VBD contact damping: one-sided vs bidirectional

Seven Franka and Allegro cloth manipulation cases from the collision benchmark, each run twice with the same source, settings, seeds and hand commands. The only difference is the new SolverVBD option contact_damping_bidirectional: False (one-sided: damping only while a contact gap closes; the constructor default when these runs were made) versus True (bidirectional: damping also while the gap opens, with the matching Hessian; the constructor default on the branch since the afternoon of 2026-11-04).

Standard VBD elasticity, fast math off, rigid-soft DAT on, merged OGC edge rule, contact stiffness capped at 10x the membrane coefficient, 30 Hz frames, ten substeps (dt = 1/300 s), 60 s hold. Contact damping coefficient kd = 0.1 in every scene. Single runs; they do not establish repeatability.

Update, 2026-11-04 afternoon

Anka decided to make the bidirectional damping the default. The branch now carries the option (b989e846c), the default change (36590004e) and the rebase onto upstream main 7f0c13af6. Two more commits followed: the tuning-guide additions (81fd7250a) and an objective gap rate for the self-contact and rigid-rigid dampers (68c18328a), which the repo's rigid-motion test required once the two-sided gate became the default. The arm labels on this page were changed to match: One-sided (closing only; previous default) is the run with contact_damping_bidirectional=False, Bidirectional (new default) the run with True. The runs themselves were made at commit c49851e53, before these follow-up commits, and are unchanged. The poker-card results below were added in the same update.

Poker cards: the issue #3494 scene

The shipped example_cloth_poker_cards is the scene behind issue #3494 (the card stack never settles). Two measurements, both at commit c49851e53, with the same two damping modes.

1. The demo itself with the ball disabled (issue-3494 harness)

52 cards, 60 Hz, 20 substeps, 10 iterations, soft_contact_ke 1e5, soft_contact_kd 100, sphere velocity set to zero. Settle score = mean over frames 40 to 79 of the largest card-vertex speed; lower is more settled, 0.1 m/s was the threshold of the 2026-07 investigation. Two runs per mode.

Mode	Run	Settle score m/s	Mean kinetic energy J	Stack height swing m	Verdict
One-sided (closing only; previous default)	1	0.615	2.33e-3	0.0404	bouncy
One-sided (closing only; previous default)	2	0.630	2.38e-3	0.0379	bouncy
Bidirectional (new default)	1	0.0705	5.12e-5	0.0026	settled
Bidirectional (new default)	2	0.0704	4.84e-5	0.0023	settled

Reference values from the 2026-07 investigation with the same harness: Newton v1.3.0 0.0025 m/s, v1.4.0rc1 0.62 m/s, the fix proposed then 0.06 to 0.07 m/s. Re-run at the final commit 68c18328a (objective gap rate), one run per mode: one-sided 0.539 m/s, bidirectional 0.070 m/s (one-sided | bidirectional). Raw: default-run1.json | default-run2.json | bidirectional-run1.json | bidirectional-run2.json | analysis | schedule.

2. The benchmark's passive card scene s4_poker_cards_no_ball

Benchmark recipe: 60 Hz, ten substeps (dt 1/600 s), ten iterations, platform contact stiffness capped at 1e5, observation 5 to 65 s, no hold or release criteria. In both arms the deck spills off the platform within the first seconds and cards keep slipping later, as in the synced-main run. The J* and V* windows are dominated by those falls, so this scene does not separate the two modes.

Arm	J* mm	V* m/s	Windows	Overlap mm	Substep ms	Finite
One-sided (closing only; previous default)	16.36	0.421	120	0.782	4.53	yes
Bidirectional (new default)	15.61	0.422	120	0.864	6.19	yes
Synced-main campaign of 2026-11-04 (one-sided, different source revision)	19.33	0.402	n/a	0.800	n/a	n/a

One-sided (closing only; previous default)

J* 16.36 mm, V* 0.421 m/s, overlap 0.782 mm, solver 4.53 ms per substep, wall 239 s. Status completed, finite True, contact overflow False.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

J* 15.61 mm, V* 0.422 m/s, overlap 0.864 mm, solver 6.19 ms per substep, wall 289 s. Status completed, finite True, contact overflow False.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Overall

Arm	Completed runs	Retained through hold	Mean J* mm	Mean V* m/s	Mean static overlap mm	Mean solver substep ms
One-sided (closing only; previous default)	7/7	5/7	25.44	0.495	0.780	23.74
Bidirectional (new default)	7/7	5/7	26.64	0.497	1.093	18.78

Means are arithmetic means of per-case worst values over the 7 cases completed in both arms. J* is the largest cloth-vertex displacement RMS after removing a fitted linear drift over complete 0.5 s hold windows; V* is the largest cloth-vertex speed RMS over the same windows; the static overlap is the largest sampled overlap of the cloth particle shell with the declared table or ground. Hold retention uses the benchmark's conservative sampled whole-hand separation test: 'retained' means no sampled state proved full separation during the hold. The tables are measurements, not pass or fail verdicts.

Per case

Case	Robot	Rows	Arm	Hold	Release	J* mm	V* m/s	Windows	Overlap mm	Substep ms	Finite
m2_franka_pouch	Franka	512	One-sided (closing only; previous default)	separated at 4.5 s	hand clear at end (from 4.5 s)	16.96	0.695	120	0.998	57.66	yes
m2_franka_pouch	Franka	512	Bidirectional (new default)	separated at 42.0 s	hand clear at end (from 42.0 s)	28.34	0.532	120	0.998	23.29	yes
m3_franka_tote_empty	Franka	128	One-sided (closing only; previous default)	separated at 7.3 s	hand clear at end (from 7.3 s)	50.08	1.072	120	1.838	10.47	yes
m3_franka_tote_empty	Franka	128	Bidirectional (new default)	separated at 17.9 s	hand clear at end (from 17.9 s)	53.65	1.233	120	1.006	9.97	yes
r5_allegro_towel	Allegro	512	One-sided (closing only; previous default)	retained through hold	hand clear at end (from 68.6 s)	22.88	0.371	120	0.901	11.11	yes
r5_allegro_towel	Allegro	512	Bidirectional (new default)	retained through hold	hand clear at end (from 68.7 s)	21.61	0.372	120	2.938	11.20	yes
r5_franka_dress	Franka	512	One-sided (closing only; previous default)	retained through hold	hand clear at end (from 68.6 s)	24.55	0.344	120	0.209	27.20	yes
r5_franka_dress	Franka	512	Bidirectional (new default)	retained through hold	hand clear at end (from 69.4 s)	23.34	0.348	120	0.000	30.73	yes
r5_franka_patch_contact_v2	Franka	512	One-sided (closing only; previous default)	retained through hold	hand clear at end (from 68.4 s)	20.67	0.324	120	0.712	15.44	yes
r5_franka_patch_contact_v2	Franka	512	Bidirectional (new default)	retained through hold	hand clear at end (from 68.4 s)	21.16	0.334	120	2.710	20.44	yes
r5_franka_towel_contact_v2	Franka	512	One-sided (closing only; previous default)	retained through hold	hand clear at end (from 68.3 s)	24.45	0.359	120	0.799	16.31	yes
r5_franka_towel_contact_v2	Franka	512	Bidirectional (new default)	retained through hold	hand clear at end (from 69.0 s)	20.50	0.355	120	0.000	12.46	yes
r5_franka_trousers	Franka	64	One-sided (closing only; previous default)	retained through hold	hand clear at end (from 68.7 s)	18.52	0.301	120	0.000	27.98	yes
r5_franka_trousers	Franka	64	Bidirectional (new default)	retained through hold	hand clear at end (from 69.4 s)	17.85	0.301	120	0.000	23.38	yes

Rows: self-contact vertex and edge row storage per primitive. The towel cases stopped on the benchmark's 64-row capacity guard during release in one arm each (after a complete hold); the tables use their 512-row repeats in both arms, listed with every attempt below. The solver substep time is a secondary measure here: four cases ran at the same time on four GPUs of one host, and a dropped cloth changes the contact load.

All attempts

Case	Arm	Storage	Status	Stopped at s	Hold	J* mm	V* m/s	Windows
m2_franka_pouch	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	separated at 4.5 s	16.96	0.695	120
m2_franka_pouch	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	separated at 42.0 s	28.34	0.532	120
m3_franka_tote_empty	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	separated at 7.3 s	50.08	1.072	120
m3_franka_tote_empty	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	separated at 17.9 s	53.65	1.233	120
r5_allegro_towel	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	retained through hold	22.91	0.374	120
r5_allegro_towel	One-sided (closing only; previous default)	512-row repeat	completed	completed	retained through hold	22.88	0.371	120
r5_allegro_towel	Bidirectional (new default)	as selected by the synced-main campaign	error	67.90	run error	23.60	0.375	120
r5_allegro_towel	Bidirectional (new default)	512-row repeat	completed	completed	retained through hold	21.61	0.372	120
r5_franka_dress	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	retained through hold	24.55	0.344	120
r5_franka_dress	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	retained through hold	23.34	0.348	120
r5_franka_patch_contact_v2	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	retained through hold	20.67	0.324	120
r5_franka_patch_contact_v2	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	retained through hold	21.16	0.334	120
r5_franka_towel_contact_v2	One-sided (closing only; previous default)	as selected by the synced-main campaign	error	68.67	run error	24.11	0.358	120
r5_franka_towel_contact_v2	One-sided (closing only; previous default)	512-row repeat	completed	completed	retained through hold	24.45	0.359	120
r5_franka_towel_contact_v2	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	retained through hold	22.77	0.351	120
r5_franka_towel_contact_v2	Bidirectional (new default)	512-row repeat	completed	completed	retained through hold	20.50	0.355	120
r5_franka_trousers	One-sided (closing only; previous default)	as selected by the synced-main campaign	completed	completed	retained through hold	18.52	0.301	120
r5_franka_trousers	Bidirectional (new default)	as selected by the synced-main campaign	completed	completed	retained through hold	17.85	0.301	120

Context: the synced-main campaign (2026-11-02)

Same recipes, one-sided damping, upstream main 351b9d9f (43 commits before this branch's base), Warp 1.19.0+cu12. Shown only to judge run-to-run sensitivity; it is a different source revision.

Case	Hold	J* mm	V* m/s	Overlap mm	Substep ms
m2_franka_pouch	separated at 49.5 s	27.72	0.458	0.998	14.18
m3_franka_tote_empty	separated at 13.0 s	53.47	1.330	2.683	11.68
r5_allegro_towel	retained through hold	23.05	0.369	0.892	11.00
r5_franka_dress	retained through hold	25.84	0.383	0.098	22.34
r5_franka_patch_contact_v2	retained through hold	20.47	0.328	0.822	12.30
r5_franka_towel_contact_v2	retained through hold	24.67	0.356	0.592	11.64
r5_franka_trousers	retained through hold	18.21	0.301	0.277	19.17

m2_franka_pouch (Franka)

M2 | Franka hand and pouch: prepared rim hold, carry, release

One-sided (closing only; previous default)

Hold: separated at 4.5 s. Release: hand clear at end (from 4.5 s). J* 16.96 mm, V* 0.695 m/s, overlap 0.998 mm, solver 57.66 ms per substep, wall 1266 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: separated at 42.0 s. Release: hand clear at end (from 42.0 s). J* 28.34 mm, V* 0.532 m/s, overlap 0.998 mm, solver 23.29 ms per substep, wall 533 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 40,
 "gravity_m_s2": [
  0.0,
  0.0,
  -9.81
 ],
 "density_kg_m2": 0.22,
 "mass_kg": 0.01537800021469593,
 "tri_ke": 3000.0,
 "tri_ka": 3000.0,
 "tri_kd": 0.025,
 "edge_ke": 0.025,
 "edge_kd": 0.0005,
 "particle_radius_m": 0.001,
 "effective_shell_thickness_m": 0.002,
 "self_contact_enabled": true,
 "self_contact_interaction_margin_m": 0.002,
 "self_contact_extra_detection_gap_m": 0.0015,
 "topological_contact_filter_rings": 2,
 "rest_shape_contact_exclusion_radius_m": 0.0,
 "soft_contact_ke": 50000.0,
 "soft_contact_kd": 0.1,
 "soft_contact_mu": 1.5,
 "robot_shape_mu": 1.5,
 "table_shape_mu": 0.6,
 "rigid_shape_margin_m": 0.0,
 "robot_shape_gap_m": 0.002,
 "ground_shape_gap_m": 0.002,
 "table_shape_gap_m": 0.1,
 "rigid_soft_extra_detection_gap_m": 0.003,
 "rigid_soft_full_surface_contact": true,
 "friction_epsilon": 0.001,
 "mesh_dimensions_xyz_m": [
  0.15,
  0.018,
  0.2
 ],
 "mesh_cells_xyz": [
  24,
  4,
  24
 ],
 "mesh_vertex_count": 1469,
 "mesh_triangle_count": 2880,
 "nominal_closed_finger_opening_m": 0.0015,
 "robot_pose_integration": "Input FK at t; output FK at t+substep_dt"
}

m3_franka_tote_empty (Franka)

M3 | Franka hands: tote | 0 payloads

One-sided (closing only; previous default)

Hold: separated at 7.3 s. Release: hand clear at end (from 7.3 s). J* 50.08 mm, V* 1.072 m/s, overlap 1.838 mm, solver 10.47 ms per substep, wall 268 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: separated at 17.9 s. Release: hand clear at end (from 17.9 s). J* 53.65 mm, V* 1.233 m/s, overlap 1.006 mm, solver 9.97 ms per substep, wall 257 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 40,
 "density_kg_m2": 0.22,
 "tri_ke": 3000.0,
 "tri_ka": 3000.0,
 "tri_kd": 0.025,
 "edge_ke": 0.025,
 "edge_kd": 0.0005,
 "particle_radius_m": 0.003,
 "payload_density_kg_m3": 180.0,
 "payload_k_mu": 8000.0,
 "payload_k_lambda": 8000.0,
 "payload_k_damp": 0.5,
 "payload_particle_radius_m": 0.0025,
 "self_contact_enabled": true,
 "self_contact_interaction_margin_m": 0.005,
 "self_contact_extra_detection_gap_m": 0.005,
 "topological_contact_filter_rings": 2,
 "rest_shape_contact_exclusion_radius_m": 0.0,
 "soft_contact_ke": 50000.0,
 "soft_contact_kd": 0.1,
 "soft_contact_mu": 1.5,
 "robot_shape_mu": 1.5,
 "table_shape_mu": 0.6,
 "robot_shape_gap_m": 0.002,
 "table_shape_gap_m": 0.1,
 "rigid_shape_margin_m": 0.0,
 "rigid_soft_extra_detection_gap_m": 0.01,
 "rigid_soft_full_surface_contact": true,
 "nominal_closed_finger_opening_m": 0.004,
 "robot_pose_integration": "External rigid VBD: input FK at t; output FK at t+dt"
}

r5_allegro_towel (Allegro)

R5 | Allegro: original-size towel

One-sided (closing only; previous default)

Hold: retained through hold. Release: hand clear at end (from 68.6 s). J* 22.88 mm, V* 0.371 m/s, overlap 0.901 mm, solver 11.11 ms per substep, wall 301 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: retained through hold. Release: hand clear at end (from 68.7 s). J* 21.61 mm, V* 0.372 m/s, overlap 2.938 mm, solver 11.20 ms per substep, wall 302 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 30,
 "density_kg_m2": 0.22,
 "membrane_stretch_coefficient_N_m": 1800.0,
 "membrane_area_coefficient_N_m": 1800.0,
 "membrane_damping": 0.025,
 "bending_coefficient_N": 0.08,
 "bending_damping": 0.0005,
 "bending_rest_angles": "unchanged original source",
 "particle_radius_m": 0.003,
 "self_contact_enabled": true,
 "self_contact_margin_m": 0.002,
 "self_contact_gap_m": 0.001,
 "self_contact_rest_exclusion_m": 0.005,
 "self_contact_topological_rings": 1,
 "friction_robot": 1.0,
 "friction_table": 1.0,
 "friction_soft_self": 1.0,
 "contact_stiffness_N_m": 50000.0,
 "contact_damping": 0.1,
 "shape_margin_m": 0.001,
 "rigid_soft_detection_gap_m": 0.005,
 "full_surface_rigid_soft_contact": true,
 "mesh_vertices": 2180,
 "mesh_triangles": 4228,
 "original_scene_mesh_scale": 1.0
}

r5_franka_dress (Franka)

R5 | Franka: original-size dress

One-sided (closing only; previous default)

Hold: retained through hold. Release: hand clear at end (from 68.6 s). J* 24.55 mm, V* 0.344 m/s, overlap 0.209 mm, solver 27.20 ms per substep, wall 701 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: retained through hold. Release: hand clear at end (from 69.4 s). J* 23.34 mm, V* 0.348 m/s, overlap 0.000 mm, solver 30.73 ms per substep, wall 756 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 30,
 "density_kg_m2": 0.22,
 "membrane_stretch_coefficient_N_m": 1800.0,
 "membrane_area_coefficient_N_m": 1800.0,
 "membrane_damping": 0.025,
 "bending_coefficient_N": 0.08,
 "bending_damping": 0.0005,
 "bending_rest_angles": "unchanged original source",
 "particle_radius_m": 0.003,
 "self_contact_enabled": true,
 "self_contact_margin_m": 0.002,
 "self_contact_gap_m": 0.001,
 "self_contact_rest_exclusion_m": 0.005,
 "self_contact_topological_rings": 1,
 "friction_robot": 1.0,
 "friction_table": 1.0,
 "friction_soft_self": 1.0,
 "contact_stiffness_N_m": 50000.0,
 "contact_damping": 0.1,
 "shape_margin_m": 0.001,
 "rigid_soft_detection_gap_m": 0.005,
 "full_surface_rigid_soft_contact": true,
 "mesh_vertices": 4602,
 "mesh_triangles": 8972,
 "original_scene_mesh_scale": 0.6000000238418579
}

r5_franka_patch_contact_v2 (Franka)

R5 | Franka: original-size patch

One-sided (closing only; previous default)

Hold: retained through hold. Release: hand clear at end (from 68.4 s). J* 20.67 mm, V* 0.324 m/s, overlap 0.712 mm, solver 15.44 ms per substep, wall 435 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: retained through hold. Release: hand clear at end (from 68.4 s). J* 21.16 mm, V* 0.334 m/s, overlap 2.710 mm, solver 20.44 ms per substep, wall 532 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 30,
 "density_kg_m2": 0.22,
 "membrane_stretch_coefficient_N_m": 1800.0,
 "membrane_area_coefficient_N_m": 1800.0,
 "membrane_damping": 0.025,
 "bending_coefficient_N": 0.08,
 "bending_damping": 0.0005,
 "bending_rest_angles": "unchanged original source",
 "particle_radius_m": 0.003,
 "self_contact_enabled": true,
 "self_contact_margin_m": 0.002,
 "self_contact_gap_m": 0.001,
 "self_contact_rest_exclusion_m": 0.005,
 "self_contact_topological_rings": 1,
 "friction_robot": 1.0,
 "friction_table": 1.0,
 "friction_soft_self": 1.0,
 "contact_stiffness_N_m": 50000.0,
 "contact_damping": 0.1,
 "shape_margin_m": 0.001,
 "rigid_soft_detection_gap_m": 0.005,
 "full_surface_rigid_soft_contact": true,
 "mesh_vertices": 3205,
 "mesh_triangles": 6244,
 "original_scene_mesh_scale": 6.0
}

r5_franka_towel_contact_v2 (Franka)

R5 | Franka: original-size towel

One-sided (closing only; previous default)

Hold: retained through hold. Release: hand clear at end (from 68.3 s). J* 24.45 mm, V* 0.359 m/s, overlap 0.799 mm, solver 16.31 ms per substep, wall 421 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: retained through hold. Release: hand clear at end (from 69.0 s). J* 20.50 mm, V* 0.355 m/s, overlap 0.000 mm, solver 12.46 ms per substep, wall 331 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 30,
 "density_kg_m2": 0.22,
 "membrane_stretch_coefficient_N_m": 1800.0,
 "membrane_area_coefficient_N_m": 1800.0,
 "membrane_damping": 0.025,
 "bending_coefficient_N": 0.08,
 "bending_damping": 0.0005,
 "bending_rest_angles": "unchanged original source",
 "particle_radius_m": 0.003,
 "self_contact_enabled": true,
 "self_contact_margin_m": 0.002,
 "self_contact_gap_m": 0.001,
 "self_contact_rest_exclusion_m": 0.005,
 "self_contact_topological_rings": 1,
 "friction_robot": 1.0,
 "friction_table": 1.0,
 "friction_soft_self": 1.0,
 "contact_stiffness_N_m": 50000.0,
 "contact_damping": 0.1,
 "shape_margin_m": 0.001,
 "rigid_soft_detection_gap_m": 0.005,
 "full_surface_rigid_soft_contact": true,
 "mesh_vertices": 2767,
 "mesh_triangles": 5384,
 "original_scene_mesh_scale": 1.0
}

r5_franka_trousers (Franka)

R5 | Franka: original-size trousers

One-sided (closing only; previous default)

Hold: retained through hold. Release: hand clear at end (from 68.7 s). J* 18.52 mm, V* 0.301 m/s, overlap 0.000 mm, solver 27.98 ms per substep, wall 723 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Bidirectional (new default)

Hold: retained through hold. Release: hand clear at end (from 69.4 s). J* 17.85 mm, V* 0.301 m/s, overlap 0.000 mm, solver 23.38 ms per substep, wall 596 s. Status completed, finite True, contact overflow False, source guard unchanged True.

Contact sheet

Raw: results.json | task-observations.json | dat-ogc-provenance.json

Scene physics parameters

{
 "frame_dt_s": 0.03333333333333333,
 "substeps": 10,
 "substep_dt_s": 0.003333333333333333,
 "iterations": 30,
 "density_kg_m2": 0.22,
 "membrane_stretch_coefficient_N_m": 1800.0,
 "membrane_area_coefficient_N_m": 1800.0,
 "membrane_damping": 0.025,
 "bending_coefficient_N": 0.08,
 "bending_damping": 0.0005,
 "bending_rest_angles": "unchanged original source",
 "particle_radius_m": 0.003,
 "self_contact_enabled": true,
 "self_contact_margin_m": 0.002,
 "self_contact_gap_m": 0.001,
 "self_contact_rest_exclusion_m": 0.005,
 "self_contact_topological_rings": 1,
 "friction_robot": 1.0,
 "friction_table": 1.0,
 "friction_soft_self": 1.0,
 "contact_stiffness_N_m": 50000.0,
 "contact_damping": 0.1,
 "shape_margin_m": 0.001,
 "rigid_soft_detection_gap_m": 0.005,
 "full_surface_rigid_soft_contact": true,
 "mesh_vertices": 4570,
 "mesh_triangles": 8990,
 "original_scene_mesh_scale": 0.7058823704719543
}