Communications of the ACM
Framework Allows Robots to Perform Interactive Tasks in Sequential Order
Interactive navigation requires a robot to reach a goal location while interacting with obstacles on the way, which has proven to be the most difficult for robots to learn.
Credit: Georgia Tech Research
A framework developed by Georgia Institute of Technology Ph.D. student Niranjan Kumar allows quadruped robots to perform tasks that get progressively more complex without having to relearn motions.
The Cascaded Compositional Residual Learning (CCRL) framework acts as a library where each new skill learned by the robot is added and then accessed to achieve more complex skills.
The framework was demonstrated by a robot using energy transfer to open a heavy door.
Currently, a robot can learn and deploy 10 skills using the CCRL.
Kumar said, "It just takes longer to train as you keep adding more skills because now the policy also has to figure out how to incorporate all these skills in different situations. But theoretically, you can keep adding more skills indefinitely as long as you have a powerful enough computer to run the policies."
From Georgia Institute of Technology
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA
No entries found