Nelson Mandela University researcher Xolile Zepe
The study, completed in December 2025 at Mandela University, found no South African rehabilitation facility currently using socially assistive robots for lower limb therapy.
He found that 80% therapists surveyed across four provinces did not know about the technology. Zepe’s work sets out nine evidence-based guidelines to help the sector assess adoption, address legal uncertainty and manage risk in a system already under strain.
“It goes back to even before I came to the University,” Zepe says. “My sister was involved in an accident … they said that she will never walk again.”
Issued with a wheelchair at a public hospital in Gqeberha, his older sister was advised to attempt exercises at home if possible and the family immediately went to work on her rehabilitation programme.
Zepe, who attended Tyhilulwazi Senior Secondary School, remembers walking around the backyard with her after school each afternoon.
“We were sometimes in the dark because we didn’t know some of the exercises, but we were just trying something that she can do,” he says. “Eventually she managed to walk.”
The family persisted and, today, his sister walks to the shops without crutches. She still turns carefully and her gait is not perfect, but she is mobile and independent. For Zepe, that is proof of what sustained post-operative therapy can achieve without the formal support of a private physiotherapy rehabilitation programme.
“If you have medical aid, then there’s a chance that you can walk even if you have a stroke,” he says. “But if you rely on the public health system it often becomes a challenge to recover.”
His research responds to that inequity. South Africa faces persistent shortages of physiotherapists and biokineticists in the public sector, particularly outside major metros.
Lower limb disability, whether caused by accidents, stroke or diabetes-related complications, often calls for sustained, supervised exercise to restore function.
Socially assistive robots, such as the one analysed for this research, do not strap onto the body like exoskeletons. Instead, they guide exercises, demonstrate correct movements, record performance data and provide feedback. Therapists can programme the exercises and supervise use, retaining clinical authority.
“One of the questions from the survey was about the robot autonomy,” Zepe says. “A lot of the therapists would opt for supervised autonomy, which means that they at least want to have some sort of control, even if the robot is able to work on its own.”
The attraction of the robot assistant, he argues, is less about novelty than scalability. A robot can coach repetitive exercises consistently, record progress and share data with clinicians. It does not lose patience, tire or rush a session.
In facilities where one therapist may be responsible for dozens of patients, that consistency could extend professional oversight rather than dilute it. It also may help many more patients than a single therapist could cope with.
His supervisor, Dr Sue Petratos Senior Lecturer in the School of Information Technology, says the absence of local use was unexpected. “We couldn’t find any studies that actually focused specifically on lower limb therapy, and also specifically in South Africa,” she says. “We think we’ve uncovered something new here.”
The preference Zepe found for using a robot only under supervision, says Petratos, reflects a broader unease about new technologies. Respondents raised concerns about liability, data protection and professional accountability.
In addition, existing legal frameworks in South Africa, such as the National Health Act and the Protection of Personal Information Act were not drafted with AI-enabled, context-aware healthcare technologies in mind.
“If the robot makes a mistake or injures someone, who is responsible, is that going to be the therapist? Is it the manufacturer?” Zepe asks. “Those kind of regulations need to be in place.”
The guidelines he has developed through design science research methodology address regulatory compliance, training, infrastructure readiness, cost-sharing models and human–robot interaction. They also begin with awareness.
“If you don’t know the socially assistive robots, then you won’t know where to start,” Zepe says.
The core argument remains practical rather than futuristic. The robots are not proposed as replacements for clinicians, but as tools to extend limited capacity.
“The robots are not there to replace you,” both Zepe and Dr Petratos says. They are there to work with the therapy team, not to replace it.
With his master’s degree complete, he is now preparing doctoral research to hone in on regulatory gaps and expand awareness within the healthcare sector.
From a backyard in Gqeberha to a research agenda with national policy implications, the project positions Nelson Mandela University at the centre of an emerging debate: how to introduce intelligent systems into public healthcare without eroding trust, accountability or professional oversight.