II-NEW: Robot Arms for Interactive Perception of Highly Non-Rigid Objects

点击翻译按钮获取中文摘要

This infrastructure proposal, acquiring two Kinova JACO Research arms to upgrade aging mobile manipulators, will enable fundamental research in interactive perception with non-rigid objects. Traditionally, robotics has focused upon rigid objects. The PI team proposes to enable study of the problem of sifting through a cluttered pile of non-rigid objects, many of which are occluding one another, and properly classifying and manipulating each object. The proposed work is not merely concerned with computing grasp points of rigid objects for which training examples are available for off-line learning; rather the focus is upon segmenting, grasping, and interacting with non-rigid objects for which models are difficult to construct. The aim is to merge ideas from interactive perception with those of the textile manipulationn to learn object properties over time and interact with objects in unstructuredenvironments.Non-rigid objects are important in a number of highly-visible industries, such as textiles and automotive manufacturing, so broader impacts are numerous. Moreover, automated laundry folding will have a significant impact on service robots, where products such as vacuum cleaning robots and lawn mowing robots have already reached the commercial sector. The findings of the research will be made available to the research community via conference proceedings and journal publications, and the research will be integrated into existing courses to reach undergraduate and graduate students. Presentations to nearby schools to introduce disadvantaged and underrepresented high school students to STEM careers is included in the outreach plan.