Robot Assisted Dressing: From Buttonhooks to ABB

Getting dressed is one of the first daily jobs people lose when a stroke, arthritis or a spinal injury takes away strength and reach, and it is one of the hardest for a machine to take over. Robot assisted dressing is the research field that tries anyway: a robot arm guides a sleeve over a hand, past the elbow and up to the shoulder while the person stays seated. This post follows the job from carers and simple dressing aids to the lab robots that pull on gowns and jackets, looks at where ABB cobots could fit, and lists the problems still open.

PR2 research robots at a lab, the model Georgia Tech used to slide hospital gowns onto people's arms. Photo: Oleg Alexandrov / Wikimedia Commons (CC BY-SA 3.0)

From carers' hands to buttonhooks and shoehorns

For most of history, help with dressing came from another person: a family member at home, a servant in a wealthy house, a nurse or orderly in a hospital ward. The work is close and physical. The helper has to read how far a stiff shoulder will move, when a hand needs to be guided through a cuff and when the person would rather do it alone. When doctors began to measure independence in older patients, dressing was on the list. The Katz Index of Activities of Daily Living, published in 1963, scores dressing next to bathing, toileting, transferring, continence and feeding.

Nurses talk with patients at an American hospital in Queensland, Australia, 1943, when daily care such as dressing was done by hand. Photo: State Library of Queensland via Wikimedia Commons (no known copyright restrictions)

Before anyone thought of robots, people built small tools for the hardest parts. The buttonhook is the best known. In the late nineteenth and early twentieth centuries, boots and gloves often closed with rows of small buttons, and households kept steel hooks with bone or silver handles to pull each button through its hole. Long shoehorns did the same job for feet that could not be reached. Occupational therapists still hand out modern versions today, along with sock aids and reachers, because a cheap tool that lets someone dress alone beats waiting for help every morning.

Small glove buttonhooks beside a longer boot buttonhook, tools for fastening rows of tiny buttons. Photo: TrudiJ / Wikimedia Commons (CC0)

Most robot dressing research dates from the 2010s. At Georgia Tech, Charlie Kemp's Healthcare Robotics Lab used a Willow Garage PR2 to slide a hospital gown onto a person's arm. In work presented at the ICRA conference in 2018, the team led by PhD student Zackory Erickson had the robot learn from nearly 11,000 simulated dressing attempts in a single day. The PR2 relied on touch rather than vision, predicting the forces the gown would put on the arm and adjusting its pull, and it dressed one arm in about 10 seconds. Georgia Tech noted that more than a million Americans need daily help getting dressed.

The PR2's two grippers working at a table, the touch based arms used in early dressing research. Photo: Willow Garage / Wikimedia Commons (CC BY-SA 3.0)

Lab robots that pull on sleeves today

Erickson moved on to Carnegie Mellon University, where his lab and David Held's group took the work further. In 2023 their paper "One Policy to Dress Them All" described a robot that learned in simulation, with randomised clothing, and then dressed real people. Across 510 trials with 17 participants of different body shapes and arm poses, using five garments, it pulled sleeves on with an average success of 86 percent. A later CMU paper used a Sawyer arm's built in force sensing so the robot could keep dressing while the person moved their arm. Both arms at once and overhead garments such as T shirts remain future work.

A Rethink Robotics Sawyer arm, the robot model used in later CMU dressing research, here at a factory in Schramberg, Germany. Photo: Jeff Green / Rethink Robotics via Wikimedia Commons (CC BY 4.0)

The news this autumn comes from homes rather than labs. On 29 September 2026, Hello Robot announced a Phase IIB grant of nearly $3 million from the US National Institutes of Health, through the National Institute on Aging, with the University of Illinois Urbana-Champaign. Over three years, starting 22 September, researchers will test the Stretch 4 mobile manipulator with 60 older adults, their care partners and care staff, looking at cognitive assistance, social contact and safety. Dressing is not on that list, yet the study is building the trust data any home dressing robot will need. Hello Robot was cofounded by the same Charlie Kemp who ran the Georgia Tech dressing work.

A long shoehorn helps a foot into a shoe, a job robots still find harder than sleeves. Photo: Infrogmation of New Orleans / Wikimedia Commons (CC BY-SA 4.0)

ABB does not sell a dressing robot, and no ABB dressing project has been announced. Its arms do appear in cloth research: UC Berkeley's SpeedFolding project in 2022 used a dual arm ABB YuMi IRB 14000 to fold garments. For dressing, the closer fit is the GoFa CRB 15000, a cobot built to work beside people that limits its power and force and stops when it senses contact. A lab could build a dressing cell on GoFa, test the arm path first in RobotStudio and use SafeMove to cap speed near the face. ABB has not announced anything like it, so treat this as an option a lab could try.

ABB's dual arm YuMi IRB 14000, the robot model UC Berkeley used to fold garments, working beside a lab technician. Photo: ABB

What home dressing robots still need

The first real use is likely to be in care homes and rehabilitation wards, where trained staff can supervise and the same few garments come up every day: gowns, cardigans, loose jackets. A robot that does the sleeve while a carer handles buttons would already save backs and time. In homes, the target users are people who live alone with weak arms. For them a robot that helps with a jacket or compression stockings could mean dressing on their own schedule. Feeding robots went down the same road, as our post on the assistive feeding robot describes.

A GoFa cobot at ABB's healthcare research hub. Photo: ABB

Clothes fold, bunch and snag, so the robot cannot predict exactly where a sleeve will be, and the CMU team names deformable fabric as a core difficulty. Safety comes next: an arm pulling on cloth around a frail shoulder must never use too much force, which is why the research leans on force sensing. Shoes and socks are harder still, since feet are low, hard to see and sometimes swollen. Then there is cost: research arms like these are expensive, and nobody yet sells a dressing robot at a price a family or a care home would pay.

A modern buttoning aid, the kind occupational therapists give out so people can dress alone. Photo: Mautpreller / Wikimedia Commons (CC BY-SA 4.0)

For factories, the same skills matter in a different form. Robots that can handle soft, floppy cloth are useful in laundries, garment making and textile recycling, and ABB cobots already do machine tending and packing next to people. Gains in cloth handling and force control in those settings will feed back into care research, and the reverse is true too. The skills question is real on both sides: care staff will need simple ways to teach a robot, such as no code programming, while engineers will need to learn how a dressing task feels to the person wearing the clothes.

ABB's GoFa cobot is guided by hand and stops when it senses contact. Photo: ABB

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com