ForceBody: Force-Paired Parametric Body Motion with Torque Uncertainty
Organizations: Michigan State University · Independent Researcher · University of North Carolina at Chapel Hill
Abstract
Forces and torques drive human motion, central to biomechanics, robotics, and physics-aware animation, yet they are often absent from learning-based motion pipelines for body reconstruction and generation. Some motion datasets provide force annotations, but the paired motion sequences are incompatible with the parametric body models used in modern learning pipelines. In addition, existing datasets typically obtain joint torques through inverse dynamics, which is inherently sensitive to the quality of the input kinematics; however, they do not quantify the reliability of these torque estimates. We close this gap with ForceBody, the first dataset to bring ground truth force supervision into a parametric body representation, shipped with per-sample torque uncertainty. ForceBody pairs the SKEL body model with ground reaction forces and inverse-dynamics joint torques across 10,386 motion trials (9.7M frames, 26.9 hours) from 140 subjects. Each torque label is paired with a per-frame, per-joint uncertainty annotation obtained by Monte Carlo sampling through the inverse dynamics pipeline. We benchmark six architectures (MLP, Conv1D, LSTM, GRU, Mamba2, Transformer) on ground reaction force and joint torque prediction from SKEL motion. As one example use of the released uncertainty, finetuning a Transformer with per-sample uncertainty weighting reduces the MAE of joint torque estimation by 5%. Code and data are at https://github.com/DominoAI-Lab/ForceBody-NeurIPS-ED-2026.
Figures & tables
| Dataset | Parametric | Force Supervision | Scale | |||
| Body Model | GRF source | Torque | Uncertainty of Torque | Subjects | Hours | |
| UnderPressure [ 24 ] | ✗ | Insole ∗ | ✗ | ✗ | 10 | 5.5 |
| VidSole [ 14 ] | ✗ | Insole+Force plate | ✓ † | ✗ | 52 | 6.6 |
| GroundLink [ 8 ] | SMPL-X | Force plate | ✗ | ✗ | 7 | 1.5 |
| ImDy [ 18 ] | SMPL | Simulated | ✓ | ✗ | 435 | 152.3 |
| AddBiomechanics [ 42 ] | ✗ | Force plate | ✓ | ✗ | 273 | 57.6 |
| Dataset | Body Model | Architecture | GRF (N/kg, ) | GRM (Nm/kg, ) |
| AddB | Rajagopal [ 30 ] | MLP | 2.091 | 0.278 |
| GLN | 1.712 | 0.275 | ||
| ForceBody | SKEL [ 15 ] | MLP | 1.165 0.011 | 0.168 0.009 |
| GLN | 1.136 0.040 | 0.149 0.011 |
| MLP | Conv1D | LSTM [ 10 ] | |||||||
| R | M | P | R | M | P | R | M | P | |
| Upper mean | 0.008 0.000 | 0.004 0.000 | 0.644 0.002 | 0.007 0.000 | 0.004 0.000 | 0.700 0.003 | 0.007 0.000 | 0.004 0.000 | 0.691 0.004 |
| Lower mean | 0.241 0.006 | 0.168 0.004 | 0.833 0.003 | 0.210 0.003 | 0.141 0.002 | 0.870 0.003 | 0.214 0.004 | 0.145 0.002 | 0.864 0.003 |
| All joints | 0.148 0.003 | 0.103 0.002 | 0.757 0.002 | 0.129 0.002 | 0.086 0.001 | 0.802 0.002 | 0.131 0.002 | 0.088 0.001 | 0.795 0.003 |
| Without Uncertainty | With Uncertainty | |||||
| R | M | P | R | M | P | |
| Upper mean | 0.008 0.000 | 0.005 0.000 | 0.651 0.014 | 0.008 0.000 ( ) | 0.004 0.000 ( ) | 0.666 0.014 ( ) |
| Lower mean | 0.253 0.015 | 0.176 0.011 | 0.833 0.007 | 0.243 0.012 ( ) | 0.166 0.011 ( ) | 0.842 0.011 ( ) |
| All joints | 0.155 0.009 | 0.107 0.007 | 0.760 0.010 | 0.149 0.007 ( ) | 0.101 0.007 ( ) | 0.772 0.012 ( ) |
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
| Source | Trials | Subj. | M/F/U | Height (m) | Mass (kg) | Activity |
| Carter2023 [ 1 ] | 25/25/0 | Running / walk | ||||
| vanderZee2022 [ 39 ] | 0/0/10 | Walking | ||||
| Moore2015 [ 23 ] | 8/4/0 | Walking | ||||
| Tiziana2019 [ 16 ] | 14/11/0 | Walking | ||||
| Han2023 [ 8 ] | 3/4/0 | Mixed | ||||
| Uhlrich2023 [ 38 ] | 4/6/0 | Walk / jump |
| Term | Symbol | Weight |
| Joint position | ||
| Bone length | ||
| Shoulder width | ||
| Foot height | ||
| Marker | ||
| Pose reg. |
| Source | Trials | Sigma-Med [Nm/kg] | Rel-Med |
| Hamner2013 | |||
| Carter2023 | |||
| Tiziana2019 | |||
| Li2021 | |||
| Uhlrich2023 | |||
| vanderZee2022 |
| Joint | Run | Jump | Sit-to-stand | Others | Walk | Stair | Standing |
| Hip | |||||||
| Knee | |||||||
| Ankle | |||||||
| MTP | |||||||
| Elbow | |||||||
| Radioulnar |
| Positional Err. | Physical-plausibility Err. | |||||
| MPJPE | PA-MPJPE | Vel. err. | Acc. err. | Foot slid. | Ground pen. | |
| [mm] | [mm] | [mm/f] | [mm/f 2 ] | [mm] | [mm] | |
| ForceBody | ||||||
| Metric | Mean | Median | P | P |
| MPJPE (mm) | ||||
| PA-MPJPE (mm) | ||||
| Velocity error (mm/f) | ||||
| Acceleration error (mm/f 2 ) | ||||
| Foot sliding (mm) | ||||
| Ground penetration (mm) |
| Joint | MLP | Conv1D | LSTM | GRU | Mamba2 | Transformer | ||||||||||||
| R | M | P | R | M | P | R | M | P | R | M | P | R | M | P | R | M | P | |
| Upper body | ||||||||||||||||||
| elbow | 0.022 | 0.013 | 0.842 | 0.020 | 0.010 | 0.867 | 0.020 | 0.010 | 0.867 | 0.019 | 0.010 | 0.871 | 0.019 | 0.010 | 0.875 | 0.023 | 0.014 | 0.830 |
| pro_sup | 0.003 | 0.001 | 0.253 | 0.003 | 0.001 | 0.371 | 0.003 | 0.001 | 0.342 | 0.003 | 0.001 | 0.354 | 0.003 | 0.001 | 0.385 | 0.003 | 0.001 | 0.315 |
| wrist_flex | 0.003 | 0.002 | 0.792 | 0.003 | 0.001 | 0.814 | 0.003 | 0.001 | 0.809 | 0.003 | 0.001 | 0.816 | 0.002 | 0.001 | 0.826 | 0.003 | 0.002 | 0.767 |
| wrist_dev | 0.003 | 0.002 | 0.689 | 0.003 | 0.002 | 0.746 | 0.003 | 0.002 | 0.748 | 0.003 | 0.001 | 0.764 | 0.003 | 0.001 | 0.772 | 0.003 | 0.002 | 0.692 |
| Condition | Inputs | Upper mean | Lower mean | All joints | ||||||||
| poses | shape | motion | R | M | P | R | M | P | R | M | P | |
| (A) poses only | ✓ | 0.008 | 0.005 | 0.568 | 0.260 | 0.174 | 0.746 | 0.159 | 0.106 | 0.675 | ||
| (B) +parametric body | ✓ | ✓ | 0.009 | 0.005 | 0.546 | 0.267 | 0.178 | 0.745 | 0.163 | 0.109 | 0.666 | |
| (C) +motion (full) | ✓ | ✓ | ✓ | 0.008 | 0.005 | 0.651 | 0.253 | 0.176 | 0.833 | 0.155 | 0.107 | 0.760 |