Admittance Versus Impedance Control Architecture Guide

When designing joints for a humanoid robot or cobot, choosing the right force control architecture is critical. The debate of admittance versus impedance control isn't just a software problem—it is fundamentally dictated by your hardware's mechanical properties, such as backdrivability, gear ratio, and sensor placement.

Use the interactive architecture selector below to determine the optimal control scheme for your specific robotic joint hardware.

Interactive Architecture Selector

Control Architecture Selector

Screen admittance versus impedance control from the hardware facts that usually decide the architecture: reduction, sensing, compliance, and backdrivability.

Load sample Clear

Gear RatioSelect reduction rangeHigh ratio, strain wave/cycloidal >50:1Low ratio, QDD/planetary <20:1Direct drive or cable drive, near 1:1

Use the output reduction seen by the joint, not just motor gearhead data.

External Force/Torque SensorSelect force sensingYes, measured external force is availableNo, current/encoder signals only

Count joint torque sensors or wrist/end-effector 6-axis F/T sensors.

Physical ComplianceSelect mechanical complianceStiff / rigid transmission pathCompliant joint or SEA element

Describe the mechanical path before software compensation.

BackdrivabilitySelect backdrivabilityLow / hard to move by handHigh / transparent at the output

Use measured reflected inertia/friction if you have test data.

Selector result

Admittance Control

Best fit for stiff, non-backdrivable hardware that can measure external force and command a position loop.

91% screening confidence, not safety certification

Conditions and limits

• Use the force signal as the input and output a compliant position reference.
• Keep the inner position loop stiff but monitor contact latency.
• Validate virtual mass/damping against the stiffest expected environment.

Main failure mode

Unexpected impacts can arrive faster than the force-sense, compute, and position-track chain can react.

Next action

Specify F/T sensor range, noise, and sampling rate before finalizing gearbox and controller tuning.

Email notes WhatsApp reviewDiscuss integration

---

Executive Summary: Key Conclusions

When evaluating admittance versus impedance control, keep these four conclusions in mind. Literature checked on July 24, 2026 supports the causality distinction; the bandwidth ranges below are engineering screening targets that still require actuator-level measurement.

1. Hardware dictates the control scheme, not the other way around. You cannot force high-bandwidth impedance control onto a heavily geared, non-backdrivable harmonic drive without severe stability issues.
2. Bandwidth and latency boundaries must be validated on the exact joint. Torque-loop impedance can target higher contact bandwidth when the actuator is transparent. Admittance is usually lower because force sensing, filtering, admittance calculation, and inner position tracking happen in series.
3. Admittance requires sensors; Impedance requires transparency. Admittance control fundamentally relies on high-quality external force/torque sensors (6-axis F/T). Impedance control relies on backdrivability and torque transparency (e.g., QDD motors with <20:1 gear ratios).
4. Safety responses differ fundamentally. Impedance control is physically yielding by nature (like a spring). Admittance control is simulated yielding and is bottlenecked by sensor-to-motor latency during unexpected, high-frequency impacts, making it prone to "bouncing" against stiff environments.

Visualizing the Control Loops

Understanding the signal flow is the easiest way to grasp the difference between these two paradigms.

Evidence and Core Methodologies

1. Admittance Control (Force $\rightarrow$ Motion Causality)

In admittance control, the robot is fundamentally position-controlled. It reacts to forces and yields by moving.

• The Input: The system measures external forces using a 6-axis force/torque sensor or joint torque sensors.
• The Math: The controller applies an admittance model (usually M*x'' + B*x' + K*x = F_ext) to translate that force into a new desired position.
• The Execution: A stiff, high-gain inner position loop drives the motor to that new position.
• Applicability & Limits: Ideal for systems with high friction or high gear ratios (e.g., >50:1 strain wave or cycloidal reducers). These systems are difficult to backdrive, so measured force is converted into a motion command. The serial chain of measure force $\rightarrow$ filter $\rightarrow$ compute $\rightarrow$ actuate position limits contact response and must be measured on the real drive.

2. Impedance Control (Motion $\rightarrow$ Force Causality)

Introduced fundamentally by Neville Hogan (1985), impedance control treats the robot as an equivalent mass-spring-damper system.

• The Input: The system calculates the error between the desired trajectory and the actual position.
• The Math: The controller applies an impedance model (F_cmd = K(x_des - x) + B(x_des' - x')) to compute a required restorative force.
• The Execution: The motor driver directly applies this calculated torque (often using Field Oriented Control and current sensing).
• Applicability & Limits: Ideal for direct drive or quasi-direct drive (QDD) systems (e.g., planetary gears <20:1). Low friction makes motor current a more useful output-torque proxy, enabling faster force interaction than a sensor-to-position admittance chain when torque calibration and damping are done correctly.

What the Selector Is Actually Testing

The selector intentionally shows boundary and instrumentation gap states. Those states are not failures of the tool; they are the cases where a procurement or controls team should pause before selecting motors, reducers, sensors, and drive electronics.

Selector Output to Validation Record

Do not copy the selector answer into an RFQ as a final decision. Convert it into a validation record that a supplier or controls team can reproduce on the actual joint. Pair this table with the humanoid joint RFQ checklist before freezing a motor, reducer, encoder, or force-sensor package.

Quantitative Comparison and Risk Trade-offs

Contextual Risks & Mitigation Strategies

1. High-Frequency Impact (Legged Robotics / Locomotion) If you are building a legged robot that experiences harsh, sudden impacts with the ground, pure admittance control is highly risky. The latency of the sensor-compute-actuate loop cannot react fast enough to protect rigid gearboxes from shock loads. Impedance control (or mechanical compliance like Series Elastic Actuators) is strongly recommended for impact-heavy applications.

2. Stiff Environment Interaction (Precision Assembly/Machining) When a robot makes contact with a rigid surface (e.g., metal-on-metal insertion), admittance control often struggles with stability ("bouncing"). The delay in force sensing and position tracking causes the controller to overshoot. Mitigation: Lower the virtual mass and increase damping in the admittance model, though this makes the robot feel sluggish.

3. Unconstrained Space (Free Space Motion) Impedance control can become unstable or underdamped in free space if not tuned correctly, because it relies on the environment to provide counter-forces. Mitigation: Controllers often implement a hybrid approach—stiff position control in free space, smoothly blending into impedance control upon contact detection.

Evidence Scope, Sources, and Limits

This report was reviewed on July 24, 2026. The sources below support the control-causality framing and the known stability/passivity concerns. The bandwidth numbers in this page are not universal standards; they are early architecture screening targets that must be replaced by measured data from the selected actuator, drive, sensor, controller rate, and contact fixture.

• Hogan, N. (1985). Impedance Control: An Approach to Manipulation: Part I - Theory. Journal of Dynamic Systems, Measurement, and Control. ASME DOI: 10.1115/1.3140702.
• Keemink, A. Q. L., van der Kooij, H., & Stienen, A. H. A. (2018). Admittance control for physical human-robot interaction. The International Journal of Robotics Research. DOI: 10.1177/0278364918768950.
• Design-screening assumption used here: admittance contact response is penalized by external force sensing, filtering, virtual admittance integration, and position-loop tracking; impedance response is penalized by actuator transparency, current-to-torque calibration, damping, and thermal limits. Validate both with closed-loop bandwidth, impact, and hard-contact stability tests.

Decision Limits and Procurement Checklist

Before choosing a joint module for admittance versus impedance control, ask for these data points from the actuator supplier:

Next Steps for Your Hardware Sourcing

Choosing between admittance versus impedance control dictates your BOM (Bill of Materials). If you choose admittance, you must budget for high-resolution force sensors and stiff gearboxes. If you choose impedance, you need high-torque-density motors and transparent, low-ratio gearboxes.

If you are currently evaluating hardware for your next robotic platform, we can help you align your mechanical constraints with your control software goals.

• Email us: [email protected]
• WhatsApp: Request an engineering sales review

(Looking for broader context? Read our comprehensive guide on Impedance Control in Robotics, or compare the older terminology variant admittance control vs impedance control.)

FAQ: Admittance Versus Impedance Control

What is the shortest practical difference?

Admittance control measures force and outputs motion. Impedance control measures motion error and outputs force or torque.

When should a high-ratio humanoid joint use admittance?

Use admittance when the reducer is stiff and non-backdrivable, and when a reliable F/T sensor can provide the force input. Without that force input, the control loop has little direct evidence of contact force.

When should a low-ratio QDD joint use impedance?

Use impedance when the actuator is backdrivable, torque/current calibration is reliable, and the drive can command torque with enough bandwidth for the contact task.

Can the architecture be hybrid?

Yes. A common pattern is position control in free space, impedance behavior during fast contact, and admittance behavior when a stiff transmission needs force-guided motion.

Are 20-30 Hz and >100 Hz guaranteed bandwidth values?

No. They are screening values for early discussion. The only defensible value is the measured closed-loop contact bandwidth of your actuator, sensor, drive, and environment.

What is the first test before RFQ approval?

Run a low-energy contact test that logs force, motor current, output position, command latency, and temperature. The test should include the stiffest expected environment and at least one free-space/contact transition.