Intelligent Data Fusion and Multi-Agent Coordination for Target Allocation
Round 1
Reviewer 1 Report
Some suggestions:
1.In abstract:
The keywords are not matching with the content:
It should be: multi-sensor data, military surveillance robots, fusion rules, identification of targets.
The original keywords in the paper:
Target Allocation: Not exist
Coordinated Decision-Making: Not exist
Modeling and Simulation for Distributed Multiple Robots:
It is too long and did not exist.
2,English: Please do not use first person, use third person (the paper).
For example in in conclusion: We proposed a set of fusion operators to combine multi-sensor data ……
(a)The paper proposed…
(b) A combine with B….
And so on in the other place of the paper.
3.The author uses mean, product and DS theory three operation as the basic of mathematics in the paper., and use experiments to prove the final results, it is a good idea.
I think the paper is focused on software simulation. Please point out the kinds of program, and to emphasize is it self-developed or not, if it is self-developed, please plot the flowchart to prove it is the creative place by author.
4.Limitation and future work did not appear in Conclusion. Please recheck
5.REF[16] is published at 1990, it is too far away, or it is the Bible in the field. Please recheck.
Author Response
1. In abstract:
- The keywords are not matching with the content:
- It should be: multi-sensor data, military surveillance robots, fusion rules, identification of targets.
- The original keywords in the paper:
- Target Allocation: Not exist
- Coordinated Decision-Making: Not exist
- Modeling and Simulation for Distributed Multiple Robots: It is too long and did not exist.
Answer) From line 15 to line 16, in keywords, we have rewritten the keywords. However, this paper addresses not only the fusion process but also the coordinated decision-making to monitor targets, which is the mapping from multiple robots to multiple targets. Further, we have designed and developed the intelligent military simulator as a test bed. So, we would like to include target allocation, coordinated decision-making, and modeling and simulation as keywords.
2. English: Please do not use first person, use third person (the paper).
- For example in conclusion: We proposed a set of fusion operators to combine multi-sensor data ……
(a) The paper proposed…
(b) A combine with B….
And so on in the other place of the paper.
Answer) As you recommended, we have revised our paper as possible as we can. In particular, we have rechecked every sentence starting with ‘we.’
3. The author uses mean, product and DS theory three operation as the basic of mathematics in the paper, and use experiments to prove the final results, it is a good idea.
I think the paper is focused on software simulation. Please point out the kinds of program, and to emphasize is it self-developed or not, if it is self-developed, please plot the flowchart to prove it is the creative place by author.
Answer) We have designed and developed all of the programs for ourselves. Of course, it is self-developed.
From line 203 to line 265, in section 4, we have explained the unit simulator in Figure 1 and the integrated simulator in Figure 3, respectively. In particular, in section 4.2, we have explained the event flow of intelligent simulator for data fusion and target allocation, as depicted in Figure 2.
4. Limitation and future work did not appear in Conclusion. Please recheck
Answer) From line 378 to line 388, in the third paragraph of conclusion, We have added the limitation and future work in details.
5. REF[16] is published at 1990, it is too far away, or it is the Bible in the field. Please recheck.
Answer) We would like to refer to the original paper that Dempster and Shafer proposed their framework. As you mentioned, it is the Bible in this field.
Reviewer 2 Report
The paper focuses on identifying the specific target, then follow its movement. The paper also proposes three operators which was used to combine information and a decision-theoretic framework. The simulation experiments illustrate the efficiency of the proposed method. However, there are some points the author needs to consider.
Comments:
- Check the grammatical and description. Such as:
The reference number is disorder.
Line 71: citeDSfusion…
Line 366: …while they are contnuously moving.
- Why the agent coordination is dynamic coordination. How the dynamic is reflected.
- The real world application of the proposed method should be explained.
- The authors can compare their method to the other fusion method used in distributive agents.
- In section 5.2., the author mentioned that when the probability of a target passed a certain threshold value, the robots will change from search mode into tracking mode, what is the concrete value of threshold and the reason for set it.
- Why the LWIR sensors give better fusion result than MWIR sensors.
- After reading the whole article, it is not clear why the author proposed three operators. What are the advantages of proposed fusion method compared with the classic Dempster’s combination rule?
In conclusion, my opinion is to reconsider this work after the above issues are well addressed.
Author Response
The paper focuses on identifying the specific target, then follow its movement. The paper also proposes three operators which was used to combine information and a decision-theoretic framework. The simulation experiments illustrate the efficiency of the proposed method. However, there are some points the author needs to consider.
Comments:
Check the grammatical and description. Such as:
- The reference number is disorder.
Answer) The reference has been rearranged in order.
- Line 71: citeDSfusion…
Answer) Done.
- Line 366: …while they are contnuously moving.
Answer) Done.
- Why the agent coordination is dynamic coordination. How the dynamic is reflected.
Answer) From line 162 to line 167, in the first paragraph of section 3.2, we have explained "dynamic" property of coordination.
- The real world application of the proposed method should be explained.
Answer) From line 378 to line 388, in the third paragraph of conclusion, we have added the possibility of "real world" application.
- The authors can compare their method to the other fusion method used in distributive agents.
Answer) From line 64 to line 82, in the first paragraph of related work, we have mentioned a variety of fusion method. And we compared our method to other fusion methods, i.e., probabilistic fusion, evidential belief reasoning, fuzzy reasoning, Bayesian method, and so on, in distributive environments.
- In section 5.2., the author mentioned that when the probability of a target passed a certain threshold value, the robots will change from search mode into tracking mode, what is the concrete value of threshold and the reason for set it.
Answer) From line 297 to line 299, 306 to 308, and 314 to 315, in sections 5.1 and 5.2, we have added the summary of experimental results, the reason for the threshold, and the value itself.
- Why the LWIR sensors give better fusion result than MWIR sensors.
Answer) From line 319 to line 322, in section 5.2, we have added the reason.
- After reading the whole article, it is not clear why the author proposed three operators. What are the advantages of proposed fusion method compared with the classic Dempster’s combination rule?
Answer) From line 131 to line 156, in section 3.1, we have explained why three operators are proposed and have added the additional reason. Actually, the fusion rule using DS theory in the paper is identical to the classic Dempster’s combination rule.
- In conclusion, my opinion is to reconsider this work after the above issues are well addressed.
Round 2
Reviewer 1 Report
After few days to read the paper, I find the paper had modified the contents according to th original review's opinions.
Only one suggestion, the paper still exist too many first-person style.
For example,
line 30: We propose a set..., please change into: The paper proposes...
line 160: we endow them with..., please change into: it endows...
and so on.
Please recheck
.
Reviewer 2 Report
The paper focuses on identifying the specific target, then follow its movement. The paper also proposes three operators which was used to combine information and a decision-theoretic framework. The authors also implemented a military simulator to compare the capabilities of them and have experimented the framework in distributed and uncertain battlefield environments. However, there are some points the author needs to consider as follows:
- In equation (4), does the function EU_k equal to EU in equation (3)? If it does, what does k mean? If it doesn’t, how are EU_k defined?
- In section 4, “Implementation of Intelligent Simulator for Data Fusion and Target Allocation”, the author only constructed the unit simulator for individual fusion process in detail. The description about target allocation is poor. What actions can the agent take? What is the set of world states? How are the utilities of each states defined?
- In section 4.2, the author said “After identifying the type of targets, a group of surveillance robots coordinate each other to monitor targets even if some of them are hidden by obstacles.”. Is the obstacle moving or fixed? If the obstacles are moving, how do robots distinguish obstacles from targets? If the obstacles are fixed, the application of military simulator is very limited.
- In lines 241-242, the author said “our robots are beginning to search possible targets within its range”. It means that the system doesn’t knows the number of targets before starting simulation. The possible targets searched by each robot may be repeated, so how to determine the number of targets?
Based on these comments, I suggest MAJOR REVISIONS before its acceptance.
Author Response
The paper focuses on identifying the specific target, then follow its movement. The paper also proposes three operators which was used to combine information and a decision-theoretic framework. The authors also implemented a military simulator to compare the capabilities of them and have experimented the framework in distributed and uncertain battlefield environments. However, there are some points the author needs to consider as follows:
1. In equation (4), does the function EU_k equal to EU in equation (3)? If it does, what does k mean? If it doesn’t, how are EU_k defined?
Answer) The EU_k is equal to EU in equation (3) but k indicates one of the EU’s in equation (4). From line 187 to line 189, we have added what k means. Since k is also used in equation (3), k is substituted with l in equation (4).
2. 2.1) In section 4, “Implementation of Intelligent Simulator for Data Fusion and Target Allocation”, the author only constructed the unit simulator for individual fusion process in detail. The description about target allocation is poor.
Answer) In section 4.2, the event flow of the integrated simulator and the simulator itself has been explained. From line 248 to line 254, we have added the description about target allocation.
2.2) What actions can the agent take? What is the set of world states? How are the utilities of each states defined?
Answer) From line 174 to line 192, we have explained the coordinated actions for our agents. In particular, the probabilities and the utilities in our target allocation model have been mentioned.
3. In section 4.2, the author said “After identifying the type of targets, a group of surveillance robots coordinate each other to monitor targets even if some of them are hidden by obstacles.”. Is the obstacle moving or fixed? If the obstacles are moving, how do robots distinguish obstacles from targets? If the obstacles are fixed, the application of military simulator is very limited.
Answer) In our simulator, the obstacles are fixed. Actually, in a real-world military application, this feature could be a limitation. From line 369 to line 379, in the third paragraph of conclusion, we have mentioned our future work to extend our simulator.
4. In lines 241-242, the author said “our robots are beginning to search possible targets within its range”. It means that the system doesn’t knows the number of targets before starting simulation. The possible targets searched by each robot may be repeated, so how to determine the number of targets?
Answer) From line 226 to line 234, we have explained the overall process of target identification and target allocation.
From line 301 to line 307, in the second paragraph of section 5.2, we have explained the process of determining the number of targets in each state. At first, the robots were in search mode, searching for targets while rotating 360 degrees. When the probability that a target was considered to be a real threat passed a threshold value, each robot switched into tracking mode. Then, the robots could recognize the targets as real, and determine the number of targets being monitored.
Based on these comments, I suggest MAJOR REVISIONS before its acceptance.
Round 3
Reviewer 2 Report
The manuscript was correspondingly revised according to the comments and required changes and it can be published.
