Stochastic Port-Hamiltonian Gaussian Processes for Learning from Noisy Dynamics on Lie Groups
Organizations: Department of Computer Science, University of Texas at Austin · Oden Institute of Computational Engineering & Sciences, University of Texas at Austin
Abstract
Learning predictive models for dynamics is difficult because observations are often sparse and noisy, while the underlying dynamics evolve on nonlinear configuration manifolds and are affected by environmental dissipation and stochastic disturbances. Standard neural dynamics models can fit short-term trajectories but often violate physical structure, accumulate error over long horizons, or rely on parameterizations that introduce singularities such as gimbal lock. We introduce LieSPHGP, a stochastic Gaussian-process framework for learning rigid body dynamics with port-Hamiltonian structure and stability-aware robustness. LieSPHGP represents motion on Lie groups and Lie algebras, enabling coordinate-consistent optimization without singular parameterizations. Learned mass, potential energy, dissipation, and control components retain physical roles and a separate diffusion term represents stochastic forcing. Each component is modeled by Matérn kernel Gaussian processes which capture uncertainty induced by noisy observations. To improve physical validity, Lie group exponential map updates preserve rotational constraints, while positive-definite damping ensures nonnegative deterministic energy dissipation. The implicit-explicit integrator additionally provides unconditional stability for the frozen linear damping subsystem. The resulting model learns predictive dynamics that are robust to noise, respect rigid body geometry, and capture both intrinsic motion and environmental dissipation. Experiments on various noisy rigid body benchmarks show improved prediction accuracy, stronger long-horizon stability, and better satisfaction of manifold constraints compared with neural dynamics baselines. Code, data and interactive results are available on the project website: https://thribhuvanrapolu.github.io/liesphgp/
Figures & tables
| Method | Dynamics Type | Architecture | Geometry | DE Type |
|---|---|---|---|---|
| NODE ( Chen et al., 2018 ) | No Hamiltonian | Neural Network | Euclidean | ODE |
| HNN ( Greydanus et al., 2019 ) | Hamiltonian | Neural Network | Euclidean | ODE |
| NSDE ( Liu et al., 2019 ) | No Hamiltonian | Neural Network | Euclidean | SDE |
| VIN ( Saemundsson et al., 2020 ) | Lagrangian | Neural Network | Euclidean | ODE |
| PHNN ( Desai et al., 2021 ) | Port-Hamiltonian | Neural Network | Euclidean | ODE |
| InfGAN ( Kidger et al., 2021 ) | No Hamiltonian | Neural Network | Euclidean | SDE |
| 5-point (0.2 seconds) windows | 41-point (2 seconds) windows | |||
| Error | PH-NODE | LieSPHGP ( ) (Ours) | PH-NODE | LieSPHGP ( ) (Ours) |
| Rotation (rad) | ||||
| Velocity (rad/s) | ||||
| Energy | ||||
| Det. ( ) | ||||
| Orth. ( ) | ||||
| Error | PH-NODE | NeuralSDE | Lie-PH-NN-SDE | LieSPHGP |
|---|---|---|---|---|
| Rotation (rad) | ||||
| Velocity (rad/s) | ||||
| Energy | ||||
| Det. | ||||
| Orth. | ||||
| Rotation CRPS |
| Gain scale | PH-NODE | LieSPHGP (Ours) | True model | |
|---|---|---|---|---|
| Success (%) | ||||
| Final angle (rad) | ||||
| Metric | PH-NODE | LieSPHGP (Ours) | True model |
|---|---|---|---|
| Position RMSE (m) | |||
| Linear velocity RMSE (m/s) | |||
| Attitude RMSE (rad) | |||
| Angular velocity RMSE (rad/s) | |||
| Energy RMSE | |||
| Valid tracking time (s) | – |
| Metric | PH-NODE | LieSPHGP (Ours) | True model |
|---|---|---|---|
| Position error (m) | |||
| Attitude error (rad) | |||
| Angular velocity error (rad/s) | |||
| Position CRPS (m) | |||
| Position coverage | |||
| Thrust difference (/ hover) |
| Model | (0.02 s) | (0.2 s) | (2 s) | VPT (s) |
|---|---|---|---|---|
| Persistence | ||||
| Physics (published parameters) | ||||
| Koopman (EDMDc) | ||||
| LieSPHGP (ours) |
| System | Metric | Embedded | Intrinsic |
|---|---|---|---|
| Pendulum | Test EKF NLL | ||
| Rotation error (rad) | |||
| Rotation CRPS | |||
| Damping error | |||
| Training time (min) | |||
| Quadrotor | Test EKF NLL |
| Determinant | Orthogonality | |||
| Model | Lie-IMEX | RK4 | Lie-IMEX | RK4 |
| Deterministic pendulum | ||||
| LieSPHGP ( ) | ||||
| PH-NODE | ||||
| Windy pendulum | ||||
| LieSPHGP | ||||
Appendix figures & tables18 assets
Supplementary material from the paper’s appendix.
Appendix
| Model | Error (m) | Error (m/s) | Error (rad) | Error (rad/s) |
|---|---|---|---|---|
| Naïve ( Busetto et al., 2026 ) | 0.3536 | 0.7745 | 0.1639 | 0.5797 |
| Physics ( Busetto et al., 2026 ) | 0.0448 | 0.2094 | 0.1014 | 0.5797 |
| Res-MLP ( Busetto et al., 2026 ) | 0.0795 | 0.2488 | 0.1210 | 0.7097 |
| Hybrid ( Busetto et al., 2026 ) | 0.0456 | 0.2051 | 0.1228 | 0.5782 |
| Res-LSTM ( Busetto et al., 2026 ) | 0.1182 | 0.3314 | 0.1129 | 0.7149 |
| PH-NODE ( Duong et al., 2024 ) | 0.0393 | 0.2262 | 0.1451 | 0.7849 |
| Model | ( s) | ( s) | Position – s (m) | VPT (s) |
|---|---|---|---|---|
| Persistence | ||||
| Physics (published parameters) | 1.19 | |||
| Koopman (EDMDc) | ||||
| LieSPHGP (ours) |
| Product | LieSPHGP (ours) | Lie-PH-NN-SDE (ours) | PH-NODE (prior work) |
|---|---|---|---|
| Damping | |||
| Control gain | |||
| Gravity | |||
| Diffusion | – |
| Quantity | Truth | LieSPHGP (ours) |
|---|---|---|
| Deterministic pendulum | ||
| Obs. noise | ||
| NIS per dimension (ideal ) | ||
| Windy pendulum | ||
| Obs. noise | ||
| Diffusion | ||
| Variant | Rotation | Rot. CRPS | cov. | Swing-up | ||
|---|---|---|---|---|---|---|
| LieSPHGP | / | |||||
| Loss | ||||||
| Rollout MSE, noisy start | – | / | ||||
| Rollout MSE, learned start | – | / | ||||
| GP weights | ||||||
| Point estimate, MLE | / | |||||
| Metric | SPH-NN | SPH-NN-mb | LieSPHGP (ours) |
|---|---|---|---|
| Rotation error (rad) | |||
| Velocity error (rad/s) | |||
| Energy error | |||
| Rotation CRPS | |||
| Rotation coverage |
| Product | LieSPHGP (ours) | PH-NODE (prior work) |
|---|---|---|
| Thrust row of | ||
| Torque rows of ( ) | ||
| Gravity | ||
| Translational damping | ||
| Rotational damping | ||
| Diffusion | – |
| 5-sample windows | 101-sample windows | ||||
| Metric | PH-NODE | LieSPHGP | PH-NODE | LieSPHGP | True model |
| Closed-loop position RMSE (m) | |||||
| Closed-loop velocity RMSE (m/s) | |||||
| Rotor saturation share | |||||
| Model-in-the-loop position error (m) | |||||
| Model-in-the-loop coverage | |||||
Dynamical Systems
Dynamical systems studies how states evolve over time, including stability, feedback, chaos, and transitions in physical, biological, and engineered systems.
See all 512 papers