A "blind" bipedal robot trained in a simulator by Oregon State University (OSU) researchers can negotiate varying terrain, including the climbing of stairs.
The researchers applied sim-to-real Reinforcement Learning to establish how Agility Robotics' Cassie robot would ambulate.
The OSU team taught Cassie virtually to manage various situations, including stairs and flat surfaces. In real-world tests, the robot could handle curbs, logs, and other uneven terrain it had never encountered before, and ascended and descended stairs with 80% and 100% efficiency, respectively.
According to the researchers, "This work has demonstrated surprising capabilities for blind locomotion and leaves open the question of where the limits lie."
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA