ETH Zurich RAMBO robot Unitree Go2
by Terrymiller: Science
ETH Zurich researchers have created RAMBO, a control system that lets legged robots move and manipulate objects at the same time. By merging a model-based controller with reinforcement learning, the Unitree Go2 can push a shopping cart, balance a plate, and grip soft objects.
Transcript (en)
Researchers at ETH Zurich have developed Rambo, a new control system that enables lagged robots to move and manipulate objects simultaneously. Rambo combines a model-based controller, which plans precise movements, with a reinforcement learning-based feedback controller that enhances the robot's stability and responsiveness. With the help of Rambo, the Unitree Go 2 successfully performed real-world tasks such as pushing a shopping cart, balancing a plate, and gripping soft objects, operating in both quadrupedal and bipedal modes.
