mlsanityData to 7 Oct 2026

Safety-Critical Contextual Control via Online Riemannian Optimization with World Models

arXiv 2604.19639 ↗First posted 21 Apr 2026eess.SY cs.AI cs.SYcited by 0

On the map: World models → Latent dynamics models and model-based RL

Abstract (arXiv)

Modern world models are becoming too complex to admit explicit dynamical descriptions. We study safety-critical contextual control, where a Planner must optimize a task objective using only feasibility samples from a black-box Simulator, conditioned on a context signal xit. We develop a sample-based Penalized Predictive Control (PPC) framework grounded in online Riemannian optimization, in which the Simulator compresses the feasibility manifold into a score-based density hatp(u mid xit) that endows the action space with a Riemannian geometry guiding the Planner's gradient descent. The barrier curvature kappa(xit), the minimum curvature of the conditional log-density -lnhatp(·midxit), governs both convergence rate and safety margin, replacing the Lipschitz constant of the unknown dynamics. Our main result is a contextual safety bound showing that the distance from the true feasibility manifold is controlled by the score estimation error and a ratio that depends on kappa(xit), both of which improve with richer context. Simulations on a dynamic navigation task confirm that contextual PPC substantially outperforms marginal and frozen density models, with the advantage growing after environment shifts.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Latent dynamics models and model-based RL, one of 11 directions in World models. Builds on 22 of its 38 references that are in the map. First posted on arXiv 21 Apr 2026.

Builds on

22 of 38 references.

Closest ancestors · 2

Same direction, newest first

  1. 2018Recurrent World Models Facilitate Policy Evolution4 Sep
  2. 2018Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models30 May

Other references · 20

  1. 2015Variational Inference with Normalizing Flowscited by 2,161
  2. 2020Learning to Simulate Complex Physics with Graph Networkscited by 867
  3. 2024Genie: Generative Interactive Environmentscited by 706
  4. 2019Control Barrier Functions: Theory and Applicationscited by 332
  5. 2016Control Barrier Function Based Quadratic Programs for Safety Critical Systemscited by 261
  6. 2019End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Taskscited by 256
  7. 2019PRECOG: PREdiction Conditioned On Goals in Visual Multi-Agent Settingscited by 178
  8. 2025Ctrl-World: A Controllable Generative World Model for Robot Manipulationcited by 149
  9. 2018A predictive safety filter for learning-based control of constrained nonlinear dynamical systemscited by 91
  10. 2020Sample-efficient Cross-Entropy Method for Real-time Planningcited by 90
  11. 2023Video Prediction Models as Rewards for Reinforcement Learningcited by 76
  12. 2019Learning for Safety-Critical Control with Control Barrier Functionscited by 66
8 more
  1. 2019Data-Driven Model Predictive Control with Stability and Robustness Guaranteescited by 25
  2. 2024Maximum Entropy Reinforcement Learning via Energy-Based Normalizing Flowcited by 25
  3. 2020Learning-Based Predictive Control via Real-Time Aggregate Flexibilitycited by 7
  4. 2023Data-driven MPC with stability guarantees using extended dynamic mode decompositioncited by 7
  5. 2025Landing with the Score: Riemannian Optimization through Denoisingcited by 7
  6. 2025Data-Driven Input-Output Control Barrier Functionscited by 2
  7. 2025Learning State-Action Control Barrier Functions for Model-Free Control under Constraintscited by 2
  8. 2026Robust adaptive NMPC using ellipsoidal tubescited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Guided Riemannian Optimization (GuRO): Bridging Model Predictive Control and Decision Transformers
  2. 2026λ-Reachability: Geometric-Horizon Safety Bellman Equations for Humanoid Safety
  3. 2021Uncertainty Estimation Using Riemannian Model Dynamics for Offline Reinforcement Learning
  4. 2026Pixels to Proofs: Probabilistically-Safe Latent World Model Control via Parallel Conformal Robust MPC
  5. 2026Finite-Sample Metric Non-Collapse for Geometrically Supervised Latent World Models in Control
  6. 2025Generalizing Safety Beyond Collision-Avoidance via Latent-Space Reachability Analysis
  7. 2026Learning from World Feedback: Why Model Uncertainty Fails as a Risk Signal in Model-Based RL
  8. 2026Beyond Euclidean Clipping: Overcoming Exploration Collapse in LLM RL via Riemannian Isometric Policy Optimization