--- In SeattleRobotics@yahoogroups.com, "David Buckley" <david@...> wrote:
>
> Hi
> Bimbo Town, bit of a mystery.
...........
> Just looked at your link to BimboTown, more pictures of me again, Rebecca in
the 'smalltalk' photo was quite stunning.
>


Following link calls Bimbotown "the world's only robotic nightclub". Sounds
pretty decriptive. Has some good pictures of the robotic drummer.

http://www.extremearts.co.uk/

The more I look at your site, the more amazed I am! I notice robot designs you
built back in 1993 that I am just beginning to try to emulate today [on a much
smaller scale however, LOL]. They look like great platforms to build robotic-AI
upon. Were/are they ROV or autonomous? Has anyone done extensive research with
them? Martin Smith?

http://davidbuckley.net/DB/ImperialLadies.htm



>
> The flexibility and soft approach to being right, necessary to perform
usefully in a human (animal) society will be needed in robots. They will look
generally predictable (as sane humans are) although we won't be able to predict
on a fine scale. Condor my best walking robot has just undergone a dual hip
replacement, it was showing signs of wear. Now it performs better and it gave me
an oportunity to get the program to dynamically adjust the side to side rock.
However now sometimes Condor appears to stop and ponder which way to go and at
other times appears nervous as if it isn't quite sure. These are just artifacts
of the interplay of program and the mechanical resonance and spatio temporal
changes in the perceived >environment.
>


Yes, this is similar to what Rolf Pfeifer talks about in his books as
"morphological computation", and Rbt Full calls mechanical "preflexes". When I
was testing one of hexapod walkers at different speeds, I noticed the faster I
tried to make it go, the more it became jerky and unstable. From reading Pfeifer
and Full, I think the frame was far too rigid, and eventually plan to build
better flexibility in future walkers, so the frame can self-adjust to hard
foot-impacts.


>
> Once we abandon all the inverse kinematics and pretending humanoid robots are
just human looking machine tools which WILL/MUST work as in the screen
simulation then stuff like being ablr to do the Woz test will naturally follow.
Something I realised the other day is that on the occasions when Asimo has
fallen down there is no recovery procedure for it to follow, once it gets out of
the enveloppe, it locks into the last command position and CRASH.
>
> David
>


This is an interesting thought - abandoning IK. Opposite stance to the totally
computational approach of the past 50-years. Of course, this is what Pfeifer and
Full are also saying. After all, brains don't compute diff-eqs, they largely
reproduce learned patterns of muscle movements, and use sensors and feedback for
real-time adjustments.

OTOH, the motor emulation stuff of Rick Grush is very interesting. Ie, evolved,
predictive motor circuitry in brains.

http://mind.ucsd.edu/