Powered exoskeletons: revision 1 to 3

Revision 1

Revision 3

categories: ["bodies", "enhancement"]

tags: ["robotics", "rehabilitation", "mobility", "wearables", "human augmentation", "actuators"]

summary: "Wearable powered frames that apply force in parallel with the wearer's own limbs, used for gait rehabilitation, industrial load support, and proposed strength augmentation."

summary: "Wearable powered frames that apply force in parallel with the wearer's own limbs, used for gait rehabilitation, industrial load support, and proposed strength augmentation."

updated: "2026-07-27"

updated: "2026-10-30"

humanEvidence: "People with spinal cord injury or stroke walk in these devices under supervision, though reviews find the evidence too thin to show an advantage over conventional gait training; the clearest measured gains are metabolic savings in healthy walkers."

humanEvidence: "People with spinal cord injury or stroke walk in these devices under supervision, though reviews find the evidence too thin to show an advantage over conventional gait training; the clearest measured gains are metabolic savings in healthy walkers."

access: "Cleared for clinical use and, in the United States since 2014, for personal use; medical devices cost tens of thousands of dollars or more with limited payer coverage, while passive industrial exosuits are sold outright."

access: "Cleared for clinical use and, in the United States since 2014, for personal use; medical devices cost tens of thousands of dollars or more with limited payer coverage, while passive industrial exosuits are sold outright."

reversibility: "reversible"

{ "label": "First US personal-use clearance", "value": "2014" },

{ "label": "Actuation", "value": "Electric, hydraulic, pneumatic, passive spring" },

{ "label": "Binding constraint", "value": "Actuator and battery power density" },

{ "label": "Best lab metabolic saving", "value": "Roughly 15–20% for walking" },

{ "label": "Best lab metabolic saving", "value": "Roughly 15-20% for walking" },

{ "label": "Readiness", "value": "TRL 8 (medical), TRL 5 (full-body augmentation)" }

]

}

**Powered exoskeletons** are wearable machines whose structure runs alongside the body's own

skeleton and whose actuators apply torques across the wearer's joints. They are used clinically to

move the legs of people with paralysis, industrially to offload weight from a worker's back and

shoulders, and militarily — mostly in prototype — to carry loads. The engineering problem is

shoulders, and militarily, mostly in prototype, to carry loads. The engineering problem is

deceptively hard, because an exoskeleton must produce more benefit than the metabolic and

deceptively hard, because an exoskeleton must produce more benefit than the metabolic and mechanical

mechanical cost of carrying it. Unlike the implanted devices grouped under

cost of carrying it. Unlike the implanted devices grouped under [[neuroprosthetics]], or the

[[neuroprosthetics]], or the skeletal anchorage described in [[osseointegration]], an exoskeleton

skeletal anchorage described in [[osseointegration]], an exoskeleton touches the body only at straps

touches the body only at straps and cuffs, which is both its principal safety advantage and the

and cuffs. That is both its principal safety advantage and the source of most of its engineering

source of most of its engineering difficulty.

difficulty.

```figure

{"key": "exoskeleton-hal", "caption": "HAL reads surface electromyography at the thigh and drives the joint the wearer is already trying to move."}

{"key": "exoskeleton-hal", "caption": "HAL reads surface electromyography at the thigh and drives the joint the wearer is already trying to move."}

Intent detection is the control problem. Options range from simple state machines triggered by

weight shift and joint angle, through force sensors in the shoe, to surface electromyography that

reads the wearer's muscle activity before movement begins — the approach taken by Japan's HAL

reads the wearer's muscle activity before movement begins. Japan's HAL system takes the last

system, which uses skin-surface bioelectrical signals as its control input and shares its signal

approach, using skin-surface bioelectrical signals as its control input, and shares its signal chain

chain with [[myoelectric-prosthetics]]. Cortical control through a

with [[myoelectric-prosthetics]]. Cortical control through a [[brain-computer-interface]] has been

[[brain-computer-interface]] has been demonstrated in single participants and adds surgical risk to

demonstrated in single participants and adds surgical risk to a problem that peripheral sensors

a problem that peripheral sensors mostly solve; the same intent-estimation methods used in

mostly solve; the same intent-estimation methods used in [[neural-decoding]] apply, but the signal

[[neural-decoding]] apply, but the signal is easier to obtain at the muscle.

is easier to obtain at the muscle.

The distinction between *rehabilitation* devices, which move a limb the wearer cannot move, and

*augmentation* devices, which assist a limb the wearer can move, matters more than the hardware

nothing when the wearer needs no help, a much harder requirement, because an actuator that fights

the user is worse than no actuator at all.

A third use turns the device around. In some operator stations for [[robotic-avatars]],

upper-body and hand exoskeletons capture the wearer's movements to drive a remote robot and return

force and haptic feedback, adding information rather than strength.[^behnke2023]

## Development history

```timeline

[

  { "year": "1965–1971", "title": "Hardiman", "text": "General Electric builds a full-body hydraulic master-slave suit weighing around 680 kg. Only one arm is ever operated, and it moves violently enough to be considered unsafe with a person inside." },

  { "year": "1965-1971", "title": "Hardiman", "text": "General Electric builds a full-body hydraulic master-slave suit weighing around 680 kg. Only one arm is ever operated, and it moves violently enough to be considered unsafe with a person inside." },

  { "year": "Late 1960s", "title": "Active exoskeletons in Belgrade", "text": "Miomir Vukobratović's group at the Mihajlo Pupin Institute builds legged exoskeletons for rehabilitating paraplegic patients, developing the zero-moment-point criterion that later underpins bipedal robotics." },

  { "year": "Late 1960s", "title": "Active exoskeletons in Belgrade", "text": "Miomir Vukobratović's group at the Mihajlo Pupin Institute builds legged exoskeletons for rehabilitating paraplegic patients, developing the zero-moment-point criterion that later underpins bipedal robotics." },

  { "year": "2000–2008", "title": "DARPA funding and BLEEX", "text": "A US programme on exoskeletons for human performance augmentation funds the Berkeley Lower Extremity Exoskeleton and spawns Ekso Bionics and Lockheed's HULC." },

  { "year": "2000-2008", "title": "DARPA funding and BLEEX", "text": "A US programme on exoskeletons for human performance augmentation funds the Berkeley Lower Extremity Exoskeleton and spawns Ekso Bionics and Lockheed's HULC." },

  { "year": "2011–2016", "title": "Clinical clearances", "text": "Medical exoskeletons for spinal cord injury and stroke gain regulatory clearance in Japan, Europe, and the United States; the first personal-use clearance in the US comes in 2014." },

  { "year": "2011-2016", "title": "Clinical clearances", "text": "Medical exoskeletons for spinal cord injury and stroke gain regulatory clearance in Japan, Europe, and the United States; the first personal-use clearance in the US comes in 2014." },

  { "year": "2015–2022", "title": "Soft suits and optimized assistance", "text": "Textile-based exosuits and human-in-the-loop optimization of assistance profiles produce the largest measured reductions in the metabolic cost of walking." },

  { "year": "2015-2022", "title": "Soft suits and optimized assistance", "text": "Textile-based exosuits and human-in-the-loop optimization of assistance profiles produce the largest measured reductions in the metabolic cost of walking." },

  { "year": "2019–2025", "title": "Military programmes retrench", "text": "The US special operations powered-armour programme ends without a fielded suit, and several full-body industrial exoskeleton efforts are scaled back, while passive industrial exosuits find a durable market." }

  { "year": "2019-2025", "title": "Military programmes retrench", "text": "The US special operations powered-armour programme ends without a fielded suit, and several full-body industrial exoskeleton efforts are scaled back, while passive industrial exosuits find a durable market." }

]

```

walk in a supervised setting, and the devices are safe in trained hands. What is not established is

that they produce better functional outcomes than conventional gait training. Systematic reviews of

exoskeleton-assisted walking after spinal cord injury and stroke repeatedly conclude that the

evidence base is small, heterogeneous, and insufficient to demonstrate superiority; secondary

evidence base is small, heterogeneous, and insufficient to demonstrate superiority. The secondary

benefits often claimed — bowel function, spasticity, bone density, cardiovascular fitness — are

benefits often claimed (bowel function, spasticity, bone density, cardiovascular fitness) are

supported mainly by small studies and self-report.

Practical constraints compound this. Walking speeds achieved with lower-limb exoskeletons are

crutches and therefore intact upper-limb function, donning takes minutes and often requires

assistance, and users cannot navigate the stairs, kerbs, and uneven ground that make up ordinary

environments. For most people with paraplegia a wheelchair remains faster, cheaper, and more

reliable, which is not a failure of the technology so much as a statement of the standard it has to

reliable; that is less a failure of the technology than a measure of the standard it has to beat.

beat. Disability advocates have also questioned the emphasis on restoring upright walking over

Disability advocates have also questioned the emphasis on restoring upright walking over improving

improving wheelchair access, an argument set out under

wheelchair access, an argument set out under [[disability-rights-and-enhancement]]. Implanted

[[disability-rights-and-enhancement]]. Implanted alternatives that stimulate the user's own spinal

alternatives that stimulate the user's own spinal circuitry attack the same problem from inside the

circuitry attack the same problem from inside the body, and epidural stimulation trials have

body, and epidural stimulation trials have produced volitional stepping in participants who could

produced volitional stepping in participants who could not achieve it in a powered frame.

not achieve it in a powered frame.

Where exoskeletons appear more clearly useful is as therapy delivery: a device that lets one

therapist provide high-repetition, high-dose gait practice instead of two or three therapists

## Industrial and military use

Industrial exoskeletons have quietly become the largest deployed category, and most are passive.

Spring-loaded shoulder supports for overhead work and back supports that store energy during

Spring-loaded shoulder supports for overhead work and back supports that store energy during forward

forward bending reduce measured muscle activity in the assisted region, which is well established.

bending reduce measured muscle activity in the assisted region, which is well established. Whether

Whether they reduce injury rates over years is not, and there is a recognized concern that load

they reduce injury rates over years is not, and there is a recognized concern that load removed from

removed from one joint reappears at another — a back support can increase demand on the hips or

one joint reappears at another; a back support, for instance, can increase demand on the hips or

alter gait. Long-term field trials are scarce.

Military programmes have a consistent record of ambition followed by retrenchment. The US special

operations "tactical assault light operator suit" effort was terminated in 2019 without producing a

fielded system, and several powered load-carriage exoskeletons have been discontinued. The

fielded system, and several powered load-carriage exoskeletons have been discontinued. The recurring

recurring failure is not strength but endurance: a soldier's mission profile demands hours of

failure has been endurance rather than strength: a soldier's mission profile demands hours of

untethered operation, and no battery chemistry supplies that at acceptable weight. Lighter,

narrower-purpose devices such as powered knee assists for load carriage have fared better.

## The power and actuator problem

The physics is unforgiving. An exoskeleton that assists walking must overcome the metabolic penalty

of its own mass, and mass carried distally — at the ankle or foot — costs several times more than

of its own mass, and mass carried distally, at the ankle or foot, costs several times more than mass

mass at the waist. Electric motors have high power density but low torque density, so they need

at the waist. Electric motors have high power density but low torque density, so they need

gearboxes, which add mass, friction, and backdrive resistance. Hydraulics deliver torque but need a

pump, fluid, and heat rejection. Batteries are the hard ceiling: lithium-ion cells store on the

order of a few hundred watt-hours per kilogram, well over an order of magnitude below liquid fuel,

popular image of powered armour.

> [!key] Assistance has to be personalized

> The largest metabolic gains come not from stronger actuators but from tuning the timing and shape

> The largest metabolic gains come from tuning the timing and shape of assistance to the individual

> of assistance to the individual, discovered automatically by measuring the wearer's energy use

> rather than from stronger actuators. The best profile is discovered automatically by measuring the

> across many candidate profiles. A generic assistance pattern can make walking harder rather than

> wearer's energy use across many candidate profiles. A generic assistance pattern can make walking

> easier.

> harder rather than easier.

## Outlook

The demographic argument is the most consequential and least discussed. Age-related loss of muscle

mass and power is the mechanism behind most mobility loss in old age, and no drug reliably reverses

it; the best evidence remains for training itself, discussed under

it; the best evidence remains for training itself, discussed under [[exercise-and-aging]], and

[[exercise-and-aging]], and mobility is one of the functional domains any serious measure of

mobility is one of the functional domains any serious measure of [[healthspan]] has to capture. A

[[healthspan]] has to capture. A light, cheap, comfortable hip or ankle assist that lets an eighty-year-old

light, cheap, comfortable hip or ankle assist that lets an eighty-year-old keep walking would affect

keep walking would affect far more people than any spinal cord injury device, and it is a much

far more people than any spinal cord injury device, and it is a much easier engineering target than

easier engineering target than powered armour. Whether such a device can be made unobtrusive enough

powered armour. Whether such a device can be made unobtrusive enough that people will actually wear

that people will actually wear it is a design question, not a physics one, and it is where the

it is a design question rather than a physics one, and it is where the field's near-term value most

field's near-term value most likely lies.

likely lies.

## See also

- [[disability-rights-and-enhancement]]

- [[exercise-and-aging]]

- [[brain-computer-interface]]

- [[robotic-avatars]]

## References

[^collins2015]: `paper` Collins, S. H., Wiggin, M. B., Sawicki, G. S. "Reducing the energy cost of human walking using an unpowered exoskeleton." *Nature*, 2015. {The saving is in healthy adults walking on level ground in a laboratory, and says nothing about impaired gait.}

[^collins2015]: `paper` Collins, S. H., Wiggin, M. B., Sawicki, G. S. "Reducing the energy cost of human walking using an unpowered exoskeleton." *Nature*, 2015. {The saving is in healthy adults walking on level ground in a laboratory, and says nothing about impaired gait.}

[^slade2022]: `paper` Slade, P., Kochenderfer, M. J., Delp, S. L., Collins, S. H. "Personalizing exoskeleton assistance while walking in the real world." *Nature*, 2022.

[^slade2022]: `paper` Slade, P., Kochenderfer, M. J., Delp, S. L., Collins, S. H. "Personalizing exoskeleton assistance while walking in the real world." *Nature*, 2022.

[^behnke2023]: `paper` Behnke, S., Adams, J. A., Locke, D. "The $10 Million ANA Avatar XPRIZE Competition: How It Advanced Immersive Telepresence Systems." *IEEE Robotics & Automation Magazine*, 2023. doi:10.1109/MRA.2023.3310864 {A competition report co-written by Sven Behnke of the University of Bonn group behind the winning team and David Locke of the XPRIZE Foundation; it describes operator stations but is not a controlled study of them.}