Online Shielding for Stochastic Systems

TL;DR

This paper introduces an online synthesis algorithm for runtime safety enforcers, called shields, in reinforcement learning systems, enabling real-time safety guarantees in stochastic environments.

Contribution

It presents a novel online shield synthesis method that reduces offline computation and memory needs, applicable to diverse planning problems with stochastic dynamics.

Findings

Effective safety enforcement in a stochastic Snake game.

Reduced offline computation compared to traditional methods.

Real-time safety guarantees demonstrated in a complex game environment.

Abstract

In this paper, we propose a method to develop trustworthy reinforcement learning systems. To ensure safety especially during exploration, we automatically synthesize a correct-by-construction runtime enforcer, called a shield, that blocks all actions that are unsafe with respect to a temporal logic specification from the agent. Our main contribution is a new synthesis algorithm for computing the shield online. Existing offline shielding approaches compute exhaustively the safety of all states-action combinations ahead-of-time, resulting in huge offline computation times, large memory consumption, and significant delays at run-time due to the look-ups in a huge database. The intuition behind online shielding is to compute during run-time the set of all states that could be reached in the near future. For each of these states, the safety of all available actions is analysed and used for shielding as soon as one of the considered states is reached. Our proposed method is general and can be applied to a wide range of planning problems with stochastic behavior. For our evaluation, we selected a 2-player version of the classical computer game SNAKE. The game requires fast decisions and the multiplayer setting induces a large state space, computationally expensive to analyze exhaustively. The safety objective of collision avoidance is easily transferable to a variety of planning tasks.