Physics-Informed Causal Reasoning for Nonstationary Robot Dynamics
Organizations: Department of Electrical and Computer Engineering, Seoul National University, Seoul 08826, Korea · Graduate School of AI, Seoul National University, Seoul 08826, Korea
Abstract
Reasoning about what an action will cause requires internalizing how a system behaves. As conditions seldom stay fixed, this must often happen within a few interactions. Yet systems that look identical in a snapshot can respond differently to the same action, depending on physical conditions the snapshot hides. Context-based models capture this separation by summarizing recent transitions into a latent context, but do not specify how that context should change the equations of motion. Physics-informed neural networks build those equations into the model, but describe only a single system and cannot recognize when it has changed. We present Physics-Informed Causal Reasoning (PICR), which bridges the two. Only a context, re-inferred from recent interactions, is updated; it conditions a fixed physics-informed network that determines the robot's inertia and potential energy, so the robot adapts to new physical conditions from its own experience without retraining or a hand-specified robot model. On a Franka Panda arm in NVIDIA Isaac Sim, PICR predicts motion under changing and unseen payloads, reducing error on the latter by 1.6-3.8x over baselines, and extends to counterfactual queries under alternative torques. Our analysis shows that the context, inferred solely from interaction, encodes payload physical properties, tracks their changes, and governs the arm's response. The gravity and inertia terms induced by the context further demonstrate its physical validity, each agreeing with the simulator at the level of individual Lagrangian components.
Figures & tables
| Metric | Setting | PICR | RWM | CaDM | CaDeLaC | DeLaN | Neural-Fly |
|---|---|---|---|---|---|---|---|
| ADD (cm) | Seen (0.1–2 kg) | 4.93 | 7.45 | 6.06 | 8.75 | 6.96 | 8.05 |
| Unseen (2–4 kg) | 6.76 | 11.15 | 11.40 | 21.41 | 25.78 | 22.27 | |
| Gain | Joints 1–3 (mean) | 0.90 | 0.75 | 0.65 | 0.17 | 0.35 | 0.84 |
| RMSE (N m) | Gravity | 2.15 | n/a | n/a | 11.29 | 9.99 | n/a |
| Inertial | 2.31 | n/a | n/a | 12.49 | 5.83 | n/a | |
| Coriolis | 1.59 | n/a | n/a | 2.15 | 1.67 | n/a |
Appendix figures & tables7 assets
Supplementary material from the paper’s appendix.
Appendix
| Method | Context from history | Output | Parameters | Optimizer |
|---|---|---|---|---|
| PICR | MLP encoder, , FiLM | (Lagrangian) | 6,545,957 | Adam , cosine |
| RWM | GRU state, 256-d | next state | 956,184 | Adam , clip 1 |
| CaDM | MLP encoder, | state change | 231,008 † | Adam |
| CaDeLaC | LSTM encoder, , concat | (Lagrangian) | 6,719 | Adam , clip 1 |
| DeLaN | none | (Lagrangian) | 12,125 | Adam |
| Neural-Fly | least-squares weights | 133,719 † | Adam |
| Figure-eight ADD [cm] | |||
|---|---|---|---|
| 0.1–2 kg | 2–4 kg | / / RMSE [N m] | |
| PICR | 4.58 | 6.92 | 1.84 / 1.63 / 1.46 |
| PICR w/o contrastive | 6.07 | 9.78 | 2.11 / 2.87 / 1.15 |
| Figure-eight ADD [cm] | |||
| 0.1–2 kg | 2–4 kg | Gain at 20 N m | |
| RWM | 10.58 | 13.57 | 0.73 |
| RWM + contrastive | 12.48 | 13.21 | 0.55 |
| CaDM | 9.76 | 10.56 | 0.63 |
| CaDM + contrastive | 11.06 | 11.66 | 0.67 |
| CaDeLaC | 11.33 | 22.16 | 0.16 |
| Figure-eight ADD [cm] | |||
|---|---|---|---|
| 0.1–2 kg | 2–4 kg | / / RMSE [N m] | |
| FiLM (PICR) | 2.15 / 2.31 / 1.59 | ||
| Input concat | 4.02 / 2.27 / 2.09 | ||
| Shift only | 3.33 / 2.29 / 1.96 | ||
| Low-rank hypernet | 5.85 / 1.91 / 3.02 | ||
| Figure-eight ADD [cm] | |||
| Parameters | 0.1–2 kg | 2–4 kg | |
| PICR | 6.55M | 4.58 | 6.92 |
| PICR, width 256 | 0.56M | ||
| RWM | 0.96M | 7.09 | 11.63 |
| CaDM | 0.23M | 6.79 | 10.90 |
| CaDeLaC | 6.7k | 9.07 | 24.75 |
| Set | Motion and control | Payload condition | Size and use |
|---|---|---|---|
| Training | Joint excitation; PID with OU torque perturbation | – , 8 levels; CoM lattice | 318 training / 40 validation episodes |
| Payload sweep | PPO figure-eight tracking | Mass ramp from to | 33 episodes / 32,967 windows |
| Torque intervention | PPO replay with torque offsets | – ; grasp offset along hand | 1,887 of 5,616 pairs |
Robot Dynamics Learning
Robot dynamics learning builds models that predict how robots and objects move under actions, enabling control, planning, and adaptation.
See all 113 papers