arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.06239cs.LGstat.ML
收藏

基于线性函数逼近的安全强化学习

Safe Reinforcement Learning with Linear Function Approximation

University of British Columbia(不列颠哥伦比亚大学);University of California, Los Angeles(加利福尼亚大学洛杉矶分校)

Sanae Amani, Christos Thrampoulidis, Lin F. Yang

摘要:

近年来,强化学习的安全性变得越来越重要。然而,现有的解决方案要么不能严格避免选择不安全的行动,这可能会导致灾难性的结果,在安全关键的系统,或无法提供遗憾的保证设置安全约束需要学习。在本文中,我们解决这两个问题,首先建模安全作为一个未知的线性成本函数的状态和行动,必须始终低于一定的阈值。然后,我们提出的算法,称为SLUCB-QVI和RSLUCB-QVI,情节马尔可夫决策过程(MDPs)与线性函数逼近。我们表明,SLUCB-QVI和RSLUCB-QVI,而与\emph{no safety violation},实现了$\tilde{\mathcal{O}}\left(κ\sqrt{d^3H^3T}\right)$遗憾,几乎匹配的国家的最先进的不安全的算法,其中$H$是每个情节的持续时间,$d$是特征映射的维度,$κ$是一个常数表征的安全约束,和$T$是动作播放的总数。我们进一步提出了数值模拟,证实了我们的理论研究结果。

英文摘要:

Safety in reinforcement learning has become increasingly important in recent years. Yet, existing solutions either fail to strictly avoid choosing unsafe actions, which may lead to catastrophic results in safety-critical systems, or fail to provide regret guarantees for settings where safety constraints need to be learned. In this paper, we address both problems by first modeling safety as an unknown linear cost function of states and actions, which must always fall below a certain threshold. We then present algorithms, termed SLUCB-QVI and RSLUCB-QVI, for episodic Markov decision processes (MDPs) with linear function approximation. We show that SLUCB-QVI and RSLUCB-QVI, while with \emph{no safety violation}, achieve a $\tilde{\mathcal{O}}\left(κ\sqrt{d^3H^3T}\right)$ regret, nearly matching that of state-of-the-art unsafe algorithms, where $H$ is the duration of each episode, $d$ is the dimension of the feature mapping, $κ$ is a constant characterizing the safety constraints, and $T$ is the total number of action plays. We further present numerical simulations that corroborate our theoretical findings.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑