CAREER: Quantifying Humanlike Enveloping Grasps

点击翻译按钮获取中文摘要

Over the past decade, our ability to produce graphicalimages has improved to the extent that we can createimaginary scenes that are virtually indistinguishable fromreality. Digital humans have been called the last frontierin this march to graphical realism, and the area of humananimation has also seen dramatic developments. Increasinguse of motion capture data and new techniques formanipulating that data allow us to reproduce human motion atan extremely high level of fidelity. Graphically generatedcharacters in video games and films can seem uncannily real.Pending the development of easy-to-use tools for directingdigital humans, we should soon see digital humans asplausible user interfaces, and animated characters willbecome much more prevalent in education, demonstration, andtraining applications. If digital humans are the lastfrontier in realistic computer graphics, the last frontierin realistic digital humans is generating believable handmotion. Human hands are beautiful and complex mechanisms,amazing in their utility and adaptability. It is argued thatit is our hands that make us human, and that hand evolutionwas a primary factor in the development of intelligence.Hand use in autonomous digital human characters, however, isgenerally quite unconvincing. Hands may be placed in asingle frozen pose, and interaction between characters andobjects is avoided when possible. The main problem is thatgeometric models of the human hand have far too muchflexibility. This flexibility makes working with handsdifficult even for trained animators, and it poses atremendous challenge for creating autonomous characters thatmust interact with their environment. I believe that the keyto making further progress in hand motion for digitalcharacters is much more detailed consideration of theanatomy of the human hand. In analysis of human grasps, forexample, critically important considerations include theamount of contact between finger pads, palm, and object; theability of muscles to produce or resist task force; and thestabilization roles of fingers and muscles, yet none ofthese issues have been explored in grasp synthesis researchin either the robotics or computer graphics communities. Inpursuit of the goal of believable hand use for digitalcharacters, we propose an anatomy-based model of humangrasping. In particular, we propose a tendon-based qualitymeasure for humanlike enveloping grasps, and we plan toevaluate this quality measure (1) for ability todiscriminate between grasps, (2) as a predictor of graspforces, and (3) for use in modeling grasp acquisition.Because of the strong emphasis on human anatomy, thisresearch has the potential for additional impact outsidegraphics and animation in areas including ergonomics (tooldesign), robotics (robot hand design), and anthropology(research in human hand evolution and tool use). Theeducational portion of this proposal focuses on teaching andmentoring of undergraduates. The research ideas, techniques,and results will be incorporated into a course at Brown thatattracts both un-dergraduate and graduate students, and willprovide them with an opportunity to learn and experiment ina problem domain that is a nice mix of computationalgeometry and numerical optimization, grounded in humananatomy and supported by data. A special effort will be madeto include undergraduate women in the research program, forexample, through the CRA Distributed Mentor Program.Research results will include a library of example graspsand applied forces, as well as a tool for adapting theexamples to new hand and object geometries. Once thisresearch is published, the data and tools will be madeavailable to other researchers on the web, and should serveas a useful resource for creating digital characters foreducation, entertainment, and training applications.