Navigating Non-Playable Characters Based on User Trajectories with Accumulation Map and Path Similarity

Round 1

Reviewer 1 Report

The paper proposes a novel method for navigating non-playable characters (NPCs) in video games. The method utilizes user input via an accumulation map, based on which viable paths for NPCs are generated.

The idea is interesting, while the algorithm does not seem computationally expensive, which differentiates it from most of the available solutions for path planning. Furthermore, the potential applications for the method are not only limited to video games. Nevertheless, in my opinion, the paper has to be improved before it warrants publication. In the current version, the method and its verification procedure are not fully explained. Also, it would be advisable to provide a comparison between the proposed approach and some selected common approaches. Please find specific comments below.

1. The introduction is acceptable and provides an overview of the approaches used for path planning in interactive media. Nevertheless, most of the cited works are from 2008 or older. Newer works should also be cited. Furthermore, the authors focused mostly on grid based approaches. In my opinion, to make it more complete, it would be advisable to mention methods, which can work in continuous domains – RRT (widely applied in robotics), optimization, potential fields.
I would also suggest to provide a clear description of a path planning problem – what is given, what has to be obtained, etc. This will make the paper more accessible to a wider audience.

2. Section #2 – Related Works, feels like a continuation of Introduction. I would suggest to merge the two, possibly with subsections – Introduction already provides some of related works. Also, section #3 should be renamed to method, while section #4 – Results and Discussion. This would make the paper easier to follow.

3. Section #3.1 – this part is problematic, its name is “Collecting User’s Path Data” and it begins with an equation, which looks like a path generator. Based on the title, I expected a description of test environments, the test users, etc. Did you collect actual user-data or did you artificially generate user-like paths for verification?

The subsection should be rewritten. Furthermore, if these are artificial paths, some of them should be compared to actual user-created ones, to make sure that the data generation procedure is viable. Lastly, this subsection should be later on the paper. At this point you should explain your main novelty – the path planning procedure – give basic definitions of accumulation maps, how to obtain an NPC path from it, etc. – in other words: extend the subsection #3.2 and move it so that it appears before #3.1.

4. [L195-L199] This part belongs in Results and Discussion. Also, please explain the low success rate in cond. #1 and #2 - how often does this appear in practical applications, what causes failure.

5. Section #3.2 – I understand that you employed Unity in your approach. Nevertheless, this part of the paper would benefit from a block diagram with an overview of the algorithm with more data on how to implement it. As mentioned in my comment #3 this section lacks some definitions. Furthermore, please explain the NPC path generation procedure based on an accumulation map. Is it equation (1)? How did you measure path-similarity? Provide an equation for it.

6. [Regarding the paths used to validate the approach] The collision spaces seen in the figures are relatively simple – most algorithms would do well in such conditions. Did you test the approach in more difficult environments? Provide more samples of your collision spaces.

7. Results – in my opinion, there should be a comparison to more common grid based approaches. The comparison should feature some metrics regarding: the time required to plan a path, success rate, etc. As of this version, some of the claims made in the results do not feel properly grounded.

8. This is more of a general comment, but why not just have NPCs traverse the actual paths collected from users (assuming your procedure requires a number of paths anyways)? In other words, what is the main benefit of your approach in this regard?

Author Response

1. The introduction is acceptable and provides an overview of the approaches used for path planning in interactive media. Nevertheless, most of the cited works are from 2008 or older. Newer works should also be cited. Furthermore, the authors focused mostly on grid based approaches. In my opinion, to make it more complete, it would be advisable to mention methods, which can work in continuous domains – RRT (widely applied in robotics), optimization, potential fields.
   I would also suggest to provide a clear description of a path planning problem – what is given, what has to be obtained, etc. This will make the paper more accessible to a wider audience.

2. Section #2 – Related Works, feels like a continuation of Introduction. I would suggest to merge the two, possibly with subsections – Introduction already provides some of related works. Also, section #3 should be renamed to method, while section #4 – Results and Discussion. This would make the paper easier to follow.

3. Section #3.1 – this part is problematic, its name is “Collecting User’s Path Data” and it begins with an equation, which looks like a path generator. Based on the title, I expected a description of test environments, the test users, etc. Did you collect actual user-data or did you artificially generate user-like paths for verification?

The subsection should be rewritten. Furthermore, if these are artificial paths, some of them should be compared to actual user-created ones, to make sure that the data generation procedure is viable. Lastly, this subsection should be later on the paper. At this point you should explain your main novelty – the path planning procedure – give basic definitions of accumulation maps, how to obtain an NPC path from it, etc. – in other words: extend the subsection #3.2 and move it so that it appears before #3.1.

4. [L195-L199] This part belongs in Results and Discussion. Also, please explain the low success rate in cond. #1 and #2 - how often does this appear in practical applications, what causes failure.

5. Section #3.2 – I understand that you employed Unity in your approach. Nevertheless, this part of the paper would benefit from a block diagram with an overview of the algorithm with more data on how to implement it. As mentioned in my comment #3 this section lacks some definitions. Furthermore, please explain the NPC path generation procedure based on an accumulation map. Is it equation (1)? How did you measure path-similarity? Provide an equation for it.

6. [Regarding the paths used to validate the approach] The collision spaces seen in the figures are relatively simple – most algorithms would do well in such conditions. Did you test the approach in more difficult environments? Provide more samples of your collision spaces.

7. Results – in my opinion, there should be a comparison to more common grid based approaches. The comparison should feature some metrics regarding: the time required to plan a path, success rate, etc. As of this version, some of the claims made in the results do not feel properly grounded.

8. 8. This is more of a general comment, but why not just have NPCs traverse the actual paths collected from users (assuming your procedure requires a number of paths anyways)? In other words, what is the main benefit of your approach in this regard?

Author Response File: Author Response.pdf

Reviewer 2 Report

First of all, I would like to congratulate the authors for the clarity and quality of the work presented.

The title of the article suggests the topic.

The abstract reflects important aspects of the content of the article, highlighting a method to effectively control the path of non-playable characters (NPCs) in an interactive virtual environment. I recommend the authors to reorganize the writing of the abstract highlighting: 1) the purpose 2) methodology 3) the findings 4) the implication of the research 5) the originality of the paper.

The introduction provides a clear picture of the need to develop and expand existing static navigation networks and improve the NPC route (inefficient in the virtual environment). In the virtual world, users can control and interact with certain NPCs at a certain level. I recommend expanding and improving the section by adding more contributions.

The purpose of the paper is obviously - to propose a new method of adaptive navigation based on user interaction data.

The organization of the article is a classic one, scientifically correct. However, I recommend the authors to introduce a paragraph on how to structure the paper as follows: section 2 - RelatedWorks, section 3 - Our Framework section 4 -....

The RelatedWorks section is well done.

In the Our Framework Section are presented in detail the modeling procedures followed.

The results were clearly presented in a logical sequence, the authors highlighting the advantages obtained by the proposed method.

The Conclusions section is too short. I recommend its extension by highlighting the advantages found in relation to other research, the limitations identified in the modeling process, the implications for different users and researchers.

I recommend the authors to complete and include after conclusions the following aspects: Author Contributions, Funding and Conflicts of Interest.

The article is interesting and after the changes, I recommend it for publication.

Author Response

The abstract reflects important aspects of the content of the article, highlighting a method to effectively control the path of non-playable characters (NPCs) in an interactive virtual environment. I recommend the authors to reorganize the writing of the abstract highlighting: 1) the purpose 2) methodology 3) the findings 4) the implication of the research 5) the originality of the paper.

The introduction provides a clear picture of the need to develop and expand existing static navigation networks and improve the NPC route (inefficient in the virtual environment). In the virtual world, users can control and interact with certain NPCs at a certain level. I recommend expanding and improving the section by adding more contributions.

The purpose of the paper is obviously - to propose a new method of adaptive navigation based on user interaction data.

The organization of the article is a classic one, scientifically correct. However, I recommend the authors to introduce a paragraph on how to structure the paper as follows: section 2 - RelatedWorks, section 3 - Our Framework section 4 -....

The RlatedWorks section is well done.

In the Our Framework Section are presented in detail the modeling procedures followed.

The results were clearly presented in a logical sequence, the authors highlighting the advantages obtained by the proposed method.

The Conclusions section is too short. I recommend its extension by highlighting the advantages found in relation to other research, the limitations identified in the modeling process, the implications for different users and researchers.

I recommend the authors to complete and include after conclusions the following aspects: Author Contributions, Funding and Conflicts of Interest.

The article is interesting and after the changes, I recommend it for publication.

Author Response File: Author Response.pdf

Reviewer 3 Report

This is a very interesting research work! expanding the research to study scenarios, where dynamics of the agents are taken into consideration would be an interesting addition in my view. Also, incorporating some AI/ML techniques, to learn from experience might add to the credibility of the proposed method.

The paper is well written. It was a joy to read it. Also, figures and equations are well explained and self explanatory. 

I would recommend this for publication for no reservation. 

There are, however, a few typos that need to be fixed (not many). 

Author Response

This is a very interesting research work! expanding the research to study scenarios, where dynamics of the agents are taken into consideration would be an interesting addition in my view. Also, incorporating some AI/ML techniques, to learn from experience might add to the credibility of the proposed method.

The paper is well written. It was a joy to read it. Also, figures and equations are well explained and self explanatory. 

I would recommend this for publication for no reservation. 

There are, however, a few typos that need to be fixed (not many). 

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have significantly improved the paper and provided clear responses to comments made in the initial review. I have no further remarks.

Reviewer 2 Report

No additional comments