Multi-Task Vehicle Platoon Control: A Deep Deterministic Policy Gradient Approach

Round 1

Reviewer 1 Report

Abstract should be more precise summarizing the proposed work and results.

Add related works as a separate section.

Why deep Reinforcement learning algorithm is used?

Authors claim reward function as the proposed system. Whether this type of reward function is not used anywhere?

What kind of hyperparameter selection methodology is used?

Provide more explanations for the figures in results section

What are the limitations of the proposed work?

 

Author Response

We gratefully thank you for your thorough review, insightful comments, and remarks. We have carefully revised the manuscript to address your issues and concerns. The major revisions are marked in green color in the revised manuscript. Please see our point-to-point reply to your comments in the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper investigates the multi‐task platooning control of connected vehicles. A Deep Deterministic Policy Gradient Approach is proposed for reliable and fast gap‐closing/opening actions, as well as the vehicle movement tracking. The topic is interesting. Below are my comments.

 

1.      The specific technical challenges should be clearly stated. In addition, the superiority of the current research over the existing ones should be explained through comparisons.

 

2.      It is better to clearly state what the multi-task is, and how they combined together.

 

3.      There lacks rigorous proof of the control method. How to guarantee the convergence of the vehicle tracking errors?

 

4.      Some related results could be considered for a more complete research background, such as ‘Secure and collision-free multi-platoon control of automated vehicles under data falsification attacks’, and ‘Secure distributed adaptive platooning control for automated vehicles over vehicular ad-hoc networks under denial-of-service attacks’.

 

5.      The quality of the figures in the presented simulation results should be further improved for the benefit of readers.

 

6.      The labels in Figure 7 could be moved to left side since the final convergence is covered by them.

Author Response

We gratefully thank you for your thorough review, insightful comments, and remarks. We have carefully revised the manuscript to address your issues and concerns. The major revisions are marked in green color in the revised manuscript. Please see our point-to-point reply to your comments in the attached file.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this work, the authors developed a multi-task deep deterministic policy gradient car‐following algorithm in a platoon system. The proposed approach combines gap closing/opening into a unified platoon control strategy. The topic is interesting. However, there are some problems which need to be addressed.

1. I am interested in the computation complexity of the deep reinforcement learning algorithm. As we know, in the practical transportation, the real-time control policy is needed.

2. What perform did you use to train the neural network?

3. The contributions of the work should be further highlighted.

4. The Figures in the simulation section should be further improved.

Author Response

We gratefully thank you for your thorough review, insightful comments, and remarks. We have carefully revised the manuscript to address your issues and concerns. The major revisions are marked in green color in the revised manuscript. Please see our point-to-point reply to your comments in the attached file.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The raised problems have been addressed. I have no further comments.