mlsanityData to 14 Oct 2026

RL-Guided PAC-NMPC for Probabilistically-Safe Perception-Based Navigation in Unknown Environments

arXiv 2609.39854 ↗First posted 30 Sep 2026cs.ROcited by 0

On the map: Deep reinforcement learning methods → RL for robot navigation and UAV control also fits: Safe and constrained RL

Abstract (arXiv)

In this paper, we present an approach for combining stochastic nonlinear model predictive control (SNMPC) and reinforcement learning (RL) to enable probabilistically-safe perception-based navigation in unknown environments. Our method first uses RL to train probabilistic actor-critic and sensor prediction models. We then leverage these probabilistic models in a sampling-based SNMPC framework known as Probably Approximately Correct (PAC)-NMPC, which uses hard constraints to enforce finite-time statistical guarantees on the probability of collision and value function improvement. By ensuring that our finite-horizon SNMPC policies decrease the value function in expectation, we can approach the long-horizon performance of the RL approach while satisfying probabilistic safety constraints. Through simulation experiments, we show that our approach can improve the safety of perception-based RL navigation policies and scale to high dimensional systems with large sensor input spaces and complex nonlinear dynamics. We also demonstrate our approach through hardware experiments, showing improved performance for vision-based navigation with an agile fixed-wing aerial vehicle in unknown environments.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RL for robot navigation and UAV control, one of 15 directions in Deep reinforcement learning methods. Builds on 46 of its 77 references that are in the map. First posted on arXiv 30 Sep 2026.

Builds on

46 of 77 references.

References · 46

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,448
  4. 2018Addressing Function Approximation Error in Actor-Critic Methodscited by 2,417
  5. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,333
  6. 2018Soft Actor-Critic Algorithms and Applicationscited by 1,240
  7. 2017Constrained Policy Optimizationcited by 983
  8. 2018Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Modelscited by 774
  9. 2018Reward Constrained Policy Optimizationcited by 417
  10. 2022Temporal Difference Learning for Model Predictive Controlcited by 397
  11. 2021Safe Learning in Robotics: From Learning-Based Control to Safe Reinforcement Learningcited by 371
  12. 2019End-to-End Safe Reinforcement Learning through Barrier Functions for Safety-Critical Continuous Control Taskscited by 289
34 more
  1. 2018A predictive safety filter for learning-based control of constrained nonlinear dynamical systemscited by 118
  2. 2018Plan Online, Learn Offline: Efficient Learning and Exploration via Model-Based Controlcited by 113
  3. 2019Safe Reinforcement Learning Using Robust MPCcited by 85
  4. 2017OptLayer - Practical Constrained Optimization for Deep Reinforcement Learning in the Real Worldcited by 70
  5. 2020The Marathon 2: A Navigation Systemcited by 70
  6. 2020Learning Off-Policy with Online Planningcited by 53
  7. 2019Combining Optimal Control and Learning for Visual Navigation in Novel Environmentscited by 52
  8. 2022DiffStack: A Differentiable and Modular Control Stack for Autonomous Vehiclescited by 37
  9. 2018Deep Drone Racing: Learning Agile Flight in Dynamic Environmentscited by 34
  10. 2016Funnel Libraries for Real-Time Robust Feedback Motion Planningcited by 29
  11. 2019Large Scale Model Predictive Control with Neural Networks and Primal Active Setscited by 29
  12. 2018PAMPC: Perception-Aware Model Predictive Control for Quadrotorscited by 28
  13. 2021Where to go next: Learning a Subgoal Recommendation Policy for Navigation Among Pedestrianscited by 28
  14. 2018PAC-Bayes Control: Learning Policies that Provably Generalize to Novel Environmentscited by 22
  15. 2019Deep Value Model Predictive Controlcited by 20
  16. 2023Actor-Critic Model Predictive Control: Differentiable Optimization meets Reinforcement Learning for Agile Flightcited by 19
  17. 2019Information Theoretic Model Predictive Q-Learningcited by 18
  18. 2024Transformer-based Model Predictive Control: Trajectory Optimization via Sequence Modelingcited by 16
  19. 2019A computationally efficient robust model predictive control framework for uncertain nonlinear systems -- extended versioncited by 14
  20. 2022CACTO: Continuous Actor-Critic with Trajectory Optimization -- Towards global optimalitycited by 12
  21. 2023Constraint-Informed Learning for Warm Starting Trajectory Optimizationcited by 7
  22. 2020Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motioncited by 4
  23. 2023Learning-Aided Warmstart of Model Predictive Control in Uncertain Fast-Changing Trafficcited by 4
  24. 2024TransformerMPC: Accelerating Model Predictive Control via Transformerscited by 4
  25. 2025Deep Learning Warm Starts for Trajectory Optimization on the International Space Stationcited by 4
  26. 2022Aerobatic Trajectory Generation for a VTOL Fixed-Wing Aircraft Using Differential Flatnesscited by 3
  27. 2020Direct NMPC for Post-Stall Motion Planning with Fixed-Wing UAVscited by 1
  28. 2022Post-Stall Navigation with Fixed-Wing UAVs using Onboard Visioncited by 1
  29. 2022Tube Stochastic Optimal Control for Nonlinear Constrained Trajectory Optimization Problemscited by 1
  30. 2022Probably Approximately Correct Nonlinear Model Predictive Control (PAC-NMPC)cited by 1
  31. 2023Robust Perception-Based Navigation using PAC-NMPC with a Learned Value Functioncited by 1
  32. 2025Trajectory Planning and Control for Differentially Flat Fixed-Wing Aerial Systemscited by 1
  33. 2025Dense Fixed-Wing Swarming using Receding-Horizon NMPCcited by 1
  34. 2026Warm-Starting Collision-Free Model Predictive Control With Object-Centric Diffusioncited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Lightweight Safe Reinforcement Learning for End-to-End UAV Navigation
  2. 2023Adaptive Stochastic Nonlinear Model Predictive Control with Look-ahead Deep Reinforcement Learning for Autonomous Vehicle Motion Control
  3. 2026NavRL++: A System-Level Framework for Improving Sim-to-Real Transfer in Reinforcement Learning-Based Robot Navigation
  4. 2026High-Speed Vision-Based Flight in Clutter with Safety-Shielded Reinforcement Learning
  5. 2026SafePG: Safe and Globally Optimal Reinforcement Learning with Hard Constraints
  6. 2024NavRL: Learning Safe Flight in Dynamic Environments
  7. 2025SIGN: Safety-Aware Image-Goal Navigation for Autonomous Drones via Reinforcement Learning
  8. 2024A Dynamic Safety Shield for Safe and Efficient Reinforcement Learning of Navigation Tasks