Multistage Dynamic Optimization with Different Forms of Neural-State Constraints to Avoid Many Object Collisions Based on Radar Remote Sensing
Round 1
Reviewer 1 Report
As radar remote sensing getting popular, this manuscript will be interested to the readers of MDPI Remote Sensing. In my opinion, the work is worthy of a scientific journal, however, the manuscript is not well-structured. It needs a revision before it can be accepted for publication.
- Abstract: I found the abstract little informative about the main results and findings of this work. In my opinion, it must be improved in order to show clearly these aspects.
- Keywords: they must be keywords, so short.
- The Introduction: section needs to be rewritten in order to show clearly: i) a critical overview of state of the art (many more references need to be inserted), ii) the author’s contribution to this sate of art.
- The figures must be on one page only, for example figure 4 and figure 6 are wrong. I suggest you to insert (a),(b)… for each panel so that you can indicate them in the caption. You need to better align the figures.
- Replace figure 2, must be a block diagram.
- Equations needs to be revised for mathematical consistency and notation (for example number (16)). All terms of the equations must be explained and reported in the text.
- “6. Research Results”: only “Results”
- I suggest you split the methods better from the results. Create a section for the test case, one for methods, one for results, discussions and conclusions. Part of the discussions are done in results, I suggest you make a separate section of discussions. I suggest you use sub-sections where you need them.
- Avoid writing conclusions by points
Author Response
Response to Reviewer 1 Comments
Point 1: Abstract: I found the abstract little informative about the main results and findings of this work. In my opinion, it must be improved in order to show clearly these aspects.
Response 1: Abstract have been corrected:
“The article presents the possibility of helping the navigator in directing the movement of an object, while safely passing through other objects using an artificial neural network and optimization methods. It has been shown that the best trajectory of the object in terms of optimality and security, among many possible, can be determined by the method of dynamic programming, with the simultaneous use of an artificial neural network, depicting the encountered objects as moving forbidden domains. Analytical considerations are illustrated by examples of simulation studies of the developed calculation program on examples of real navigational situations at sea. The research took into account both the number of objects encountered and the different shape of domains assigned to the objects encountered. Finally, the optimal value of the safe object trajectory time was compared depending on the setpoint value of the safe passing distance of objects in given visibility conditions at sea and the degree of discretization of calculations determined by the density of the location of nodes along the route of objects.”
Point 2: Keywords: they must be keywords, so short.
Response 2: Keywords have been corrected:
“radar; safe control; optimization; neural network; computer simulation.”
Point 3: The Introduction: section needs to be rewritten in order to show clearly: i) a critical overview of state of the art (many more references need to be inserted), ii) the author’s contribution to this sate of art.
Response 3: The Introduction made a critical assessment of the state of the art referring to a larger number of 20 works and the author's contribution in this subject was shown.
Point 4: The figures must be on one page only, for example figure 4 and figure 6 are wrong. I suggest you to insert (a),(b)… for each panel so that you can indicate them in the caption. You need to better align the figures.
Response 4: The presentation of Figures 4, 5 and 6 has been improved and the marking: a, b, c, d for each of their parts has been introduced.
Point 5: Replace figure 2, must be a block diagram.
Response 5: Figure 2 has been replaced with the corresponding block diagram.
Point 6: Equations needs to be revised for mathematical consistency and notation (for example number (16)). All terms of the equations must be explained and reported in the text.
Response 6: Formula 16 has been corrected and the description of the quantities occurring in them supplemented.
Point 7: “6. Research Results”: only “Results”.
Response 7: Fixed to "Results".
Point 8: I suggest you split the methods better from the results. Create a section for the test case, one for methods, one for results, discussions and conclusions. Part of the discussions are done in results, I suggest you make a separate section of discussions. I suggest you use sub-sections where you need them.
Response 8: The whole text of the article was divided into new sections and sub-sections, and Section 5 "Discussions" was created.
Point 9: Avoid writing conclusions by points.
Response 9: The Conclusions were removed for points.
Author Response File: 
Author Response.docx
Reviewer 2 Report
- The logic of the flow chart is not clear.
- The logical relationship between formula (1) and formula (2) is not clearly described.
- There have been relatively similar studies on optimization methods. The novelty of this manuscript is a big concern.
- The data of the training set of neural network is just from about 300 experienced navigators during ARPA training courses. How does the author ensure the accuracy of the data？
Author Response
Response to Reviewer 2 Comments
Point 1: The logic of the flow chart is not clear.
Response 1: Figure 2 has been replaced with the corresponding block diagram.
Point 2: The logical relationship between formula (1) and formula (2) is not clearly described.
Response 2: The following explanation has been introduced:
“where x, u-respectively are state and control of object, which is describe by state equation:”
Point 3: There have been relatively similar studies on optimization methods. The novelty of this manuscript is a big concern.
Response 3: The essence of the article is the use of an artificial neural network to map the navigator's subjectivity in the assessment of a collision situation, and the optimization method in the form of dynamic programming is only a tool in the synthesis of the entire calculation algorithm.
Point 4: The data of the training set of neural network is just from about 300 experienced navigators during ARPA training courses. How does the author ensure the accuracy of the data?
Response 4: The article on page 7 has been supplemented with the following sentence:
“To ensure data accuracy, the network learning process was based on several standard scenarios for navigational situations at sea. For each situation, each navigator chose the best according to himself, i.e. subjectively in accordance with good maritime practice, an anticollision maneuver to change course and/or speed of the ship. In this way, the learned network represents the average experience of a larger population of navigators.”
Author Response File: Author Response.docx
Reviewer 3 Report
General Comments:
- As shown in title, and the conclusion “Presentation of encountered-objects movement in the form of moving neural domains of variable size depending on the distance and time of approaching objects reflects the navigator’s subjectivity in the assessment of collision risk”. What’s the function of radar remote sensing? What does the trajectory of met objects determine?
- The methodology of ANN related to the conclusion “The use of navigator officers in the learning process of the artificial neural network represents the valuable averaged practical knowledge of an experienced navigator” is not well introduced.
- Figure 2 is not clear and needs redraw to make it visible.
- Symbols in equations (1), (2) and (5) need explanation.
- Some mistakes occurs in the description after equation (8), where t is time, but t is missing. And where is the Inequality (2). Besides, some terminologies such as “Dependence (12) and priority Law (8) are not well defined previously.
- After equation (17), is an average square error of neural-network learning, here average should be sum?
- In figure 4, I’m confused on the trajectory of three objects, especially when two objects overlapped. Moreover, some explanations are needed for section 6.1.
- “However, the more encountered objects there are, the more nodes in the dynamic programming grid are rejected by the ANN procedure, and the shorter the calculation time is. This is one of the advantages of using Bellman’s practical optimality principle—the more restrictions, the faster a successful result is achieved.” It is against the common sense.
Author Response
Response to Reviewer 3 Comments
Point 1: As shown in title, and the conclusion “Presentation of encountered-objects movement in the form of moving neural domains of variable size depending on the distance and time of approaching objects reflects the navigator’s subjectivity in the assessment of collision risk”. What’s the function of radar remote sensing? What does the trajectory of met objects determine?
Response 1: The content of section 6 Conclusions has been supplemented with the following sentence:
“The use of radar remote sensing to identify objects movement parameters allows the synthesis of an appropriate algorithm to support the navigator in determining the safe trajectory of the object as a sequence of subsequent changes in its course and speed.”
Point 2: The methodology of ANN related to the conclusion “The use of navigator officers in the learning process of the artificial neural network represents the valuable averaged practical knowledge of an experienced navigator” is not well introduced.
Response 2: The article in sub-section 3.2 on page 7 has been supplemented with the following sentence:
„To ensure data accuracy, the network learning process was based on several standard scenarios for navigational situations at sea. For each situation, each navigator chose the best according to himself, i.e. subjectively in accordance with good maritime practice, an anticollision maneuver to change course and/or speed of the ship. In this way, the learned network represents the average experience of a larger population of navigators.”
Point 3: Figure 2 is not clear and needs redraw to make it visible.
Response 3: Figure 2 has been replaced with the corresponding block diagram.
Point 4: Symbols in equations (1), (2) and (5) need explanation.
Response 4: The following explanations have been introduced:
“where x, u-respectively are state and control of object, which is describe by state equation:”
“where L represents the minimum value of the control quality index I, i.e. in the safe ship control considered in the article, the minimum time needed to safely reach the nearest turning point on a given cruise trajectory.”
Point 5: Some mistakes occurs in the description after equation (8), where t is time, but t is missing. And where is the Inequality (2). Besides, some terminologies such as “Dependence (12) and priority Law (8) are not well defined previously.
Response 5: The article on page 4 has been corrected and supplemented with the following sentence:
“where gj is the function describing the shape of the domain of the encountered object (for example, circles, hexagons, parabolas, ellipses), (Xj,Yj) is the position of the encountered object, and k is period of time discretization, ensuring compliance with the following condition of safe object traffic control:
Dj min > Ds (9)
where Ds is a safe distance of approaching objects.”
Point 6: After equation (17), is an average square error of neural-network learning, here average should be sum?
Response 6: Fixed to "sum".
Point 7: In figure 4, I’m confused on the trajectory of three objects, especially when two objects overlapped. Moreover, some explanations are needed for section 6.1.
Response 7: The explanation that the object domains do not overlap is included in the revised sentence at the beginning of sub-section 4.1:
“Figure 4 shows safe and optimal object trajectories, especially the movement of all domains in subsequent moments of time is presented to show the change in their size depending on the degree of danger of approaching objects, just like the navigator estimates in practice.”
Point 8: “However, the more encountered objects there are, the more nodes in the dynamic programming grid are rejected by the ANN procedure, and the shorter the calculation time is. This is one of the advantages of using Bellman’s practical optimality principle—the more restrictions, the faster a successful result is achieved.” It is against the common sense.
Response 8: This sentence has been corrected to be understood more comprehensively:
„However, the more objects encountered, the more unacceptable nodes in the dynamic programming grid are rejected by the ANN procedure and the shorter the calculation time. This is one of the advantages of using Bellman's practical principle of optimality - the more limitations, the more the search area of acceptable solutions decreases and the faster the optimal solution to the problem is found.”
Author Response File: 
Author Response.pdf
Reviewer 4 Report
This paper presents a methodology for optimizing the safe management procedure of own-object movement passing among many other objects at sea. The subject of the paper is very interesting and applicable in the real world. However, it seems introduction section was poorly organized and written and needs more attention.
In addition, figures need to be more clearly adjusted specially Figure 2 which is not possible to easily be followed.
Author Response
Response to Reviewer 4 Comments
Point 1: English language and style are fine/minor spell check required.
Response 1: I sent Manuscript to check the correctness of the English language to MDPI.
Point 2: Introduction section was poorly organized and written and needs more attention.
Response 2: Introduction has been completely rewritten and corrected:
The International Maritime Organization (IMO) has introduced the requirement for objects to have installed an automatic radar plotting device called Automatic Radar Plotting Aid (ARPA). The system automatically initiates and continues the tracking of detected echoes, generating alarms in hazardous situations, which were presented by Graziano et al. and Huang et al. in [1-2]. The ARPA computer calculates the distance and time for critical approach of objects, and then compares the obtained values with their recommended values in the current situation at sea. If the calculated values exceed the set limits, the dangerous target alarm is activated. It is also possible to simulate a trial maneuver of the object, but only for one target, namely, the most dangerous encountered object. This process helps the navigator evaluate the effects of the planned anticollision maneuver at an accelerated timescale. Therefore, as stated by Bist in [3], the ARPA is a source of support for the work of the navigator, increasing navigational safety. The main task of ARPA to be prepared maneuvering decisions made by the navigator, especially in situations of concentrated object traffic in restricted waters was presented in [4-6].
In practice, there are many possible safe collision-avoidance maneuvers; ideally, the navigator chooses the optimal maneuver, which is the one that, in addition to minimizing the collision risk, results in deviation from the original path. The possibilities of supplementing the system with appropriate methods to support maneuvering decisions taken in an uncertain navigational situation that occurs in a short time in relation to the greater number of objects encountered are described in [7-12]. Reducing the uncertainty of assessing the real navigational situation of the object by using an artificial neural network is shown in [13-15]. Lenart in [16] proposed the parameter "time for a safe distance" after detecting dangerous objects as a potentially important parameter, accompanied by the display of possible evasive maneuvers. Borkowski in [17] presented acceptable solutions for altering course range in compliance with the International Regulations for Preventing Collisions at Sea (COLREGs Rules). Liu et al. and Lisowski [18-19] proposed a different approach to preventing object collisions at sea, through use of game-theory methods to determine the safe trajectory of ships by considering elements of an indefinite nature in real navigational situations at sea.
In previous works, the subjectivity of the navigator making a maneuvering decision in a collision situation was omitted in the calculation algorithms, this subjectivity is in 88% the cause of accidents at sea. That is why the purpose of this article is the synthesis of an algorithm for safe and optimal object control in situations where several objects pass by using the artificial neural network generating subjective domains of passing objects and the Bellman dynamic-programming method.
Point 3: In addition, figures need to be more clearly adjusted specially Figure 2 which is not possible to easily be followed.
Response 3: I replaced Figure 2 with a new, more transparent and understandable one.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Accept
Author Response
I sent Manuscript to check the correctness of the English language to MDPI.
Author Response File: Author Response.docx
Reviewer 3 Report
All my comments previous have been well replied and corrections have been made. It can be accepted as this version.
Author Response
I sent Manuscript to check the correctness of the English language to MDPI.
Author Response File: Author Response.docx
