The Swarm Game
The Swarm Game
Written by Mike James   
Saturday, 19 October 2013

Can you control a swarm of robots with a single controller that has the same effect on all of them? You can find out by playing a game and helping develop new control algorithms.

Imagine that you have a lot of nearly identical robots and you need a way to control them. Controlling each separately is possible, but it would take a lot of computing power and a big communications bandwidth. The question is, can they be controlled and made to do useful things using a single central signal? For example, what could you achieve if the robots simply all followed the same rule - move towards the light?

Could you make them do useful things by simply changing the position and intensity of the light?

The following video should convince you that you can:

 

If you watch carefully at the beginning, you can see some of the principles of control, in particular the use of an obstacle to make one of the swarm behave differently. What isn't as easy to see is that the control mechanism depends on each robot having a slightly different response to the control stimulus.

In a paper to be presented next month at IROS 2013, researchers, Aaron Becker and James McLurkin, prove that as long as the robots are all slightly different you can create an algorithm that will move each one to a different target position. 

The overall idea is to use an attractive, or repulsive, point source to move the swarm in such a way that their combined flow moves other physical objects or changes the environment in some way. The problem is very similar to herding sheep with a single dog. 

You can see a convincing demonstration of 100 robots assembling and delivering a component in the next video: 

   

Notice that having algorithms of this sort would allow nanobots to move atoms and build useful nanomachines. There are many other possibilities, including using controlled swarms in medicine and research - for example sorting all the cells in a Petri dish. It is also worth noticing that the same algorithms might be relevant to morphogenesis, i.e the way living cells create regular structures. 

Despite having some theoretical results on how to control swarms, the research team would like you to help by playing some fun games. Your task is to play sheepdog, or perhaps that should be robot dog, and herd some robots into performing some task or another. As well as being fun, it is also very instructive and in no time at all you start to try out heuristics to see if you can improve your performance. 

 

swarmcontrol

 

When you think of the potential of this research there is no doubt that it could be the most important robot research going on at the moment. 

 swarmicon

Banner


Are Developer Bootcamps Dying Out?
16/08/2017

The announcement of the impending closure of Dev Bootcamp, and news that The Iron Yard had already stopped accepting new registrations  has led to some speculation about the demise of coding boot [ ... ]



SQLite 3.20 Improves Query Planner
08/08/2017

There's a major upgrade to SQLite, the embeddable SQL database library that is found in many memory constrained gadgets such as cellphones, PDAs, and MP3 players. The new release has an improved query [ ... ]


More News

 

 

Last Updated ( Saturday, 19 October 2013 )
 
 

   
Banner
RSS feed of news items only
I Programmer News
Copyright © 2017 i-programmer.info. All Rights Reserved.
Joomla! is Free Software released under the GNU/GPL License.