Enhance ABA-based FD computation by adding motor dynamics #45
Conversation
|
Do you have some reference for the math implemented? In particular, I was a bit mislead by the title "adding motor dynamics", as in the code it does not seem that any new state is added to the simulated dynamic system. |
|
The problem formalization and the related pseudocode can be found at https://github.com/ami-iit/element_rl-for-codesign/issues/61#issuecomment-1658609058. The title refers to the fact that this ABA implementation takes into account motor inertias, viscous friction and transmission ratio. I can change the title to "Enhance ABA-based FD computation by considering motor parameters" if that makes things more clear |
No problem, once you clarify it in the description I think it is ok! |
flferretti
force-pushed
the
motor_dynamics
branch
2 times, most recently
from
3836a3c
to
a43274e
Compare
This comment was marked as outdated.
This comment was marked as outdated.
flferretti
force-pushed
the
motor_dynamics
branch
2 times, most recently
from
b2e7343
to
551fc7d
Compare
|
xref of this interesting comment full of references ROAM-Lab-ND/spatial_v2_extended#6 (comment), we might have a deeper look at it (original issue ROAM-Lab-ND/spatial_v2_extended#5). |
flferretti
force-pushed
the
motor_dynamics
branch
2 times, most recently
from
c87ca07
to
b0657e7
Compare
flferretti
force-pushed
the
motor_dynamics
branch
from
ec3d4a1
to
5078b36
Compare
|
Changing the base branch to |
flferretti
force-pushed
the
motor_dynamics
branch
2 times, most recently
from
3ebcaad
to
fb61a64
Compare
flferretti
force-pushed
the
motor_dynamics
branch
from
fb61a64
to
a974ea5
Compare
flferretti
force-pushed
the
motor_dynamics
branch
from
8ab569d
to
408c9e1
Compare
flferretti
force-pushed
the
motor_dynamics
branch
from
a466981
to
1009885
Compare
There was a problem hiding this comment.
First round of review, it looks good even if I still have some doubts on the updated ABA and RNEA algorithms. If we want to merge into main and the new methods are not yet 100% tested and benchmarked, let's consider to commit them for the time being as aba_motors.py and rnea_motors.py. We can fix the remaining problems in the future without affecting current functionality.
| vJ = S[i] * qd[ii] * (qd.size != 0) | ||
| m_vJ = m_S[i] * qd[ii] * (qd.size != 0) |
There was a problem hiding this comment.
Did you validate jit compilation? I remember I had to use the if/else statement in this way otherwise jax was complaining
There was a problem hiding this comment.
Yes, it did not complain, but we can further investigate on this
|
Also this has been closed once merged |
This pull request introduces the possibility for the simulator to accept motor inertia, motor viscous friction and gear ratio, which will be saved in the already existing JAX dataclasses:
high_level.jointhigh_level.modelmodel.physics_modelto be ready for Forward Dynamics computation through a new ABA version which incorporates these motor parameters into the simulation process.
As of now, the PR is marked as a draft since even though it successfully passes the tests against iDynTree, it requires further validation with comprehensive motor dynamics to ensure its robustness and accuracy.
C.C. @diegoferigo @CarlottaSartore @S-Dafarra @traversaro