Recent advancements in robotics, highlighted in Science Robotics, have led to the development of more agile and adaptive legged robots capable of navigating unstructured and cluttered environments. A hybrid control architecture that combines reinforcement learning with trajectory optimization has enabled the ANYmal robot to improve its foot placement and agility, as well as recover from slipping. Research on bipedal robots has also made significant strides, with new controllers allowing for robust performance in various tasks such as 400m dashes, running over different terrains, responding to perturbations, targeted jumping, and compliant walking. Additionally, a fully autonomous four-legged robot capable of navigating cluttered environments has been unveiled. Moreover, an open source collision avoidance learning system has been introduced, which aims to ensure that high-speed legged locomotion remains safe by avoiding collisions with obstacles or humans.
“Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion” A new open source collision avoidance learning system. https://t.co/TPVZbUfavA
WATCH: A team of researchers have unveiled a fully autonomous four legged robot capable of navigating cluttered environments https://t.co/HcM34Lqlx8
This is walking robot has six legs https://t.co/FADmvRB0ad
Agile But Safe Learning Collision-Free High-Speed Legged Locomotion https://t.co/gMknUewy7G Legged robots navigating cluttered environments must be jointly agile for efficient task execution and safe to avoid collisions with obstacles or humans. Existing studies either… https://t.co/FGaM7mH9dY
Interested in making your bipedal robots to be athletes? We summarized our RL work to create robust & adaptive controllers for general bipedal skills. 400m-dash, running over terrains/against perturbations, targeted jumping, compliant walking, not a problem for bipeds now.🧵👇 https://t.co/xAqcejqq1T
A hybrid control architecture combining reinforcement learning and trajectory optimization let the ANYmal robot improve foot placement and agility on unstructured terrain—and recover from slipping. Read more in @SciRobotics: https://t.co/m2tL2HKBhI https://t.co/Ikb4fvz8cB