Multi-UAV Collaboration to Survey Tibetan Antelopes in Hoh Xil
Round 1
Reviewer 1 Report
This paper aims to detect, count, and survey endangered animals, namely Tibetan antelopes, using a generic few-shot detector algorithm with the use of multiple UAVs.
1. What is the advantage of using four UAVs instead of multiple combinations, for example, two, three, or five UAVs? Please mention the UAV specification?
2. In the proposed model using ResNet-50 backbone instead of another deeper network. Is there any specific reason to use ResNet-50? Please explain.
3. Defined the combined adaptation losses used in during the test. But two losses are different in training and testing, why?
4. Compared with old techniques in path planning for multiple UAVs and detection methods.
5. In figure 4, Please describe clearly.
6. Mentioned FamNet is not clear in this paper. Could you please explain it in deep?
7. Mentioned Algorithm 2 in this paper. Could you please check it once?
8. The language has to be thoroughly revised and rewritten.
9. Besides, some general mistakes
a. Enhancement in columns of tables 2 and 3 are mentioned twice.
b. In section 3.1.3, please avoid the repeated sentence.
c. The references 37 and 38 are wrong please check them once.
Author Response
Dear Reviewer,
Thank you for your careful replies and thoughtful comments on our previous draft. We have carefully taken the comments into consideration. Aimed at each comment that the reviewer provided, I have revised my manuscript entitled “Multi-UAV Collaboration to Survey Tibetan Antelopes in Hoh Xil” (drones-1839322). The following are my specific responses and revisions to the reviewer.
Notes: The yellow background mark indicates the reply and modification to each question of the reviewer.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The manuscript presents methods and algorithms for calculating the number of animals using unmanned aerial vehicles (UAVs) and offers an attractive study with a real and new perspective on the current scientific problem. During the work, these algorithms were implemented in a set of UAVs, and experiments were performed in the field. The results showed that the method proposed by the authors can be used to perform aerial surveys with shorter trajectories and mission time. This discussion provides the authors with an interesting contribution to the motivation for this study. The study is important in its applied sense. The research presented in this article seeks novelty. The research methodology is replicable. The choice of research methodology is appropriate, logical, and well-described and allows other researchers to replicate the results in their own work. Keywords and sources from the cited literature are suitable.
Specific comment:
The conclusion section of the paper is very condensed. If the authors' conclusions provided a more interesting contribution to the motivation for this study and some interesting open questions for future research, it would encourage the reader to take more interest in the topic.
My overall opinion is positive, and I would suggest that the authors continue the work.
Author Response
Dear Reviewer,
Thank you for your careful replies and thoughtful comments on our previous draft. We have carefully taken the comments into consideration. Aimed at each comment that the reviewer provided, I have revised my manuscript entitled “Multi-UAV Collaboration to Survey Tibetan Antelopes in Hoh Xil” (drones-1839322). The following are my specific responses and revisions to the reviewer.
Notes: The yellow background mark indicates the reply and modification to each question of the reviewer.
Author Response File: Author Response.docx
Reviewer 3 Report
If we focus on the application (Survey Tibetan Antelopes in Hoh Xil) the work is extremely interesting. If we focus on the scientific aspect (multi-UAV trajectory planning method was proposed for various population counting and surveys, which can fully use UAVs’limited flight time to conduct aerial counting surveys over large areas), the work is a bit weaker, in the sense that the methodology is not extremely solid and well explained. In addition the authors are missing recent works also appearing on MDPI such as: A semi-physical platform for guidance and formations of fixed-wing unmanned aerial vehicles; Convergence analysis of path planning of multi-UAVs. These are relevant due to the focus on formations of unmanned aerial vehicles. But I believe we should focus on the application and support the work.
I agree with the motivation provided by the authors: It is critical to reduce the total mission time and flight distance in a wildlife survey because of environmental impacts and animal movements. Even from the scientific point of view, there is some innovation: Unlike current coverage path planning methods based on sweep or polygon, the path planning problem is encoded as a Satisfiability modulo theory using a one-hot encoding scheme. Maybe this innovation is not extremely strong, but acceptable. Summarizing:
Each instance generates a set of feasible trajectories at each iteration and flexibly optimizes the trajectories based on the number of available UAVs, endurance, and other parameters.
I add the following comments
- it seems that the authors use rotary wing UAVs. Why nt using fixed-wing UAVs which typically can fly for longer time? this point should be explained
- why airframe is used? which autopilot is used to control the UAV? These points about the hardware/software used should be explained
- what is the computational complexity of the proposed multi-UAV CCPP? and which software (optimization software, I guess) is use to solve the multi-UAV CCPP problem?
Then it comes the strongest lart of the work, that is: The planning algorithm was implemented with four UAVs to conduct several wildlife aerial survey experiments in areas around Zonag Lake, the Tibetan antelope calving place, which investigates more than 6 km2 within approximately 2 hours.
This application is very strong and I also support the conclusions: In contrast, the previous manually controlled survey method with a single UAV requires more than 4 days to complete the task in the same area. A generic few-shot detector that can perform effective counting without training on the target object is utilized in this paper, which can achieve an accuracy of over 97%. The results are impressive
Author Response
Dear Reviewer,
Thank you for your careful replies and thoughtful comments on our previous draft. We have carefully taken the comments into consideration. Aimed at each comment that the reviewer provided, I have revised my manuscript entitled “Multi-UAV Collaboration to Survey Tibetan Antelopes in Hoh Xil” (drones-1839322). The following are my specific responses and revisions to the reviewer.
Notes: The yellow background mark indicates the reply and modification to each question of the reviewer.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors have addressed all my concerns, and this paper looks good now.
