feat(assets): local_jar_lid — the operator's own printed screw jar

#28
by jxxsteven7 - opened
.gitattributes CHANGED
@@ -889,3 +889,4 @@ assets/articulated/artvip/microwave_3/usd/model.usd filter=lfs diff=lfs merge=lf
889
  assets/articulated/artvip/rice_cooker_3/usd/model.usd filter=lfs diff=lfs merge=lfs -text
890
  assets/articulated/artvip/rice_cooker_7/usd/model.usd filter=lfs diff=lfs merge=lfs -text
891
  assets/articulated/artvip/scissors_22/usd/model.usd filter=lfs diff=lfs merge=lfs -text
 
 
889
  assets/articulated/artvip/rice_cooker_3/usd/model.usd filter=lfs diff=lfs merge=lfs -text
890
  assets/articulated/artvip/rice_cooker_7/usd/model.usd filter=lfs diff=lfs merge=lfs -text
891
  assets/articulated/artvip/scissors_22/usd/model.usd filter=lfs diff=lfs merge=lfs -text
892
+ assets/local_jar_lid/Jar__Lid_source.stl filter=lfs diff=lfs merge=lfs -text
assets/local_jar_lid/Jar__Lid_source.stl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:122c103674a6e61fe04bc0a48f5fd078747b3d3a2610b029aac97a8528963391
3
+ size 628284
assets/local_jar_lid/README.md ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # local_jar_lid — the operator's own 3D-printed screw jar
2
+
3
+ Project-authored screw asset (the `local_bottle_cap` / `local_juice_shot`
4
+ family), built from the operator's own print plate `Jar__Lid_source.stl`
5
+ (both parts on one plate, mm units).
6
+
7
+ ## Provenance / reproduction
8
+
9
+ - `Jar__Lid_source.stl` — the original upload, unmodified (12 564 triangles,
10
+ body at x ∈ [−45, 45], lid at x ∈ [77, 168] on the plate).
11
+ - Split by triangle-centroid x < 60 mm, each part recentred (xy on its AABB
12
+ centre, base at z = 0) and scaled mm → m into `jar_body_vis.obj` /
13
+ `jar_lid_vis.obj` (visual-only meshes; collision in the MJCF is procedural).
14
+ - **The thread is MEASURED from this mesh, not guessed**: fitting the helix
15
+ through the lid's inner-skirt crest vertices and the neck's outer crest
16
+ vertices independently, both converge on a **3-start thread, lead
17
+ 13.35 mm/turn** (phase concentration 0.73 / 0.65). Engagement spans ~7.5 mm
18
+ of skirt ⇒ the real lid disengages in ~0.56 turn.
19
+ - Geometry: body 90 × 90 × ~110 mm chamfered square + Ø77.4 × 11 mm neck
20
+ (total 121.83 mm); lid 90 × 90 × 33 mm, 15 mm skirt, 0.6 mm diametral
21
+ clearance.
22
+
23
+ ## Mechanism (`jar_lid.xml`)
24
+
25
+ `lid_slide` (prismatic, FIRST joint — parrot_sim overlay contract) +
26
+ `lid_twist` (coaxial hinge) coupled by a joint equality at the measured lead
27
+ (`polycoef[1] = 0.01335 / 2π`). Dry thread friction 0.055 N·m + 5 N skirt
28
+ friction (both `solreffriction`-stiffened) ⇒ straight-pull threshold ≈ 31 N;
29
+ `<contact><exclude>` between lid and jar (the equality IS the thread).
30
+
31
+ **ESTIMATED, pending measurement**: part masses (PLA; authored 0.20 kg jar /
32
+ 0.08 kg lid). Weigh the real print and correct.
33
+
34
+ ## USD
35
+
36
+ `usd/jar_lid_render.usda` — render-grade conversion via
37
+ `RoboVerse/scripts/mjcf_screwcap_to_usd.py`; the screw ratio rides the
38
+ prismatic joint as `wuji:screwPitchMPerRad` customData (UsdPhysics has no
39
+ screw-coupling schema; wiring a PhysX mimic drive is the Isaac-side follow-up).
40
+
41
+ ## Consumers
42
+
43
+ - task `hand_track.open_jar_lid`
44
+ (`RoboVerse/roboverse_pack/tasks/hand_track/hand_track_open_jar_lid.py`)
45
+ - physics regressions
46
+ `RoboVerse/roboverse_pack/tasks/hand_track/tests/test_jar_lid_asset.py` (8)
assets/local_jar_lid/jar_body_vis.obj ADDED
The diff for this file is too large to render. See raw diff
 
assets/local_jar_lid/jar_lid.xml ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <mujoco model="jar_lid">
2
+ <!--
3
+ 3D-printed square storage jar with a screw lid, built from the user's own
4
+ print (Jar__Lid.stl, one plate with both parts side by side; split, recentered
5
+ xy-on-centre / base-at-z=0, mm units -> scale 0.001).
6
+
7
+ THE THREAD IS MEASURED FROM THE MESH, not guessed (the open_juice_shot lesson:
8
+ model the LEAD, not the ridge pitch). Fitting the helix through the lid's
9
+ inner-skirt crest vertices and the neck's outer crest vertices independently:
10
+ both converge on a 3-START thread, joint best fit LEAD = 13.35 mm/turn
11
+ (phase concentration 0.73 / 0.65 on a 12.5k-triangle mesh). Thread engagement
12
+ spans ~7.5 mm of the skirt, so the real lid disengages in ~0.56 turn - inside
13
+ the one-wrist-rotation regime open_juice_shot proved collectable (its v1, at
14
+ 1.67 turns, failed 22 consecutive real takes).
15
+
16
+ Measured geometry: body 90 x 90 x ~110 mm square (chamfered corners, max
17
+ half-diagonal 61.15 mm), neck OD 77.4 mm x ~11 mm tall on top (total jar
18
+ 121.83 mm); lid 90 x 90 x 33 mm with a 15 mm-deep skirt, thread crests at
19
+ inner r ~38.1 mm (0.6 mm diametral clearance to the neck).
20
+
21
+ Mechanism (the local_bottle_cap v2 design): the lid rides a vertical slide
22
+ (lid_slide, FIRST joint on the body - parrot_sim's articulation_joint_frames
23
+ takes the first joint, and the overlay must draw the slide-travel gate) plus a
24
+ coaxial hinge (lid_twist), tied by a joint equality at the measured lead
25
+ (polycoef[1] = 0.01335 / 2pi = 0.00212473 m/rad). Dry friction on the twist
26
+ joint makes a straight pull back-drive the screw: F_pull = 2pi*frictionloss /
27
+ lead = 2pi*0.055/0.01335 ~ 25.9 N, ~9.4x the 0.28 kg assembly's weight, so a
28
+ yank lifts the whole jar instead of opening it. Breakaway torque 0.055 N*m is
29
+ ~11x under the torque a priority-2 grip delivers at the lid's 45 mm half-width
30
+ - the cure for "hard to turn" is grip friction, never a weaker thread.
31
+ The huge slide armature is the screw's reflected inertia, I_twist/lead_rad^2
32
+ ~ 1.1e-4 / (0.00212473^2) ~ 24 kg-eq. The equality's near-rigid solimp is the
33
+ juice_shot elastic-leak fix (a soft equality let a 20 N pull deflect 3.2 mm
34
+ WITHOUT the thread slipping). lid<->jar contacts are excluded: the equality IS
35
+ the thread, and MuJoCo's automatic parent-child filter is skipped when the
36
+ parent is welded to the world, i.e. exactly in fixed-base standalone loads
37
+ like the asset tests.
38
+
39
+ ESTIMATED (not measured): masses - PLA print, jar 0.20 kg / lid 0.08 kg;
40
+ weigh the real parts and correct. The closed lid sits at z = 110 mm (its
41
+ skirt over the neck, resting at the square body's top edge).
42
+
43
+ Collision is procedural (box body + cylinder neck; box lid), matching the
44
+ board_nail / bottle_cap philosophy; visuals are the real print meshes. The
45
+ box corners overstate the chamfered print by ~2.5 mm at the diagonals.
46
+ -->
47
+ <compiler angle="radian" meshdir="."/>
48
+ <asset>
49
+ <mesh name="jar_body_mesh" file="jar_body_vis.obj"/>
50
+ <mesh name="jar_lid_mesh" file="jar_lid_vis.obj"/>
51
+ </asset>
52
+ <worldbody>
53
+ <body name="jar">
54
+ <geom name="jar_visual" type="mesh" mesh="jar_body_mesh" group="2"
55
+ contype="0" conaffinity="0" density="0" rgba="0.96 0.95 0.92 1"/>
56
+ <!-- square body: 90 x 90 x 110 mm; grasp-priority friction (family value) -->
57
+ <geom name="jar_body_col" type="box" size="0.045 0.045 0.055" pos="0 0 0.055"
58
+ mass="0.20" friction="3.0 0.15 0.01" priority="2" rgba="0 1 0 0" group="3"/>
59
+ <!-- neck: OD 77.4 x 11 mm, z 110..121 -->
60
+ <geom name="jar_neck_col" type="cylinder" size="0.0387 0.0055" pos="0 0 0.1155"
61
+ mass="0.0" friction="3.0 0.15 0.01" priority="2" rgba="0 1 0 0" group="3"/>
62
+ <body name="lid" pos="0 0 0.110">
63
+ <!-- slide FIRST (overlay contract), then the coaxial twist -->
64
+ <!-- The slide carries its own dry friction: the 15 mm skirt rides the neck
65
+ with 0.6 mm diametral clearance, and that rubbing is real. It also
66
+ closes the last elastic leak: 20 N of pull reached the twist friction
67
+ only THROUGH the equality, whose regularization let 3.4 mm creep
68
+ past; 5 N of slide friction (with the same stiff solreffriction)
69
+ holds the direct path shut. Total pull threshold ~ 25.9 + 5 = 31 N. -->
70
+ <joint name="lid_slide" type="slide" axis="0 0 1" range="0 0.045"
71
+ armature="24" damping="0.5" frictionloss="5.0" solreffriction="0.004 1"/>
72
+ <!-- solreffriction stiffens the dry-friction CONSTRAINT itself: with the
73
+ default-soft value a constant load creeps through it regardless of the
74
+ frictionloss magnitude (measured here: a 20 N pull walked the slide
75
+ 4.9 mm in 3 s) - the hammer_insert_nail sag fix. -->
76
+ <joint name="lid_twist" type="hinge" axis="0 0 1" range="0 21.2"
77
+ frictionloss="0.055" damping="0.02" solreffriction="0.004 1"/>
78
+ <geom name="lid_visual" type="mesh" mesh="jar_lid_mesh" group="2"
79
+ contype="0" conaffinity="0" density="0" rgba="0.97 0.96 0.93 1"/>
80
+ <geom name="lid_col" type="box" size="0.045 0.045 0.0165" pos="0 0 0.0165"
81
+ mass="0.08" friction="3.0 0.15 0.01" priority="2" rgba="0 1 0 0" group="3"/>
82
+ </body>
83
+ </body>
84
+ </worldbody>
85
+ <equality>
86
+ <!-- slide = lead/(2pi) * twist ; lead = 13.35 mm/turn, 3-start, MEASURED -->
87
+ <joint joint1="lid_slide" joint2="lid_twist" polycoef="0 0.00212473 0 0 0"
88
+ solref="0.002 1" solimp="0.999 0.9999 0.00001 0.5 2"/>
89
+ </equality>
90
+ <contact>
91
+ <exclude body1="jar" body2="lid"/>
92
+ </contact>
93
+ </mujoco>
assets/local_jar_lid/jar_lid_vis.obj ADDED
The diff for this file is too large to render. See raw diff
 
assets/local_jar_lid/usd/jar_lid_render.usda ADDED
The diff for this file is too large to render. See raw diff