Hydraulic Hexapod Walker

One low-tech approach that I've seen (not on hydraulic or pneumatic, but
on electric motors) is to use a clutch that links the two sides. In your
case, your clutch would have a cam face machined into it so that it will
"click" into place in only one position of the 360 circle. That position
would be the exact tripod gait for the two sides in synchronism. Given
the size and weight of your robot, maybe a magnetic clutch would work
best. On the motors I've seen it was just a spring loaded mechanism,
where the clutch would allow slip if there was enough speed difference
between the two sides.

I once worked up a chain-driven hexapod using an electric motor and a
large aluminum frame (about 18" by 24"), and the walking efficiency is
not great. Something with 2DOF independent control of each legs works
better. But linked gait walking does function, and if the machine is
large and noisy, it looks/sounds impressive.

It would be great if this could be done with steam!

-- Gordon

Hi Gordon-

Thanks for your response. This sounds very much like what I'm looking
for, as I'm not necessarily after a high tech solution. I'll have to
dig up some info on a suitable clutch.

Do you have any pics of your hexapod? Mine has about a 6 inch "step"
(actually a cycloid motion).

A steampunk version would indeed be cool

Thanks again
Paul

(Replying here to questions also in e-mail).

One version of such a clutch was used in the Milton Bradley Big Trak.
The motors were once a common find on the surplus market but they've
long since sold out. Basically the clutch was composed of a spring and a
N/S magnet on the inside of the motor shaft (the wheel was mounted on
the other side of the motor shaft). The inside shafts faced one another.

If the motors are going the same direction, the two magnets will
re-align and tend to "catch" one another. If one motor goes in reverse,
or substantially different speed, the magnetic pull is broken, and the
sides are no longer in sync. I recall the magnets used in the Big Trak
had north and south on the same face, which is unusual. However, you can
make the equivalent out of two magnets for each side. Flip one magnet so
its N is facing out; flip the other so its S is facing out.

I doubt on a big machine that permanent magnets would work, but maybe
larger electromagnets? As you're working with hydraulics maybe the motor
running the pump could also run a small DC generator for the
electromagnets.

The other approach would be a completely mechanical version using a
clutch with a shape machined into each face so that it will align only
one way (kind of a cam). You'd still probably want a magnetic clutch to
release the clutch (rather than the tension of a spring) or else I think
eventually you've have a lot of wear or breakage on the cam.


It's in my book Robot Builder's Bonanza:
http://www.amazon.com/exec/obidos/ASIN/0071362967

Amazon lets you read though (some) pages. Starts on page 335 (chapter
22). You can also search the book using the phrase 'Walkerbot.' The book
is also available at most libraries. Walkerbot is not a great example of
a hexapod, and it's made for people with only access to a hardware
store, but it might give you some ideas. Again, to really do a hexapod
right, you want 2DOF on each leg, which means 12 cylinders and 12
valves. Yikes!

-- Gordon

That's really clever.  It lets you make a machine that actually goes in
a straight line, despite the unavoidable slight differences in motor
speed, without having to have encoders and a complex control scheme.


I'm partial to hexapod walkers with 3 DOF *total*.  One moves the front
and back left leg forward and back; another does the same on the right;
and a third tilts the center legs up and down (so that one is always up
while the other is down).  With that, you can walk forwards, backwards,
and turn either direction.  It's not graceful, but it's pretty effective.

Best,
- Joe

It is clever, though not the only way of doing it.

I've also seen on some old mechanical toys that use two spring-loaded
crown gears facing each other. The WAO robots from OWI work this way (or
did). The gears would be cheaper, a little easier to engineer (no funky
selection of magnets so get just the right pull/push), but probably not
as sturdy over the long run. The teeth of the gear can strip out over
time.

The same idea might work with a dimpled clutch plate and one or more
spring-loaded ball plunger.  Wouldn't suffer the wear problem so much.
You'd have to select the ball plungers with just the right springiness,
but fortunately you can get them pretty cheap -- under a buck each for
the small ones with a plastic press-fit housing.

-- Gordon

This thing will have a 6 HP gasoline motor and likely two variable
displacment pumps (as transmissions), and two hydraulic motors. Final
drive will be very slow, 40-50 RPM at the legs. I'd imagine torque is
going to be immense, so anything I make or hack together is going to
have to be pretty tough.

Maybe manual skid steer isn't such a bad idea...

I do like the idea of some RPM sync so it WILL go in a straight line;
the servos getting some feedback as to what the other side is doing.
The Futaba servos do have a port on them for some pulse input...

Thanks
Paul

OK, that's got me thinking about a clutch (almost like a sprag) with a
ramp on one face, maybe a spring loaded pin in the other. In one
direction it will freewheel & slip, the other way the pin will catch,
and off we go...maybe a modified motorcycle clutch. Plenty of spring
pressure there.

Thanks a lot
Paul

<snip>

OK, that's got me thinking about a clutch (almost like a sprag) with a
ramp on one face, maybe a spring loaded pin in the other. In one
direction it will freewheel & slip, the other way the pin will catch,
and off we go...maybe a modified motorcycle clutch. Plenty of spring
pressure there.

Thanks a lot
Paul

    I see the problem.  What you really need to control is the
range of motion on each side, not the speed.  You always want
to keep both tripods in phase, but for turning, you need to
shorten the stride on one side.   The problem is that
one motor controls each tripod, not each side.  Right?

    So it really only needs one motor.  But it needs some
way to tweak the motion of at least two legs.

    Suppose you spring-loaded the middle leg on each side
in the fore-aft direction, so it can be stopped without
damaging the mechanism, then provided it with a
hydraulic cylinder which keeps it from moving.
When you want to turn, you close the block valves for
the inside cylinder, which locks the fore-aft movement
of the middle leg.  The machine then pivots around that
leg for half a cycle.  No synchronization is really necessary,
though it would probably work better if you locked the leg
at the midpoint of its travel.

    This won't let you turn in place; the best you can do is
a tight circle.  (Sherman tanks had the same problem;
they couldn't run one track forward and one backwards, since
they just had clutches.)

    While you're adding springs, it would help if the
rear legs were given some flexibility in the transverse
direction.  That keeps them from being overstressed in
turns.  The "skid steer" thing is tough on pointy legs.

    Personally, I'd do the whole thing with hydraulic
cylinders, lose the motors and chains, and synch
everything in software, but then I'm a controls guy.

            John Nagle

Paul wrote:

That's a good summation.  I don't care abuot zero-turn or anything like
that but I would like the ability to turn and keep everything in phase.

But no, I foresaw this as each side controls each side, not each
tripod. One pump & motor on a side, the legs synced via the gears &
chain (see URL of first post).
I could arrange it in "per tripod" fashion easily enough with a
jackshafts, and sync wouldn't be a problem, but turning would still
require some special arrangement.

Can you elucidate a bit on what you mean by "spring loaded"?

Thanks
Paul

John Nagle wrote: