Admittance Control

• Also called: position-based impedance control
• Category: Control › Force & interaction control
• Tags: force-control, compliance, interaction-control, human-robot-interaction, manipulation

A force-control scheme that measures the force applied to a robot and converts it into a motion command, making a position-controlled robot behave like a chosen mass-spring-damper.

Definition

Admittance control makes a robot respond to external forces with a prescribed dynamic behavior, usually that of a virtual mass-spring-damper. The controller measures the interaction force, typically with a force-torque sensor at the wrist or at a handle, passes it through a model such as M ẍ + D ẋ + K x = F_ext, and integrates the result to obtain a position or velocity offset. That offset is added to the reference trajectory and sent to an inner position or velocity controller. In mechanical terms an admittance maps force to motion, and that is exactly what the controller does: it reads force and outputs motion.

Admittance control is the counterpart of impedance control. Impedance control measures motion, as position and velocity error, and commands force or torque, so it needs a robot with good torque control and low friction, such as a torque-controlled or backdrivable arm. Admittance control measures force and commands motion, so it suits robots with stiff, high-gain position control, including most industrial arms with high-ratio gearboxes, whose joint friction would mask small forces. Both aim at the same target behavior, but their practical strengths are opposite. Impedance control performs well when rendering low stiffness and in free motion, and poorly when asked to be very stiff. Admittance control renders stiff behavior well but can oscillate or chatter when the virtual mass and damping are low and the robot touches a stiff environment.

Tuning centers on that trade-off. A low virtual mass and damping make the robot feel light when a person guides it, but they raise the risk of oscillation against rigid contacts and amplify sensor noise. Higher values are stable but feel sluggish. Implementations filter the force signal, compensate for the weight of the tool mounted beyond the sensor, add a deadband to reject sensor drift, and often vary the parameters online depending on whether the robot is in free space, being guided or in contact.

Why it matters

Most industrial and many collaborative arms are position-controlled and do not expose accurate joint torque control. Admittance control is the usual way to give such robots compliant behavior for hand guiding, assembly and contact tasks with an added force sensor, and understanding its stability limits prevents oscillation when the robot meets a rigid surface.

Key applications

• Hand guiding and lead-through teaching of industrial and collaborative arms
• Peg-in-hole and other assembly with force feedback
• Rehabilitation robots and exoskeletons that respond to the user's effort
• Physical human–robot collaboration, such as carrying a load together
• Polishing, sanding and other surface finishing with regulated contact
• Interaction control for drones and mobile robots, where external forces shift a commanded position

In practice

An integrator adds hand guiding to a position-controlled industrial arm so that operators can teach waypoints by pushing the tool. A six-axis force-torque sensor sits between the flange and a handle. In each control cycle, the software subtracts the sensor bias and the gravity load of the handle, applies a small deadband, and passes the remaining force through a virtual mass-damper model with no spring, so the arm moves in the direction it is pushed and stops when released. The resulting velocity is integrated into a position offset and streamed to the robot's position interface. Tests show that the arm chatters when the operator presses the tool against a steel fixture, so the team raises the virtual damping when contact forces exceed a threshold and lowers it again in free motion.

Common misconceptions

• Admittance and impedance control are the same thing — they target the same behavior but in opposite directions: admittance reads force and commands motion, impedance reads motion and commands force
• Lower virtual mass and damping always give better compliance — they make the robot feel lighter but reduce stability in stiff contact and amplify force-sensor noise
• Admittance control needs torque-controlled joints — it is designed for position- or velocity-controlled robots and relies on an external force measurement instead

Who should know this

• Force-control and manipulation engineers
• Integrators adding hand guiding to industrial arms
• Rehabilitation and exoskeleton developers
• Physical human–robot interaction researchers

Related technologies

• Six-Axis Force-Torque Sensor
• Virtual Mass-Spring-Damper Model
• Tool Gravity Compensation
• Position and Velocity Control Interfaces
• Variable Admittance Control

Related terms

• Impedance Control
• Force/Torque Sensor
• Physical Human-Robot Interaction
• Hybrid Position/Force Control
• Kinesthetic Teaching
• Backdrivability

Related projects

• LeRobot: HuggingFace ML library for real-world robotics in PyTorch
• Isaac Lab: NVIDIA robot learning framework: one task API across Newton, PhysX and Isaac Sim
• NVIDIA Isaac GR00T: NVIDIA open foundation model for humanoid robot reasoning and skills
• OpenVLA: 7B-parameter Vision-Language-Action model for robotic manipulation
• Drake: MIT robotics toolbox for model-based dynamics, optimization, and control system design
• OpenArm: 7DOF humanoid arm for physical AI research in contact-rich environments