Research map / Probabilistic and scientific modeling / Dynamical systems
Extending Lagrangian and Hamiltonian Neural Networks with Differentiable Contact Models
Yaofeng Desmond Zhong, Biswadip Dey, Amit Chakraborty
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add regularization R=εI (CMr) to the contact model, vs unregularized CM (ablation) | CM-CD-CLNN · BP5-e | relative L1 trajectory error: 2 orders of magnitude | Hurt |
| add regularization (CMr) vs CM (ablation) | CM-CD-CLNN · Rope | learned friction μ (true 0.000): 0.026 -> 0.037 | Hurt |
| add regularization (CMr) vs CM (ablation) | CM-CD-CLNN · Gyro | learned restitution e_P (true 0.800): 0.800 -> 0.892 | Hurt |
| add regularization (CMr) vs CM (ablation) | CM-CD-CLNN · CP3-e | learned friction μ (true 0.000): 0.000 -> 0.002 | Hurt |
| add regularization (CMr) vs CM (ablation) | CM-CD-CLNN · CP3-e and Gyro-e | energy conservation | Hurt |
| use smaller regularization ε (e.g. 0.001) in CMr (ablation) | CMr · contact property learning | accuracy of learned contact properties | Improved |
| make ε a learnable parameter in CMr (ablation) | CMr · contact property learning | accuracy of learned contact properties | No clear effect |
| replace CLNN with CHNN backbone | CM contact model · 8 dynamics tasks | relative L1 trajectory error | No clear effect |
| proposed contact model vs MLP/IN baselines | MLP-CD-CLNN, IN-CP-CLNN, IN-CP-SP · 8 dynamics tasks, 5-step prediction | relative L1 trajectory error | Improved |
| proposed contact model vs baselines on long trajectories | baselines · 50-step test trajectories | relative L1 error | Improved |
| train on LCP-generated data | CM · Gyro (3D) | learned restitution e_P (true 0.800): 0.800 -> 0.822 | Hurt |
| train on noisy data | CM · dynamics tasks | performance | No clear effect |
Closest papers by meaning
- Geometric Contact Flows: Contactomorphisms for Dynamics and Control (ICML 2025 · similarity 0.90)
- Hamiltonian Neural PDE Solvers through Functional Approximation (NeurIPS 2025 · similarity 0.86)
- A Riemannian Framework for Learning Reduced-order Lagrangian Dynamics (ICLR 2025 · similarity 0.86)
- Learned Lagrangian Models of PDEs via Euler–Lagrange Residual Minimization (NeurIPS 2026 · similarity 0.86)
- Poisson-Dirac Neural Networks for Modeling Coupled Dynamical Systems across Domains (ICLR 2025 · similarity 0.85)
- Hamiltonian Score Matching and Generative Flows (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
