An mjlab integration for the Booster K1: robot model, training tasks and an AMP motion-prior pipeline, from simulation to the real robot.
Locomotion from the AMP velocity task, acrobatics from motion tracking
AMP velocity-tracking policy on the Booster K1 after 30k training steps
The AMP velocity policy running live in MuJoCo, steered with a joystick, keyboard or gamepad
vx 0.00 m/s vy 0.00 m/s yaw 0.00 rad/s
Push the stick forward to walk and sideways to turn, or click the scene and use WASD to move and QE to turn. Gamepads work too. Drag with the mouse to orbit the camera.
Locomotion clips from LAFAN1, retargeted onto the parallel-ankle K1. They drive the AMP motion prior, and the built-in motion viewer lets you browse, play back and edit clips before training.
booster_mjlab is a whIRLwind project. Our mission is to push the boundaries of robotics and AI in robot football. We are always looking for interested students or collaboration partners to join us. Learn more at whirlwind.team.
