| |
Hey everyone.
I have built a realistic grasping end effecter that can feel when something in
is its grasp (without touch sensors). I have attached a deconstructed version
that I use.
http://www.brickshelf.com/gallery/BuckeyeShep/ArticulatedArm/end_effector.jpg
The principle is pretty simple. As the end effecter closes and touches the
object, the worm gears cause the fingers to securely grasp the object. On the
power train up to the end effecter, there are two rotation sensors, one with a
rigid spur gear and one with a slip clutch gear. As long as the rotation
sensors relative position (in this case, a 3:1 ratio) remains the same (i.e.,
the clutch hasn't begun to slip), there is nothing sensed and the motor
continues to close the fingers. As soon as something is securely grasped, the
clutch begins to slip and the relative positions of the rotation sensors will no
longer match. This condition would be sensed and the program would stop the
motor.
Opening the fingers would work in the same fashion. When the linkage hits
something so the fingers can't open any further, the clutch again spins and the
motors would be turned off. I haven't shown it, but on the bottom side I have
constructed a mechanical stop.
The end effecter is slightly rickety, but effective none the less. The
non-reversing nature of the worm gears gives a firm and non releasing grip. The
fingers give a really nice parallel motion that consistently meets at the same
position every time.
I prefer to use an old 9v motor for a couple of reasons. The gear ratio that is
shown gives a nice smooth motion that is not too slow or too fast at full speed.
I also like it because the motor is relatively light so it won't add a bunch of
weight at the end of a long arm.
It uses to input ports, which many won't like. I always use two programmable
bricks when I build articulated arms, so for me it doesn't hurt (much).
I like this design because the end effecter resembles a lot of hobby-like and
commercially available types
Chris
|
|
| |
In lugnet.technic, Chris Shepherd wrote:
> I have built a realistic grasping end effecter that can feel
> when something in is its grasp (without touch sensors).
Nice. And I agree, nice-looking. I once Bricklinked some of those same "rubber
skirt" pieces for this, but never did get them in an effecter that I liked (I
was trying smaller at the time). Now I use the flexible 1x2 studless beams that
first showed up in some of the old hockey sets:
http://www.peeron.com/inv/parts/45590
These seem to make ideal, compliant "fingers" to pick things up.
> there are two rotation sensors, one with a
> rigid spur gear and one with a slip clutch
> gear.
I must be missing something - why use two rotation sensors? Just monitor the one
on the clutch gear - if it stops advancing and the program is still commanding
the motor to turn, you know that the jaws have stalled and the fingers are on
something. If you're careful about how you use the rotation readings, you can
even tell the rough size of whatever the gripper is holding. I've used this on
the NXT on several occasions (although not for a gripper), with the nice feature
that there it requires *no* sensor ports (the rotation sensor is built into the
motor).
Here's an example of a way to "DetectStall" on the NXT:
http://www.brickshelf.com/cgi-bin/gallery.cgi?i=2036021
--
Brian Davis
|
|
| |
In lugnet.technic, Brian Davis wrote:
> In lugnet.technic, Chris Shepherd wrote:
>
> > I have built a realistic grasping end effecter that can feel
> > when something in is its grasp (without touch sensors).
>
> Nice. And I agree, nice-looking. I once Bricklinked some of those same "rubber
> skirt" pieces for this, but never did get them in an effecter that I liked (I
> was trying smaller at the time). Now I use the flexible 1x2 studless beams that
> first showed up in some of the old hockey sets:
>
> http://www.peeron.com/inv/parts/45590
>
> These seem to make ideal, compliant "fingers" to pick things up.
>
> > there are two rotation sensors, one with a
> > rigid spur gear and one with a slip clutch
> > gear.
>
> I must be missing something - why use two rotation sensors? Just monitor the one
> on the clutch gear - if it stops advancing and the program is still commanding
> the motor to turn, you know that the jaws have stalled and the fingers are on
> something. If you're careful about how you use the rotation readings, you can
> even tell the rough size of whatever the gripper is holding. I've used this on
> the NXT on several occasions (although not for a gripper), with the nice feature
> that there it requires *no* sensor ports (the rotation sensor is built into the
> motor).
>
> Here's an example of a way to "DetectStall" on the NXT:
>
> http://www.brickshelf.com/cgi-bin/gallery.cgi?i=2036021
You are not missing anything, I missed your idea, one sensor will do just fine.
I was comparing sensor reading to sensor reading and your point compares sensor
reading to motor on/off condition, which is even better.
I have never considered your idea of motor stall. That is probably because I am
notorious for high gear ratios. When I use worm gears especially, I usually
avoid stalling because of the high forces generated. But using a 1/10th second
stall, not many bad things will happen.
Great ideas, thanks.
|
|
| |
Hi, this is a very good solution! I really like the way this is done, even if it
uses 2 sensor ports.
The use of the double gear boxes to close the fingers at the same time is very
nice!
Really great work, excellent end effector solution.
Eric Sophie
|
|
|
