Autism Therapy Robot: KASPAR, QTrobot and ABB Lessons

An autism therapy robot is a tool that a therapist, teacher or parent uses during a session. It plays turn taking games, shows simple facial expressions and does the same thing the same way every time, which many autistic children find easier to read than a person. That clinical use is a separate job from the companion robots sold to families as playmates, and it is the subject here. The post traces the research from the first university robots in the late 1990s to NAO and QTrobot in clinics and homes, then looks at what comes next and at what ABB's collaborative robots can and cannot offer the field.

Two KASPAR robots, child sized humanoids built to help autistic children practise social skills, on display in 2021. Photo: 384 / Wikimedia Commons (CC BY-SA 4.0)

How therapists and the first social robots worked with autistic children

Before robots, social skills teaching was done by people, usually one to one. A therapist or teacher sat with a child, modelled a greeting or a turn in a game, prompted and praised, and repeated the exercise many times over weeks and months. That approach still works, but it depends on trained staff, and families on waiting lists can go a long time without enough hours of support. Researchers began asking whether a simple, predictable machine could hold a child's attention and act as a bridge to a person, with the adult still running the session.

A teacher works one to one with a pupil at a training school for deaf children, 1917, the hands on style of teaching that therapy still relies on. Photo: Lewis Hine, Library of Congress via Wikimedia Commons (public domain)

At the University of Hertfordshire, Professor Kerstin Dautenhahn and her team began the Aurora project in 1998, studying robots as therapeutic toys for children with autism. In Japan, Hideki Kozima's small yellow Keepon, a robot that could only nod, turn and bob, was used in playroom studies with young children. These early machines were limited on purpose. A child could predict what they would do next, and researchers could watch how children reacted to a partner that never got tired or impatient.

Researcher Hideki Kozima with a group of Keepon robots at Wired NextFest, 2007. Photo: Matt Schley / Wikimedia Commons (CC BY-SA 3.0)

The Hertfordshire group built the first KASPAR in 2005. It is the size of a small child and has a simplified face with no wrinkles or hair detail, plus a neutral expression that is neither clearly male nor female. The team kept it plain because earlier work suggested that a robot which looks too lifelike can be unnerving. KASPAR has been redesigned many times after trials in schools and family homes, and the university says 28 prototypes now exist.

A KASPAR robot for autistic children in the Robots exhibition at the Science Museum, London, 2017. Photo: Matt Brown / Wikimedia Commons (CC BY 2.0)

NAO, QTrobot and robots in clinics and homes today

The robot most labs and clinics actually bought was NAO, a small humanoid from the French company Aldebaran that reached universities in 2008. In 2012 donated NAO robots were used to teach autistic children at a school in the UK, and NAO has since turned up in many published studies. One randomised trial in Romania, run between 2017 and 2018, compared robot enhanced therapy with standard behavioural therapy for 70 children aged 3 to 6, each getting eight sessions of 45 minutes.

A NAO humanoid robot from Aldebaran at the Innorobo show in Lyon, 2015. Photo: Xavier Caré / Wikimedia Commons (CC BY-SA 4.0)

Therapy is moving into homes as well. In a 2018 Yale study published in Science Robotics, 12 autistic children aged 6 to 12 used an autonomous social robot with a caregiver for 30 minutes a day for a month. Afterwards they showed better joint attention with adults even when the robot was not in the room. LuxAI's QTrobot, an expressive tabletop humanoid, is now being tested at home with the Luxembourg Institute of Health and the University of Birmingham, in a study of 69 families with children aged 2.5 to 4.5 that is due to finish by the end of 2026.

A NAO robot reads a story to children and parents at a library in Mexico, 2015. Photo: Elisa Jarquin / Wikimedia Commons (CC BY-SA 4.0)

The research is growing faster than anyone can evaluate it. A scoping review in the Journal of Autism and Developmental Disorders, reported on 2 October 2026, counted 72 studies from 2019 to 2025, 23 of them on socially assistive robots. Social communication was the target in 49 studies, while daily living skills came up in five and sensory processing in three. A week earlier Bioengineer.org ran a piece headlined "Robot Therapies for Autistic Children Face a Hard Ethical Reckoning". The hardware side is fragile too: NAO's maker went into receivership in 2025, and a new owner now keeps NAO6 on sale.

An early NAO robot in a university lab, 2008, the year NAO reached research groups. Photo: Jiuguang Wang / Wikimedia Commons (CC BY-SA 3.0)

What comes next for autism therapy robots and where ABB fits

The next generation of these robots will probably adapt more closely to each child. Only five of the 72 studies in that review used closed loop systems that read signals such as heart rate or brain activity and change the session in response. Better speech recognition and vision should let a robot notice when a child is overwhelmed and slow down. For clinics, the goal is a robot that a therapist can program without an engineer, much like the no code tools now used to set up industrial arms.

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

ABB does not make autism therapy robots, and we found no source showing its arms in autism trials. Its collaborative robots are relevant in a narrower way. The GoFa CRB 15000 and the dual arm YuMi are built to work next to people, and GoFa stops when it senses contact. ABB has also shown GoFa at its healthcare research hub. A research team that needs a safe arm for tabletop activities, such as passing blocks or sorting shapes with a child, could plausibly use one, but for now that is only a possibility.

Students working with an ABB YuMi dual arm robot in an Italian school lab. Photo: ABB

Most of the open problems sit outside the hardware. Therapy robots are a large purchase for a small clinic, and someone on staff has to learn the software. Safety includes emotional safety, and some autistic adults argue that therapy should respect autistic ways of communicating instead of training children to look neurotypical. Robots for rehabilitation therapy faced similar questions about evidence and cost before hospitals bought them. Whether autism robots follow will depend on results like the QTrobot home trial, and on families deciding that the robot helps their child.

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

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com