From ScienceDaily:
A way of making hundreds or even thousands of tiny robots cluster to carry out tasks without using any memory or processing power has been developed. Engineers have programmed extremely simple robots that are able to form a dense cluster without the need for complex computation, in a similar way to how a swarm of bees or a flock of birds is able to carry out tasks collectively.
The original experiment/simulation is based on sight, that is, whether the robot can sense each other, if it can't it will simply rotate or move about until it senses another of its kind. This kind of 'behaviour' eventually allows them to form a cluster formation:
Each robot uses just one sensor that tells them whether or not they can 'see' another robot in front of them. Based on whether or not they can see another robot, they will either rotate on the spot, or move around in a circle until they can see one. In this way they are able to gradually form and maintain a cluster formation. The system's ingenuity lies in its simplicity. The robots have no memory, do not need to perform any calculations and require only very little information about the environment.
Looks pretty interesting but I'm wondering how does it scale up? Is it solving problems by just being in a certain formation or is it like a 'Divide & Conquer' strategy where the task at hand is broken down into its smallest task which a single robot can tackle?
(Score: -1, Offtopic) by Anonymous Coward on Saturday April 19 2014, @01:49PM
Ask Soylent: how can we encourage more (quality) submissions?
(Score: 1, Insightful) by Anonymous Coward on Saturday April 19 2014, @02:02PM
How does huddling together equal carrying out tasks?
(Score: 1) by jorl17 on Saturday April 19 2014, @03:04PM
Isn't this just a very simple reactive agent which does no useful work at all?
(Score: 4, Interesting) by stormwyrm on Saturday April 19 2014, @06:06PM
You'd be surprised at what even purely reactive agents can do. You could build a very simple, purely reactive robot essentially capable of following the shoulder to the wall method of solving a simple maze, all without building up any representation of the world it's moving in or anything more complicated than sensors and some comparatively trivial logic. Just have one sensor capable of detecting the wall, have another to detect collisions to its front. Wire it up so that while it detects a wall to the right, just keep moving forward. If it stops seeing the wall, turn to the right and move forward until it can see the wall again. If the collision sensor triggers, make it turn to the right until it collides no more. No need for planning or even very complicated sensors. This is a very simple example of the subsumption architecture [wikipedia.org] advocated by Professor Rodney Brooks [mit.edu] and his colleagues at MIT.
Numquam ponenda est pluralitas sine necessitate.