Communications of the ACM
AI Researchers Challenge a Robot to Ride a Skateboard in Simulation
A new framework for controlling four-legged robots is more energy-efficient and adaptable than traditional model-based gait control.
Credit: Starleth
Researchers from Nvidia, the California Institute of Technology, the University of Texas, Austin, and the Vector Institute at the University of Toronto in Canada have developed a framework for controlling four-legged robots that is more energy-efficient and adaptable than the traditional model-based gait control.
The researchers demonstrated that the framework adjusts to conditions in real time by causing the system to slip on frictionless surfaces, ride a skateboard, and climb on a bridge while walking on a treadmill.
The quadruped model is trained in simulation, and that training is transferred to a Laikago robot in the real world.
The framework combines a high-level controller that uses reinforcement learning with a model-based lower-level controller.
Said the researchers, "By leveraging the advantages of both paradigms, we obtain a contact-adaptive controller that is more robust and energy-efficient than those employing a fixed contact sequence."
From Venture Beat
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found