A humanoid may need to walk toward an object, shift its stance, keep balance, reach with both arms and manipulate the object. Whole-body control coordinates those movements under shared physical constraints.
Why is whole-body control difficult?
Every movement changes balance, reachability and contact forces. A useful controller must handle many joints, changing contact with the floor, obstacles and objects while continuously updating from sensor feedback.
Whole-body control vs manipulation
Manipulation can focus on an arm, hand or gripper. Whole-body tasks add locomotion and posture: the robot may have to move its feet or torso simply to place its hand where it needs to be.
What changed in 2026?
Google DeepMind says Gemini Robotics 2 expands from upper-body tabletop tasks to controlling full humanoids from feet to fingertips, including walking, crouching, stretching and manipulation. Figure’s Helix 2.5 similarly evaluates long-horizon tasks that combine locomotion, bimanual manipulation, active perception and whole-body control across unseen homes.
Why does it matter?
General-purpose robots need to operate in spaces built for people. Coordinated whole-body intelligence is therefore a key step from fixed industrial automation toward robots that can perform varied tasks in homes, factories and logistics environments.
Related terms
Sources
Physical-AI terminology is evolving quickly. We use the broader technical meaning and distinguish it from individual vendors’ model names and claims.