Adaptive Collision Avoidance for Multiple UAVs in Urban Environments
and Hao Liu 2,4
Round 1
Reviewer 1 Report
Congratulation, very good paper.
Table 1. is broken by page.
Figure 8: the difference between ddpg and improved ddpg should be analysed. When the difference is less than 5%? Is the legend same for the second figure?
Figure 9: Which obstacle or which UAS is the basis of comparison? The nearest? Can the minimum distance lowered?
Could this papers be cited?
Bagdi, Z., Csámer, L., & Bakó, G. (2023). The green light for air transport: sustainable aviation at present. Cognitive Sustainability, 2(2). DOI: https://doi.org/10.55343/cogsust.55
Bauer, P., Bokor, J. (2008) “LQ Servo control design with Kalman filter for a quadrotor UAV”, Periodica Polytechnica Transportation Engineering, 36(1-2), pp. 9–14. https://doi.org/10.3311/pp.tr.2008-1-2.02
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The aim of this article is to propose a collision avoidance method for multiple UAVs in order to prevent the increased UAVs in low altitude area from threatening the urban environment security. The proposed method aims to provide an adaptive solution using deep reinforcement learning and collaborative collision avoidance models.
The alleged contribution of this article is that adaptive collision avoidance method provides safe guidance for multi-UAV traffic in low altitude area. The proposed method aims to avoid collisions and increase safety in the urban environment by making real-time decisions. It also offers technical contributions such as providing fast convergence and shortening the training time by using an improved training algorithm.
In this article, alternative methods to the proposed method are not clearly specified, the statement that "the method may not be fully adapted to other models and algorithms" confirms this situation. For example, why was no comparison made with another deep learning model or optimization algorithm? Also, different strategies and computational techniques for collision avoidance can be considered as alternative methods. However, it was stated that these alternative methods were not evaluated within the scope of the article and only a macro-level comparison was made with the relevant methods. For a computer-only study, it is a reasonable expectation to present the results of alternative methods.
In the Introduction part, a paragraph consisting of many sentences is written. In the introduction, the literature is very crude and far from defining the problem. More clear and specific examples should be presented and provided as additives. The important thing here should be comparison with competitor and similar studies. Each contribution defined in the Introduction should be presented comparatively as a subheading in the results section.
The article is a trending topic in its current form, in this respect, it will attract the attention of the reader and it is possible to get many references in the future. However, the authors did not provide adequate benchmarking and valuable comparision tables because they did not evaluate competing studies sufficiently. In this respect, there is no content to be expected from a computer-based study, and it is not at a level that can be published in a journal with SCI-E scope in its current form.
There are short many paragraphs or single sentence paragraphs, these should either be combined or written as separete bullets.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Dear Authors!
The work is interesting and definitely useful!
It is of great interest from the theoretical and practical aspects, in particular for aviation designers and specialists in the field of control of unmanned aerial vehicles, and airspace security services in urban environments!
And despite the overall positive impression, there are a number of comments that you should pay your attention to before moving forward with the manuscript:
1. A similar problem was successfully solved by Chinese colleagues in 2022 - using the Particle Swarm algorithm and the MyFly method (based on the hierarchy analysis method). Russian scientists in 2016 - based on the ranked FIPS algorithm. I recommend that in the review, focus on the novelty of the method you proposed and evaluate its effectiveness in comparison with those previously proposed on a specific test example.
2. In the text of the document there is no reference to table 1. despite its informativeness.
3. The sample volume is not sufficiently consecrated in the work - what is the maximum number of UAVs that can be guaranteed to be controlled using the stated approach.
4. Please pay attention to the pairing of proposals. I recommend checking the work with a native speaker.
5. From the theory of development and operation of aircraft, it is known that in order to ensure flight stability, it is necessary to construct and solve a system of 16 differential equations with 16 derivatives. Is this included in your proposed model? And how can this be aggregated with the solutions you proposed?
6. How can one justify the effectiveness of the proposed solutions for stochastic fluctuations, in particular, with a variation in the strength of the crosswind?
7. In conclusion, it is required to formulate the final model first.
8. For me, the question remained open - how will the control objects eventually behave under linear active perturbations of a fractional order?
Good luck with revisions!
Please pay attention to note #4
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Paper can be accepted in present form
Reviewer 3 Report
Dear Authors!
Thank you for the quality and detailed answers.
Considering that you have taken into account all my wishes, I am ready to support you!
I wish you success and continued work in this direction.
