Re: [beam] RoboSapiens mechanics (bipeds part I)
- Hi Bruce,In http://groups.yahoo.com/group/beam/message/41633 I suggested ( and hereby slightly editted):>RoboSapien appears to use one motor in the abdomen (1) to generate the swinging inverted pendulum>motion of the torso to altenately lift each leg and a motor in each "ankle" (2) which drives a linkage connected to>the hip joint to extend the lifted "foot" forward and the planted foot backwards with respect to the hip. With both >feet planted, and both ankle motor rotating in the same direction, the torso (remains vertical but) moves forward and >backward with respect to the feet. (showroom mpg)>The "ankle" appears to use a large partial spur gear and a pinion gear (don't know the right term).Did you see the size of those feet? For the batteries to be sure but I suspect the ankle motors (which are linked to the hips) are down there too. In the showroom mpg, there appears to be an arched ridge across the top of each foot which may hide a section of a spur gear that the motor pinion gear engages.>The other 4 motors manipulate the arms. The shoulders have one motor each (2).agreed.>The head appears to rotate with the difference in the two shoulder motor positions facing toward the arm which is >lifted higher.agreed.>The elbows are controlled by one motor each (2). The elbow motion rotates the lower arm with respect to the upper >arm to change the angle of the elbow.From what I could see, straightning the arms at the elbows is always accompanied by a rotation of the forearm.>Each gripper is opened and closed by the elbow motors by turning/rotating the elbow beyond a limit to rotate >the wrist and open the grippers. I am curious to know where the elbow motors are located.When the arm is straighten out by rotating, the end of rotation opens the claw.wilf----- Original Message -----From: Bruce RobinsonSent: Sunday, February 29, 2004 10:26 PMSubject: [beam] RoboSapiens mechanics (bipeds part I)G'day, everyone.
As Wolfgang requested, here are some thoughts on the mechanics of
RoboSapiens. A special note of thanks to Ori and Wolfgang for offering
to send me a CD of the videos.
First point. The robot is underactuated, meaning the number of degrees
of freedom (I count at least 12) exceeds the number of motors (7
according to Tilden). That means there will be some interesting
mechanical linkages between various joints. Give me one to hack, and
that's the first thing I'd be looking at.
Second point. Where are the motors? I'd say one at the center of the
torso, two at the hips, two at the shoulders, two at the elbows. That's
also what Tilden seems to imply when he talks about the hexagon
arrangement. I suspect the elbow motors will be GM10's. The ankles will
be connected to the hips by a linkage. I'll guess the grippers are
connected to the elbows or shoulders, or both. The head seems to be
connected to both shoulder joints by some sort of differential linkage.
The grippers are especially interesting, and oh, for a high quality,
slow motion video. Pure guesswork here, but I wouldnt' be surprised if
they're operated by a combination of shoulder and elbow joint positions.
You could use several different pairs of joint angles to put the "hand"
in any desired position. But you could make one of these joint angle
pairs open the grippers through a linkage, while the other angle pairs
left them closed.
The comments about a non-Cartesian coordinate system are interesting. In
purely functional terms, a limb attached to a motor is most easily
described by polar co-ordinate system (or by a cylindrical coordinates
if you want three dimensions).
From a design point of view, most of us with formal training in
mechanical design use Cartesian coordinates because that's the way we
were trained to design things. It's actually a function of the drawing
instruments and manufacturing machines we use. I've found it very useful
to deliberately use an unfamiliar coordinate system to design something
-- it can lead to very interesting insights (try designing a three motor
walker using cylindrical coordinates.)
When Tilden talks about a system based on triangles, I suspect (and this
is pure guesswork) he is talking linkages that work in a plane -- a
triangle will always define a plane; the plane in turn is rotated by a
motor or shaft. The main advantage is in design terms. Once again, give
me one and I'll tell you pretty quick.
There is a great deal of compliance in the robot. Tilden grabs it's arms
and twists and turns things without appearing to break anything.
I especially like the idea that it will react to the sensors in a limb
only when that limb is "activated" by the control program. This sort of
thing allows you to forget about programming sensors. Program the limb
-- the sensor is enabled. Very Tildenesque.
Springy feet, I'd say. It seems to rely on the momentum of the torso to
lift a foot for walking. Slightly springy feed would help a lot. I've
gone a bit bug-eyed trying to figure out from the video if the robot
lifts a foot when the torso reaches the end of its motion, or if it's a
sudden reversal of the torso motor that lifts the feet. In any case, the
robot is remarkably stable considering how much body and arm waving it
I'm infering that there are several linkages. I'd love to see them
because I suspect there'll be a good deal of BEAMish minimalism
involved. In other words, how simple can you make something that will
give you complex motion.
Sorry I can't be more specific. Just lend me one for a couple of hours
and I'll give you a whole lot more ("It'll work fine when I put it back
together, Mom, honest." :)
More thoughts on bipeds to follow.
Message<<Wilf wrote . . .>>>Did you see the size of those feet? For the batteries to be sure but I suspect the ankle motors (which are linked to the hips) are>>down there too. In the showroom mpg, there appears to be an arched ridge across the top of each foot which may hide a section>>of a spur gear that the motor pinion gear engages.I don't think there are any motors in the feet, its all done with linkages from the motors in the hips - I think the linkage is a parallelogram (??) so the feet can move backwards and forwards without rotating, or if you do it without activating the waist then the body will shift forwards and backwards over the feet but whilst staying vertical.Take a look at the animation I did here: http://boardroom.solarbotics.net/rosap/page2.htmlIt shows what I think the leg mechanism is like, although I've realised another possible variation where the linkages at the foot are wider apart than where they attach to the body - the result (and I'll try and post an animation) is that when the body is pushed forward the hips move further than the head, producing a neat pelvic thrust, with the added advantage that the centre of gravity doesn't shift nearly as much as the perceived movement of the torso.I don't know how this would affect walking though but it might make it walk better - as the foot is pushed back the front end is pushed down in a similar way to the way we push with out toes a little when we walk.<<Bruce wrote>>>> It seems to rely on the momentum of the torso to lift a foot for walking.Its using a pendulum action, very similar to the way we walk (but we don't need to swing out torso from side to side to get us going, we use out super articulated legs and ankles instead ;)One interesting observation on the way it walks, when it was shown walking normally with its arms by its side there way quite a lot of side to side movement in the torso to keep the pendulum motion going. When Tilden demonstrates it walking with both arms raised there appears to be much less side to side motion, except when it initiates walking - when the robot has its arms up in the air it requires less energy to maintain the pendulum motion because of the extra mass near the top of the robot. it looks like the motors or motor oscillators are dynamically adjusting their activity to suit the current configuration (and conserve energy).>>There is a great deal of compliance in the robot. Tilden grabs it's arms and twists and turns things without appearing to break>>anything.All the major motors appear to be GM9's so they have a low gear reduction and are easily back drivable, I could swear I heard a 'click' as one point in one of the videos - was it one of the motor safety clutches popping? . . .Bill---------
- Much appreciated, Bruce. This promises to be an interesting thread.
J Wolfgang Goerlich
Bruce Robinson wrote:
> As Wolfgang requested, here are some thoughts on the mechanics of
> RoboSapiens. A special note of thanks to Ori and Wolfgang for
> offering to send me a CD of the videos.
- Was there a part 2? I miss it?
Learn how to help protect your privacy and prevent fraud online at Tech
Hacks & Scams. http://special.msn.com/msnbc/techsafety.armx