Research on Multi-Target Path Planning for UUV Based on Estimated Path Cost
Round 1
Reviewer 1 Report
The paper presents a method for multi-target path planning that consists of two phases. In a first stage, the cost is estimated by a CNN. This cost is the input of a planning method which is obtained by deriving a TSP problem where a GA is used in order to obtained a quasi-optimal order of pass. I think that the system can be of interest to the reader and also taking into account the real experimentation presentd. The paper is in general well written, but there are some lacks in it that do not allow me to accept it its current form.
Some comments:
1. Introduction.
You claim that "UUVs are recognized for their distinctive attributes of stealth and artificial intelligent". This state is somewhat unprecise. Even though there are intelligent UUVs there are also remotely operated ones, in which the operator is in charge of the more relevant decisions. In addition, these affirmation should be supported with references.
Which relates to the next comment: the references are scarce and there are some omissions such as:
Enric Galceran, Narc´ıs Palomeras and M. Carreras. Profile
Following for Inspection of Underwater Structures. Paladyn Journal of Be-
havioral Robotics. Volume 4, Issue 4, December 2013, Pages 209-220.
Rubio-Sierra, C.; Domínguez, D.; Gonzalo, J.; Escapa, A. Path Planner for Autonomous Exploration of Underground Mines by Aerial Vehicles. Sensors 2020, 20, 4259. https://doi.org/10.3390/s20154259
Li, Daoliang & Wang, Peng & Du, Ling. (2018). Path Planning Technologies for Autonomous Underwater Vehicles-A Review. IEEE Access. PP. 1-1. 10.1109/ACCESS.2018.2888617.
Regarding to the contributions, the first one is quite obvious because the problem is very similar (is a TSP). Then, the third one is just an implementation detail, which I assume is used in many methods in the literature. It is not clear whether the Informed-RRT* is a contribution or not in the text, please add the proper reference in Section III to clearly state it.
In addition, in Section 3.2, you claim that Informed-RRT* algorithm inherits the asymptotic optimality, without a proof or a hint (it is the exact same algorithm at the first stages and then reduces the sampling area to improve convergence). Please clearly state why you do these affirmations and if not, please include the reference where it is proven clearly (see [...] section ...])
In my opinion: the structure of the paper could be improved: there are parts of the method (cost estimation) that are explained in the Simulations Section, they should be explained, in my view, in the previous section. Then, you put two different scenarios and the parameters, they are nice in Section IV.
In Section 3, the results of the estimation given by your NN are not very well presented. There should be much many situations to train the network, and they should be divided at least into training and test situation. You should train with the former and then see if your CNN network can infer proper results in the test cases.
Regarding to the results, why don't you use the same algorithm with the actual cost Matrix obtained by Informed-RRT*? In my opinion, the number of scenarios is not adequate to claim that both methods will always give the same order of pass.
In addition, I think a nice addition to the paper is to generate several random scenarios (not only two) to further compare the results between the two methods. In this way, we can compare them with much more data.
Last, the experimental part is interesting, but is not clearly related to the algorithm. You clearly should add what were the inputs of your algorithm (the locations of the mines and the obstacles) and then which were the paths given by it. Then you could include the reference and actual paths.
To sum up, the paper is interesting, but in my view there are plenty of areas in which it can be improved.
The English use is appropriate and I have not found any major errors or typos.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
jmse-2506730
In this article, the authors presented a path cost estimation strategy based on neural networks. This strategy swiftly generates an accurate cost matrix, ensuring the attainment of high-quality traversal sequences when utilized as input for the traveling salesman problem, thereby yielding a globally optimal path. I have some comments and suggestion as follows:
There are numerous grammatical mistakes and typos that must be corrected with careful revision of overall manuscript.
Further increase the font size of the text in the figures, such as figure 1 and so on. Moreover, increase the resolution of figures.
The literature review is very limited. The authors are suggested to add further literature, especially from the latest years.
To know more about underwater communication, the authors can also refer to "Localization and Detection of Targets in Underwater Wireless Sensor Using Distance and Angle Based Algorithms,” IEEE ACCESS", "A Review of Underwater Localization Techniques, Algorithms and Challenges”, Journal of Sensors"
The network area seems very limited as the authors mentioned 80 and 100. Is this approach applicable for limited area? Or also applicable for large cross-section area?
As the underwater environment posses several challenges, such as water current and other obstacle. So, is this model is influenced with such challenges? Moreover, this model is tested in a lake area but is this model also suitable for sea environment too?
The authors have added a comparison of time costs under three cost matrix and a comparison of path costs of three cost matrix algorithms. But the authors missed the comparison with other state-of-the-art approaches. Therefore, it is suggested to add a proper comparison and would be better to add in a separate table, also add their references.
References are very limited according to the paper length. More references should be added, especially from the latest years.
Extensive editing of English language required
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The work described in this paper is about the forth an innovative neural network-based strategy for path cost estimation. The principles and methodologies of multi-target path planning and the training process for the neural network aimed at path cost estimation are thoroughly expounded. The experiment has already obtained information on obstacles and suspected target 82 points in the mission sea area by other reconnaissance methods. Simulation experiments demonstrate that while maintaining high-quality solutions, the proposed strategy significantly enhances the computational efficiency of the algorithm.
Some remarks regarding the content of paper:
- It is not enough overview of existing approaches and relevant references in state of the art (Introduction). There are only several references after 2020 year. What about research after this year in present time?
- There is no no explanation what it is Informed-RRT* Algorithm.
- Some abbreviations are entered several times (for example, TSP).
- Formulas are not located on the same line as the text.
- The plots in Figure 10 are too small for understanding.
- The Figures 2, 3, 4, 6, 7, 8, 9, 12 should be in the middle of page.
Authors should carefully examine and correct syntactic errors.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments to authors:
As a reviewer of the first version, I've found that you have clarified many of the issues I detected in
the first version. The description of the training process to reach your NN is upgraded and the
structure of the paper is better suited. Moreover, the study of the state of the art has significantly
improved, but still there are missing references such as:
- Wu X, Gao Z, Yuan S, Hu Q, Dang Z. A Dynamic Task Allocation Algorithm for Heterogeneous UUV Swarms. Sensors (Basel). 2022 Mar 9;22(6):2122. doi: 10.3390/s22062122. PMID: 35336293; PMCID: PMC8951437.
- S. Mahmoudzadeh, D. M. W. Powers and A. Atyabi, "UUV’s Hierarchical DE-Based Motion Planning in a Semi Dynamic Underwater Wireless Sensor Network," in IEEE Transactions on Cybernetics, vol. 49, no. 8, pp. 2992-3005, Aug. 2019, doi: 10.1109/TCYB.2018.2837134.
Authours should explain in which parts their algorithm differs and presents upgrades to these references.
Unforutnately, in this version I've found some typos in the new sections that should be corrected before publication.
A review of the paper, if possible from a native English speaker, is encouraged.
In the experimental part, a clarification that we are using the paths of Scenario A is encouraged. Also, the deviations of the trajectories w.r.t. the plan should be included to certify that your plans are suitable to your experimental part (it seems so from the aerial images, but a quantitative analysis is encouraged).
Some typos:
- Section 2: Information on obstacles and suspected target points ... have been (has -> have)
- Missing space: line 237
- Comma should be replaced by a dot --> line 317
While in the first submission the language was very polished, the insertions in this versions seem more rushed. Please see the first part (comments to authors) to see the typos I've found. A revision of the whole text is encouraged.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
NA
Minor editing of English language required
Author Response
Please see the attachment
Author Response File: Author Response.pdf
