AeroFlex™ (Coming Soon)
A modular pneumatic artificial muscle system.
Pneumatic artificial muscles (PAMs) are not new. Classic McKibben-style muscles have been around for decades; there are plenty of DIY tutorials floating around, and companies like Festo sell polished versions if you can afford them..
So why are we bothering?
We believe the core tech is good, but the ecosystem around it is not. PAMs are strong, light, and inherently safe around humans, yet most people still reach for a servo or stepper long before they consider a muscle.
AeroFlex™ is our attempt to fix that.
We are trying to build a standardized family of braided pneumatic muscles, hardware, and control tools so you can treat a pneumatic muscle like just another part in your toolkit.
This page is the 101: what PAMs are, why they are useful, and how AeroFlex pushes them further.
What is a pneumatic artificial muscle?
At the simplest level, a classic PAM (and our AeroFlex muscles) is:
A soft inner tube or bladder
Wrapped in a braided sleeve
Hooked up to compressed air
When you pump in air, the inner tube wants to expand in diameter. The braid resists that change in length. The result:
The diameter increases
The length decreases
You get a strong pulling force along the axis
Think "Chinese finger trap, but in reverse."
This basic geometry is sometimes called a McKibben muscle. We are not reinventing that concept. We are standardizing, instrumenting, and packaging it so it is actually practical.
Why PAMs are interesting
High power-to-weight
For their size and mass, braided pneumatic muscles can pull hard and move fast.
Lots of force in a compact package
Great for joints, linkages, and mechanisms where motors and gearboxes can't fit.
If your alternative involves bolting something the size of a car alternator to your robot, PAMs start looking pretty good.
Inherently compliant
PAMs are soft and squishy by default.
Safer interaction with humans
Less damage during collisions
Fewer shattered gear trains when something hits a hard stop
You get mechanical compliance built into the actuator, instead of trying to fake it with springs, flexures, or control loops.
Silent(ish) actuation
Most of the noise in a PAM system comes from the compressor, not the muscle itself.
Compressors can be loud, but you can park them off-board, put them in another room, or bury them in an enclosure so the noisy part is physically separated from the robot or wearable
A buffer tank lets you run short bursts of motion with the compressor off, which makes things much quieter in practice
With good valve selection, reasonable pressures, and a sane plumbing layout, airflow and valve noise stay much lower than a typical motor+gearbox setup
"Robot moves without sounding like a blender" is underrated.
Simple mechanical construction
You do not need a full machine shop to build a basic PAM:
Soft tubing or a custom bladder
Braided sleeve (PET, Nylon, etc.)
End fittings
This makes them friendly to labs, classrooms, and garage projects.
Great for "body scale" motion
PAMs shine in applications that look more like biomechanics than CNC machines:
Exoskeleton joints
Animatronic limbs and faces
Soft grippers and tentacles
Tails, spines, and other creature mechanics
Assistive and wearable devices
If you want movement that feels more like muscle than motor, PAMs belong on your list.
Decoupled power source
The heavy stuff can live somewhere else:
Compressor
Tanks
Regulators
Filtration
You can park those off-board or in a backpack, then run light, flexible air lines to the muscles themselves. That is a big deal for wearables, mobile robots, and animatronics where mass at the joint is painful.
What is wrong with PAMs today?
If PAMs are so great, why is everything still running on motors?
Some pain points in the current ecosystem:
No standard parts: Everyone ends up having to make their own.
Awkward contraction ratios: Many DIY builds get mediocre stroke per length.
Bulky pneumatics: Valves, regulators, and sensors are often sized for industrial pneumatics, not wearables.
Cost and availability: Good commercial muscles are expensive and not stocked like standard actuators. DIY approaches are cheap but don't have the performance
Poor integration story: Mounting hardware is improvised. Sensing and control are ad hoc. Every project feels like a custom job.
AeroFlex exists to attack these problems directly.
Enter AeroFlex
AeroFlex is our family of braided pneumatic muscles and supporting hardware that we are standardizing at Delta Robotics.
The core idea:
Take proven McKibben-style tech, and make it cheap, repeatable, and usable in real projects.
Our design goals
Known, repeatable geometry
Defined braid angle, inner and outer diameters, nominal lengths, or make custom
Predictable contraction ratios with a target around 50 percent stroke
Tunable variants for force vs stroke tradeoffs
Known pressure and force ranges
Documented operating and maximum pressures
Force curves vs pressure and displacement
Characterization data you can drop into a simulator or controller
Integrated sensing
Pressure sensing at or near the muscle
Force sensing via in-line load cells or end-effector measurement
Length / contraction sensing via encoders, linear sensors, or indirect models
In other words: we want you to be able to do more than "add air until it feels right."
Portable pneumatics
Smaller, more efficient valves sized correctly for AeroFlex muscles
Compact regulators and manifolds that make sense on a wearable or small robot
Path toward backpack or on-board pneumatic modules, not just benchtop rigs
Real integration hardware
Standard end fittings and mounting patterns
Brackets, clevises, and link hardware that can be reused across projects
Thoughtful routing and strain relief for air lines and sensor cables
What we are actively researching and improving
Better contraction ratio
Target: roughly 50 percent contraction, within sane pressures and lifetimes.
We are exploring:
Low-angle, high-aspect-ratio braids
Different combinations of inner tube and sleeve stiffness
Multi-filament bundles where several smaller muscles are grouped to act like a single unit
The goal is to get more stroke per unit length so you do not have to build absurd linkages just to move a joint.
Smaller, more efficient pneumatics
We want AeroFlex setups that are:
Wearable, not cart-mounted
Optimized for the flow and pressure ranges that AeroFlex actually uses
That means:
Valve selection and characterization specifically for AeroFlex muscles
Experiments with compact compressors, tanks, and regulators
Layouts that can realistically live in a backpack, torso shell, or robot base
Integrated sensing and control
To control AeroFlex muscles like real actuators instead of "vibes and PSI," we are:
Embedding or closely coupling pressure sensors to each muscle or muscle group
Benchmarking force output across pressure and strain
Experimenting with length estimation through sensors and modeling
On the control side:
Modeling muscle behavior for feedforward control
Exploring closed-loop pressure control, force control, and position control
Firmware and software that make AeroFlex act like a clean API, not a tangle of hoses
Materials, durability, and repeatability
We are testing:
Different inner tube materials and wall thicknesses
Various braided sleeves (PET, aramid, etc.) and weave patterns
End-fitting designs that survive repeated cycling, bending, and abuse
The goal is to publish not just "it works once," but:
Cycle life expectations
Failure modes
Reasonable derating guidelines
Standardization and documentation
AeroFlex is not just a pile of parts. We are building:
A small number of standard muscle sizes with documented performance
Reference designs for exoskeleton joints, animatronics, and robotic limbs
Integration guides and CAD for mounts, brackets, and manifolds
If you want a "just works" path, there will be off-the-shelf options. If you want to tweak and customize, there will be recipes.
Where AeroFlex is headed
In the AeroFlex family and related docs, you will see us working toward:
A catalog of standard AeroFlex muscles with known specs
Control stacks that integrate muscles, valves, and sensors cleanly
Example applications: exoskeleton joints, robot arms, animatronics, and prosthetic components
Both pre-made muscles and DIY-friendly kits
Our endgame is simple:
Make it as normal to drop a pneumatic muscle into your project as it is to drop in a servo.
Who this is for
AeroFlex is for you if:
You have built or tried McKibben muscles and want something more predictable
You have looked at Festo muscles and decided your budget would rather not
You wanted to try PAMs but got scared off by compressors, fittings, and random forum posts
If you have experience, horror stories, or hard-earned lessons with pneumatic muscles, we want those. If you are new and full of "dumb" questions, those are welcome too.
We are all collectively arguing with air pressure here.
Last updated
Was this helpful?