UAV Network Path Planning and Optimization Using a Vehicle Routing Model
Round 1
Reviewer 1 Report (Previous Reviewer 2)
Hi,
The English writing in this new submission has been improved from your previous two submissions. Please address the comments we made in your last submission (remotesensing-2218485.v2) regarding the assumption of the endurance time for the UAVs simulated in your paper, and the claim of "collision avoidance". The new comments is also added in the reviewed pdf file. Please have a check.
Thanks.
Comments for author File: 
Comments.pdf
Author Response
Thank you for your valuable comments and suggestions on our manuscript once again. We have carefully considered your feedback and made the following responses to our manuscript:
- Regarding the endurance time for the UAVs, we chose 2 hours as our endurance time based on some existing UAV models that can fly for more than 2 hours with a petrol-electric hybrid system, such as Thor 2101. This model has an endurance time of 3 hours according to its official website (http://www.foxtechcn.com/mobile/arc/show/id/272.html). We have used this kind of UAV in previous flight experiments, so we set the corresponding simulation parameters.
If we assume that the endurance time is 0.6 h, it will change the task allocation results, i.e., the number of UAVs and the network observation time will vary with the endurance time. However, it will not affect the usability of our algorithm as we have mentioned in Section 3.2 of our manuscript. Additionally, the endurance time value is a parameter in the constraint （equation 5 in section 2.2） such that the observation time of each UAV does not exceed its endurance, to ensure operational safety.
- Regarding the claim of “collision avoidance”, we recognized that the term “collision avoidance” that we used in initial manuscript was narrowed down to avoid collision between UAVs, and may be not the right terminology for the proposed methods. We have revised the terminology to “prevent route crossings” in the latest manuscript based on your last suggestion.
- Regarding the definition of “observation time” and “operation time”, we have revised our expression to make it clearer and added instructions in section 2.2. To clarify the difference between them, we have the definitions of these terms as follows:
-
- Observation time: The time when all the UAVs in the network complete their observation tasks over the mission area and collect data using their onboard sensors. This is called network observation time that depends on the preparation time and flight operation time of each UAV.
- Operation time: The time when a single UAV completes its flight operation over its assigned subarea within the mission area. This can be understood as the flight operation time for each UAV.
Moreover, we have carefully read and replied to each of your comments highlighted in the PDF. Please check in the attachment.
We hope that these revisions address your concerns and improve our manuscript’s quality.
Author Response File: Author Response.pdf
Reviewer 2 Report (Previous Reviewer 1)
The authors have corrected the article and considered the suggestions point by point
Author Response
Thank you for taking the time to review our article. We appreciate your feedback and suggestions, and have carefully considered each point in making revisions to the manuscript. We are glad to hear that our efforts to address your concerns have been successful. Thank you again for your valuable feedback.
Reviewer 3 Report (New Reviewer)
1. The Vehicle Routing Problem (VRP) method is widely used in transportation task assignments. In this paper, the authors apply algorithms purely to the assignment of UAV flight networks, but they ignore the flight characteristics of UAVs, such as :
(1) Payload
(2) Battery charging place
(3) Battery charging time
(4) Weather conditions (wind profile)
(5) Environment condition (obstacle, building)
(6) Remote radio
(7) Automatous control
(8) Detect and Avoid (DAA)
(9) UTM system to monstering in real-time
If the authors ignore the above conditions and without practice flight, the feasibility of the simulation model cannot be verified.
2. Regarding UAV cross-flight, it can be dealt with by UTM system and resolve the DAA problem.
Author Response
To reviewer 3
Thank you for your suggestions regarding the effectiveness of our algorithm. As you pointed out, our algorithm focuses mainly on task assignment in UAV networks, and lacks consideration for potential situations that may arise during actual flight processes. However, our algorithm is designed for UAV self-organizing networks for remote sensing applications, and the application scenarios mainly target low-cost and efficient UAV networking remote sensing observation tasks, such as terrain mapping, engineering measurement, agriculture and forestry investigation, emergency mapping, and so on. We have carefully considered the UAV flight characteristics that you mentioned and made the following responses:
- The mission payload is determined by the size and loading capacity of the UAV as well as mission requirements, which usually include additional cameras, sensors or delivery packages. Different UAVs are selected according to different remote sensing tasks, e.g., UAVs equipped with laser radar sensor are needed for terrain mapping tasks to extract high-precision DEM and rapidly generate terrain maps. Our path planning algorithm is applicable to UAVs with different payloads.
- Regarding the battery charging place and time, we assume that all UAVs have been charged in the hangar or field station before takeoff, and the duration constraint in the algorithm requires that the observation time of each UAV must be less than its endurance to ensure operational safety. Therefore, we didn’t consider the case where the UAVs would require mid-flight charging. In addition, the charging time for different types of UAVs varies, but it does not affect the effectiveness of our algorithm.
- Regarding the weather conditions, most UAVs are currently marked with wind resistance performance, and generally, normal shooting and flight can be ensured in 4-5 level winds. However, due to the precision requirements of the mapping task, it is generally not suitable for operation in strong wind weather. We appreciate your suggestions and will consider the possibility of encountering sudden changes in weather and wind strength in our future research.
- Regarding the environment conditions, especially the obstacle and buildings that you pointed out. The DEM or DSM of the mission area will be obtained in advance before path planning for remote sensing observations, and set a reasonable flying height accordingly to avoid obstacles that may exist in the flying environment, such as tall trees and buildings.
- The wireless communication link system is an important component of the UAV system, used to complete remote control, telemetry, and task information transmission by the ground control station to the UAV over long distances. However, this does not fall under the scope of remote sensing-based UAV path planning, and therefore, our discussion of pre-flight path planning does not include research on real-time UAV flight control.
- The DAA problem. In order to fly safely in national airspace alongside noncooperative aircraft and other non-broadcasting dynamic and static obstacles, an unmanned drone needs another form of risk mitigation, which is where Detect-and-Avoid (DAA) comes in. We find that an electro-optical based DAA solution, which rely on computer vision techniques to classify visible objects and estimate their positions and velocities, is the best form of risk mitigation for autonomous UAS. As we mentioned earlier, obtaining a digital surface model (DSM) of the mission area allows for collision avoidance with large static obstacles through flight altitude settings. In addition, applying for airspace clearance prior to actual flight can prevent collisions with other UAVs performing different tasks, and the flight altitudes of remote sensing observation do not interfere with commercial aircraft. While encountering flying birds is an uncontrollable obstacle during UAV flights, the probability is relatively low, and processing collision avoidance with birds through computer vision techniques would increase algorithmic costs.
- The UTM system is an air traffic management system designed by NASA for the FAA, and is a critical safety system that allows many drones to fly simultaneously in low-altitude airspace without collision. This software system, which can interact with both drones and existing air traffic control systems, can effectively ensure the safety of drone flights. However, it may be somewhat unnecessary for our remote sensing tasks, as our application requirements mainly involve quickly covering a large area of ​​the operating area, obtaining high-frequency, large-scale, ultra-high-resolution data, and do not require real-time monitoring. Therefore, we only considered the collision avoidance problem among the UAVs when considering the obstacle avoidance system, and achieved this by adding constraints to the path planning algorithm. Our algorithm can achieve low-cost and efficient completion of UAV networking tasks for remote sensing applications.
Thank you very much for your valuable feedback which has broadened our research ideas. We believe that it will have a significant guiding impact on our future research direction.
Reviewer 4 Report (New Reviewer)
1. This manuscript is more like a research proposal than a research paper
2. The authors transform UAV network path planning into a vehicle routing problem. Why choose this method？What are the advantages of this method over existing methods？The authors have not explained.
3. Is it reasonable that the path planning of unmanned aerial vehicles is three-dimensional and the author transforms it into two-dimensional
4. In simulation, the simulation conditions are too simple and have not been compared with existing methods. Thus, the simulation results are not convincing enough.
5. The authors set the cruise speed of the UAV to 7.5 km/h and 6 km/h in the simulation conditions, which is obviously not consistent with the actual situation.
Author Response
Thank you for taking the time to review our manuscript. We have prepared a point-by-point response to your comments, which we hope will address your concerns and questions.
Comment 1: This manuscript is more like a research proposal than a research paper
Response: Thank you for your feedback on our manuscript. We appreciate your comments on the need for more empirical research to support our proposed algorithm. However, we would like to clarify that our manuscript is a research paper that presents original research findings based on our study objectives, methodology, data analysis, and conclusions. We have revised the abstract and introduction part in our manuscript to emphasize the research problem and our contribution. We hope that these changes address your concerns and meet the requirements of a research paper.
Comment 2: The authors transform UAV network path planning into a vehicle routing problem. Why choose this method？What are the advantages of this method over existing methods？The authors have not explained.
Response: Thank you for your comment and for bringing up an important point. In the introduction, we briefly reviewed the existing UAV network path planning methods for task allocation, which can be divided into three types. The first two methods allocate tasks to sub-regions based on UAV performance or using region segmentation algorithms, and then plan paths within each sub-region. However, these methods are prone to the risk of UAV collisions between sub-regions and are difficult to apply in practice. The third method is numerical optimization, which directly allocates tasks without dividing them into sub-regions.
We apologize for your confusion caused by the unclear expression. We will revise the manuscript to make it more logical and improve clarity. For example, the Vehicle Routing Problem (VRP) can be used to allocate tasks in UAV network path planning. VRP is a classical optimization problem in operations research that can transform the task allocation problem in UAV network path planning into a numerical optimization problem by using mathematical programming techniques to find the optimal solution under specific objective functions. This method has several advantages over existing methods such as being able to handle multiple objectives, constraints, and uncertainties. By using VRP, we were able to effectively solve the task allocation problem in UAV network path planning while considering UAV performance. However, there are still some issues that need to be optimized such as task allocation objectives and route crossing problems. Therefore, we chose this method for further research on UAV network path planning algorithms.
Thank you for your feedback! We appreciate your comments and will modify the relevant expression of the manuscript accordingly.
Comment 3: Is it reasonable that the path planning of unmanned aerial vehicles is three-dimensional and the author transforms it into two-dimensional.
Response: Thank you for your comment. We need to explain that it is necessary to strictly follow the aerial photography standards in UAV path planning for remote sensing observation tasks. The standard requires that the ground resolution of the image acquired by a mission should be consistent in theory, so the relative flight altitude of the aircraft must remain unchanged. When the terrain of the observation area is flat and the terrain undulations are small, the impact of changes in ground elevation can be ignored, and two-dimensional path planning of UAV networking is required. However, for observation areas with large terrain undulations, the relative height of the drone is affected and high-precision three-dimensional path planning is required.
Compared with the constraints of two-dimensional path planning, the constraints of photography flight altitude and overlap rate in three-dimensional path planning are different. For fixed-wing drones that cannot adjust flight altitude, terrain undulations will cause changes in the overlap degree between adjacent images in the flight direction, and a decrease in overlap degree will affect the stitching accuracy of images. Therefore, it is necessary to adjust the constraints on overlap degree. For variable-height flying drones, their actual flight height changes correspondingly according to changes in terrain undulations, and the photo shooting elevation value also changes. Additionally, this problem is generally solved by DEM interpolation.
After calculating the basic parameters of UAV track based on the aerial photography constraints, task allocation for UAV network can be performed, which is the focus of our research. Currently, research on three-dimensional path planning for UAV network is still based on two-dimensional path planning according to sensor detection range, followed by bilinear interpolation to obtain the elevation value corresponding to two-dimensional path points. Finally, a multi-objective integer programming model was constructed using genetic algorithms to complete task allocation. Therefore, whether it is a two-dimensional or three-dimensional drone networking task allocation method, they are essentially the same. Specific factors should be considered according to specific problems and some optimization algorithm should be used for task allocation. That is why we chose to focus on two-dimensional path planning in this study. We hope this explanation helps clarify our approach.
Comment 4: In simulation, the simulation conditions are too simple and have not been compared with existing methods. Thus, the simulation results are not convincing enough.
Response: Thank you for this valuable feedback. In the simulation experiment, we considered a variety of factors such as the endurance of the drone, the distance between the starting point and the field station, the total time of the networking observation, the number of UAVs, the performance indicators of the UAVs, and the takeoff position of the UAVs while ensuring the safety. In addition, we also considered searching for the optimal path for UAVs to fly from the field station to the observation area. Our simulation conditions are feasible for completing UAV network path planning tasks quickly and cost-effectively. However, our algorithm for path planning was completed before UAV networking flight and may have ignored some unexpected weather conditions or environmental factors that could arise during actual flight. In future research, we plan to introduce computer vision technology to further study real-time flight control issues, but this is not particularly necessary for the application requirements of our current research.
Comment 5: The authors set the cruise speed of the UAV to 7.5 km/h and 6 km/h in the simulation conditions, which is obviously not consistent with the actual situation.
Response: Thank you for your comment. Regarding the cruise speed of the UAV, we set a lower flight speed for the UAV based on the size of the observation area to obtain higher quality aerial photography data. The setting of relevant parameters was also based on flight parameters that had been set in other previous experiments. We used a large, long-endurance multi-rotor drone called Thor 210, which is equipped with a 6 kW hybrid power system that can adapt to our flight parameter settings. Different task requirements will have different flight speed settings and different trajectory planning results, but this does not affect the effectiveness of our algorithm.
Thank you again for your valuable comments. We will carefully consider and make the necessary revisions. Please let us know if there is anything else we can do to improve our manuscript.
Round 2
Reviewer 3 Report (New Reviewer)
Nil
Reviewer 4 Report (New Reviewer)
None
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
In this article, the authors focuses on the fast networking path planning methods of UAVs for remote sensing observation and proposed an optimized algorithm of UAV networking path planning based on the vehicle routing problem (VRP) model. The algorithm transforms the task assignment problem of UAV networking into VRP on the basis of the optimized coverage path of one single UAV, and takes the shortest observation time of UAV network as the goal to carry out task assignment.
I have the following remarks
1. In the summary the results should be introduced.
2. It is recommended to review your abstract because it is difficult to see your problematic.
3. I suggest that you revise your introduction as well. Please try to make three sub-sections. One section describing the problem studied, another for the work done and one for your motivations and contribution.
3. More recent articles are needed for a better performance. you can add this article but not in the obligation.
Ouamri MA, OteÅŸteanu M-E, Barb G, Gueguen C. Coverage Analysis and Efficient Placement of Drone-BSs in 5G Networks. Engineering Proceedings. 2022; 14(1):18.
4. Several of your parameters are not defined. add a table describing the parameters.
5. Your English is poor. for example this sentence
The results of solvers CPLEX and Gurobi are the same. The results shown in the graphs show that the number of UAVs solved by the same method with the same parameters and the total time for the networking operation are the same for both solvers.
6. you say that The results of solvers CPLEX and Gurobi are the same. what is your contribution then. A comparison with other methods is necessary.
7. Your figures are misinterpreted and need to be explained.
8. the article needs to be revised in its entirety
Author Response
Dear reviewer,
We would like to thank you for your valuable comments and suggestions regarding our manuscript. We have carefully reviewed each of your comments and made the necessary revisions to improve the quality of our paper. Our response is as follows.
Comment 1. In the summary the results should be introduced.
Response：We appreciate your comment regarding the introduction of results in the summary. We have revised the abstract to better highlight the main result of our study, which is the development of an optimized algorithm for UAV network path planning based on the VRP model. We acknowledge that the initial introduction of results in the abstract was not clear enough, and we have further highlighted the optimization of the proposed method.
Comment 2. It is recommended to review your abstract because it is difficult to see your problematic.
Response：Thank you for your suggestion to review our abstract. We agree that the problematic of our study was not clearly introduced in the initial version of the abstract. We have revised the abstract to better explain the research problem, which is the lack of effective methods for UAV network path planning in remote sensing observations. We believe that the revised abstract better conveys the objectives and contributions of our study.
Comment 3. I suggest that you revise your introduction as well.
Response：We appreciate your suggestion to revise our introduction. We have recognized that the sub-sections in the introduction were not well-organized, and we have restructured them according to your suggestions. The first section now introduces the background of the research, while the second section focuses on the research status of task assignment. The third section presents the problem, pointing out that the UAV group network path planning problem is still worth studying. We have also included a final paragraph that explains the motivations and contributions of our study.
Comment 4. Several of your parameters are not defined. add a table describing the parameters.
Response：Thank you for pointing out that several of our parameters were not defined clearly. We have rechecked all the constants and variables definitions, and we agree that they were unclear due to awkward wording. We have revised the paper to provides a clear and concise definition of all the parameters used in our study.
Comment 5. Your English is poor.
Response：We regret that there were problems with the English in our paper. We have addressed this issue by having the paper carefully revised by a professional language editing service. We are confident that the revised paper now meets the required standards of clarity and readability.
Comment 6. You say that the results of solvers CPLEX and Gurobi are the same. What is your contribution then? A comparison with other methods is necessary.
Response：Thank you for your insightful comment. We compared our optimization method with the original method by using the control variate method, and the use of same solvers (CPLEX and Gurobi) was to demonstrate the effectiveness of our optimization. However, it is just one of the results observed from the simulations and does not represent our full contribution. We agree that comparing our method with more optimization techniques would enhance the contribution of our research, and we will consider incorporating a comparison with other methods in future work.
Comment 7. Your figures are misinterpreted and need to be explained.
Response：Thank you for bringing this to our attention. We acknowledge that our figures need improvement and further explanation to avoid any misinterpretation. We have carefully revised and clarified the meaning of our figures to ensure that they are correctly interpreted.
Comment 8. The article needs to be revised in its entirety
Response：Thank you for your feedback. We appreciate your honesty and recognize the need to improve the overall structure and clarity of our article. We have thoroughly revised the manuscript to simplify the language, eliminate repetition, and clarify the meaning of our research. We sincerely thank you for your constructive comments and suggestions which helped us to improve the quality of our paper.
Once again, we would like to express our sincere gratitude for your helpful comments and suggestions. We believe that the revisions we have made will significantly improve the quality of our paper, and we look forward to your feedback on the revised version.
Reviewer 2 Report
This paper focuses on the fast networking path planning methods of UAVs for remote sensing operations and proposes an optimized algorithm of UAV networking path planning based on the Vehicle Routing Problem (VRP) model. The simulation work has been done in the MATLAB; however, the simulation results are not promising. For example, how is collision avoidance handled? how are UAVs taken off when there are more than two UAVs in the simulation (from logistics centers?). In addition, the English writing of the paper should be improved in future submissions. Please visit your university English writing center or a professional English service. The detailed comments have been added and highlighted in the reviewed PDF file as attached, so please check them in detail for possible revision.
Comments for author File: Comments.pdf
Author Response
Thank you for your comments and feedback regarding our paper. We have carefully reviewed each of your comments and made the necessary revisions to improve the quality of our paper.
Comment 1. How is collision avoidance handled?
Response：In our UAV network operation, we have considered the constraint of collision avoidance in the task allocation process. Equations (12) and (13) were introduced as constraints to prevent UAVs from conducting invalid operations and ensure the safety of the UAV network operations. Specifically, the constraints require each route to be adjacent to the next one when the UAV flies, and UAVs cannot fly from the first route to a route other than the second one. Equation (12) is the constraint for even number of routes, while Equation (13) is the constraint for odd number of routes.
Comment 2. How are UAVs taken off when there are more than two UAVs in the simulation (from logistics centers?)
Response：In our simulations, we have set up a fixed field station from which all UAVs take off. Each UAV has a preparation time and must wait for other UAVs to take off.
Comment 3. The English writing of the paper should be improved in future submissions
Response：We regret that there were problems with the English in our paper. We have addressed this issue by having the paper carefully revised by a professional language editing service. We are confident that the revised paper now meets the required standards of clarity and readability.
Comment 4. The detailed comments in the reviewed PDF.
(1) Undefined acronyms
Response：Regarding the undefined acronyms, we apologize for the confusion. The undefined acronyms like “YALMIP” “CPLEX” are not acronyms, but rather the names of the toolbox and solver tool used in MATLAB.
(2) The endurance time is 2 hours - it may be too long for the UAV.
Response：As far as we know, the performance of UAV power systems can vary depending on the type of engine used. Electric rotorcraft UAVs typically have an endurance of around 30 minutes, while those with internal combustion engines can last longer. The fixed wing UAV uses an oil-powered generation system, which is why we believe a 2-hour endurance time is reasonable.
(3) Several confusing terms in the manuscript. (“observation time” and “operation time”, and "operation time" and "observation time").
Response：We appreciate your feedback regarding the confusing terms in our manuscript. We have thoroughly revised the manuscript to avoid such confusions, and clarify the meaning of our research.
Once again, we would like to express our sincere gratitude for your helpful comments and suggestions. We believe that the revisions we have made will significantly improve the quality of our paper, and we look forward to your feedback on the revised version.
Round 2
Reviewer 1 Report
The authors did not respect the remarks given by the reviewers, no action was taken unless the authors submitted the old version of the article
Author Response
Dear Reviwer,
We appreciate your time and effort in reviewing our manuscript once again. Your feedback is essential in improving the quality of our work. We notice that you found no apparent changes in our revised version. However, we would like to kindly note that all modifications were made using the track changes function, as shown in the MS Word version we uploaded. We also have responded to your valuable comments point by point in the Author's Reply module. We sincerely apologize if the unmarked PDF version we provided caused any confusion or misunderstanding. For your convenience, please refer to the attached document for confirmation of the changes made. Thank you again for your valuable insights, and we look forward to your further guidance.
Best Regards,
Sutton Chen
Reviewer 2 Report
The English writing of the paper has been improved a lot from the last submission, but the English writing in the current revised paper is still required further improvement as commented and highlighted in the reviewed paper. Please read through the revised paper before the next submission since several broken sentences exist in the current revised paper. In addition, please submit a version with tracked changes in the next submission so that the reviewers can easily know the changes that have been made.
Regarding a few comments listed in the last submission, let's have more discussions.
(1): How is collision avoidance handled?
As explained in your response, it seems you only consider "obstacle avoidance" among the UAVs in your tasks. The "obstacle" may include trees, buildings, or other aircraft flying into your task area. Thus, your "obstacle avoidance" may be not the right terminology for the methods or considerations in your study.
(2) The endurance time is 2 hours - it may be too long for the UAV.
As the UAV flight speed assumed in the paper (e.g., 7.5 km/h or 6 km/h), the UAVs discussed are most likely rotary-wing aircraft: rotorcraft (helicopters) or multirotors. Currently, in most applications, multirotors are dominant and they are often powered by batteries (because of the use of electric motors).
On the other hand, the fixed-wing aircraft are usually flying faster than 10 km/h.
(3) Several confusing terms in the manuscript. (“observation time” and “operation time”).
Please help define the "observation time" and the "operation time" (what are their differences?) since they are both used in the current revised paper as highlighted in the reviewed paper.
Comments for author File: Comments.pdf
