An Improved Probabilistic Roadmap Planning Method for Safe Indoor Flights of Unmanned Aerial Vehicles

Round 1

Reviewer 1 Report

The manuscript entitled “An Improved Probabilistic Roadmap Planning Method for Indoor Safe Flight of Unmanned Aerial Vehicle” proposes an improved probabilistic roadmap (PRM) planning method for indoor safe flight of the UAV. The paper is well-written and the theory, proposal, discussion, and results seem convincing. In this reviewer’s opinion, the manuscript seems interesting and worthy of attention. However, some aspects should be addressed or clarified before further consideration:

1. The introduction section needs to be supported by the related literature. 
For example “Line 25-line 26: The UAV has rapidly developed and it is applied in many aspects of the industry as well as human life, e.g., agriculture, monitoring, transportation, delivery and rescue.” the necessary published articles need to be cited.
Apolo-Apolo, O. E., J. Martínez-Guanter, G. Egea, P. Raja, and M. Pérez-Ruiz. (2020)"Deep learning techniques for estimation of the yield and size of citrus fruits using a UAV." European Journal of Agronomy 115 126030.
Ghelichi, Zabih, Monica Gentili, and Pitu B. Mirchandani. (2021) "Logistics for a fleet of drones for medical item delivery: A case study for Louisville, KY." Computers & Operations Research 135 105443.
Liu, Chang, and Tamás Szirányi. (2021) "Real-Time Human Detection and Gesture Recognition for OnBoard UAV Rescue." Sensors 21, no. 6 2180
It would be useful if the authors could review such techniques and put them into the context of this work.

2. The algorithms for route planning need to be reviewed and summarized in the first chapter, and not only the indoor scenarios but also the outdoor scenarios should be presented.

3. Figure 2 should have the class label: class 1 is an obstacle and class 0 is free values.

4. Equation 1 should be clarified, what is x-min what is y-min and where does the formula come from? How to proof?

5. Pseudo-code of improved PRM needed.

6. The details of Figure 6 should be described.

7. The quality of the writing is not very high, not only in the formatting, but also in the spelling of the text, for example, the introduction of the formula in line 511 is wrong.

8. I am interested in the comparison figures of the results of the basic PRM and the improved PRM methods, and it would be a very convincing thing if the authors could present it in this paper.

9. The authors should make their experimentation fully reproducible, hence I encourage the authors to make their data and source code publicly available.

10. The main contributions and innovations need to be summarized at the end.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Good work with a detailed quantitative and qualitative data. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper proposes an improved probabilistic roadmap (PRM) planning method for indoor safe flight based on the assumption of the UAV in quadrotor model. From the results, it can be seen that the method ensures indoor safe flight of the UAV while also significantly improving computational efficiency.

Firstly, the author describes how to transform three-dimensional space into two-dimensional space, that is, the method of generating reduced-dimensional raster map from three-dimensional indoor environment. This greatly simplifies the environment while retaining the necessary environmental information such as boundaries and obstacles.

Secondly, the author introduces the basic PRM algorithm and the improvement strategy of PRM algorithm, including setting the "connection distance" parameter to reduce the number of edge generations between distant nodes. At the same time, based on the strategy of "constructing first and checking later", the method of path local check and incremental update is proposed, which reduces ineffective collision checks on edges, decreases consuming time and improves algorithm efficiency.

After that, the author uses the path post-processing optimization method to reduce unnecessary access to redundant nodes.

Finally, the author verifies the effectiveness and efficiency of this method from the aspects of path-finding success rate, planning time and path length.

Generally speaking, the article is fluent in writing, natural in connection, clear in logic and comprehensive in content.

Reviewer 4 Report

The paper presents a UAV path planning strategy for indoor flight, leveraging an improved PRM path planning and dimensionally-reduced model of the environment (from a 3D to a 2D one).

Although the authors accurately described their approach and provided several results supporting it, the contribution of the proposed method is too weak compared to current state of the art in UAV path planning.

The dimensionality reduction oversimplifies  the path planning problem thus making the proposed approach more suited to a ground vehicle rather than an aerial one.

Furthermore, as stated by the author themselves, the proposed 3D projection method can be applied only to simple 3D models, i.e., those mainly including vertical and sharp obstacles without any overhanging structure. Therefore, the 3D projection method cannot be considered a relevant contribution.

The enhancements performed on the PRM path planning seem mainly to be technical improvements of the PRM implementation available in Matlab. Moreover, since the method has been tested only on a classical desktop PC without any particular computing resource limit, the intention of the authors to obtain an efficient algorithm is hardly applicable to typical hardware available onboard of a UAV.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

The authors partially addressed the raised issues. Concerning the multi-layer grid they just briefly mentioned the "overlapping area" approach in the experiments, rather than clearly explain their method in Section 3. Therefore, it is unclear how the graph used by the PRM is generated, how the overlapping area is discovered and, finally, how the whole path is automatically created.

The authors should report how they have migrated the Matlab based implementation of their algorithm to the one running on the onboard PC (as I assume they have not installed Matlab on the DJI Manifold PC).

Furthermore, it is unclear why in Fig. 9a caption the authors say "Initial path before optimization". Is it the green line a single path?

Finally, the paper could definitely be improved by including some experimental trial with a real quadrotor (possibly including a video). In fact, in my opinion the multi-layer grid approach is still weak compared to current state of the art in quadrotor path planning as it limits the path planning reasoning to a "layered" approach, thus potentially lead to sub-optimal path. However, if the authora are able to prove the efficiency and quickness of their method in a real-world setting, this could definitely enhance the quality of the paper.

Minor comments:

- Please check carefully the English of your manuscript (especially in the newly added parts)

- Please change "segementation" to "segmentation" in Fig. 3

Author Response

Please see the attachment.

Author Response File: Author Response.docx