Understanding framelinvel

[nilsiism]

Hello Community,

I have a problem understanding the framelinvel sensor

I added three sensors

<framelinvel name="fooVelBase" objtype="site" objname="Foo" reftype="body" refname="Bar"/>
<framelinvel name="fooVelWorld" objtype="site" objname="Foo"/>
<framelinvel name="barVelWorld" objtype="body" objname="Bar"/>

Next I read their sensor values and take the norm

bar_lin_vel_W = mjdata.sensor("barVelWorld").data
norm_bar_lin_vel_W = np.linalg.norm(bar_lin_vel_W)

foo_lin_vel_W = mjdata.sensor("fooVelWorld").data
norm_foo_lin_vel_W = np.linalg.norm(foo_lin_vel_W)

foo_lin_vel_B = mjdata.sensor("fooVelBase").data
norm_foo_lin_vel_B = np.linalg.norm(foo_lin_vel_B)

Due to the triangle inequality I would have assume that the velocity of bar is bound by

norm_foo_lin_vel_W - norm_foo_lin_vel_B <= norm_bar_lin_vel_W <= norm_foo_lin_vel_W + norm_foo_lin_vel_B

However I get the following values

norm_bar_lin_vel_W = 0.576
norm_foo_lin_vel_W = 1.290
norm_foo_lin_vel_B =  0.249

Could someone help me where I'm misunderstanding the velocities?
Any help is much appreciated

[yuvaltassa]

Your assumption is incorrect. Consider the following scenario

1. Body b is at the origin, spinning around the z axis at 1 rad/sec, no linear velocity.
2. Site s is at (10, 0, 0), fixed to the world.
3. The linear velocity of both b and s is 0 in the world frame. However in the (rotating) reference frame of the body, the fixed site appears to be moving with a linear velocity of 10 m/s.

Makes sense?