Boosting verification of deep reinforcement learning via piece-wise linear decision neural networks
Formally verifying deep reinforcement learning (DRL) systems suffers from both inaccurate
verification results and limited scalability. The major obstacle lies in the large overestimation …
verification results and limited scalability. The major obstacle lies in the large overestimation …
Stochastic omega-regular verification and control with supermartingales
A Abate, M Giacobbe, D Roy - arXiv preprint arXiv:2405.17304, 2024 - arxiv.org
We present for the first time a supermartingale certificate for $\omega $-regular
specifications. We leverage the Robbins & Siegmund convergence theorem to characterize …
specifications. We leverage the Robbins & Siegmund convergence theorem to characterize …
[PDF][PDF] Safe Exploration in Reinforcement Learning by Reachability Analysis over Learned Models
We introduce VELM, a reinforcement learning (RL) framework grounded in verification
principles for safe exploration in unknown environments. VELM ensures that an RL agent …
principles for safe exploration in unknown environments. VELM ensures that an RL agent …