Holy Hopping Humanoids, Batman!

Yeah, it looks neat.  However, I didn't see it making anything that looked
like activate balance movements in the jump so I don't think it is actually
keeping itself balanced in any sense.  It seems to just be a wind-up toy
with large enough feet that it doesn't fall over when executing
pre-programmed walk and jump movements.  It could I guess have some feet
sensors that it's reacting to in such subtle ways that it doesn't really
show in the motions.


Could be.  Humanoid robots have a certain appeal to them that is missing in
the other robots.


--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

...

     It looks like the move is mostly canned, but when the robot settles,
it does seem to have some compliance.  You'll see it wobble a bit from
side to side with both feet in contact with the ground.  There's
some feedback there.

     If you can get force feedback, you can do so much more.
This has been known in industrial robotics for two decades, and
it's taken way too long to filter down to the hobbyist level.
The dumb output-only R/C servo interface has held back the
whole field.  Finally, things are moving beyond that.

                John Nagle

Yeah, good point.  Just some force feedback data from active servos would
give it a lot to work with without any other sensors.

Looking at the video again more carefully, I think I do see it reacting in
ways with the legs that doesn't seem to be caned when it lands.  But it's
hard to tell whether the reaction is the result of some spring built into
the legs and joints (which would be reasonable for a bot designed to jump
like that to reduce stress on the servos), or whether it's the result of
some active feedback algorithms at work.  A think a simple PID controller
with a lower gear ratio than typical hobby servos system could be using the
feedback to create a similar effect and it would only need an encoder to
get position feedback and not force feedback.

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

NeutrinoX probably has some active balancing system that works during
normal operation, and also stabilizes it when it lands after jumping.

I've tried getting some of my walkers to jump - or rather lift part of
the frame off the ground - but the standard servos are just too slow to
accomplish this. They just don't move fast enough to give the frame
enough momentum to lift-off. Neutrino probably has the fastest digital
servos as a start. Then, you notice it uses every body joint, from
flipping the arms upwards to bending and straightening at ankles,
knees, and hips, to produce upward momentum. The addition of springs to
the joints to give an additional boost would help a lot.

Regards my walkers, I was thinking about the possiblity of implementing
some sort of run-gallop type gait, which is the only way to get speed
beyond a fast trot. To do this requires lifting the front-end
completely off the ground and springing forward. The only way I figured
I might get this to happen is to add some sort of springs to the legs,
and use stored energy in the springs for the jump. But haven't gotten
that far yet.

You'll notice that animals in general don't get all of their lift-off
energy from muscle activation, but probably 1/2 of it comes from energy
stored in elastic parts, like tendons, ligaments, and mechanical part
of muscle fibers.

Regards other robots that could lift off, you can take a look at Scout
I + II. It gets its boost from springs and proper resonance timing ...

http://www.cim.mcgill.ca/~arlweb/scouti/scout1.htm
http://www.cim.mcgill.ca/~arlweb/scoutii/

     Humans recover about 60-70% of the energy from stride to stride in
the springyness of muscles.  For some of the specialized running
animals, like a cheetah, it's above 90%.

     Almost all the research robots that run have energy storage
in springs.  (ASIMO doesn't, but ASIMO's "run" has about 1cm of
ground clearance in the "flight" phase, which is very short.)

     The problem with springs is excess flexibility makes it hard
to do stable stuff, like standing.  An actuator followed by a
spring isn't enough.  Nobody has a really good mechanical solution
to this.  It can be done with pneumatics; if you can control
the pressure of both sides of a pneumatic cylinder separately,
you have a spring with a variable spring constant and a
variable zero point, which is what you want.  There's a robot
at CWRU that works that way.  See

    http://biorobots.cwru.edu/projects/robot3/robot3.htm

     The Stanford 4-legged running machine (Waldron) has a leg supported
by a spring, plus a motor that winds up a cable to tension the
spring.  The motor can release the cable very fast, faster
than it can tighten it, so you can wind up a leg and then,
at a chosen time, release it.  That gets you takeoff power
for running and jumping.  When the leg is partly wound
up, with the motor pulling against the spring, it's relatively
stiff, so the thing can stand without wobbling.  See

    http://www.stanford.edu/group/locolab

     Some combination of motor, spring, and maybe a clutch or brake
is probably the way to go, but the mechanical designs need to improve.

                John Nagle
                Animats

It has two gyro sensors and an accelerometer. The analog sensor signals
are processed by the the robots RCB-3 controller and used to control
multiple servos - usually four or six hip/leg/foot servos in a
parallelogram fashion.

Thanks for the info, Lem.  You seem to have more info on some of these
bots than I've seen posted anywhere.  A great feature for your site
would be, now and then, a detailed report on a particular robot -- what
sort of servos it uses, what sensors it has, how it's done in recent
competitions (if any), anything the builder's especially proud of,
whatever you can find out.  That would really be interesting and helpful
for a newbie like me.

In addition to this great jumper, I'd love to learn more about this red
"900" (or is it "006") robot shown in these images:

<http://www.flickr.com/photos/95721430@N00/245293733/in/set-7215759428758
6087/>

For example, what's inside those neat little cubes at the joints, and
why do there appear to be no wires connecting them?!?

Whatever you do, please keep up the good work with robots-dreams... it's
become one of my best resources for information on current hobby
robotics!

Best,
- Joe

Sheesh, what a newbie -- that bot is none other than the famous
OmniZero!  Don't I feel like a schmuck.  Anyway, I'd still love to see a
detailed profile of it, with as much detail as you can dig up.

Best,
- Joe

You're very welcome. It's my pleasure.


Well, it helps that I've lived here and been involved in the technology
business in Japan since late 1982.


I agree, and will try to expand and enhance Robots Dreams to include as
much of that as time permits. Still, there are only so many hours in
the day... !:')


As you figured out in your later post, that's the latest version of
OmniZero. It's number '006' in the ROBO-ONE official database.


If you're talking about the knee joints, those are servos. Maeda likes
the double servo type approach since it responds much more rapidly than
the more common single servo designs.

As for 'no wires' - they're there, it's just that Maeda is a
perfectionist and takes the time to dress and route things properly.
He's definitely a 'Pro', but then working for Vstone he should be.

Later,

Lem

Very cool!

--
Curt Welch                                            http://CurtWelch.Com/
curt@kcwc.com                                        http://NewsReader.Com/

I'm a newbie, with a question about RC servos.

Do servos use power when they are stationary?
Do they use power all the time?

If so, is this a big problem?


--
Martin Sondergaard,
London, UK.

    If there's no mechanical load on the servo, and
it's not moving, the current drops to a very low level.
If the servo is pushing against a load and can't make
it to the desired position, it will draw current even
when stationary.  This is likely to happen in leg servos
for robots, since some force has to be applied to oppose
gravity.

            John Nagle

If you want a servo to stop the best way is to no longer send it pulses.
It then uses virtually no current. Note that some serial servo
controllers are designed to always send pulses. I'd avoid those
actually, and go with a microcontroller or servo controller that allows
you to stop sending the pulses.

As John notes if the servo is not under load, even with pulses
continuously sent the current draw is low ... usually just a couple
milliamps.

-- Gordon

Thank you for your answers, John and Gordon.

--
Martin Sondergaard,
London.