2000 character limit reached
Safety verification of Neural-Network-based controllers: a set invariance approach (2312.11352v2)
Published 18 Dec 2023 in eess.SY and cs.SY
Abstract: This paper presents a novel approach to ensure the safety of continuous-time linear dynamical systems controlled by a neural network (NN) based state-feedback. Our method capitalizes on the use of continuous piece-wise affine (PWA) activation functions (e.g. ReLU) which render the NN a PWA continuous function. By computing the affine regions of the latter and applying Nagumo's theorem, a subset of boundary points can effectively verify the invariance of a potentially non-convex set. Consequently, an algorithm that partitions the state space in affine regions is proposed. The scalability of our approach is thoroughly analyzed, and extensive tests are conducted to validate its effectiveness.
- “A Neural Network Architecture to Learn Explicit MPC Controllers from Data” 21st IFAC World Congress In IFAC-PapersOnLine 53.2, 2020, pp. 11362–11367 DOI: https://doi.org/10.1016/j.ifacol.2020.12.546
- “Continuous control with deep reinforcement learning” In arXiv preprint arXiv:1509.02971, 2015
- “Neural networks for control systems—A survey” In Automatica 28.6, 1992, pp. 1083–1112 DOI: https://doi.org/10.1016/0005-1098(92)90053-I
- “Reachnn: Reachability analysis of neural-network controlled systems” In ACM Transactions on Embedded Computing Systems (TECS) 18.5s ACM New York, NY, USA, 2019, pp. 1–22
- Kyle D. Julian and Mykel J. Kochenderfer “A Reachability Method for Verifying Dynamical Systems with Deep Neural Network Controllers”, 2019 arXiv:1903.00520 [cs.SY]
- Joseph A. Vincent and Mac Schwager “Reachable Polyhedral Marching (RPM): A Safety Verification Algorithm for Robotic Systems with Deep Neural Network Components” In 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 9029–9035 DOI: 10.1109/ICRA48506.2021.9561956
- “Synthesizing barrier certificates using neural networks” In Proceedings of the 23rd international conference on hybrid systems: computation and control, 2020, pp. 1–11
- “Verifying neural network controlled systems using neural networks” In Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control, 2022, pp. 1–11
- S.Miani F.Blanchini “Set-Theoretic Methods in Control” Birkhäuser, 2008, pp. 103–106
- “On the number of linear regions of deep neural networks” In Advances in neural information processing systems 27, 2014
- “Automatic differentiation in machine learning: a survey” In Journal of Marchine Learning Research 18 Microtome Publishing, 2018, pp. 1–43
- “Pytorch: An imperative style, high-performance deep learning library” In Advances in neural information processing systems 32, 2019
- Kurt Hornik, Maxwell Stinchcombe and Halbert White “Multilayer feedforward networks are universal approximators” In Neural Networks 2.5, 1989, pp. 359–366 DOI: https://doi.org/10.1016/0893-6080(89)90020-8
- Arne Brondsted “An introduction to convex polytopes” Springer Science & Business Media, 2012
- “Exact and consistent interpretation for piecewise linear neural networks: A closed form solution” In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2018, pp. 1244–1253
- “Complexity of linear regions in deep networks” In International Conference on Machine Learning, 2019, pp. 2596–2604 PMLR
- Louis Jouret (1 paper)
- Adnane Saoud (20 papers)
- Sorin Olaru (8 papers)