I-Corps: A Smart Context-Aware Multi-Fingered System for Dexterous Grasping

这个i-Corps项目的更广泛的影响/商业潜力是为许多行业和公司开发一个潜在的平台，如先进制造、电子商务、零售商、邮寄和物流。该平台可以革新和自动化不规则形状物体的分拣和包装过程，例如，为美国邮政分拣不同尺寸和重量的塑料邮袋；为沃尔玛和亚马逊包装不规则形状的商品。该项目还可能使在先进制造中依赖通用工具的便携式自动化发生革命性变化，例如，使飞机组装在精神航空系统公司的有限空间中占用大量空间；实现需要各种操作的制造过程自动化，如放置、抓取和盖帽。这个i-Corps项目基于平台系统的开发，将机器人抓取扩展到多指手的灵巧策略，并通过融合新型传感器提高环境感知能力。为了实现灵巧的抓取和操作，将拟人的抓取机制、上下文感知和面向任务的抓取规划紧密结合起来，设计了一种灵巧的机械手。机器人手通过融合多模式传感器实现对对象负担的原位感知，包括对象尺寸和对象特性，如材料、刚性和质量。这是第一个使用红外传感器和机器学习方法感知物体特征的机器人抓手。为了针对不同的任务需求和目标需求部署有效的策略，提出了一种知识驱动模型来解决灵巧操作运动规划问题。该模型将人类知识表示为手拓扑结构，并学习面向任务的操作计划以适应不同的工作环境。由于知识和学习的一体化结构，该模型以高效的学习效率产生了稳健和适应性的计划。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of a potential platform for many industries and companies, such as advanced manufacturing, e-commerce, retailer, mailing, and logistics. The platform could revolutionize and automate the process of sorting and packing irregular-shaped objects, e.g., sorting plastic mail bags with various dimensions and weights for USPS; and packing irregular-shaped goods for Walmart and Amazon. The project could also revolutionize portable automation that relies on general tools in advanced manufacturing, e.g., enabling the large footprint of aircraft assembly in confined space access for Spirit AeroSystems; and automating manufacturing processes that require various kinds of manipulation such as placing, grasping, and capping.This I-Corps project is based on the development of a platform system extending robotic grasping towards dexterous strategies for multi-fingered hands, and increasing the capability of context perception by fusing novel sensors. To accomplish dexterous grasping and manipulation, a dexterous robotic hand was designed by tightly integrating anthropomorphic gripper mechanics, context perception, and task-oriented grasp planning. The robotic hand achieves in-situ perception of object affordances by fusing multimodal sensors, including object dimensions and object characteristics such as materials, rigidity, and mass. This is the first robotic gripper that perceives object characteristics by using infrared sensors and machine-learning methods. To deploy efficient strategies for various task requirements and object affordances, a knowledge-driven model is developed to resolve the planning of dexterous manipulation motion. The model represents human knowledge as hand topology and learns task-oriented manipulation plans to adapt to different work context. Due to the integrated structure of knowledge and learning, the model produces robust and adaptive plans with a high efficiency in learning.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.