Online shielding for reinforcement learning

Abstract

Besides the recent impressive results on reinforcement learning (RL), safety is still one of the major research challenges in RL. RL is a machine-learning approach to determine near-optimal policies in Markov decision processes (MDPs). In this paper, we consider the setting where the safety-relevant fragment of the MDP together with a temporal logic safety specification is given, and many safety violations can be avoided by planning ahead a short time into the future. We propose an approach for online safety shielding of RL agents. During runtime, the shield analyses the safety of each available action. For any action, the shield computes the maximal probability to not violate the safety specification within the next k steps when executing this action. Based on this probability and a given threshold, the shield decides whether to block an action from the agent. Existing offline shielding approaches compute exhaustively the safety of all state-action combinations ahead of time, resulting in huge computation times and large memory consumption. The intuition behind online shielding is to compute at runtime 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 approach is well-suited for high-level planning problems where the time between decisions can be used for safety computations and it is sustainable for the agent to wait until these computations are finished. For our evaluation, we selected a 2-player version of the classical computer game Snake. The game represents a high-level planning problem that requires fast decisions and the multiplayer setting induces a large state space, which is computationally expensive to analyse exhaustively.

Guided Hierarchical Reinforcement Learning for Safe Urban Driving

“…Online shielding lacks worst-case computation time guarantees, potentially allowing the agent to reach the next decision state before the shield determines which action to block. It is suitable in scenarios where alternative actions, like "waiting," can be taken if safety analysis is not completed promptly [43].…”

“…In contrast, the shielding technique [37] deploys a shield to directly forestall the agent from taking actions that might potentially breach safety regulations during the exploration phase of DRL [38]. However, the shield's strictness may sometimes hinder the learning agent's ability to effectively explore the environment and discover its optimal policy [43].…”

“…Moreover, it circumvents the pitfalls associated with cumbersome shielding techniques, which can lead to exponential complexity in both state and action dimensions. Additionally, unlike methods requiring extensive computational time, our approach offers a more efficient and real-time decision-making process [43].…”