Swarm Exploration and Communications: A First Step towards Mutually-Aware Integration by Probabilistic Learning

Round 1

Reviewer 1 Report

The paper is well organized and well explained. Presents a very interesting topic today, including semantic communications and its relationship to space issues, which brings novelty and greater innovation.

Author Response

Dear Reviewer 1, we would like to thank you for the really positive feedback on our manuscript. We will improve the introduction as desired.

Attached, you will find a reviewer full response and an article version with highlighted changes.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript considers swarm robot planning and multi-agent optimisation, where the main aim is to facilitate exploration over a given area. The manuscript considers two main problems; i)integration of exploration and communication into a single framework, ii) ensure mutual awareness among the agents. Although the second aim is of interest and the formulation considered is interesting, as it considers the problem from and information theoretic perspective, the first aims unclear. The main reason that communication is not considered is that communication protocol are very strong and their effect on learning is minuscule. The other issue of with the simulations section, specifically results in Fig. *. The main application is implied as mars exploration, whereas the range of SNR is not realistic for this application. Moreover, for the ranges of SNR less than 2, the error rate seems excessive. Finally, the main theme of manuscript is in line with distributed decision making in "Distributed Kalman Filtering and Control Through Embedded Average Consensus Information Fusion," IEEE Transactions on Automatic Control, vol. 64, no. 10, pp. 4396-4403, Oct. 2019, which I recommend as a reference. 

 

Overall, the manuscript is well written and has valid contribution. I have made some recommendations for the authors and look forward to seeming the revised manuscript.  

Author Response

Dear Reviewer 2, thank you very much for your positive and constructive feedback. We will now elaborate on your three points of criticism.

Attached, you will find a reviewer full response and an article version with highlighted changes.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors propose a new approach to couple exploration-aware communication and communication-aware exploration using factor graphs. A particular focus is the use of semantic communication to achieve performance gains, where the authors introduce the application of this methodology for exploration using a swarm of agents. 

 

General comments:

- Figures should support the understanding of the text, but be comprehensible in isolation. In general, the figures were difficult to comprehend without searching through the text and referencing back.

- There appears to be a disconnect between the discussion of semantic communication and the proposed application of swarm exploration on Mars. References to semantic communication discuss how "neat" and "fine" are close in semantic space, but it is not clear how this will map to the exploration scenario. The paper appears to present a generic communication framework rather than one that is clearly explained to be well-suited for the exploration scenario.

- Section 5.3 could be restructured to more clearly introduce the simulation and why this single simulation provides sufficient evidence of the value of your approach. The mixing of methodology, results, and discussion in this section making it difficult for the reader.

 

Specific comments:

- Figure 2, it is not clear why communications is blue in the probabilistic triangle and whether this is related to integrated design. In general, figures benefit from being understandable in isolation. It is also not clear until you read the text that the graph is a hypothesized improvement to exploration mean square error. 

- Line 60, the first reference in the text to Figure 2 does not provide enough information or context to understand the figure. Why do you not have ideal links? What is integrated design? 

- Line 79, It would be helpful to the reader to provide some context for why the approach of Xie et al. [10] is thought to be promising.

- Line 201, not familiar with "s." may be typo.

- Line 107, new paragraph should standalone and not use "In particular" to refer back to the previous paragraph.

- Line 121, not clear why "guide" is in quotes.

- Line 123, introduces exploitation a definition included in Figure 2 without context.

- Line 130, stats that RL movement strategies often fail or perform unreliably. Can you provide a citation? Or reword to say "such approaches are likely to fail ..."

- Line 142, erroneous links now introduced, again essential to understand Figure 2. Perhaps Figure 2 can come after these key concepts are introduced.

- Line 219, should state "In summary" not "In conclusion"

- Figure 3, not clear what the layer of nodes represent nor what value the figure is adding to the text without first finding the reference to the figure in the text. Please mention that it represents gas concentration in the figure caption. The figure appears before being referenced in the text which adds to the confusion. The arrows are only explained in the text and the dots above the gas concentrations are never explained.

- Line 258, inconsistent reference to factors graphs as FG or factor graphs and then later FGs

- Line 261, "a possible new design approach" why is it only a possible approach?

- Figure 4 includes a legend but key information such as the node s can only be found from searching the main text. Legend is incomplete with no reference to edges between variable and factor nodes. Not clear what is meant by "cutting edge to insert communication in FG" until you get to Figure 5.

- Figure 6, again it would be useful to the reader to define more that just c and Omega without having to search through the text.

- Line 390, it would be helpful to introduce variational message passing.

- Figure 7 appearing before reference in the text.

- Line 425, an alternative integration strategy is proposed based on a reference. Please provide the reader with some insight into this strategy beyond pointing them to the reference.

- Line 427, introduce RV as an acronym

- Line 459, the Mars scenario has only been mentioned once in the main text and 12 pages prior to this line it may be worth repeating your interest in the Mars/Moon exploration scenario or making it a stronger theme throughout the paper.

- Line 469 reword "allows to model"

- Line 490, would be worth mentioning what the similar problems are that affect MILBO and ELBO

- Figure 8, appears before text reference. I would still suggest introducing SNR acronyms within the paper and preferably in the figure caption.

- Line 565, this would be better as part of the conclusion. As mentioned in general comments, the mixing of discussion, results, methodology, and now conclusion in this single section makes it difficult for the reader.

- Line 565, what is the definition of "manageable effort" in this context?

Author Response

Dear Reviewer 3, thank you very much for your very detailed and constructive feedback.

Attached, you will find a reviewer full response and an article version with highlighted changes.

Author Response File: Author Response.pdf

Reviewer 4 Report

In this paper,  the first steps towards a framework that unifies both of these realms by a "tight" integration are presented. The comments are as follows.

1. The authors are suggested to make some improvements on the presentation of figures, such as the font size and proportion in the figure.

2. There are some grammar errors that the authors should check carefully and make the paper in better quality.

3.  There are a lot of theoretical derivations in this paper, but the authors do not clearly explain what are the challenges encountered in this paper and what methods are used to solve them?

4. There are many results on multi-agent systems control, such as Mean square leader-following consensus of second-order nonlinear multi-agent systems with noises and unmodeled dynamics. Please compare the above results with the method of this paper.

 

 

Author Response

Dear Reviewer 4, thank you very much for your valuable comments.

Attached, you will find a reviewer full response and an article version with highlighted changes.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

All my concerns have been addressed satisfactorily. Thus, this paper can be acceptable.