Path Optimization for Aircraft Based on Geographic Information Systems and Deep Learning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Please find attached a pdf file wih my comments. Thank you.

Comments for author File: Comments.pdf

Author Response

Please find the attached file as a response.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper focuses on agricultural UAV navigation, yet all trajectory samples are taken from a commercial passenger flight (Baghdad–Istanbul). This mismatch weakens the motivation, as the aerodynamic behavior, operational constraints, and environmental interactions of agricultural drones are fundamentally different from those of commercial aircraft. The authors should either justify this substitution in a more rigorous manner or redesign the dataset so that it reflects UAV operations directly.

In addition, the integration between GIS layers and the deep neural network remains insufficiently explained. The manuscript lists GIS tools and datasets, but does not clarify how spatial layers, coordinate transformations, elevation data, or atmospheric parameters are encoded and incorporated into the model. Without a clear pipeline showing how GIS-derived features influence the predictive accuracy of the network, it is difficult to assess the technical contribution of the work. The current explanation of the DNN architecture is also very brief, offering hyperparameters without a discussion of why the authors selected this particular configuration or how it compares with alternative models.

The evaluation methodology also requires improvement. The study only contrasts assigned routes, actual trajectories, and predicted routes, without comparing the proposed DNN to any baseline models such as polynomial regression, LSTM architectures, or reinforcement learning approaches commonly used in UAV navigation. Benchmarking against even one simpler or well-established model would allow readers to determine whether the proposed approach is justified or whether similar results could have been obtained with a less complex method. Furthermore, key details required for reproducibility are missing, including the exact number of training samples, preprocessing steps, normalization procedures, and justification for the stopping criteria in training.

The literature review, although extensive, lacks a structured synthesis that allows readers to understand the novelty of the present study. The discussion lists several works but does not clearly identify where existing models fall short. A comparative summary of previous studies would greatly strengthen this section. For example, a table comparing UAV navigation studies in terms of their AI models, data sources, application domains, and limitations would help position the current contribution more clearly. Such a comparison would also reveal the main knowledge gaps left by prior research, such as the limited use of real flight disturbances, the lack of dynamic GIS integration into navigation models, and the absence of unified frameworks that combine atmospheric data with deep learning in an operational setting. Highlighting these gaps explicitly would make the contribution of the paper more convincing.

Overall, while the topic is promising and the integration of GIS with deep learning is relevant, significant improvements are required in methodological clarity, dataset consistency, evaluation rigor, and the framing of the study within existing literature. A more coherent alignment between the research goal, data source, and proposed model is essential for the manuscript to reach its full potential.

Author Response

Please find the attached file, which is the response letter.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Please find my further comments in the attached file.

Comments for author File: Comments.pdf

Author Response

Thank you for your second round constructive comments, please find attached the response letter.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

All comments have been properly addressed in the manuscript, and it is ready for acceptance.

Author Response

Thank you for accepting the revision of our manuscript. We have further updated the manuscript depending on the second round of review of the first reviewer for your reference. 

Yours,

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

There are still minor points to be clarified/corrected, as follows:

1. “Aircraft” is a collective noun that can refer to a single machine or a group of flying vehicles. No plural is needed. Please correct the title and other instances of “aircrafts” in the text, that should read “aircraft” (Line 421).
2. Figure 8. The term “Route” is still standing in the legend of Fig. 8a, instead of “Path”. Moreover, the circumferences are not labeled with the same scale (0.1-05 for Fig. 8a, 1-8 for Fig. 8b). This is misleading for a fair comparison of the two figures.
3. More on distances in Tables 5 and 6. The shortest distance between IST and BGW (great circle arc) is 1612 km (https://www.greatcirclemap.com/?routes=IST-BGW). The actual flight routes are consistent with this distance (they are greater since they include maneuvers and deviation from the great circle distance), but, again, it is impossible to establish assigned and DNN-predicted routes that are shorter that the shortest distance! These routes simply cannot be less than 1612 km. In Table 5, the difference in fuel consumption between actual and DNN-predicted path is 4938-4302=636 kg, not 610. Finally, how can 1067 km (the assigned ATC-constrained path) be covered in less time than 1040 km (the DNN path). Are the paths covered at different speed? If so, why? And how could these routing problem be scaled to an agricultural UAV mission?

Author Response

Kindly check the attached file.

Author Response File: Author Response.pdf