Penetrating Radar on Unmanned Aerial Vehicle for the Inspection of Civilian Infrastructure: System Design, Modeling, and Analysis^†

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article addresses a topical issue of interest to various fields.

Some notable achievements worth noting:

1. The integration of a low-SWaP (Size, Weight, and Power) ultra-wideband (UWB) impulse radar on UAVs addresses the growing demand for non-invasive infrastructure inspection in hard-to-access areas.
2. The use of Finite Difference Time Domain (FDTD) solvers, detailed antenna modeling, and multilayer media simulation provides a robust foundation for system validation.
3. Calibration using lab measurements and synthetic data improves the accuracy of the simulation results, demonstrating good alignment with physical tests.
4. The authors developed custom algorithms for range profile processing, noise reduction, distance calibration, and image correction, all contributing to clearer subsurface imaging
5. The work moves beyond simulations with lab experiments and outdoor tests on actual concrete blocks, demonstrating the system’s practical feasibility.
6. Detailed figures and step-by-step derivations aid clarity, particularly in sections on electromagnetic modeling, imaging geometry, and ML pipeline.
7. The paper outlines specific future improvements, such as domain adaptation, transfer learning, and UAV flight testing in realistic environments.

Some aspects could be improved because the article has some small gaps:

1. Limited Real-World Testing Scope - Although real-world testing is mentioned, it is preliminary. The paper relies heavily on simulations and synthetic datasets, limiting current field applicability.
2. Simplified ML Use Case - The machine learning component is currently constrained to a single object class (rebars), in a highly controlled simulation environment with limited variability (e.g., no cracks, heterogeneous materials, or complex interference).
3. Lack of Benchmark Comparison - The paper does not compare the proposed system’s performance (accuracy, depth, resolution, etc.) against existing commercial or academic UAV-GPR systems, which would contextualize its effectiveness.
4. No Discussion on Regulatory or Operational Constraints - Real-world deployment of UAVs, especially with active radar payloads, faces regulatory, safety, and logistical constraints that are not addressed.

Author Response

Please see attached responses to comments.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The article addresses a highly relevant and timely topic by exploring the use of Ground Penetrating Radar (GPR) mounted on Unmanned Aerial Vehicles (UAVs) for inspecting civilian infrastructure. This approach represents a promising and innovative solution for applications in hard- to-reach or hazardous areas, offering significant potential for practical impact. The motivation for the work is clearly presented and well contextualized, emphasizing the importance of non-invasive inspection in urban and industrial settings. Furthermore, the manuscript adopts a complete system perspective, covering aspects of design, modeling, and experimental validation of the proposed solution. The inclusion of simulations using the gprMax tool demonstrates careful attention to the electromagnetic characterization of the environment and signal behavior, which adds technical value to the work. Finally, the article stands out for its multidisciplinary application, integrating communication engineering, remote sensing, and automation into a single coherent system.

 

 

However, despite these strengths, the manuscript still requires several improvements:

 

 

1  - The manuscript does not clearly articulate what sets this work apart from existing UAV-based GPR systems. For instance, while the system integration is discussed, it is unclear whether new antenna configurations, signal processing methods, or flight control strategies are proposed. A comparative table would help highlight the novelty.

 

 

2  - The validation is limited to a short proof-of-concept and lacks quantitative metrics. Key performance indicators such as detection depth, signal-to-noise ratio, resolution, or error rates must be provided to substantiate the claims. Reproducibility and robustness under varying environmental conditions should also be considered.

 

 

3  - The manuscript treats integration aspects (e.g., mechanical vibration, electromagnetic interference, GPS accuracy) at a high level. These are critical challenges in practical UAV-GPR systems. For instance, mechanical decoupling of the antenna and timing synchronization between radar pulses and flight trajectory are missing from the discussion.

 

 

4  - The modeling using gprMax is interesting but somewhat disconnected from the experimental work. A more detailed explanation of the simulation setup (e.g., soil parameters, dielectric properties, target geometry) is required. Additionally, there is no discussion of how these simulations correlate with real-world scenarios.

 

 

5  - Several figures (Figs 26, 27, 30, and 31) lack sufficient resolution or annotation. The signal plots should include axis labels, scale bars, and clearer legends.

 

 

6  - There is no comparison between the proposed system and similar UAV-GPR platforms or conventional terrestrial systems. A benchmark analysis or at least a qualitative comparison would allow readers to evaluate the benefits and limitations of the approach.

 

 

7  - The conclusion section summarizes the work but misses the opportunity for critical reflection. It should clearly state the key findings,   limitations, and next steps.

Author Response

Please see attached responses to comments.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The work is undoubtedly interesting, since it attempts to combine numerical modeling with experimental results. Unfortunately, its significance is offset by the fact that, as the experience of several decades of experimental research in subsurface radar shows, natural results rarely justify the authors' expectations based only on theoretical research and even laboratory experiments. This is explained by the fact that in subsurface radar one has to deal with complex and inhomogeneous media and poorly predictable properties of their surface. In laboratory experiments, the medium is usually homogeneous, and the surface is equal. I would like to wish the authors to move on to flight experiments as soon as possible. In this case, their research will be of greater interest and significance.

Author Response

Please see attached responses to comments.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript presents simulation study of a low-SWaP UWB impulse radar system integrated on a UAV for non-invasive inspection of civilian infrastructure. The study combines high-fidelity electromagnetic simulations, realistic antenna modeling, signal processing, experimental validation, and a YOLOv4-tiny-based machine learning pipeline for rebar detection in concrete.

Authors uses Finite Difference Time Domain (FDTD) solvers and realistic models for antennas (Vivaldi and Horn) and layered media (air/concrete) development and experimental validation of simulation results. Paper shows development of image correction and processing algorithms for better anomaly detection. Interesting is also implementation of a synthetic dataset-based deep learning algorithm (YOLOv4-tiny) for automated detection of rebars in concrete.

There are some limitations in the proposed method, as the object detection model was trained using synthetic images. The methodology was tested only on relatively shallow targets and small concrete blocks. It would be good to include details regarding the general overview of non-sinusoidal signals with narrow pulse widths (in the nanosecond range) used for data transmission (see DOI: 10.1109/EIECS59936.2023.10435495), a deeper understanding of the signal's behavior is critical. It is worth to  include references on Ultra-Wideband (UWB) signal shapes and their corresponding spectra, such as in DOI: 10.1109/MSM49833.2020.9202219. Additionally, considering the signal's limitations, Gaussian derivatives, which are well-localized bandpass waveforms in both time and frequency domains, like DOI: 10.1109/ISPA58351.2023.10278706, and could provide important context for proposed methodology.

Despite the achievements of the proposed method, it is recommended that the authors:

1. Include additional performance metrics such as precision, recall, and F1-score, alongside average precision (AP).
2. Consider robustness testing of image correction, specifically with respect to variations in parameters like θᵢ, εᵣc, and σ, to assess the model's performance under different conditions.
3. Incorporate a discussion on regulatory limitations, particularly regarding the use of radar payloads on drones, such as the constraints imposed by the FCC on Ultra-Wideband (UWB) signals.

Additionally, the authors might consider take into account performing a quantitative comparison of detection rates, resolution, or depth performance against other Ground Penetrating Radar (GPR) systems deployed on UAVs (Unmanned Aerial Vehicles) to show   strengths of the proposed methodology.

Manuscript is recommended for publishing after including above suggestions.

From an editorial  point of view  figures properly illustrate key aspects like  antenna geometry, signal propagation, image processing steps as well as outcomes of machine learning models.

Author Response

Please see attached responses to comments.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

All of my comments have been addressed by the authors, and the manuscript has been significantly improved. I only have one remaining remark concerning Figure 28 (Matlab result), where the gray background should be changed. Aside from this detail, in my opinion, the manuscript is suitable for publication.

Author Response

Please see attached responses to comments.

Author Response File: Author Response.docx