The four-legged DribbleBot robot developed by Massachusetts Institute of Technology (MIT) researchers can dribble soccer balls on varying terrain using onboard sensors, computing, and reinforcement learning.
DribbleBot learns in a digital twin of the natural world, where 4,000 versions of the robot are modeled in parallel in real time.
MIT's Gabe Margolis said the robot learns successful dribbling behavior by receiving an appropriate reward, then practices for hundreds of days in simulation, or a couple of days in real time.
Over time it has learned to navigate unfamiliar terrain and to recover from falls through a built-in recovery controller.
MIT's Pulkit Agrawal said, "Our goal in developing algorithms for legged robots is to provide autonomy in challenging and complex terrains that are currently beyond the reach of robotic systems."
From MIT News
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA
No entries found