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 2203.05960cs.LGcs.AIcs.RO
收藏

控制中组合泛化的策略体系结构

Policy Architectures for Compositional Generalization in Control

Stanford University(斯坦福大学);Meta AI (FAIR)(Meta人工智能(基础人工智能研究院))

Allan Zhou, Vikash Kumar, Chelsea Finn, Aravind Rajeswaran

摘要:

控制、机器人和规划中的许多任务可以使用环境中各种实体的期望目标配置来指定。学习目标制约策略是解决此类任务的自然范式。然而,目前的方法很难学习和概括的任务复杂性的增加,如环境实体或目标的组成的数量的变化。在这项工作中,我们引入了一个框架,用于在任务中建模基于实体的组合结构,并创建合适的策略设计,可以利用这种结构。我们的策略利用了Deep Sets和自注意力等架构,非常灵活,可以进行端到端的训练,而不需要任何动作原语。当使用标准强化和模仿学习方法对一套模拟机器人操作任务进行训练时,我们发现这些架构以更少的数据实现了更高的成功率。我们还发现,这些架构可以实现更广泛的组合泛化,产生的策略可以外推到与训练中不同数量的实体,并以新颖的方式将学习到的技能缝合在一起(即组合)。结果的视频可以在https://sites.google.com/view/comp-gen-rl上找到。

英文摘要:

Many tasks in control, robotics, and planning can be specified using desired goal configurations for various entities in the environment. Learning goal-conditioned policies is a natural paradigm to solve such tasks. However, current approaches struggle to learn and generalize as task complexity increases, such as variations in number of environment entities or compositions of goals. In this work, we introduce a framework for modeling entity-based compositional structure in tasks, and create suitable policy designs that can leverage this structure. Our policies, which utilize architectures like Deep Sets and Self Attention, are flexible and can be trained end-to-end without requiring any action primitives. When trained using standard reinforcement and imitation learning methods on a suite of simulated robot manipulation tasks, we find that these architectures achieve significantly higher success rates with less data. We also find these architectures enable broader and compositional generalization, producing policies that extrapolate to different numbers of entities than seen in training, and stitch together (i.e. compose) learned skills in novel ways. Videos of the results can be found at https://sites.google.com/view/comp-gen-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。

↑