At first glance it looks like a fancy leg brace.

But the “exoskeleton” system developed by a group of researchers at Carnegie Melon University in Pittsburgh could open the door to a new, more customized way of approaching human-robot interaction.

Led by computer scientist Juanjuan Zhang, the team created a new method that could have widespread impact on the way human performance is enhanced by devices of many kinds, including those that help people with disabilities, paralysis or amputations.

That’s because the method referred to as “human in the loop” puts the person in control of the kind of assistance they get from the exoskeleton, Zhang told CBC News. 

That can include helping people to walk who wouldn’t otherwise be able to do so, or helping people to walk faster or to expend less energy when they’re on the move. The exoskeleton could also be used to help the user lift a heavy load.

Previously a therapist or other expert would be in charge of how the device worked, said Zhang.

Her team’s findings appear in the journal Science Thursday.

Machine and human learn each other

“In theory it’s very simple. In order to walk forward, you want ground reaction force to push you forward. If you push the ground harder, the ground will push you back harder. We help with the pushoff,” said Zhang, now an associate professor in Nankai University in China.

That kind of assistance is called torque.

exoskeleton-research

A group of researchers from Carnegie Melon University in Pittsburgh has developed a new kind of exoskeleton system known as ‘human in the loop optimization.’ It puts the person in control of the kind of assistance they get from the devices, which can be used to help people with disabilities to walk or to walk with less effort. (Carnegie Mellon University)

Working with Steven Collins, a mechanical engineering professor at CMU working on the development of medical devices, Zhang’s team developed an algorithm to test varying kinds of help from the exoskeleton and see how their 11 healthy subjects responded.   

This “optimization algorithm” periodically changes its pattern of assistance and evaluates the metabolic cost to the subject — the amount of energy expended —which is established by measuring respiration.

Compared to using exoskeletons that the patients do not control, energy exertion was shown to be an average of 24 per cent lower.

Getting to the sweet spot of machine assistance is tricky business.

Patients have difficulty getting used to wearing an exoskeleton. As fun as it might sound to have the Ironman-like experience of wearing robotic armor, the body actually resists the help, said Zhang.

“People try to fight it. You don’t feel safe. Something is pushing you and you try to fight back,” she said. “It’s a process of robot learning and human learning.”

The future

The core of these research findings does not lie in the exoskeletons themselves, or even in their team’s particular algorithm, said Zhang.

The real importance is in the human-controlled method, which could be built upon to enhance and personalize all kinds of devices that make life better for people, such as prostheses for amputees or exoskeletons that helps the paralyzed to walk again.

Read more