Communications of the ACM
One Giant Leap for the Mini Cheetah
The researchers used reinforcement learning to train the high-level controller.
Credit: MIT News
A team of researchers from the Massachusetts Institute of Technology (MIT), Arizona State University, and the University of Massachusetts at Amherst developed a new control system that enhances the speed and agility of legged robots as they leap across gaps.
The control system algorithmically processes and translates real-time forefront video input into instructions for bodily movement.
The researchers combined the best elements of controllers that do not incorporate vision into a separate module that handles vision in real time, and trained the controller through reinforcement learning.
Tests of the system when installed into MIT's mini cheetah robot found it outperformed other systems that use a single controller, enabling it to successfully cross 90% of physical terrains.
From MIT News
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found