# Safe Reinforcement Learning via Probabilistic Shields

arXiv: https://arxiv.org/abs/1807.06096 · HTML: https://mlsanity.com/paper/1807.06096/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Jul 2018.

## Built on by (most cited first, 7 of 7)

- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [Learning Barrier Certificates: Towards Safe Reinforcement Learning with Zero Training-time Violations](https://mlsanity.com/paper/2108.01846/) (2021)
- [Model-Free Learning of Safe yet Effective Controllers](https://mlsanity.com/paper/2103.14600/) (2021)
- [PAC Statistical Model Checking of Mean Payoff in Discrete- and Continuous-Time MDP](https://mlsanity.com/paper/2206.01465/) (2022)
- [Conformal Safety Shielding for Imperfect-Perception Agents](https://mlsanity.com/paper/2506.17275/) (2025)
- [Neuro-Symbolic Injection of LTLf Constraints in Autoregressive Reinforcement Learning Policies](https://mlsanity.com/paper/2606.08312/) (2026)
- [Trust Is Not a Score: Runtime Assurance Contracts for High-Risk AI Agents](https://mlsanity.com/paper/2609.39717/) (2026)

## Abstract (arXiv)

This paper targets the efficient construction of a safety shield for decision making in scenarios that incorporate uncertainty. Markov decision processes (MDPs) are prominent models to capture such planning problems. Reinforcement learning (RL) is a machine learning technique to determine near-optimal policies in MDPs that may be unknown prior to exploring the model. However, during exploration, RL is prone to induce behavior that is undesirable or not allowed in safety- or mission-critical contexts. We introduce the concept of a probabilistic shield that enables decision-making to adhere to safety constraints with high probability. In a separation of concerns, we employ formal verification to efficiently compute the probabilities of critical decisions within a safety-relevant fragment of the MDP. We use these results to realize a shield that is applied to an RL algorithm which then optimizes the actual performance objective. We discuss tradeoffs between sufficient progress in exploration of the environment and ensuring safety. In our experiments, we demonstrate on the arcade game PAC-MAN and on a case study involving service robots that the learning efficiency increases as the learning needs orders of magnitude fewer episodes.