NRI/Collaborative Research: Robotic Disassembly of High-Precision Electronic Devices

国家机器人计划（NRI）项目解决了越来越多的废弃高精度电子产品，如手机、平板电脑和笔记本电脑。由于拆卸电池的人工成本高，目前的回收方法依赖于电池拆卸后的粉碎。结果，许多有价值的部件被埋在垃圾填埋场，无法回收利用。拆卸是回收的第一步，比组装更复杂，因为产品类型的可变性更大，因此，再制造通常无利可图。该奖项支持研究，以提供开发新型机器人系统所需的基本理解，该系统可以有效地执行高精度拆卸操作，并使其在实际和经济上可行。这项工作有可能缓解回收行业的劳动力短缺，减少电子废物，并彻底改变高精度电子产品的再制造。这项研究涉及多个学科，包括3D传感、深度学习和机器人技术。多学科研究将纳入一系列教育和推广活动，这将增加代表性不足的群体对研究的参与，并对工程教育产生积极影响。与组装产品的机器人装配线不同，由于废弃的高精度电子产品种类繁多，为重复操作编程的机器人不是拆卸的可行解决方案。因此，高精度电子产品的拆卸比装配复杂得多，对机器人的适应性、灵巧性和精度要求很高。该研究旨在通过对3D传感、深度学习、机械手设计和高精度操作的集成和融合研究，实现一种能够准确地看到、解释和拆卸高精度电子设备的新型机器人系统。特别是，研究团队将(1)对具有广泛光学特性和反射率变化的复杂表面进行精确的3D传感；(2)设计并优化三维点云解译深度学习架构设计；(3)设计了一种新型的轻型缆索驱动机械手，并开发了一种高精度的操作算法，实现了高效的经验学习。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The National Robotics Initiative (NRI) project addresses the increasing quantity of discarded high-precision electronics such as cell phones, tablets, and laptops. Current recycling methods rely on shredding after battery removal, due to high labor costs for disassembly. As a result, many valuable components are buried in landfills and not recycled. Disassembly, the first step of recycling, is more complex than assembly since there is much more variability in product type and, as a result, remanufacturing is usually not profitable. This award supports research to provide the fundamental understanding needed for the development of a novel robotic system that can effectively perform high-precision disassembly operations and make them practically and economically viable. The work has potential to mitigate labor shortages in recycling industry, reduce electronics waste, and revolutionize the remanufacturing of high-precision electronics. The research involves several disciplines including 3D sensing, deep learning, and robotics. The multidisciplinary research will be integrated into a series of educational and outreach activities which will increase the participation of underrepresented groups in research and positively impact engineering education.Unlike the robotic assembly lines that assemble products, programming robots for repetitive operations is not a feasible solution for disassembly due to the widely varying types of discarded high-precision electronics. Therefore, disassembly of high-precision electronics is significantly more complex than assembly and requires high robotic adaptability, dexterity and accuracy. The research aims to enable a novel robotic system that can accurately see, interpret, and disassemble high-precision electronics through integrated and convergent research on 3D sensing, deep learning, robotic hand design, and high-precision manipulation. In particular, the research team will (1) perform accurate 3D sensing for complex surfaces exhibiting wide ranges of optical properties and reflectivity variations; (2) design and optimize the design of deep learning architectures for 3D point cloud interpretation; and (3) design a novel lightweight cable-driven robotic hand and develop a high-precision manipulation algorithm enabling efficient learning from experience.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.