For an amputee left unable to walk, prosthetic limbs can be a life-changer. Still, there’s no getting around the fact that conventional prosthetic legs are less than ideal. Even recent powered, robotic legs can be clunky, requiring careful fine-tuning by clinicians. People using these kind of prostheses tend to fall more often, due to a lack of sensory feedback. They also walk more slowly and exert more energy too.
Fortunately, robots can help.
“Ever since my graduate days I’ve been fascinated by the fact that we can create robots to imitate the way we walk–and maybe even help people who are unable to walk to do so,” says Dr. Robert Gregg, an assistant professor of bioengineering and mechanical engineering in the Erik Jonsson School of Engineering and Computer Science at the University of Texas at Dallas.
Over the past half decade there has been a revolution in powered prostheses–or robotic legs–aided by the arrival of new lightweight motors, sensors, and microprocessors. This has led to improved prostheses too, but the technology is still bulky, and requires significant adjusting for the user and tweaks by rehabilitation specialists.
In the design of bipedal robot legs, Dr. Gregg saw an opportunity to improve the prosthetic version, by replicating a human’s natural gait by looking at a single variable: the center of pressure on the foot, which moves from heel to toe as a person walks.
When applied to a sensor-equipped robotic prosthesis, this “control strategy” allows Gregg, on his own, to modulate settings for the leg much faster than it normally takes a team of rehabilitation specialists. This also allows for the user to more easily adapt to variations in walking speed and environmental conditions than with existing prostheses.
In a study that involved walking on a treadmill at varying speeds, people trying out the robotic prosthesis were able to walk almost as fast as a person using biological legs. The result could improve the quality of life of millions of lower-limb amputees.
One of Dr. Gregg’s key contributions to the field of powered prostheses is his application of what is called robot control theory. As he explains it, although the human body is able to execute a wide range of movements, it is a minor miracle that it is able to stand up at all. Even a task as seemingly straightforward as walking without falling down is immensely complex–and a challenge to anyone working in robotics.
“There are hundreds of variables involved in walking,” says Gregg. “There are multiple joints, all of which can bend in different directions. In addition to the skeletal system, you also have a large number of muscles. In order for our body to be able to walk, and not simply to collapse, all of these need to coordinate perfectly with each other.”
This is where previous research into powered prostheses has often faltered. The robotic legs themselves may be able to generate force using built-in motors, but they lack the necessary control to remain stable while dealing with changing terrain and other disturbances. Walking on a flat surface at a fixed speed is all well and good, but trying to extend this to more challenging tasks (like climbing stairs or navigating a ramp) regularly results in challenges.
This is partly because many previous researchers have chosen to study the so-called “gait cycle” based on events occurring over time. The gait cycle refers to the way in which humans walk and run. It comprises two distinct stages: the “stance” phase, in which part of the foot comes into contact with the ground, and the “swing” phase, during which that same foot doesn’t touch the ground at all. A time-based gait cycle can mimic the appearance of walking, but offers very little flexibility to changes in walking speed or environmental conditions.
Following in the footsteps of humanoid robot designers, Gregg proposes a new way to replicate the process of a person walking for a prosthesis: by measuring one single variable representing the human body in motion. The crucial variable here is the center of pressure on the foot, which travels from heel to toe as the gait cycle takes place. The result is a more stable, flexible approach to tasks like walking.
After initially testing his theory using computer models, Gregg recently had the opportunity to literally run his technology through its paces with the help of three above-knee amputee participants. These participants were recruited with the support of the Rehabilitation Institute of Chicago and the University of New Brunswick, both of which were involved with the work leading up to and including these experiments.
“It was amazing to work with the study’s participants, and to see the years of work we’d put in really start to pay off,” Gregg says.
Gregg configured his algorithms with the users’ height, weight, and thigh dimension. They were then asked to walk on a treadmill for 15 minutes at various speeds. At their fastest, the participants were walking at more than one meter per second–which is slightly less than the average 1.3 meters per second walking speed of able-bodied people.
“One thing we noticed is that it, unsurprisingly, takes people a while to develop trust that the leg is going to behave the way they wish it to,” says Levi Hargrove, director of neural engineering for the Prosthetics and Orthotics Laboratory Rehabilitation Institute of Chicago–who was involved in the study.
“At first, people are concerned about whether the leg will hold them up. Once they see that it does, they begin using it at slow speeds, before gradually growing more confident. It’s no different than any other technology in that way,” he says.
This element of trust is crucial–and a reason why Gregg’s research is about more than just a nifty new algorithm interesting purely to researchers. The ideal powered prosthesis needs to have a synergistic relationship with its user: cooperating with them rather than feeling like a separate entity.
“My hope is that the control strategy that we’re developing will allow patients to use prosthetic legs more naturally, like they would with biological limbs,” says Gregg. “The goal is for the human user of the leg not to have to put any extra thought into walking. This would mean seamlessly adjusting to walking speeds as the subject decides to walk slower or faster.”
Gregg says that he was extremely happy with the experiment, but there’s still plenty of work to do before his technology can be made available to the general public.
For example, while the study successfully demonstrated that the powered prosthesis can intelligently adapt to different walking speeds, it still needs to be shown that it can shift between other actions without problems. This is particularly crucial when dealing with real-world scenarios that haven’t previously been anticipated.
“That’s a challenge that our lab is looking at right now,” Gregg says. “Even if we’re 99% accurate at identifying what task the leg should be dealing with at any given time, that still means that one out of every 100 steps you could have the wrong task–which is unacceptably frequent.”
For this advance to take place, researchers must continue to develop different ways for the user to interface with their prosthesis. This could be a mechanical movement of residual muscles (think the leg equivalent of a touch-screen gesture) or more subtle muscle activation, as measured using electrodes. The goal of both would be to allow the user to subconsciously switch between gait patterns when anticipating a task change.
“When you use voice recognition on your phone, if you were to look at that voice waveform on your computer it would be incredibly noisy,” says Levi Hargrove. “Yet the algorithm and signal processing on your phone does a good job of picking up what you’re trying to say, by converting that noisy signal into useful information. When your muscles contract, they make signals that are very much like a voice signal. The information is there if you decode it properly.”
Ultimately, the work is an exciting breakthrough that suggests that prosthetic legs could one day soon could come close to matching the functionality of biological limbs.
“It’s my dream that motorized robotic limbs will be able to do all the everyday things that most of us take for granted, like running to catch a bus,” says Dr. Gregg.
It may happen sooner than you think, too.
“I think this could be available within three to five years, which is pretty near-term for the prosthetics and robotics industry,” says Hargrove. “I’m very optimistic.”