Scalable physics-based modeling of skilled movement

人类和动物在他们的世界中移动，并以健壮，优雅和高度熟练的方式与环境互动。我们还不知道如何设计逼真的动画角色，生物力学模拟人类，以及能够模仿这种技能的机器人。然而，最近的发展表明，在模拟中，或者在某些情况下，与机器人一起，正在朝着模仿人类一样的灵巧性做出重大的努力。 在这个建议中，我们使用双足步行的鲁棒模拟作为开发更丰富的运动技能集的出发点，包括高度敏捷的步行和跑步，通过城市环境的自然运动，以及灵巧的手操作。 我们的目标是下一代计算机动画软件作为我们的主要应用程序，但我们希望相同的模型也将有助于理解和纠正病理步态，步行假肢的设计，获得对人类运动控制的更深入了解，并创造更熟练的机器人。

Humans and animals move through their world and interact with their environment in robust, graceful, and highly skilled ways. We do not yet know how to design realistic animated characters, biomechanical human simulations, and robots that can mimic this degree of skill. However, recent developments suggest that significant progess is being made towards mimicing human-like dexterity in simulations, or, in some cases, with robots. In this proposal, we use robust simulations of bipedal walking as a point of departure for developing richer movement skill sets, including highly agile walking and running, natural movement through urban environments, and dexterous hand manipulation. We target next generation of computer animation software as our primary application, but we expect that the same models will also be useful for understanding and correcting pathological gaits, the design of walking prostheses, obtaining a deeper understanding of human motor control, and creating more skillful robots.