Toward Telelocomotion

Опубликовано 7 октября 2016, 19:39

Human interaction with the physical world is increasingly mediated by automation -- planes assist pilots, cars assist drivers, and robots assist surgeons. Such semi-autonomous machines will eventually pervade our world, doing dull and dirty work, assisting the elderly and disabled, and responding to disasters. Recent results (e.g. from the DARPA Robotics Challenge) demonstrate that, once a robot reaches a task area and grasps the necessary tool, handle, or wheel, they are able to plan and execute whole-body motions to accomplish complex goals. However, robots frequently lose their balance and fall en route to tasks, necessitating human supervision and intervention. Integrating legged machines in daily life will require safe and stable telelocomotion, that is, robot ambulation guided by humans. This talk tackles the telelocomotion problem from the top-down and bottom-up, leveraging the theory of human motor control to design provably-safe interfaces for nonlinear control systems, and deploying series elasticity and backdrivable actuation to design intrinsically-stable gaits and maneuvers for legged robots.

See more on this video at microsoft.com/en-us/research/v...