Safe Reinforcement Learning via Confidence-Based Filters

TL;DR

This paper introduces confidence-based safety filters for reinforcement learning that certify safety constraints using probabilistic models, enabling safe policy deployment with formal guarantees.

Contribution

It presents a control-theoretic framework that reformulates safety constraints into cost functions and extends them with hallucinating inputs to ensure high-probability safety.

Findings

Formal safety guarantees are provided.

Empirical results demonstrate effectiveness.

The approach enables safe policy adjustments in real-time.

Abstract

Ensuring safety is a crucial challenge when deploying reinforcement learning (RL) to real-world systems. We develop confidence-based safety filters, a control-theoretic approach for certifying state safety constraints for nominal policies learned via standard RL techniques, based on probabilistic dynamics models. Our approach is based on a reformulation of state constraints in terms of cost functions, reducing safety verification to a standard RL task. By exploiting the concept of hallucinating inputs, we extend this formulation to determine a "backup" policy that is safe for the unknown system with high probability. Finally, the nominal policy is minimally adjusted at every time step during a roll-out towards the backup policy, such that safe recovery can be guaranteed afterwards. We provide formal safety guarantees, and empirically demonstrate the effectiveness of our approach.