optimize my model #1578
optimize my model #1578
Conversation
tools/machine-learning/mujoco/learned-identification/pyproject.toml
Outdated
Show resolved
Hide resolved
tools/machine-learning/mujoco/learned-identification/scripts/render_best.py
Outdated
Show resolved
Hide resolved
|
|
||
| def simulate_recording( | ||
| spec: mj.MjSpec, | ||
| recorded_actuators: npt.NDArray[np.float64], |
There was a problem hiding this comment.
| recorded_actuators: npt.NDArray[np.float64], | |
| recorded_motor_commands: npt.NDArray[np.float64], |
I found it a little bit confusing. Would rename it in all places.
There was a problem hiding this comment.
actuator is the correct terminology in mujoco
There was a problem hiding this comment.
Yes, but it's just an actuator command, not a complete actuator. How about recorded_actuator_commands?
I would prefer Hulk's convention, which is motor, but I'm fine with Mujoco's terminology if you prefer that
There was a problem hiding this comment.
This is not a motor command in the HULKs nomenclature. We use this name to describe stiffnesses and positions combined, here, we are only talking about positions, in the space mujoco accepts for setting actuators.
There was a problem hiding this comment.
Yes, but I think that's still better than recorded_actuators, because that could also be some sensor value. I think it's not clear, that this is our control signal
tools/machine-learning/mujoco/learned-identification/src/learned_identification/actuator.py
Show resolved
Hide resolved
schmidma
force-pushed
the
optimize-my-model
branch
from
45aa0dd to
82d3751
Compare
Why? What?
Introduce optuna based actuator identification of the mujoco model.
ToDo / Known Issues
none
Ideas for Next Iterations (Not This PR)
many...
How to Test
uv run scripts/run.py --spec ../model/scene.xml --recording recording.mcap --study study01