMO?

Though this paper contains some new contributions, along the above, the following points should also be clarified:

1. Example of V2I data with the cause of delay relating to the role of the antenna.

2. How delay-data is generated using SUMO, as it does not have communication functionality and everything can be easily obtained.

3. Does the radiation beam continuously track the vehicles? What is the length?

4. What is the key aspect that reduces the delay and increases the accuracy of the proposed method with ML; state with justification.

Author Response

Dear Editor,

Along with the clear version of the revised manuscript, please find attached the version with the highlighted changes that have been made to address the reviewers' comments.

Please find below the detailed answers to the reviewer's comments.

Thank you very much for your valuable suggestions and comments. Below we present the changes that we have made according to them.

1. Figure 2 shows the directed radiation beams to the vehicle in a less dense intersection scenario. In real applications, V2I may need to cover up to 300 m or longer, covering a larger number of vehicles on the same road. It is necessary to provide a strong signal to all such vehicles. Provide a such illustration of beamsteering considering both longer road coverage and also four directions coverage of an intersection.

We modified Figure 2, highlighting two scenarios for the system operation: a crossroad (a) and an extended linear section (b)

2. Why are there 10 sub-sectors? How are they selected?

We added comments on how the required width and the number of sub-sectors were selected in lines 189-194.

3. What kinds of data are considered to exchange between V and I? How is such an exchange initiated? Describe the cause of the delay to understand the role of ML better.

We described the process of establishing a connection between an infrastructure object and a vehicle, as well as a description of the data transmitted in the system, in lines 206-214. We also added the information on the critical response latency of the ML algorithm in lines 296-298.

4. The details of ML, i.e., inputs and outputs, should be described clearly.

For a better description of input and output data for the ML algorithm, we added to the manuscript Table 2 containing the examples from the generated dataset.

5. What is the meaning of “dynamic” vehicle dimension L?

We specified the parameter’s name (line 282-283)

6. How “the time delay between the start of request message transmission” is obtained from SUMO?

The time delay for the request message's transmission was calculated using the signal propagation simulating model outside the SUMO software (lines 199-205).

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I have no further comments