From reviewer B: The article is well written and provides a good overview of the field. However, it can be improved and some paragraph may be better organized. On the other hand, some details about the SI applications could be removed. See below for details.
Also, I think that a brief but detailed description of a swarm intelligence study would be useful for the interested reader, in order to give a concrete example. One possibility is creating a brand new subsection in which a swarm intelligence system is briefly introduced (e.g., the foraging behaviour of ants). Another possibility could be giving more details to the natural examples of the paragraph "Studies and Applications of Swarm Intelligence".
From reviewer A: Overall, I think this is a very nicely written article. It is clear and authoritative, as I would expect from a pioneering researcher in the field. I agree with the reviewer above that an example would be useful - especially if the example had been well studied in both natural and artificial domains: either foraging or cooperative transport would be good candidates.
I do have a number of comments and suggestions for improvements.
General comments: Authors answer
We have restructured the article so that we now have a section dedicated to examples of the taxonomy. The Section "Studies and applications of SI" has been improved.
From reviewer B: Currently, the definition of swarm intelligence is somewhat vague. I would better specify what swarm intelligence is. For instance, a possible characterisation of swarm intelligence is "the science that studies natural and artificial swarm-like systems, composed of...".
From reviewer A: 1. Since this is an encyclopaedia then I think a stronger opening statement of what swarm intelligence is would be better. I.e. instead of "SI deals with systems..", "SI is the study of natural and artificial systems..." 2. Also in the initial definition I think the final sentence reads rather awkwardly with too many uses of the word 'some'. How about "... and artificial systems including swarm robotics and some computer programs..."
Initial definition: Authors answer
We do not like neither "the science that studies natural and artificial swarm-like systems, composed of...", nor "SI is the study of natural and artificial systems...". In fact, both definitions highlight the "study" aspect which is more part of the scientific than of the engineering stream of swarm intelligence. We have changed the opening statement trying to make it stronger, but at the same time maintaining "deals with".
Taxonomy of Swarm Intelligence
From reviewer B: I would better structure the paragraph by highlighting the two dichotomies in the text:
Natural vs. Artificial Swarm intelligence has a marked multidisciplinary character .... Scientific vs. Engineering An alternative and somehow more informative classification....
This will help the reader in identifying the taxonomy and will stimulate reading.
Finally, the orthogonality of the two proposed dichotomies looks a bit forced for the "natural engineering" studies. The authors are invited to provide some reference or link to such studies (maybe some of the results of the LEURRE project?). Moreover, a reference to a study belonging to the "artificial scientific" class would be informative for the reader (for instance, Kube and Bonabeau. Cooperative transport by ants and robots, Robotics and Autonomous Systems, volume 30, number 1--2, pages 85--101, 2000).
From reviewer A: I think this section on taxonomy is extremely interesting, and I strongly agree that the artificial vs. natural and scientific vs. engineering are two particularly interesting axes of the field. A slight concern is that this section may give readers the impression that the scientific and engineering streams are equally advanced. In my view the science of SI is far ahead of the engineering - especially in embodied (robotic) systems. I agree with the reviewer above that LEURRE is perhaps the best example of a natural + engineering project to date.
Taxonomy of Swarm Intelligence: Authors answer
We now highlight the dichotomies and we provide an explicit example of the taxonomy. We also added a reference to LEURRE. However, please note that LEURRE is only a first step towards a natural/engineering study. In fact, while in LEURRE work on cockroaches there is the idea of controlling a natural swarm, what is still missing is the engineering goal.
We do not agree with reviewer A on the fact that the science of SI is far ahead of the engineering: first, we do not think it is really possible to make such a comparison; second, the very fact that some SI algorithms (e.g., ACO) are already used in real-world applications shows that engineering of SI systems is well advanced, at least is some areas.
Properties of a Swarm intelligence System
From reviewer B: I would invert the bullet list with the description. I think it's better to give first an overview and then specify with some text which are the relevant aspects.
From reviewer A: I suggest it would be good to link the 3rd and 4th bullets by amending the 4th bullet to say "the overall behaviour of the system emerges from the interactions of individuals with each other and with their environment, that is the group behaviour self-organises." In fact, I'm not sure I agree with the last part of the 4th bullet "..does not depend on orders external to the system". I understand what the authors mean - that there is no conventional direct external command or control - but it is nevertheless possible to influence the way the swarm self-organises and such influence would be essential in engineering applications of SI.
Properties of a Swarm intelligence System: Authors answers
Studies and Applications of Swarm Intelligence
From reviewer B: I would reduce the details about the applications (in particular to ACO and PSO), given that these applications of SI will have their own page in scholarpedia soon.
From reviewer A: I agree that the studies and applications seems somewhat unbalanced. In particular I would like to see a slightly longer section on swarm robotics (although I note this will also have an article soon). It might, I suggest, be appropriate to state that one of the principal - and as yet unsolved - problems of swarm robotics is how to determine which individual behaviours will give rise to the desired overall group behaviours, and that this problem is one of the factors limiting the real-world engineering applications of swarm intelligence.
Studies and Applications of Swarm Intelligence: Authors answer
We have slightly reduced ACO and PSO and slightly increased the length of the swarm robotics section. However, we do not think this is the right place for the discussion suggested by reviewer A. We will use the suggestion within the Swarm Robotics entry of Scholarpedia that one of the authors is currently writing.
From reviewer B:
- change the sentence "In this section we give a few examples of scientific and engineering swarm intelligence studies" in order to avoid using "we".
- reduce the references only to the essential ones: for example, (Bonabeau et al., 1999) could be sufficient for all the natural examples.
- better align the pictures with the corresponding paragraphs
This is an excellent article that serves as a good tutorial introduction.
I have made some minor edits to the text for English and Grammar.
The article would be enhanced by references to ACO for continuous and mixed variable problems, and AntNET for Ad hoc problems (I agree with the previous reviewer on most points but not the one about reducing references - I think references are really helpful in this type of article)
As Europeans I would hope that you would use the British spelling (behaviour, organisation etc)!
From Reviewer A: 1. Note the incorrect English in 'properties of swarm intelligence' para 1. "interaction of spatially neighbor" should I think be "interaction of spatially neighboring" 2. In the External Links perhaps consider mentioning the 2004 and 2006 SAB Swarm Robotics workshops.
Minor comments: Authors answer
We have made all the minor corrections suggested except for:
- The comments of the referees about references were contradictory. We have decided for a slight increase in the number of references. The criterion chosen is to cite only seminal works or reference books/special issues.
- We have decided to stay with the American spelling. One important reason to do so is that the name of some of the best known algorithms in SI uses the American spelling (e.g., Ant Colony Optimization).
- We have added a reference to the the swarm robotics web pages where the workshops proceedings are referenced.