CHOREO: Every Humanoid Skill as a Trajectory

CHOREO addresses the fragmentation problem in humanoid robotics where skills built via reinforcement learning, motion imitation, and generative modeling remain incompatible due to divergent representations, interfaces, and controllers. The framework leverages a key observation that any learned skill can be expressed as a trajectory, enabling training-free composition across heterogeneous skill sources. This approach applies to multibody dynamics involving contact mechanics and robot motion control.