A prosthetic that’s smarter, faster, and easier than other technologies sounds like a win-win-win. That’s the three-measure goal a group of researchers from the Stanford Biomechatronics Laboratory set for their untethered exoskeleton. Any prosthetic that enables someone to regain lost mobility seems miraculous at first, but bulk and weight can burn up too much of the wearer’s energy. Conventional prosthetics typically restrict movement within generalized parameters or require extensive adjustments in a laboratory setting for any degree of personalization. Stanford engineers employed machine learning to create a boot-like exoskeleton that adapts to user movement immediately and raises the level of individualization over time.

The Stanford team described their portable ankle exoskeleton in an article in Nature. The researchers used biomechanical data recorded during previous testing with people wearing an exoskeleton in about 3,600 different circumstances. The recordings captured data from sensors that measured ankle angle and velocity in different conditions. They trained a logistic regression machine learning model with the data. The model selects device interventions that require less energy and increase motion smoothness based on how the user walks.

When a new user wears the exoskeleton, the mechanism varies the assistance each time the user walks and analyzes the motion to learn the best way to help the wearer. According to the Stanford team, customizing the exoskeleton boot for a new user takes approximately one hour of walking. The engineers reported numbers of measured improvement including 17% lower energy expenditure per distance, and 9% faster walking speed. User mileage will vary, of course, but the Stanford team equated the energy savings and speed increase to “taking off a 30-pound backpack.”

Next steps in the ankle exoskeleton development include testing with older people and with people who are losing mobility because of disabilities. Other goals include improving balance and reducing joint pain. Eventually the team plans to commercialize the boot.