The ballbot work focuses on a compact but difficult control problem: how to recover useful control inputs when the robot and ball dynamics are tightly coupled and the contact must satisfy unilateral and friction constraints.
The project uses a projected task-space formulation to decompose the robot and ball dynamics, then computes contact forces through constrained optimization before reusing those forces in the robot dynamics for control. That makes the project a strong example of the broader site theme: physically meaningful control laws that stay grounded in the contact model rather than abstract kinematics alone.
It also gives the new homepage a more complete picture of the research arc by showing that the balance-control work was not limited to humanoid walking.