Adaptive Robotic Prostheses

My name is Roman Stolyarov, and for my PhD work at the Media Lab I developed a terrain-adaptive control system for robotic leg prostheses. While modern prostheses allow people with leg amputations to get around, the way that they get around is often tiring, uncomfortable, and burdensome on their intact joints. In the worst case, irregular terrain geometries—from uneven sidewalks to rugged hiking trails—can be difficult or impossible to navigate with conventional prostheses, burdening the user’s mobility, independence, and sense of being able-bodied. My work was intended to help people with amputations feel as able-bodied and mobile as possible, by allowing them to walk seamlessly regardless of the ground terrain.

As a member of the Biomechatronics research group, I developed a way to control a robotic prosthesis such that it allows the user to walk more naturally across different ground terrains. Besides increasing mobility and comfort, this control system also makes walking more symmetrical, decreases the risk of falling, and diminishes impacts on biological joints. In turn, this decreases the risk of chronic back pain and joint osteoarthritis, conditions which affect many people with leg amputations.

In contrast to much of the previous work in this area, my graduate work resulted in a control system that is simpler, more intuitive, and more adaptable to the gait of a given individual. This allows the system to adapt quickly to different terrain geometries, and makes troubleshooting when misclassifications occur relatively straightforward. Additionally, because the user can more easily comprehend the way our system works, they are much more likely to attain consistent, reliable performance.

People with amputations who’ve never walked with robotic prostheses are often skeptical the first time they see them—compared to conventional prostheses, they add mass and volume, and feel clunky when you first put them on. But then we would ask these individuals to walk around on the devices. As they walked, we would tune the device’s software to adapt to the ground terrain in much the same way that biological limbs do. Our robotic prostheses are capable of a range of reactions, like pushing the user’s leg off the ground at just the right time during walking or pointing the toe when they descend a flight of stairs. When we would get this control just right, there was a beautiful moment of realization that you could see on the person’s face: for the first time since their amputation, they would feel not like they are limping, but like they are walking. “Ohhhhhh, I get it now!” they’d exclaim. “I actually feel like it’s helping me walk!” There is nothing more rewarding than that exclamation, that look of sheer excitement when the person can finally use their body to its full potential.

Chris Shallal, an MSRP intern who tested the prosthesis, said, "The most memorable day I’ve had in my life was working with Roman Stolyarov on Matthew Carney’s new TF-08 powered ankle prosthesis. He hooked it up on me and tested his terrain detection algorithm which allowed me to walk up and down stairs, go across elevated surfaces and walk with much greater ease." 

Probably the most challenging part of the work is dealing with the combination of human variability and an individual human’s sensitivity. People have many different ways of walking, and a single person’s gait may change depending on a variety of conditions, both internal and external. At the same time, there are many more ways to control a robotic prosthesis incorrectly, or uncomfortably, than correctly. Thus, there are the coupled challenges of designing control systems that are both applicable to a wide range of individuals and precisely adaptable to the conditions under which a given user feels comfortable and assisted.

Another challenging aspect of this work is defining and quantifying improvements. Individuals often report being less comfortable even when the numbers describing their gait look more normal, or more comfortable even though the numbers don’t quite align. As biomedical scientists and engineers, we must constantly incorporate the human’s subjective feedback regarding comfort, while at the same time remembering that comfort does not necessarily lead to the healthiest outcome. For example, a person might have gotten used to walking with a limp, but this limp might make them more likely to develop osteoarthritis in their compensating limb. Do we develop control systems that are more comfortable, or control systems that prevent disease? The answer, we have found, typically lies somewhere in between.

In the future, we would like to expand our terrain adaptive technologies to above-knee amputations (as now the technology is only implemented for below-knee amputations). We also believe there are applications of the work to the control of powered exoskeletons for defense, rehabilitative, or even recreational applications. Thinking more generally, the next several decades will see a gradual merging of the biological world (humans) and the synthetic world (robots). Although many people have internalized a socialized fear of robotics, associating them with dangerous sci-fi visions like the Terminator, eventually robotics will broadly be used in applications that assist or rehabilitate humans—not as intimidating forms outside of us, but secondary to humans, always helping humans, and always under human control. 

Related Content