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 2303.03339cs.RO
收藏

减少可证明安全强化学习中的安全干预

Reducing Safety Interventions in Provably Safe Reinforcement Learning

Technical University of Munich(慕尼黑工业大学)

Jakob Thumm, Guillaume Pelat, Matthias Althoff

AI总结:

针对可证明安全 RL 中故障安全干预导致的不良动作问题,提出主动替换与主动投影方法,在零违规前提下减少干预并保持任务性能。

摘要:

深度强化学习(RL)在解决复杂机器人挑战方面已展现出前景。在现实应用中,RL通常配有故障安全控制器,作为避免灾难性事件的最后手段。尽管这些干预对安全而言是必要的,但它们可能导致不良行为,例如突然制动或激进转向。本文提出两种减少安全干预的方法:主动替换和主动投影;当智能体的动作会导致潜在的故障安全干预时,这两种方法会更改该动作。研究在 OpenAI safety gym 基准和一项人机协作任务上,将这些方法与最先进的约束强化学习进行比较。研究表明,将本文方法与可证明安全 RL 相结合,可产生高性能策略,实现零安全违规以及少量故障安全干预。这种通用方法可应用于广泛的现实机器人任务,同时在不牺牲任务性能的情况下有效提升安全性。

英文摘要:

Deep Reinforcement Learning (RL) has shown promise in addressing complex robotic challenges. In real-world applications, RL is often accompanied by failsafe controllers as a last resort to avoid catastrophic events. While necessary for safety, these interventions can result in undesirable behaviors, such as abrupt braking or aggressive steering. This paper proposes two safety intervention reduction methods: proactive replacement and proactive projection, which change the action of the agent if it leads to a potential failsafe intervention. These approaches are compared to state-of-the-art constrained RL on the OpenAI safety gym benchmark and a human-robot collaboration task. Our study demonstrates that the combination of our method with provably safe RL leads to high-performing policies with zero safety violations and a low number of failsafe interventions. Our versatile method can be applied to a wide range of real-world robotic tasks, while effectively improving safety without sacrificing task performance.

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。

↑