A wearable robot has to fit more than a leg. It has to fit a person’s movement. Personalisation research tries to discover which assistance pattern is useful for the individual instead of assuming one setting works equally well for everyone.
A portable system learns during walking
A 2022 Nature study described a portable ankle exoskeleton and a data-driven method for adjusting assistance using wearable sensors during walking outside a conventional laboratory setup. The work showed that movement data could help tune support in naturalistic conditions. The reported efficiency improvements belong to that experimental design and its participants; they do not establish benefit for women over 50 with mobility impairments. Nature
What should the algorithm optimise?
Personalisation is an appealing word, but the important question is what is being optimised. Walking faster, using less energy and feeling more secure are not identical goals. A system could improve one measure while asking the wearer to accept additional weight or a less comfortable routine.
The learning period is part of the experience of wearing an adaptive device. How much walking is required? Can the user express discomfort or a different preference? What happens if their movement changes? These are questions about usable intelligence, not just algorithmic performance.
The numbers depend on the comparison
The 2022 study reported that assistance tuned during an hour of naturalistic walking increased self-selected speed by 9% and reduced the energy used to travel a distance by 17%, compared with normal shoes. It separately reported a 23% reduction in metabolic energy consumption during treadmill walking at a fixed speed. These are different conditions, not three interchangeable descriptions of one effect. Original paper
The distinction is useful when reading a dramatic headline. Walking at a chosen speed allows behaviour to change; a fixed-speed treadmill holds an important part of the task constant. A percentage needs its reference condition attached. Otherwise, a reader cannot tell whether it describes assistance under an imposed pace or a person’s own choice about how to walk.
Personalisation is a negotiation about the goal
A device can learn which settings improve a measured outcome without knowing which outcome the wearer values most. Someone may prefer a comfortable pace with friends to the fastest possible journey. Another person may value arriving less tired, even if the journey takes the same time. A useful system should eventually make those preferences part of its design.
This is not an objection to measuring energy use. It is a reason to place that measurement alongside the rest of the experience. An efficient stride is valuable if the equipment, learning process and daily routine leave the wearer better off overall. The experiment advances one part of that case: tuning assistance without depending entirely on a conventional laboratory setup.
Why an older study still earns attention
The study is included as a technical foundation, not as a newly announced 2026 invention. It shows how movement data can help personalise assistance outside tightly controlled conditions. That is a useful idea for assessing newer exoskeleton claims, including products that use the same language of adaptation without explaining what is actually adjusted.
Ask three concrete questions of any later system: what does it measure, what parameter can it change, and what outcome guided that change? If those answers are absent, the word personalised is doing more work than the evidence.
For readers following mobility technology after 50, the next bridge is a study involving the relevant users and tasks. A method that works in one participant group is an engineering starting point. Its value for someone with a particular walking limitation depends on further testing. The interesting frontier is a system that learns from the person while remaining accountable to her preferences.
The complete wearing experience
Look for studies that include the intended users and compare complete experiences, not isolated motor assistance. The promise is a robot that adjusts to a person; the evidence must show that adjustment remains useful across the situations in which the person actually walks.
Sources and editorial note
Source note · 1 October 2026. Research dates refer to the original work. Prepared with AI assistance from the linked sources; no hands-on testing or interviews. Methods, disclosures and corrections.
