Im looking for a rotational actuator for a robot leg (knee) joint that
would, when its power is disabled, relax the lower leg into the
vertical downward position (according to gravity) - would the best type
of actuator be a stepper motor and controller? - I would rather use a
servo for ease of control, but I cant figure out a way of disabling the
servo to relax the joint.
What kind of resistance to physical force does a stepper motor give
once its power is disabled?
a stepper would be innapropriate for that, they just sit there and resist
motion, pretty much at the rated force, listed as holding torque. you
could use muscle wire, it's rather slow. if it was a big enough project
i'd use a rotating hydraulic cylinder. but, if you have reserve power on
board then maybe the stepper would work fine, just like any other motor,
Regards a stepper motor, not much resistance at all.
Regards an r/c servo, as soon as you stop sending pulses to it, it goes
limp, and the leg its driving will typically droop under the force of
You might also take a look at the "passive" knee on Scout II ... it
actually uses a solenoid-activated locking detent mechanism (whatever
it's called) ...
You will notice from the movies that the knee never actually gets into
the "bent" position as shown in the "Scout II with passive knees as
rendered in ProEngineer" diagram. Clearance is never that good. The
foot actually just drags along the ground.
- dan michaels
Unfortunately, modern actuator technology will perform badly in respect
to your desired goal. Both steppers and servos typically incorporate
gear reduction, which tends to make slack motions pretty stiff and can
even damage the actuator if too much force if applied (not that gravity
would apply). You'd probably have better luck with pneumatics or muscle
wire in this regard, but these hold disadvantages all their own that
usually make them overall worse candidates than electric motors. One of
these days (years...decades) artificial muscle technology will finally
take off, but until then, I'll say go with servos. It won't be perfect,
but at least they're self-contained and easy to interface.
Thanks for all your replys.
I have played with servos, but the resistance to external force is too
high (plus I hear that it buggers the servo!)
The other option I have been considering is some kind of electronicly
activated clutch mechanism (similarish to the "passive" knee on Scout
II ) which would physically disable the servo gear from the lower half
of the leg joint, enabling it to swing back to vertical when the clutch
is enabled (disabled?)
Does anyone know of such a device small enough to attach to a servo
It seems that in trying to model natural leg movements, it would be a
requirement to allow the joint to be both passive and active. Anyone
seen such a thing on a robot leg?
We have done this kind of thing using Air Muscles (surprisingly). Use
one muscle as in the human thigh to pull the knee straight. When the
muscle is vented, the leg will lower under gravity. If you don't need to
hold the leg in an intermediate position, you can do it with one valve.
Sure. You just need to do more reading.
Muscles are best modelled as springs whose zero point
and spring constant are adjustable. There's still no really
good way to build equivalent hardware, but there are some
devices that come close.
All serious leg work has back-driveable
acuators with some compliance. Early workers like Raibert used
pneumatic cylinders, but usually needed an external air supply.
Shadow Robotics has played another type of air actuator.
If you have good feedback, robust control algorithms, and
keep the valves close to the actuator, precision control of
air cylinders is possible.
The combination of a true linear motor and a spring has
possibilities, but linear motors are hard to find since Aura
tanked. And you don't get much energy storage, so running
is inefficient. (In running, about 70% to 80% of the
takeoff energy is recovered on landing and stored in
the spring capacity of muscles.)
Gil Pratt's people at MIT tried screw jacks in
series with springs, which, with aggressive control, can
be made to behave like muscles. There's force sensing
at the spring. When the spring compresses,
the motor frantically cranks the jack to maintain the
desired force. These aren't back-driveable (much).
They don't store energy well either, so they're
inefficient for running. But they're buildable from
off the shelf components.
The most common solution today is a servomotor and gear
train designed to be back-driveable, plus some elastic
component to take shock loads. Look at a RoboSapiens
for an example.
I have looked into using your air muscles previously - They look like a
fantastic product. The use of compressed air was an additional
complexity that I didnt really want to have to burden, but it looks
like the air muscles do satisfy my requirement. I think I could be
converted - Thanks for the info.