Alt-BEAM Archive
Message #08954
To: beam@sgiblab.sgi.com
From: Bruce Robinson Bruce_Robinson@telus.net
Date: Sun, 09 Jan 2000 14:17:07 -0800
Subject: [alt-beam] Re: Pressure Sensors
Timothy Flytcher wrote:
>
> You know Bruce, I have seen how the honda robot walks...
> Do we really walk like that???
I don't, at least most of the time. You probably don't either. But you
certainly could.
> I mean it just looks so god awful silly... and ... unnatural...
> I grant that it is smooth but there seems to be an ??? over-
> balanced look to it... Is it just because it moves in slow motion???
It would look a little more elegant if it moved faster. The specs say 2
km/h ... humans typically walk at 5 to 6 km/h.
> It shifts all it's weight from one foot to the other... as if it
> could just stop all motion at any given instant...
Oh, I'm pretty sure it can stop on a dime.
Watch an experienced person practice Tai Chi Chuan. Watch a Shotokan
Karate class practicing basics. You'll see motion of that sort --
controlled balance. The Tai Chi class will be slower and the Karate
class will be faster. You'll also see that kind of controlled balance in
many styles of Kung Fu, and in most Japanese sword arts. Commandos learn
to walk in that manner, so they can stop their motion instantly if they
encounter something that can make a sharp noise.
Martial artists can move very quickly in a balanced mode, and it looks
much more elegant than the robot. It would still look a little strange
if they walked about in public that way. Mind you, we walk the way we do
(controlled falling) because it happened to be what worked when we were
first learning to walk, and we (most of us) haven't needed to change. It
looks normal only because it's what we're used to seeing.
You're right about the robot moving smoothly, but it doesn't move
FLUIDLY. Very mechanical looking. Keep in mind, too, that P2 had a mass
of 210 kg (460 lb) and P3 has a mass of 130 kg (290 lb), so there's a
lot of momentum there to deal with.
P2 & P3 both have 6 degrees of freedom per leg. The closest
approximation to true human movement would have 8-9 degrees of freedom,
so these models will never be able to walk like us. On the other hand,
trying to control 9 degrees of freedom per leg right off the bat would
be darn near impossible. They'd have a heck of a job just getting the
thing to stand.
We learn to move by constantly monitoring what we move, by how much, and
what the result is. Our brains are constantly refining the motion, quite
unconsciously. Human-made 'bots still rely on their makers to provide
this constant refinement, so it is a MUCH slower process. However, I
suspect that by now their are some learning algorithms in their
programs.
Senior wrote:
>
> A lot of that money went into aesthetics. They had to have all
> those plastic parts custom made.
On a typical BEAM budget, the body shell would be pretty pricey. But on
their budget, it would be a trivial amount. Don't forget who's making
this -- they have access to the kind of equipment to make those parts.
> And they probably didn't walk down to their nearest machine shop
> and hand-machine the mechanics! Since they had the money, they
> probaly had all the parts professionally made (Costly!).
True, but they own the machines AND employ the professionals that run
them.
Fact is, a great deal of that money went into labour: pure worker-hours.
And the project has been going on for a couple of decades. So don't get
TOO cocky about thinking you could do it faster with that kind of
budget. You and about 100 clones, maybe :)
Bruce
8955 Sun, 9 Jan 2000 14:27:16 -0800 (PST) [alt-beam] Re: icq? beam@sgiblab.sgi.com Daniel Grace I do. I thought my ICQ number was in my signature. Oh
well, it is:
39402143
~Daniel
8956 Sun, 9 Jan 2000 17:56:02 -0500 Re: icq? Daniel Grace
> I do. I thought my ICQ number was in my signature. Oh
> well, it is:
>
> 39402143
>
> ~Daniel
>
> --- Mike Kulesza wrote:
> >
> >
> > >
> > >hey are there any of you beam guys having icq?
> > (exept for richard he`s
> > >already in my contact list)
> >
> > I! ICQ#23333229
> >
> >
> ______________________________________________________
> >
8957 Sun, 9 Jan 2000 17:59:26 -0500 Re: Pressure Sensors Timothy Flytcher
> But a tilt sensor only comes into play AFTER the robot has tilted....
> Possibly fatally???
> Timothy...
>
>
> >When you think about it feedback of pressure isn't really important if
you
> >always extend your foot the same length or distance. Granted we use it
to
> >determain if we are planted on the ground firmly. However I don't think
it
> >has any use in a bot which can set the travel of it's legs.
> > A much more usefull sensor to consider would be a tilt sensor to
tell
> >the bot if it has extended it's foot on something too high for it to walk
> >over.
> >
>
> >
> > > >
> > > >For example, a robot foot could be made from a assembly of two PCBs
> > > >separated by four CDR sensors, one at each corner. One PCB provides
the
> > > >common contact for the sensors and the second PCB has 4 separate
> >contacts
> > > >for the other side of the sensors. Now with molded CDR sensors you
> >could
> > > >do anything.....
> > >
> > >
> > > Just a thought... It seems to me (all this talk has made my mind dwell
> >on
> > > legs) that we use TWO types of sensors when we walk... one... a simple
> > > switch (tactile sense of touch) and two... our feedback of presser ...
> >So
> > > wouldn't our bots also need two??? Well, need is not the right term...
> >maybe
> > > "better if"???
> > > Timothy...
> > > __
> ______________________________________________________
>
8958 Sun, 09 Jan 2000 15:30:48 PST [alt-beam] Re: Pressure Sensors beam@sgiblab.sgi.com "Timothy Flytcher" Oh yes... I know that if I had just a fraction of their budget I could could
do it... I could easily vacuum form all the parts needed for there
aesthetics... My only question is why they did it how they did it??? Do they
really intend mass production???
Timothy...
>Another note...
>A lot of that money went into aesthetics. They had to have all those
>plastic parts custom made. And they probably didn't walk down to their
>nearest machine shop and hand-machine the mechanics! Since they had the
>money, they probaly had all the parts professionally made (Costly!). So,
>don't be scared off just because they spent so much money, we can be
>more creative :)
>
>Kyle
>
>Timothy Flytcher wrote:
> > >If biped walking were simple, Honda wouldn't have had to spend $200
> > >million on their humanoid walker project. Of course, they went with
> > >computers instead of analog technology, so that's set them way back :)
> > >
> > >Bruce
>
______________________________________________________
8959 Sun, 9 Jan 2000 18:44:19 EST [alt-beam] Re: icq? beam@sgiblab.sgi.com CIRCITZ@aol.com hey All
ICQ me @ 14450937
Dan
8960 Sun, 09 Jan 2000 15:46:51 PST [alt-beam] Re: Pressure Sensors beam@sgiblab.sgi.com "Timothy Flytcher" I don't think that a panic response is a good thing... Can we really make
our bots jumpy and still energy efficient???
Timothy...
ICQ# 26870234
>Should still have time to panic and move in some way to aviod falling
>farther.
>
>What do you do when you've got one foot on a ledge about to fall.....panic
>:)
>>But a tilt sensor only comes into play AFTER the robot has tilted....
> > Possibly fatally???
______________________________________________________
8961 Sun, 09 Jan 2000 15:21:48 -0800 [alt-beam] Re: Pressure Sensors beam@sgiblab.sgi.com Senior Bruce Robinson wrote:
> Senior wrote:
> > A lot of that money went into aesthetics. They had to have all
> > those plastic parts custom made.
>
> On a typical BEAM budget, the body shell would be pretty pricey. But on
> their budget, it would be a trivial amount. Don't forget who's making
> this -- they have access to the kind of equipment to make those parts.
>
> > And they probably didn't walk down to their nearest machine shop
> > and hand-machine the mechanics! Since they had the money, they
> > probaly had all the parts professionally made (Costly!).
>
> True, but they own the machines AND employ the professionals that run
> them.
>
> Fact is, a great deal of that money went into labour: pure worker-hours.
> And the project has been going on for a couple of decades. So don't get
> TOO cocky about thinking you could do it faster with that kind of
> budget. You and about 100 clones, maybe :)
Hehehe allright... The work they've done is great. No other robots
really match up. It just seems to me that they spent so much money and
work and only came up with static balance. But which is better? Static
balance or dynamic imbalance? Which is harder? Apparently, for us,
static balance is harder (But not soo hard). For bots, it's easier. So
it seems to me that they just took the easy route. Maybe it's better
though. Just not as exciting, like those bots that jump around (But
can't stop!), that were on a few shows and someone posted a link to pix
a while back. I guess I'll try my hand at building a smaller version and
see how well I do.
On another note, I didn't realize how heavy those guys are! That's
impressive that it walks so elegantly with that much weight! Maybe they
just spent a good sized chunk of their funding on some nice motors? ;)
-Kyle
Home