Data-Driven Control of Humanoid Robots

点击翻译按钮获取中文摘要

Abstract IIS-9900333Hodgins, JessicaGeorgia Institute of Technology $85,061 - 12 mos.Data-driven Control of Humanoid RobotsThis is the first year funding of a three year continuing award. Robots and simulations with the complexity of a rigid body model of a human are difficult to control because of the need for coordinating many degrees of freedom. Some applications also require that the resulting motion be natural looking or appealing, further complicating control. We are exploring a data-driven approach to this problem. During the time frame of this proposal, we will collect significant quantities of data from human subjects using motion capture systems construct data-driven models of human movement for a variety of tasks,and use the models to demonstrate robust, appealing motion of simulated figures and a physical humanoid robot.