arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.00587cs.LG
收藏

安全的 Langevin 软 actor-critic

Safe Langevin Soft Actor Critic

Yardi School of Artificial Intelligence, IIT Delhi, India(IIT德里人工智能学院);Department of Mathematics, IIT Delhi, India(IIT德里数学系)

Mahesh Keswani, Samyak Jain, Raunak P. Bhattacharyya

AI总结:

SL-SAC 通过参数空间探索和分布风险控制,解决了受约束强化学习中奖励与安全性的平衡问题,实现了在多个基准任务中的低成本表现。

摘要:

在受约束的强化学习中,平衡奖励和安全性仍然具有挑战性,因为从尖锐价值极小值处的泛化能力差,以及对厚尾风险分布的处理不足。我们引入了 Safe Langevin Soft Actor-Critic (SL-SAC),一种原理性的算法,通过参数空间探索和分布风险控制来解决这两个问题。我们的方法结合了三个关键机制:(1) 自适应随机梯度 Langevin 动力学 (aSGLD) 用于奖励批评者,促进集合多样性并逃离劣质极值;(2) 通过隐式分位数网络 (IQN) 进行分布成本估计,结合条件价值-at-风险 (CVaR) 优化以缓解尾部风险;(3) 一种反应性的拉格朗日松弛方案,根据经验 CVaR 来调整约束执行。我们为 CVaR 估计误差提供了理论保证,并证明基于 CVaR 的拉格朗日更新比基于预期成本的更新产生更强的约束违反信号。在 Safety-Gymnasium 评估基准上,SL-SAC 在 10 个任务中的 7 个任务中实现了最低成本,同时保持了竞争力的回报,与最先进的基线相比,在速度任务中成本减少了 19-63%。

英文摘要:

Balancing reward and safety in constrained reinforcement learning remains challenging due to poor generalization from sharp value minima and inadequate handling of heavy-tailed risk distribution. We introduce Safe Langevin Soft Actor-Critic (SL-SAC), a principled algorithm that addresses both issues through parameter-space exploration and distributional risk control. Our approach combines three key mechanisms: (1) Adaptive Stochastic Gradient Langevin Dynamics (aSGLD) for reward critics, promoting ensemble diversity and escape from poor optima; (2) distributional cost estimation via Implicit Quantile Networks (IQN) with Conditional Value-at-Risk (CVaR) optimization for tail-risk mitigation; and (3) a reactive Lagrangian relaxation scheme that adapts constraint enforcement based on the empirical CVaR of episodic costs. We provide theoretical guarantees on CVaR estimation error and demonstrate that CVaR-based Lagrange updates yield stronger constraint violation signals than expected-cost updates. On Safety-Gymnasium benchmarks, SL-SAC achieves the lowest cost in 7 out of 10 tasks while maintaining competitive returns, with cost reductions of 19-63% in velocity tasks compared to state-of-the-art baselines.

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。

↑