[ODE] controlling ball and socket joint

Russ Smith russ at q12.org
Mon Nov 12 15:14:01 MST 2001


> Do anybody know a good way to control the ball and socket joint ?
> I'm going to use it for a hip joint and a shoulder joint. The joint
> has a desired angle between each of the three axis. The angle is
> meassured between the two bodies. But if I wan't to rotate the bodies
> around one axis, then the angle will change at the other axis as well.
> This will start an acceleration. At least I think this is the problem.
> Do I then have to change the desired angles around the other axis? I
> think this is a bit complicated especially if the body is going to
> rotate about more than one axis.
> Are there another solution ?

hi camilla,

several people have asked for powered b&s joints now (as well
as joint limits), so i think i'll move it up the todo list.
you will be able to specify a desired relative angular velocity
across the b&s joint, not the position.

specifying the desired joint angles is a problem, as there are
many ways to do this (various kinds of euler angles, quaternions,
distances from fixed points, etc etc) and the problem is that
each way is more suited to a particular application. pitch, yaw
and roll angles (i.e. euler angles) normally work okay for
robotics applications, as real robotic 3 dof joints are often
made as compound 1 dof joints. euler angles are described,
e.g. in

  http://www-2.cs.cmu.edu/~baraff/sigcourse/index.html

(i think). translating desired position to desired velocity
is a separate problem, hmmmm, i think the trick is to find
the axis along which the desired relative rotation can be
performed, and then use this as the desired relative angular
velocity. quaternions would be good for this....

russ.

--
Russell Smith
http://www.q12.org



More information about the ODE mailing list