[ODE] Reaction forces with slider joint
Gib Bogle
bogle at ihug.co.nz
Tue Aug 24 16:02:13 MST 2004
Hi,
I'm simulating a chain-like object, made up of a number of sections
(about 30) each 6 m long and with mass of about 1000 kg. The sections
are connected using ball-and-socket joints. The last link in this chain
is light (like a rod) and I want this one to have variable length. The
only way I've found to accomplish this is to make the last connection a
slider joint, and to set the limits like this:
dJointSetSliderParam(joint, dParamLoStop, d-delta);
dJointSetSliderParam(joint, dParamHiStop, d+delta);
where where d is the extension that I want, and delta = 0.001 m (for
example).
When I run the model, there is a startup period over which I ramp d from
0 up to a certain values (e.g. 10 m), then I leave the joint as it is.
The motion that the chain undergoes creates some substantial tensions in
it, i.e. large reaction forces at the joints, up to about 3.3 MN. For
debugging purposes, I'm running to a steady state condition in which the
tension forces are unchanging (in case you are interested, this is
achieved by whirling the chain around at a fixed rotation speed of about
1.5 rad/sec).
The problem I'm trying to understand is that I see a large and
apparently random variation in the tension at the end where the slider
joint is (this is near the axis of rotation). If I replace the slider
joint by a ball-and-socket joint (like the others) the forces all settle
down to steady state values. With the slider joint the random
fluctuation in tension actually increases with time. After 100 sec I'm
seeing values near the axis in the range 2.5 - 4.0 MN, i.e. randomly
fluctuating about the correct mean value. The model is not useable.
If someone knows something about the slider joint that could explain
this problem, or lead to a way to eliminate it, I'd like to hear about
it. Otherwise, maybe someone can suggest a different, more stable way
to achieve the effect that I want, i.e. the ability to vary the length
of a link (a body).
Thanks
Gib
More information about the ODE
mailing list