Whole Body Control (WBC) is a control framework for legged and humanoid robots that simultaneously optimises motion tasks, contact forces, and balance constraints across all degrees of freedom by solving a hierarchical quadratic programme or weighted task-space objective at each control cycle. By treating locomotion, manipulation, and postural balance as a unified optimisation problem, WBC avoids the sub-optimality of treating these objectives separately. It typically relies on a rigid-body dynamics model and is often combined with model predictive control or reinforcement learning policies for online adaptation to uneven terrain and external disturbances.
Semantic Classification
Content
Whole Body Control addresses the coordination challenge that arises in robots with many degrees of freedom: locomotion, postural balance, and end-effector tasks all compete for the same actuator bandwidth, and treating them with decoupled controllers leads to conflicts and sub-optimal behaviour. WBC formulates control as a constrained optimisation problem—typically a hierarchical quadratic programme (HQP)—where each task (maintain CoM height, swing foot trajectory, arm endpoint position) is assigned a priority level, and lower-priority tasks are solved in the null space of higher-priority ones.
The optimisation is grounded in the rigid-body dynamics equation τ = M(q)q̈ + C(q,q̇) + g(q), where τ are the generalised joint torques, M is the mass-inertia matrix, C captures Coriolis and centrifugal effects, and g is the gravity term. Contact constraints (foot-ground, hand-surface) are incorporated as equality or inequality constraints on the contact forces, enforced within the friction cone.
Accurate system identification of inertial parameters is a prerequisite for WBC: errors in link masses or inertia tensors cause the feedforward torque predictions to be incorrect, degrading tracking quality. In practice, WBC is run at high frequency (250–1000 Hz) with a slower outer loop—either an MPC providing reference trajectories or a learned policy providing task-space goals—forming a two-layer control hierarchy. Whole body control has been demonstrated on platforms such as the Boston Dynamics Atlas, ETH ANYmal, and MIT Cheetah robots.