Computational Models of Humans

点击翻译按钮获取中文摘要

My long term goal is to create useful computational models of how humans move and physically interact with their environment. The focus is not on black-box modeling of movement data. Rather, it is on modeling the underlying system itself, as a biological machine: the motors (muscles), sensors (esp. mechanoreceptors in muscle and skin), structures (skeleton and connective tissues), and software (neural control). By modeling the underlying sensorimotor system we could achieve better generalization to novel situations and gain fundamental insights into biological principles. In the near term, we could create more believable computer animations of human characters. Specifically, I propose to focus on the following. (1) EULERIAN SOFT TISSUE SIMULATION. Eulerian discretization has shown great promise in my previous work, but constitutes a radical departure from the standard Lagrangian approach. I will systematically explore its strengths and limitations, and identify the significant advantages that could make it the future standard for soft tissue simulation. (2) REALISTIC SKIN MODEL. It is not sufficient to have good algorithms; a model must also reflect real human physiology. My goal is to develop a simulation-ready structural and statistical model of the mechanical behavior of real human skin, based on in vivo measurements. (3) HAND SYSTEM. Modeling the hand as a sensorimotor system with many interacting parts will provide deep insights into why it is so effective. I will focus on the hand since it is our primary effector for interacting with the physical world (including computers) and the key role of skin and soft tissues in hand function. Such virtual human models are required in an enormous range of applications. A central goal of computer animation is the synthesis of human characters interacting with their environment. Realistic models of human movement are also important in biomechanics, neuroscience, and medicine, and could lead to new robot designs based on biological principles. Students and postdoctoral trainees are involved in all aspects of this research, and will acquire a broad range of skills in an interdisciplinary environment, including development of advanced simulation software systems, measurement and analysis of human soft tissues with world class instrumentation, biomechanics, and computer graphics.