Formal Synthesis of Certifiably Robust Neural Lyapunov-Barrier Certificates
Not on the map. Why?
Abstract (arXiv)
Neural Lyapunov and barrier certificates have recently been used as powerful tools for verifying the safety and stability properties of deep reinforcement learning (RL) controllers. However, existing methods offer guarantees only under fixed ideal unperturbed dynamics, limiting their reliability in real-world applications where dynamics may deviate due to uncertainties. In this work, we study the problem of synthesizing robust neural Lyapunov barrier certificates that maintain their guarantees under perturbations in system dynamics. We formally define a robust Lyapunov barrier function and specify sufficient conditions based on Lipschitz continuity that ensure robustness against bounded perturbations. We propose practical training objectives that enforce these conditions via adversarial training, Lipschitz neighborhood bound, and global Lipschitz regularization. We validate our approach in two practically relevant environments, Inverted Pendulum and 2D Docking. The former is a widely studied benchmark, while the latter is a safety-critical task in autonomous systems. We show that our methods significantly improve both certified robustness bounds (up to 4.6 times) and empirical success rates under strong perturbations (up to 2.4 times) compared to the baseline. Our results demonstrate effectiveness of training robust neural certificates for safe RL under perturbations in dynamics.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 26 of its 41 references that are in the map. First posted on arXiv 5 Feb 2026.
Builds on
26 of 41 references.
References · 26
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
- 2013Intriguing properties of neural networkscited by 5,114
- 2013Playing Atari with Deep Reinforcement Learningcited by 3,211
- 2020Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attackscited by 1,442
- 2018QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulationcited by 730
- 2017Robust Adversarial Reinforcement Learningcited by 439
- 2019Control Barrier Functions: Theory and Applicationscited by 332
- 2017Parseval Networks: Improving Robustness to Adversarial Examplescited by 288
- 2020Robust Deep Reinforcement Learning against Adversarial Perturbations on State Observationscited by 235
- 2017Deep Reinforcement Learning framework for Autonomous Drivingcited by 224
- 2018Lipschitz-Margin Training: Scalable Certification of Perturbation Invariance for Deep Neural Networkscited by 193
14 more
- 2019Action Robust Reinforcement Learning and Applications in Continuous Controlcited by 167
- 2020Neural Lyapunov Controlcited by 142
- 2021Robust Reinforcement Learning on State Observations with Learned Optimal Adversarycited by 133
- 2021Globally-Robust Neural Networkscited by 106
- 2021Safe Nonlinear Control Using Robust Neural Lyapunov-Barrier Functionscited by 91
- 2024Marabou 2.0: A Versatile Formal Analyzer of Neural Networkscited by 81
- 2020Robust Reinforcement Learning using Adversarial Populationscited by 64
- 2019Risk Averse Robust Adversarial Reinforcement Learningcited by 54
- 2021OVERT: An Algorithm for Safety Verification of Neural Network Control Policies for Nonlinear Systemscited by 37
- 2021Stability Verification in Stochastic Control Systems via Neural Network Supermartingalescited by 24
- 2022Safety Certification for Stochastic Systems via Neural Barrier Functionscited by 22
- 2022Learning Control Policies for Stochastic Systems with Reach-avoid Guaranteescited by 21
- 2024Formally Verifying Deep Reinforcement Learning Controllers with Lyapunov Barrier Certificatescited by 15
- 2022Robust Reinforcement Learning with Distributional Risk-averse formulationcited by 9
Built on by
Not cited yet.