An Intelligent UAV Path-Planning Method Based on the Theory of the Three-Dimensional Subdivision of Earth Space

Round 1

Reviewer 1 Report

Currently, research is ongoing to find new methods for autonomous UAV flights. Traditional flight path planning actually becomes ineffective and insufficient for some missions. The use of the theory of three-dimensional division of Earth's space and the proposal of an innovative method of environmental modeling based on airspace grids seem to be very interesting. Especially since it can be tested in other countries, it does not only apply to the local network. There is great potential for such a solution. Of course, it still needs to be tested further in various conditions and take into account the criterion of wind direction and strength, which is worth expanding research in the future. Congratulations on the idea and the subject of the research. The importance of the research topic is enormous, especially since it can be used not only in civilian but also military transport.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This is a well written paper with good presentation and results. I just have one doubt. The system is based in GeoSOT, so the cuboids of the grid don't have the same size. Did you know at which latitude does this becomes a problem? Or is such an extreme case that is not worth considering?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Indeed, it is an interesting paper. My overall critique is as follows:

1.       Introduction

The good start of introduction section is acknowledged.

·         Line 75: For a better understanding of the readers, please elaborate on “A* algorithm”

·         Line 86: Elaborate on “Q-Learning algorithm” and the proposed grid-based DRL model

·         Line 92: “DQN algorithm” stands for what? Please elaborate

·         Line 118: The statement, “Therefore, to address both the inefficiencies of traditional environmental organization frameworks and problems in data association and representation”.

·         Please name some traditional environmental organization frameworks. Do you mean local single-scale grid system and the longitude-latitude system?

·         As such inefficiencies of these frameworks have not been mentioned. One would expect to read the inefficiencies of the frameworks. Please add these

·         Line 119-120: “…..this paper proposes a GeoSOT (Geographical coordinate grid Subdivision by One-dimension-integer and Two to n th power) based modeling method”.

·         What is the justification for proposing GeoSOT based modeling method? Please elaborate.

·         What are its key strengths and possible weaknesses?

·         In which scenarios it would work and similarly in which not?

·         Similarly, it would be valuable to include a comparison with other baseline algorithms commonly used in UAV path planning, such as A* or RRT. This would provide a clearer perspective on the performance of the proposed grid-based DRL model.

·         Line 141: “ The second chapter elucidates the method employed, the third chapter is a typical….” Is this paper based on some thesis? If so please name the thesis and its author. Otherwise, rectify the word ‘chapter’

2.        Materials and Methods

·         Line 151: Rephrase caption of Figure 1, “ Model overall Architecture Diagram”. For example, The overall architecture diagram for the model.  Similarly, rephrase the remaining captions also to make them easier to understand and comprehend

·         Line 273: Mathematical equations e.g., formula (3) and formula (4) are ok. It is suggested to provide detailed explanations of the mathematical equations to enhance clarity and readers understanding.

·         Providing a rationale for the selection of hyperparameters, such as learning rates, discount factors, and exploration strategies, would be beneficial. This could help readers understand the sensitivity of the model to these parameters.

·         Assessing the generalization capability of the proposed model to different environments or scenarios would strengthen the research. Demonstrating that the model performs well beyond the specific experimental setup would make it more applicable.

·         The paper could benefit from a more detailed contextualization of the research within the broader field of UAV path planning and DRL. Discussing related work and highlighting the unique contributions of the proposed model would provide a clearer context for the reader.

·         The paper provides experimental results, but the interpretation of these results is somewhat limited. A deeper analysis of why the proposed model succeeds or fails in certain scenarios would be valuable. What insights can be drawn from the success rates and path lengths reported? How do these results compare to existing state-of-the-art methods?

·         Deep reinforcement learning models can be computationally intensive. Providing information on the computational resources (e.g., hardware, training time) required for the experiments would be valuable for readers interested in implementing the model.

·         Given that UAVs operate in various environments, discussing ethical and safety considerations, especially when considering autonomous navigation, would be responsible. How does the proposed model ensure safe and ethical behavior in real-world scenarios?

·         The paper lacks a section discussing potential future directions or areas of research related to grid-based DRL for UAV path planning. Identifying open challenges and avenues for future work could inspire further research in the field.

Comments for author File: Comments.pdf

Extensive editing is required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Thank you for addressing my concerns.

It is acceptable