It Takes a Lazy Robot to Move Like a Human
Honda’s Humanoid Robot Project recruited Stanford Robotics Professor Oussama Khatib to research just how lazy a robot would have to be to move like a human being.
By observing a group of Stanford students and a tai-chi master, according to the physorg article, Khatib’s team came to the conclusion that humans learn how to move by recognizing which positions and motions cause them discomfort, and then avoiding those positions.
Using these observations, they’ve developed a computationally lightweight motion model based on the principal of continuously minimizing muscular effort, without pre-planning the entire motion itself.
They’re now applying this model both to a simulated “lazy” robot known as StanBot, and hope to use it in an upcoming version of Honda’s Asimo. Because this approach is less computationally intense than current methods, Khatib’s team hopes it will give Asimo, and other robots, a greater ability to manipulate it’s environment in real-time, while moving though it.
to Mr. Future's RSS Feed