Collective Intelligence for 2D Push Manipulations With Mobile Robots
Collective Intelligence for 2D Push Manipulations With Mobile Robots
复制标题
DOI：
10.1109/lra.2023.3261751
复制
发表时间：
2022-11
影响因子：
5.2
通讯作者：
So Kuroki;T. Matsushima;Jumpei Arima;Hiroki Furuta;Yutaka Matsuo;S. Gu;Yujin Tang
中科院分区：
文献类型：
--
作者：
So Kuroki;T. Matsushima;Jumpei Arima;Hiroki Furuta;Yutaka Matsuo;S. Gu;Yujin Tang
虽然自然系统通常具有集体智慧，使它们能够自我组织和适应变化，但大多数人工系统都缺乏这种能力。我们探讨这样一个系统的可能性，在合作的背景下，使用移动的机器人的二维推动操作。虽然传统的作品展示了限制设置的问题的潜在解决方案，他们有计算和学习的困难。更重要的是，这些系统在面对环境变化时不具备适应能力。在这项工作中，我们表明，通过蒸馏来自可微的软体物理模拟器到基于注意力的神经网络的计划，我们的多机器人推操作系统实现了更好的性能比基线。此外，我们的系统还可以概括训练期间未见过的配置，并且能够在应用外部湍流和环境变化时适应任务完成。补充视频可以在我们的项目网站上找到：https://sites.google.com/view/ciom/home。
While natural systems often present collective intelligence that allows them to self-organize and adapt to changes, the equivalent is missing in most artificial systems. We explore the possibility of such a system in the context of cooperative 2D push manipulations using mobile robots. Although conventional works demonstrate potential solutions for the problem in restricted settings, they have computational and learning difficulties. More importantly, these systems do not possess the ability to adapt when facing environmental changes. In this work, we show that by distilling a planner derived from a differentiable soft-body physics simulator into an attention-based neural network, our multi-robot push manipulation system achieves better performance than baselines. In addition, our system also generalizes to configurations not seen during training and is able to adapt toward task completions when external turbulence and environmental changes are applied. Supplementary videos can be found on our project website: https://sites.google.com/view/ciom/home.