Manipulation

Manipulation tasks live in src/unilab/tasks/manipulation/.

In-Hand

  • allegro_inhand and allegro_inhand_grasp have MuJoCo and Motrix PPO owners.

uv run train --algo ppo --task allegro_inhand --sim mujoco
uv run train --algo ppo --task allegro_inhand --sim motrix training.no_play=true

Platform Balancing

stewart_balance is a 6-DOF parallel (Stewart) platform that balances a free ball on its top plate. The policy commands a 2-D plate tilt (roll, pitch); an inverse-kinematics step converts the commanded plate pose into the six prismatic leg lengths that the position actuators track. The reward combines centering, zero-velocity progress and a stillness bonus, with a fall penalty; an episode ends on a fall or on sustained-still success.

The base is welded to the world. Motrix is the validated training backend; the mujoco owner constructs and steps, but its stiff closed-loop solver is not yet training-stable under load.

uv run train --algo ppo --task stewart_balance --sim motrix training.no_play=true