Why Koopman Structure Does Not Help Differentiable World Models for Humanoid Whole-Body Tracking
cs.RO
Submitted: 2026-10-31
Updated: 2026-11-05
Comments: 8 pages, 3 figures, 2 tables. Preprint
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model
- Continual Learning and Lifting of Koopman Dynamics for Linear Control of Legged Robots
- Koopman Dreamer: Spectrally Constrained Latent Dynamics for Stable World-Model Imagination
- Bilinear World Models: Learning Representations with Structured Dynamics for Efficient Control
- On the Existence of Koopman Linear Embeddings for Controlled Nonlinear Systems
- Nonparametric Control Koopman Operators
- Universality of Linear Recurrences Followed by Non-linear Projections: Finite-Width Guarantees and Benefits of Complex Eigenvalues
- Efficient Reinforcement Learning using Linear Koopman Dynamics for Nonlinear Robotic Systems
- Structured Mechanical Models for Robot Learning and Control
- Differentiable Physics Models for Real-world Offline Model-based Reinforcement Learning
- Certifiable Safe Model-Based Reinforcement Learning with Control-Affine Dynamics Approximation
- Causally Correct Partial Models for Reinforcement Learning
- Bad Habits: Policy Confounding and Out-of-Trajectory Generalization in RL
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation