The Complete Guide to Actuator Selection for Humanoid Robotics

From Micro Actuators to Full-Scale Joint Systems

Actuator selection for humanoid robotics is the engineering process of matching each joint's force, speed, weight, and packaging requirements to a compact electromechanical actuator that can deliver human-scale motion without the weight or rigidity of industrial drives.

Designing a humanoid robot is an exercise in managing compromise. Every design choice sits on a sliding scale: Torque vs. Weight, Speed vs. Precision, and Power vs. Heat.

In traditional industrial robotics, these compromises are solved with massive stationary power supplies and heavy cast-iron gearboxes. Humanoid robotics does not have that luxury. The most consequential decision in the design process is selecting the correct actuator for each joint—finding the "sweet spot" where biological form meets electromechanical function.

This guide details the engineering framework for selecting actuators that solve the "Packing Problem" while delivering human-scale performance.

In tendon-driven humanoid joints, the actuator pulls in a straight line and the mechanism handles the geometry. Route cables so the actuator never sees side load — bending forces transmitted back into a micro actuator shorten its life far faster than peak axial load ever will.

"On a humanoid hand, the actuator should never be the heaviest thing on the limb. Move the motor up the arm, run the load through a cable, and size for peak torque, not average — a stalled motor on a finger is a thermal failure waiting to happen." — Robbie Dickson, Founder and Chief Engineer of FIRGELLI Automations

1. How do actuators drive humanoid robots?

Industrial robots rely on high-RPM rotary motors and harmonic drives. While precise, they are heavy and rigid. Human-scale robots require compliance, compactness, and distributed actuation.

The shift is moving toward linear micro-actuation, specifically using the Tendon-Driven Configuration. By utilizing FIRGELLI Micro Pen Actuators (Series FA-BS16), engineers can decouple the motor from the joint.

• Stationary Actuator Placement: Motors are mounted in the "forearm" or "calf," not the joint itself.

• Inertia Reduction: This removes weight from the distal (moving) end of the limb, allowing for faster acceleration with less power.

• Bio-Mimetic Transmission: Force is transmitted via Bowden cables or linkages, mimicking the biological flexor/extensor tendon system.

2. What are the force, speed, and control tradeoffs?

Physics imposes a strict tradeoff: Force and Speed are inversely related. You cannot optimize for both without drastically increasing motor size (and weight).

Using the FIRGELLI FA-BS16 performance data as a reference, we can categorize actuators by biological function:

Biological Role	Actuator Profile	Performance Data	Ideal Application
Fast-Twitch (Speed)	High Speed / Low Force	20N @ 15mm/s	Eyelids, Facial Expressions, Fast Fingers
Balanced Muscle	Medium Speed / Medium Force	50N @ 5.8mm/s	Thumb Opposition, Wrist Deviation
Slow-Twitch (Power)	Low Speed / High Force	100N @ 3mm/s	Locking Grips, Latches, Joint Stability

Engineering Rule: Always size your actuators for the Peak Torque requirement of the joint, not the average load. It is better to have overhead than to stall a motor and trigger a thermal cutout.

3. How do you calculate joint torque requirements?

To select the right micro-actuator, you must translate "Force" (Linear) into "Torque" (Rotational).

Formula: Torque (τ) = Actuator Force (F) × Lever Arm Distance (d)

Design Example: Robotic Index Finger

• Scenario: You need a pinch force of roughly 7N at the fingertip. Through the mechanical linkage, this requires a tendon pull of 50N.

• Actuator Selection: 50N Model (11 lbs).

• Lever Arm ($d$): The tendon attaches 15mm (0.015m) from the knuckle pivot.

• Calculation: $50N \times 0.015m = \mathbf{0.75 Nm}$

Result: A 0.75 Nm torque output matches human-scale gripping requirements for handling tools, electronics, or household objects.

4. Why do micro actuators solve the "packing problem"?

The human arm is a marvel of packaging, containing dozens of muscles within 2–3 liters of volume. Standard "box" servos are difficult to pack densely because of their protruding motors and gear trains.

FIRGELLI micro actuators solve this via Inline Geometry:

• Diameter: 16 mm (0.63").

• Stackability: The cylindrical shape allows them to be bundled together like muscle fibers. You can fit a cluster of 4-5 actuators (controlling an entire hand) inside a chassis no wider than a human wrist.

• Weight: With units weighing between 49g and 81g, a full 5-finger actuation bank weighs less than 350g—lighter than many single industrial grippers.

5. What design patterns are proven in humanoid actuation?

A. Forearm Topology (Remote Actuation)

• Design: Mount a bank of 5 actuators near the elbow.

• Routing: Run Bowden cables (ptfe-lined housings) through the wrist to the fingers.

• Benefit: Keeps the hand lightweight, reducing the torque load on the shoulder and elbow motors.

B. Feedback-Driven Gripping (Haptic Intelligence)

• Logic: Drive the actuator until the Hall Sensor pulses stop changing (indicating the finger has hit an object), but the Current rises (indicating force is being applied).

• Action: The controller detects this state and automatically switches from "Position Mode" to "Force Maintenance Mode," holding the object securely without crushing it.

What usually goes wrong?

1. Thermal stall: Sizing for average load instead of peak torque leaves no headroom. When the joint hits a hard stop or grips a stiff object, current spikes and the motor's thermal cutout triggers — interrupting motion mid-task.
2. Side-loaded micro actuators: A tendon that pulls off-axis (because of poor cable routing or a misaligned pulley) loads the actuator screw radially. Micro pen actuators are designed for axial pull, and side load shortens their life dramatically.
3. Bowden cable friction creep: PTFE-lined housings work well until they're bent too tight, contaminated, or routed across a moving joint with too small a bend radius. Friction climbs over cycles, the actuator delivers less force at the fingertip, and grip strength drops.
4. Crush failure under pure position control: Driving to a target position without monitoring current means the actuator will keep applying force into a fragile object until it stalls. Force maintenance mode (current rising while Hall pulses stop) is what prevents this.
5. Distal-mounted motors: Putting an actuator at the joint instead of remote-mounting in the forearm increases the inertia the shoulder and elbow have to accelerate — every gram on the hand multiplies upstream torque demand.

How should you test it before trusting it?

1. Cycle under real load, not bench-pull: Run the joint through its full duty cycle gripping the heaviest expected object. A pinch force that works once is not the same as 10,000 grips. Look for current creep across cycles — a sign of mechanical binding or screw wear.
2. Verify the position/force handoff: With the Hall sensor logging pulses and the controller logging current, command the actuator to close on a stiff object. Confirm that pulses stop and current rises in the same cycle — that's the signal your force maintenance mode depends on.
3. Measure the hard part of travel: Peak load on a finger usually occurs at the end of close, not the middle. Test torque output at the closed-fingertip position with the lever arm fully engaged, not at mid-stroke where geometry is most favorable.
4. Thermal soak the worst case: Run repeated grip-and-hold cycles at peak force at the upper end of expected ambient temperature. The thermal cutout protects the motor — but if it trips during normal operation, your duty cycle is wrong, not the actuator.
5. Validate cable routing at end-of-stroke: Flex the wrist through its full range while the actuator holds force. Watch for tendon binding, housing kinks, or any motion the cable transmits back as side load on the actuator body.

6. Where is humanoid actuation headed next?

Actuators are evolving from "dumb" movement sources into intelligent edge devices.

With the integration of high-resolution Hall Effect sensors (up to 1143 pulses per inch) and precise current-force linearity, engineers can now design robots that "feel" their environment.

Final Thought:

Humanoid robotics will not be won by better AI code alone. It will be won by the teams that master human-scale actuation—building bodies that are as capable and adaptable as the minds controlling them.

 

Industries: Humanoid robotics, service robotics, prosthetics and bionics, research robotics, animatronics.

Mechanisms: Tendon-driven actuation, Bowden cable transmission, micro linear actuators, Hall-effect feedback control, current-based force sensing, remote (proximal) actuator placement.