Safe Reinforcement Learning via Shielding

TL;DR

This paper presents a method to ensure safety in reinforcement learning by synthesizing a reactive shield that enforces temporal logic specifications during learning and execution phases.

Contribution

It introduces a novel shielding approach that can be integrated either during decision-making or after learning to guarantee safety properties.

Findings

The shield effectively prevents safety violations in various scenarios.

The approach preserves the convergence guarantees of reinforcement learning.

Demonstrated versatility across multiple challenging environments.

Abstract

Reinforcement learning algorithms discover policies that maximize reward, but do not necessarily guarantee safety during learning or execution phases. We introduce a new approach to learn optimal policies while enforcing properties expressed in temporal logic. To this end, given the temporal logic specification that is to be obeyed by the learning system, we propose to synthesize a reactive system called a shield. The shield is introduced in the traditional learning process in two alternative ways, depending on the location at which the shield is implemented. In the first one, the shield acts each time the learning agent is about to make a decision and provides a list of safe actions. In the second way, the shield is introduced after the learning agent. The shield monitors the actions from the learner and corrects them only if the chosen action causes a violation of the specification. We discuss which requirements a shield must meet to preserve the convergence guarantees of the learner. Finally, we demonstrate the versatility of our approach on several challenging reinforcement learning scenarios.