Whole-Body Control

Whole-body control studies how robots coordinate many joints and contacts to move, manipulate objects, and respond to disturbances as one system. Model-based methods use whole-body optimization, model-predictive control, and operational-space control, often with safety constraints; learning-based methods use reinforcement learning, imitation, and teacher–student policies. Work spans humanoids, legged mobile manipulators, and cooperative robots, with evaluations in locomotion, contact-rich manipulation, teleoperation, and long-horizon skill composition, including transfer from simulation to hardware. A central challenge is making controllers both versatile and reliable across changing loads, terrain, contact conditions, and robot embodiments without sacrificing safety or motion quality.