[task-se3-equality] Jdot_qdot term about tsid HOT 5 CLOSED

Hi @PedroDesRobots ,
m_drift contains exactly Jdot_q_dot. Does that answer your question?
Andrea

from tsid.

Thank you for your prompt answer.
I was thinking that m_drift was related to the cartesian acceleration.
To be sure, m_drift is computed with the function frameClassicAcceleration, which call

      const Frame & f = m_model.frames[index];
      frameAcceleration = f.placement.actInv(data.a[f.parent]);
      Motion v = f.placement.actInv(data.v[f.parent]);
      frameAcceleration.linear() += v.angular().cross(v.linear());

Are those lines computed the Jdot_qdot? Do you know some literature where I could learn more about that?

from tsid.

This is indeed a trick that's not very well documented in the literature, but it's very simple to explain. We know that the Cartesian acceleration ddx is related to the joint accelerations via the equation:

ddx = J*ddq + dJ*dq

We also know that we can compute ddx using the forward phase of the Recursive Newton Euler algorithm (RNEA). So, rather than trying to compute dJ, and then multiplying it times dq, we could directly compute ddx using RNEA setting ddq=0. Is that clear now?

from tsid.

Perfect, thank you very much for taking the time to explain this trick!

from tsid.

No problem!

from tsid.