Artificial Intelligence Methods in Safe Ship Control Based on Marine Environment Remote Sensing
Round 1
Reviewer 1 Report
The manuscript is clear and presented in a well-structured manner. The author demonstrates how to effectively computerize the navigator's maneuvering decisions in complex navigational situations by the appropriate control algorithms design using artificial intelligence methods and radar remote sensing of the marine environment. This work shows that combining traditional control algorithms, artificial intelligence methods, and marine radar environmental remote sensing can effectively computer support the navigator's maneuvering decisions in complex navigational situations. Based on this manuscript and previous work, the author is an expert in maritime navigation, transportation safety, optimal control, and game theory.
Approximately 40% of the scientific papers cited in the manuscript are from 2017-2022, and the number of self-citations is minimal. Not all work results are reproducible, as accurate data are not provided. However, the reader can directly conclude that the tables, figures, and schemes are appropriate - they correctly show the data and are easy to interpret and understand. The data is interpreted consistently throughout the manuscript. The conclusions are consistent with the arguments presented in the manuscript.
The ethics statements and data availability statements ensure they are adequate.
Concluding, the problem of the safe ship traffic management at sea system design is original and well-defined. Especially, determining the safe and optimal trajectory of own ship using dynamic programming with neural constraints of the state in the form of domains of encountered ships generated by a shallow feedforward neural network and game theory is interesting.
A combination of remote sensing, neural network, and game theory for safe ship traffic management at sea provides advancement of the current knowledge.
The work fits the journal's scope.
The results were interpreted appropriately and are significant. In general, the conclusions are justified and supported by the results.
Generally, the article is written appropriately, and the analysis is presented accordingly. The advantage of the article is that the calculations were carried out for the navigational situation recorded on the radar screen of the research/training ship in the actual navigational situation in the Skagerrak-Kattegat Strait.
Not all conclusions can be drawn by the reader from the data, as accurate data would have to be provided separately. The tools used by the author are well-known and famous in the academic community (Matlab/Simulink). It is a pity that the Matlab code is not described in English (or at least commented on).
The paper can attract a wide range of readers interested in the severe applications of artificial intelligence methods in decision support systems design and control. However, shipping safety is a niche, so that it will interest only a limited number of people.
Minor adjustments to the manuscript are recommended.
1. English should be improved.
2. As for maintaining the highest technical standards would have been better if the article had been written using a professional technical word processor, i.e., LaTeX editor, especially the mathematical equations (1)-(9).
3. Pages 15 and 19: The abbreviation for MPDM needs to be understood.
4. Page 18: Not "Conlusions" but "Conclusions."
5. Page 16: According to [JL]?
6. Pages 9-11: Algorithm NEURAL DOMAINS should be commented on in English.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
I. Comments, remarks, doubts and questions
The Author presents the methodology of solving ship encounter situations and avoiding collisions at sea using radar as an information source.
The examples presented demonstrate the effectiveness of the proposed methodology, however there is no information if ship dynamics of own and target ships were taken into account.
Radar is commonly used for solving collision situations at sea. A number of examples of the use of radar in collision avoidance decision support systems are presented in the literature. The lack of literature analysis indicating the gaps that the author wants to eliminate makes it difficult to identify the author's original contribution.
There is a lack of information about the originality of the proposed methodology (authors contribution).
The author presents short descriptions of four proposed algorithms for determining a safe trajectory and solutions to collision situations obtained on their basis. However, he does not compare the results: determined trajectory lengths, maneuver parameters (number of waypoints, efficiency of individual maneuvers, times to perform determined manoeuvres), deviation (shift from a previous trajectory) or other quantitative and qualitative criteria for the assessment of the performance of generated solutions.
II. Detailed remarks:
1. There is a lack of literature analysis of the methods proposed or used in navigational decision support systems.
2. Figure 1 can be considered as a graphical abstract. The figure does not show the structure of a decision support system.
3. The DT algorithm uses a neural network evaluating the risk of collision with each encountered ship and determining the size of domain of passing ship. There is no information about the parameters of the vessels used for gathering learning data set (sets). There is also no information about the size of collected data set used in supervised learning process.
4. The presented algorithm named “Algorithm.Neural domains” (page 9-11) contains incomprehensible and non-English names and descriptions. I propose to replace it with an algorithm written in pseudocode in English.
5. It is advisable to discuss the results of using each of the algorithms in more detail.
6. It is advisable to compare the results obtained using the presented algorithms.
7. It is advisable to highlight the achievements/innovative solutions proposed by the author in the reviewed article.
III Summary
The paper is interesting, however, it raises a number of comments, questions and reservations.
Major revisions.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
--
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Author considered remarks and comments included in the review. The article is more coherent after Author's corrections. There are still some language incoherencies they, however, do not prevent the Reader from understanding the article.
