Learning Barrier Certificates: Towards Safe Reinforcement Learning with Zero Training-time Violations

TL;DR

This paper introduces CRABS, an algorithm that ensures zero training-time safety violations in reinforcement learning by iteratively learning barrier certificates, dynamics models, and policies, enabling safe exploration without prior knowledge.

Contribution

The paper presents a novel safe RL algorithm that guarantees zero safety violations during training using adversarially learned barrier certificates and dynamics models.

Findings

CRABS achieves zero safety violations in simple environments.

Prior methods require hundreds of violations for similar rewards.

CRABS enables safe exploration near safety boundaries.

Abstract

Training-time safety violations have been a major concern when we deploy reinforcement learning algorithms in the real world. This paper explores the possibility of safe RL algorithms with zero training-time safety violations in the challenging setting where we are only given a safe but trivial-reward initial policy without any prior knowledge of the dynamics model and additional offline data. We propose an algorithm, Co-trained Barrier Certificate for Safe RL (CRABS), which iteratively learns barrier certificates, dynamics models, and policies. The barrier certificates, learned via adversarial training, ensure the policy's safety assuming calibrated learned dynamics model. We also add a regularization term to encourage larger certified regions to enable better exploration. Empirical simulations show that zero safety violations are already challenging for a suite of simple environments with only 2-4 dimensional state space, especially if high-reward policies have to visit regions near the safety boundary. Prior methods require hundreds of violations to achieve decent rewards on these tasks, whereas our proposed algorithms incur zero violations.