[WIP] Model Updates (Contributed by Muscle Mimic Team)#73
Conversation
…dd a detailed change log
- Corrected left-right limb asymmetries in joint definitions - Fixed muscle-jumping behaviors at joint limits - Added collision pairs for leg-leg and foot-foot self-contact
There was a problem hiding this comment.
in other models e.g. arm, myolegs_tendon.xml was integrated into myolegs_assets.xml
There was a problem hiding this comment.
It's helpful to have it separated incase one wants to create a simpler model without the tendons.
mjSpec can definitely handle such usecases. It should be handled in a separate scope, if relevant.
There was a problem hiding this comment.
it makes sense. We should have this for other models too
| <worldbody> | ||
| <body name="Full Body" pos="-.025 0.1 1"> | ||
| <include file="../torso/assets/myotorso_chain.xml"/> | ||
| <body name="Full Body" pos="-.025 0.1 1" euler="0 0 -1.57"> |
There was a problem hiding this comment.
wouldn't this be problematic for the initial condition (assuming the previous one was the correct one)?
There was a problem hiding this comment.
The previous one was in-consistent with rest of our models.
Update torso and leg geom color created myofulbody.
…etection. Updated euler angles for left arm components and removed deprecated myobody and myotorso_bimanual_chain files. Enhanced myotorso_arm_chain with new collision geometries for self-contact.
…in myofullbody.xml for improved model alignment.
…aints and improve moment arm and force-length curve calculations. Updated XML asset files for leg model adjustments and corrected joint axis orientations.
…prove consistency in naming conventions. Adjusted euler angles for various body components to ensure accurate positioning. Enhanced leg and torso collision geometry for better interaction in simulations.
|
While reviewing the contribution from the musclemimic team, we observed that the proposed changes go well beyond incremental fixes and introduce a number of structural and numerical differences relative to the current Some of the differences identified include:
These are not isolated parameter changes; they affect model behavior, numerical equivalence, and backward compatibility with the current Given this, I think we effectively have two paths forward:
At the moment I lean toward option 2, mainly because it provides a cleaner long-term direction and avoids accumulating compatibility debt. That said, there may still be additional considerations from other contributors/users that we should take into account before deciding on the migration strategy. Please, let me know your thoughts @vikashplus @cherylwang20 @CharlieLeee |
|
Duplicate and diverging models will be very hard to jointly develop as well as maintain. The need of a more structured organization has been felt for a while. We should be forward looking and make fast progress. Incase old references are needed, we do have previous tagged versions for users to resort to. Having said that, there is merit in taking some time to carefully plan for the upgrade. Modeling editing and compositions can become complex. Additionally, there are myo_sim, myo_skeleton, muscle_mimic. Getting to a common base will be very important to avoid divergence. |
sg. Is there any concern about this current design? #75 |
3 participants
NOTE: This PR is a WIP. [Accepting contribution/ issues/ feedback on all fronts]
AUTHORS: Bianca Ziliotto, Chengkun Li, Huiyi Wang
This update, contributed by the muscle mimic team, brings in important functional updates to myo_sim model. Notably,
ToDos