wu :: forums - Print Page


    
      
        wu :: forums
        (http://www.ocf.berkeley.edu/~wwu/cgi-bin/yabb/YaBB.cgi)
      

        riddles >> general problem-solving / chatting / whatever >> Selforganizing Robots.
        
(Message started by: rloginunix on Aug 15^th, 2014, 9:15am)

Title: Selforganizing Robots.
Post by rloginunix on Aug 15^th, 2014, 9:15am

In today's (August 15, 2014) "The Wall Street Journal" (US newspaper) on page A3 we find a "Tiny Robot Swarm Works On Its Own" article.

It claims that Harvard University scientists managed to convince 1024 tiny robots to whip themselves into various shapes without any central orchestrating intelligence to guide them.

Each robot, about the size of a US penny (one cent) or 19 mm in diameter, has a processor, an infrared sensor, a vibration motor to move it about and can do three things: 1). follow the edge of a swarm, 2). track its distance from origin, 3). track its relative position.

Title: Re: Selforganizing Robots.
Post by rloginunix on Aug 15^th, 2014, 9:18am

It took the little buggers 11.66 hours to form a solid 5-vertex star, 11.71 hours to form a solid letter K and 5.95 hours to form a solid "wrench" shape on their own.

The leader of the project claims that "you could have them build other robots out of themselves".

I'm thinking there must be some interesting algorithms involved to achieve this.

Title: Re: Selforganizing Robots.
Post by towr on Aug 15^th, 2014, 12:33pm

I read it on newscientist yesterday. It has video as well (sped up, because you don't want to watch it real-time).
http://www.newscientist.com/article/dn26060-watch-a-swarm-of-1000-minirobots-assemble-into-shapes.html#.U-5f77QTW3s

Maybe if they'd given the robots wheels they could have done it a bit faster.

Title: Re: Selforganizing Robots.
Post by rloginunix on Aug 16^th, 2014, 8:28am

A number of observations comes to mind.

1). It's probably too early to ask if these bots can invent new mathematics and physics theories on their own but I think it's reasonable to ask if they can come up with new shapes that we can recognize. A scary thought. I remember watching "Star Gate 1" with my kids and these "Replicators" used to send the chill down my spine. Jules Verne ...

2). Does each member of the swarm know how many members are in it? If yes, then what is the algorithm. Given N objects how, after a finite number of simple operations, can each one learn what N is.

3). It would be interesting to attempt to, khm, replicate this experiment in software. Each bot can be represented by a small circle of radius R with some number of functions attached to it. The distance D between the centers of any two circles can not be < 2R. If D == 2R then it's a collision. If D > 2R and D < (2R + delta) then these two bots are "close enough" to each other.

4). The whole theme echoes with the idea developed by Gottfried Leibniz (in the West) of taking something very simple, zero and one, and manipulating it via some number of predefined rules to achieve arbitrary complexity.