Alt-BEAM Archive

Message #03689



To: beam@corp.sgi.com
From: "John A. deVries II" zozzles@lanl.gov
Date: Tue, 25 May 1999 10:31:22 -0600
Subject: [alt-beam] Re: Spyder (was: Mr. Tilden & Co. on Discovery Channel)


At 05:16 AM 5/25/99 , George Rix wrote:
>An awesome shot of Spyder walking!
>Or trying, anyhow, Mark's hand was preventing that, as Spyder's legs
>couldn't reach the table =)

Was some reason given for not permitting Spyder to walk??

Speaking of which, Spyder is one of those robots about which you read "it
_learns_ to walk in just... (4? 10?) steps" as if some sort of magic
convergence takes place. Actually, I think what happens is the following:

(a) both the central microcore and the leg chains come up saturated.
(b) as a result of the leg chains being saturated, the leg motors make
fairly uncoordinated motions -- call them "steps" if you will, but I
wouldn't. More like twitches.
(c) until the PNC "cleans up" the central microcore, processes from it
continue to be fed to the leg chains, causing them to continue in their
relatively uncoordinated fashion.
(d) chances are there is a brief period between the time that the excess
processes are extinguished by the PNC and the PNC releases, during which
the legs calm down.
(e) finally, the PNC releases and a process is allowed to circulate around
the central microcore.
(f) it probably takes two circuits of that process around the microcore
before the leg chains are properly initialized and the legs appear to act
in a coordinated fashion.


No magic chaotic emergence, no learning, although the actual amount of time
between turn-on and coordinated movement will vary. Speaking of which,
given the arrangement of Spyder's legs (diagonally distributed at the
corners of a rectangle) how does it walk very well at all (in fact I have
seen it walk reasonably well) ? How can it be made to turn? Finally, is
Spyder indeed the robot described by the patent or has stuff been added
between the time of the patent and its current state?


Any thoughts?


Zoz


---------------------------------------------------------------
John A. deVries II
zozzles@lanl.gov

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com - Simplifying group communications



Home