arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2406.04724cs.LGcs.AI
收藏

关于在对抗性强化学习中最小化对抗性反事实误差

On Minimizing Adversarial Counterfactual Error in Adversarial RL

Singapore Management University(新加坡国立管理学院);Rutgers University(罗格斯大学)

Roman Belaire, Arunesh Sinha, Pradeep Varakantham

AI总结:

针对深度强化学习易受对抗性扰动影响的问题,提出基于信念的对抗性反事实误差(ACoE)目标及其可扩展的累积版本C-ACoE,在MuJoCo、Atari和Highway基准上显著提升鲁棒性。

摘要:

深度强化学习(DRL)策略极易受到观测中对抗性噪声的影响,这在安全关键场景中构成了重大风险。对抗性扰动固有的挑战在于,通过改变智能体观测到的信息,状态变得仅部分可观测。现有方法要么在邻近状态间强制执行一致动作,要么在对抗性扰动观测内最大化最坏情况值。然而,前者在攻击成功时会导致性能下降,而后者往往过于保守,导致在良性环境中性能次优。我们假设这些局限性源于它们未能直接考虑部分可观测性。为此,我们引入了一个新颖的目标函数,称为对抗性反事实误差(ACoE),它基于对真实状态的信念进行定义,并平衡了价值优化与鲁棒性。为了使ACoE在无模型设置中具有可扩展性,我们提出了有理论依据的替代目标函数——累积ACoE(C-ACoE)。我们在标准基准(MuJoCo、Atari和Highway)上的实证评估表明,我们的方法显著优于当前解决对抗性强化学习挑战的最先进方法,为在对抗性条件下提高DRL的鲁棒性提供了一个有前景的方向。我们的代码可在https://github.com/romanbelaire/acoe-robust-rl获取。

英文摘要:

Deep Reinforcement Learning (DRL) policies are highly susceptible to adversarial noise in observations, which poses significant risks in safety-critical scenarios. The challenge inherent to adversarial perturbations is that by altering the information observed by the agent, the state becomes only partially observable. Existing approaches address this by either enforcing consistent actions across nearby states or maximizing the worst-case value within adversarially perturbed observations. However, the former suffers from performance degradation when attacks succeed, while the latter tends to be overly conservative, leading to suboptimal performance in benign settings. We hypothesize that these limitations stem from their failing to account for partial observability directly. To this end, we introduce a novel objective called Adversarial Counterfactual Error (ACoE), defined on the beliefs about the true state and balancing value optimization with robustness. To make ACoE scalable in model-free settings, we propose the theoretically-grounded surrogate objective Cumulative-ACoE (C-ACoE). Our empirical evaluations on standard benchmarks (MuJoCo, Atari, and Highway) demonstrate that our method significantly outperforms current state-of-the-art approaches for addressing adversarial RL challenges, offering a promising direction for improving robustness in DRL under adversarial conditions. Our code is available at https://github.com/romanbelaire/acoe-robust-rl.

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。

↑