UAV Abnormal State Detection Model Based on Timestamp Slice and Multi-Separable CNN
Round 1
Reviewer 1 Report
In the current submission, the authors primarily investigate a method for detecting the abnormal state of UAV based on timestamp slice and multi separable convolutional neural network, also named as TS-MSCNN. Then, the authors have conducted concrete experimental analyses with the help of the real flight data. At last, the results show that the TS-MSCNN has higher accuracy than other traditional machine learning and the latest deep learning methods. In general, the paper is well-structured with clear study motives, so the paper can be accepted subject to the following minor issues, I list them below for boosting the paper quality.
1. For presentation issues, several aspects own some room for the enhancement: a) The figure quality should be improved by increasing the DPI number. b) The language quality should be checked to enhance the readability level.
2. In abstract, the authors can ensure to highlight with research objective, research motivation and research usefulness of the studying TS-MSCNN.
3. The study limitations should be added in both Section 1 and the last section that in terms of your future study plans.
4. The reference can be unified with the journal style, and some recent references on anomaly detection can be added, such as https://doi.org/10.1016/j.ijcce.2022.10.001, https://doi.org/10.1007/s13042-020-01230-3, https://doi.org/10.1016/j.ijcce.2022.08.002.
5. The study motives and innovation points of studying a method for detecting the abnormal state of UAV based on timestamp slice and multi separable convolutional neural network can be listed in Section 1.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
The paper presents a method for detecting the abnormal state of UAV based on timestamp slice and multi separable convolutional neural network. It is well written and propose an innovative method on this topic supported by simulations. My questions are:
1)I think that, prior to publication, since the main idea presented on paper is based on POS and SS data, the introduction section can be enanched by adding some reference on how POS and SS can be estimated starting by sensor measurements, for example see
[1] T. Hamel and R. Mahony, “Attitude estimation on SO(3) based on direct inertial measurements,” in Proc. IEEE Int. Conf. Robot. Autom., 2006, pp. 2170–2175
[2] G. Garraffa, A. Sferlazza, F. D’Ippolito and F. Alonge, "Localization Based on Parallel Robots Kinematics As an Alternative to Trilateration," in IEEE Transactions on Industrial Electronics, vol. 69, no. 1, pp. 999-1010, Jan. 2022
2)Since it is a paper related to UAV I think that also some references on how UAV are driven by control law has to be inserted, see for example
[1] Kendoul F. Survey of advances in guidance, navigation, and control of unmanned rotorcraft systems. J Field Rob 2012; 29(2): 315–378
[2] Alonge F, D’Ippolito F, Fagiolini A, Garraffa G, Sferlazza A. Trajectory robust control of autonomous quadcopters based on model decoupling and disturbance estimation. International Journal of Advanced Robotic Systems. 2021;18(2). doi:10.1177/1729881421996974
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Minor edits:
1. Abstract section mentions POS but the abbreviation is explained further below in introduction section.
2. Recommend combining introduction and related work.
Eg: Line 46 mentions rule based anomaly has low detection performance but this point is elaborated down in related work section
Line 49: It summarizes reference literature but doesn't summarize it.
3. Line 51: What are shortcomings in UAV anomaly detection? Please elaborate
4. What is ALFA? Please elaborate.
Author Response
Please see the attachment
Author Response File: Author Response.pdf