Fix taxel centers and normals #6
Conversation
L2 and L1 taxels had wrong origin and scale/rot not applied causing the pos/ori to be wrong. L3 pos/ort were already correct. Additionally all taxels were cleaned to be only faces, not closed cuboids that had too many vertices and normals, unwanted for the pos/ori computation
|
@rhaschke This is not ready for a PR yet. We wanted to discuss the need or not to change also the taxel center with projection or not. Hence me not creating a PR |
model/tactile_markers.urdf.xacro
Outdated
| <!--tmm does not exist for L1 thumb, but can be tmo which is larger and contains tmm --> | ||
| <xacro:if value="${layout=='L1' and finger=='thumb'}"> | ||
| <xacro:taxel mesh="${F}mm" idx="${tactMap.get(F+'mo', -1)}"/> | ||
| </xacro:if> | ||
| <xacro:taxel mesh="${F}mm" idx="${tactMap.get(F+'mm', -1)}"/> |
There was a problem hiding this comment.
| <!--tmm does not exist for L1 thumb, but can be tmo which is larger and contains tmm --> | |
| <xacro:if value="${layout=='L1' and finger=='thumb'}"> | |
| <xacro:taxel mesh="${F}mm" idx="${tactMap.get(F+'mo', -1)}"/> | |
| </xacro:if> | |
| <xacro:taxel mesh="${F}mm" idx="${tactMap.get(F+'mm', -1)}"/> | |
| <!--tmm does not exist for L1, but is covered by tmo as well --> | |
| <xacro:taxel mesh="${F}mm" idx="${tactMap.get(F+'mm', tactMap.get(F+'mo', -1) if layout=='L1' else -1)}"/> |
There was a problem hiding this comment.
none of the text are clear in fact. and I don't find how to do another suggestion or reject this one...
There was a problem hiding this comment.
it took me some time to ensure this optimization of the code is doing the same due mentally needing to evaluate the code to understand what it does, while the if case is straight forward because it explicitly refers to the thumb.
With the optimization we have to access an external document, to verify that there is no other ${F}mm that is not missing in another finger to be sure it works only for the thumb. For me, this opacifies the understanding for future developments, and would for sure be missed in case one new layout has a pinky tmm missing or so.
I confirm it should do the job for now, but don't like it.
rhaschke
changed the title
|
Open issues:
|
|
Latest commit should fix the center and normals for all left/right L2/L3. See attached displays |
|
Otherwise looks good to me. |
Some rot/scale were still not applied on the L2right model as well as 2 origins + rot/scale on L3right. L2left had all normals flipped and outside recomputation not active in the script.
rhaschke
force-pushed
the
TaxelCenterFix
branch
from
15906fa to
993af0c
Compare
rhaschke
changed the title
L2 and L1 taxels had a wrong origin and scale/rot not applied causing the positions/normals to be wrong. Additionally, all taxels were cleaned to be only faces, not closed cuboids. The latter had vertices and normals which shouldn't be considered for pos/normal computation.
Centers and normals were off in L2 layout.