Based on my knowledge and the projects I've worked on, Leadership formation among large number of agents is one problem. Sensor Network Leadership (SNL) protocols are among those that have been studied [1]. Reaction-diffusion techniques [2] (proposed by Alan Turing) can also be used to form patterns in multiple agents, that could potentially be used to study group formations in those agents.

With respect to using mathematical group theory [3] (especially symmetry groups) for pattern/group identification in agents, symmetry analysis on the various features shared among agents (locations, tasks, perceptions etc), can be used . However, I've only worked on using group theory for improving cognitive capabilities of only a single agent; how this transfers to multiple agents is still an area of exploration.

[1]: http://www.cs.utah.edu/~tch/publications/pub205.pdf

[2]: http://www.cs.utah.edu/~tch/publications/pub271.pdf 

[3]:http://www.cs.utah.edu/~tch/publications/pub274.pdf

It is not directly related, but in the control community there are projects like dymasos that deal with this issue in the context of interacting subsystems:

http://www.dymasos.eu

Thank you for these two helpful answers. One way of restating my original question is that I am looking for follow up work to the MIT "Nerd Herd" studies on communities of robots carried out by Maja Mataric. I suspect that there is ongoing relevant work involving military drones but I am on the wrong side of the wire to know about that...

Hi Dear,

what is your mean of no central control?!  I think that first you should define the control system level of inter-acting and inter-relating and inter-connection in hierarchy formation for . so we will discuss about details!

Best

Let me introduce my question differently. Suppose we forcibly collected 100 humans (men women children) selected entirely at random from around the world and deposited them on a previously uninhabited island (with good basic living resources but no communication with the rest of the world) in some remore part of the ocean. Suppose that we then leave the 100 humans there and come back a few years later to see in what state they are. What kind of society would you expect to have emerged? (Compare William Golding's famous novel "Lord of the Flies").

Now pose the analogous question for 100 randomly selected autonomous hardware+software robots. In what circumstances would some hierarchical social organisation ermerge? Who has recently studied this question with actual robots?

Hello, Mr. James. I believe the subject you are interested is Swarm Robotics, as it works specifically with this subject: tens, often hundreds of robots that do simple behaviors that emerge in a collective behavior.

Scholarpedia article on the subject: http://www.scholarpedia.org/article/Swarm_robotics

The article "Cohesion and segregation in swarm navigation" works specifically with group cohesion and segregation.

Another famous work is the BEECLUST, which involves a simple individual routine and is quite successful.

Other emerging behaviors are foraging (very common in literature) and area coverage. Most works, however, use simulated environments, or use a small amount of real robots in addition to simulations to achieve a proof of concept.

http://books.google.com.br/books?hl=pt-BR&lr=&id=oqwuwsdwxT8C&oi=fnd&pg=PT183&dq=beeclust&ots=UiSiWp5Xkf&sig=okgkWBODNfTL7JZGzsUoEW0pwpY#v=onepage&q=beeclust&f=false

http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=9220377&fileId=S0263574714000563

Swarm robotics clearly relevant (Many thanks to Murillo) but does not fully answer my question which was about recent work addressing HIERARCHY EMERGENCE in hardware ROBOT SETS. Yes I can trawl the swarm robotics literature (which I do know a little about) but can anyone point me directly to recent/ongoing projects in hardware Robot Hierarchy emergence – preferably NOT reflecting studies of human or animal or insect social hierarchies.

Hi Dear, Mr. James.

I think you should note that about capability of the robots learning and their instructions. Hence, you should provide and abstract view of your subjected structure. So this will need to design hardware with simulated interface to mask low level communication. So you can define three sections for robot hierarchy: Generic section, Device Interface section and Simulator Interface section. To have a good design, you should determine a behavioral dynamic for you system. You can design a cooperation system level among robots by using a cognitive architecture in a learning mechanism. So you can specialize and organize the roles of each section. I propose you to use an evolutionary method to make a hierarchical controller for this system.

Let me know are you thinking about learning and cognition viewpoint? I would like to draw your attention about the method of modeling human systems. That could help to make a behavioral hierarchical controller system with learning to control of its sub-controllers.

Best

To Mahdi Yusefi Azar Khanistan

Thank you for your interesting answer -- but  you seem to be focussing more on design of robot cognition and design of cooperation between robots than on emergence within a group of autonomous robots. You refer to a "cognitive architecture in a learning mechanism". How does your thought/design compare with the MIAP cognitive architecture I have proposed which is described in the working draft available from my profile here on ResearchGate entitled:

"The MIAP Agent Architecture: Learning, Memory, Planning, Meta-Control and Symbolic Communication"

This article can be read online or fully downloaded.

All comments and feedback welcome!

Jim

It looks as if the answer to my original question is "There are no such projects".

But ... surely technically feasible (am I wrong?) for "social" hierarchies to emerge ... someone somewhere must be looking at this ... and I keep wondering about drones ....

Jim