Robust Motion Control for UAV in Dynamic Uncertain Environments Using Deep Reinforcement Learning

Round 1

Reviewer 1 Report

The manuscript entitled “Robust motion control for UAV in dynamic uncertain environments using deep reinforcement learning” represents an original research. The research goal is the develop a controller that allows robust flying of an unmanned aerial vehicle (UAV) in dynamic uncertain environments. The idea of the research is interesting and will attract an audience in the field of UAV sensors and robust motion control.

The paper title and abstract are written correct and concise. In the entire paper, authors use standard technical and scientific terminology. After the Introduction, the authors explained the problem of UAV motion control and UAV kinematics; and developend novel UAV motion control, as well as achieved experimental results. The experiments and results were conducted according to the scientifically correct approach. The discussion and conclusions are logical and based on the results of the research. The paper topics fit in Remote Sensing aims and scope, especially in Remote sensing with unmanned aerial systems.

 

I recommend this paper can be accepted after minor revision.

Comments for authors:

Please explain more the advantage of suggested method. Please emphasis more the applicability of the proposed method. Please introduce abbreviations if you want to use it in the manuscript text (e.g. DQN etc.). Please, increase font size on figures 6, 10 and 11. Please use MDPI standard font in the figures if you can (Palatino Linotype). The variable names must have the same font style and size in equations, on figures, tables, and in the manuscript text. Please describe/introduce all variables used in equations or on figures in the manuscript text. Please, double check all references and reference style.

Author Response

Please see the attachment. Thank you very much.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript presents  a novel deep reinforcement learning (DRL) method and robust deep  deterministic policy gradient (Robust-DDPG)

 

 

Introduction is written well.

 

The analytical part of the work is quite well and interesting. However, in methodology it is difficult to find novelty.

In section 2.2 UAV motion control as a MDP

The description of the methodology used is not detailed enough, please extend it.

 

The main disadvantages of this research is the lack of statistical assessments and explanation for the chosen parameters and methods.

 

Add a statistical significance test to assess if the differences between the compared models are really significant. If your results data are not parametric, you could use for example Wilcoxon or Friedman tests. 

 

I have objections to the discussion section. The authors need to re-organize ,the results and discussion therein to better highlight to the reader what was done and what is relevant. The gain of the presented technique for the addressed application should be made more explicit in the form: What do the findings allow what was possible before. Authors should discuss the results and how they can be interpreted in perspective of previous studies and of the working hypotheses.

 

Conclusions are correct.

Author Response

Please see the attachment, thank you very much.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper content is partial aligned with Remote Sensing Journal. The article present an innovative and interesting approach to control UAV in several environmental conditions based on deep reinforcement learning. No related article was previously published in Remote Sensing, but I believe that it is an important topic to be considered in Remote Sensing Journal.

In general, the article is well-written. The authors organized it in five sections. The section 2, which is the literature review gives the support to understand the proposed approach.

Regarding the abstract, I suggest to include the name of the state-of-the-art methods that were used to compare with the proposed method. Some phrase in the abstract are long. I suggest to divide them.

The introduction present the state-of-the-art, beginning with the traditional methods and concluding with DDPG method, which is adapted in the current work. Finally, the original contributions are presented, which is very important to show the relevance of the developed work.

In section 3, I suggest to review the use of the term “in our opinion”.

In the results section, only simulated results are presented. In the paper, I would like to see a discussion about the challenges of becoming this real, with a real UAV.

The conclusion is well-written.

References section present current references.

Author Response

Please see the attachment. Thank you very much.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

All my comments have been taken into account. The authors improved the quality of their paper.