arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2206.12784cs.RO
收藏

学习用物理启发的风险意识重新安排

Learning to Rearrange with Physics-Inspired Risk Awareness

UC San Diego(加州大学圣地亚哥分校);Meta

Meng Song, Yuhan Liu, Zhengqin Li, Manmohan Chandraker

摘要:

现实世界的应用需要机器人在物理世界中操作,除了完成任务外,还要意识到潜在的风险。很大一部分的危险行为来自于在不知道启示的情况下与物体的交互。为了防止代理作出不安全的决定,我们建议训练的机器人代理强化学习执行任务的物理属性,如质量和摩擦在室内环境中的意识。我们通过一个新的物理启发的奖励函数来实现这一点,该函数鼓励代理学习识别不同质量和摩擦系数的策略。我们介绍了两个新颖的和具有挑战性的室内重排任务-可变摩擦推动任务和可变质量推动任务-允许评估的学习政策,在权衡性能和物理启发的风险。我们的研究结果表明,通过配备建议的奖励,代理是能够学习的政策,选择推动目标或目标达到的轨迹,以最小的物理成本,这可以进一步利用作为一种预防措施,以限制代理的行为,在安全批评的环境。

英文摘要:

Real-world applications require a robot operating in the physical world with awareness of potential risks besides accomplishing the task. A large part of risky behaviors arises from interacting with objects in ignorance of affordance. To prevent the agent from making unsafe decisions, we propose to train a robotic agent by reinforcement learning to execute tasks with an awareness of physical properties such as mass and friction in an indoor environment. We achieve this through a novel physics-inspired reward function that encourages the agent to learn a policy discerning different masses and friction coefficients. We introduce two novel and challenging indoor rearrangement tasks -- the variable friction pushing task and the variable mass pushing task -- that allow evaluation of the learned policies in trading off performance and physics-inspired risk. Our results demonstrate that by equipping with the proposed reward, the agent is able to learn policies choosing the pushing targets or goal-reaching trajectories with minimum physical cost, which can be further utilized as a precaution to constrain the agent's behavior in a safety-critic environment.

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。

↑