humanoidsdata.com

Search

Search datasets, articles, and glossary terms for humanoids and embodied AI.

← All glossary terms

Hardware & control

Impedance control

Impedance control regulates the dynamic relationship between a robot’s motion error and the force or torque it exerts, usually by giving the robot virtual stiffness, damping and sometimes inertia. It lets an end effector or limb yield predictably during contact instead of trying to hold an exact position regardless of interaction forces.

Also known as: Robot impedance control, Mechanical impedance control

Updated

A controlled relationship, not one variable

Neville Hogan’s original impedance-control theory argues that manipulation cannot be described by controlling position or force alone: the robot and environment are mechanically coupled. The controller instead imposes a chosen relation between motion deviation and interaction force.

A virtual spring-damper is the common simple case. Its equilibrium pose sets where the robot tends to return, stiffness sets how strongly displacement is opposed, and damping resists relative velocity. A fuller impedance can include inertia, coupling between axes and nonlinear behaviour. The official libfranka Cartesian example maps Cartesian pose and velocity errors through stiffness, damping and the Jacobian to joint torques.

Impedance, admittance and position control

Direct impedance control commonly takes motion as input and produces force or torque. Admittance control takes measured force as input and produces a motion command, often for an inner position loop. An ICRA treatment of both implementations explains why their causality and contact behaviour differ, although some literature uses “impedance control” as the umbrella term.

A position controller may feel stiff, but that alone does not specify a desired mechanical impedance. The distinction matters when comparing controllers or judging how a robot should respond to a push.

What humanoid data must include

Humanoids may use different impedance along different axes, limbs and contact phases. A position trajectory therefore captures only the equilibrium motion, not the intended compliance.

Training records should include stiffness, damping and inertia parameters where used, their coordinate frame, equilibrium targets, torque output, measured wrench, contact state, saturation and update rate. Without them, two demonstrations with the same pose path but very different contact behaviour look falsely equivalent.

Sources