[Robotgroup] More USB info

[Glenn Currie]

I read Isaac Asimov's robot stories a long time ago.  He was one of the 
first, if not
the first to portray robots in a positive way, and to my mind, in a 
believable way.
He was looking way into the future, but his robots seemed much more 
believable
that in most sci-fi stories, where the robot is a soulless machine that 
is destine to
destroy its creators, or mechanical clowns.

It seems that there is a blur between what some consider science 
fiction, and
fantasy stories.  I always thought "speculative fiction" was a more 
accurate label
for the type of science fiction I find interesting.  In such stories, 
you are allowed,
very few, maybe even a single key piece of enabling technology to spin 
the tale
around, then everything has to hang together.

Clumping together an endless string of improbable pieces of technology 
and a comic book
plot has been over done, and is only of passing interest - nice 
graphics, maybe,
but a 12 cent, overused plot is not what I find interesting.  And as for 
fantasy,
endless stories derivative of King Aurthur and space aliens that look like
cats, I will pass on.  If others enjoy them that's fine.  They seem to 
outnumber
the "hard science fiction" fans by several orders of magnitude.

In good speculative fiction, the consequences of a particular technology 
can be
explored, and in a well written novel, things fit together nicely.  
Years ago,
a film called "Independence Day" came out.  It had some nice graphics, but
the plot made no sense and was repeatedly self contradicting.

It made a lot of money, so some folks liked it.  And I will never forget 
a couple
of people that were sitting the the row in front of me, endlessly asking
"whats going on?",  "Who is that?", "Are those the bad guys?".   Even with
a shallow and fragmented plot, the producers had spent millions of dollars
to present visuals and a multichannel sound track to tell their story.
All you  had to do was watch and listen.  I can see that it would be 
difficult
to follow any story, if you were not paying attention, because you were
endlessly asking what happened the the previous scene.  I am sure these
folks all have cell phones now,  and endless "free" minutes, so they can 
continue
their inquiries into the recent past.

I was very disappointed the the film adaptation of Isaac Asimov's "I Robot".
Nice graphics, but all but but the most superficial aspects of the original
story were lost.

Men in Black, was well done - clearly a comedy from the start.

Bladerunner is one of the best sci-fi movies ever made, in my opinion.
It did not do great at the box office, and the producers insisted in
changing the ending to a "happy" one which was made from steady
cam helicopter shots originally shot for "The Shining" but never used
in that film.  Riddley Scott got his ending put back in in the 
"Directors Cut"
of the movie.

There are many technical details to be worked out before we have robots
anything like those portrayed in Asimov's stories.  One of them, which
got me off on this thread, it adaptable, interchangeable parts.  There are
the mechanical challenges, which are tough enough, but also the control
systems, are where the bulk of where the effort lies, as far as I can see.

The idea that you would have to hand code control sequences, and
manually have to splice them together to configure a robot, is only 
acceptable in a research
environment.  To have significant numbers of smart machines in the field,
requires, spare parts, and good diagnostics to keep them working -
not to mention competent technicians.

Most of the robots we hear about are university / military research devices,
and one of a kind.

So back to the control system - although you could use a single host 
controller
to integrate all the control systems I think it would be more robust to 
have
much of the control system distributed / housed in the appendage / sensor
module itself.  For example, you could have an shoulder / arm / hand 
assembly
clamped in a test jig and run it through its paces.  Motor limit 
detection, over
current / over temperature functions would be pushed down into the assembly.
This is somewhat similar to we call a "knee jerk" reaction in humans.  
If we sense heat,
certain types of pressure, or a fast moving object crossing our field of 
vision,
we act first, and do higher level thinking, once we have, hopefully, jumped
out of harms way.

Something similar to the USB bus could be useful in building such a system,
but would involve what you might call "smart hubs".  With USB there are
a few things that a hub does, but for the most part it passes most data 
upstream
to a central host to deal with.

Most modern PCs actually have multiple root hubs on the mother board, so you
could give it try.  One of the other things that would be nice is to 
have the data
sent over fiber optic cable, rather than copper.  The noise immunity and 
data rates
are much better.

Oh, well, so many projects, so little time.

Cheers,

-Glenn

Alberto Alonso wrote:
> I did buy a couple of WRT54G (pre v5) with the intention to hack
> them eventually. I don't have any of the WRTSL54GS but they also do
> look like a good match for robotics.
>
> I don't see a problem with the id changing if you implement 
> serialization (as in serial numbers) of your hardware. The
> difficulty under such scenario would be to figure out which
> arm the hand is attached to, specially since all devices would
> connect to the central brain (ie. the SBC with USB host controller).
> If you have a simple arm id that the hand would know what it is
> when connecting to it, then it can report to your driver when it asks
> for its id then the problem is solved.
>
> It does propose an interesting concept in which you could plug
> 20 hands into a single arm. Quite an intriguing control problem.
>
> I never got my USB 802.11 cards working under Linux so they've
> just been collecting dust. It seems that you have an interesting
> use for them.
>
>
> Alberto
>