The problem with exoskeletons: “The Three Legged Race” (Part 1 of 3 on Exos)
Hello! I’m Keenon, and it’s nice to meet you :) I’m experimenting with documenting my research in a blog, and this will be my first post.
I’m a 3rd year PhD student at Stanford, working on a very simple problem: I have one letter wrong in my genetic code, which leads to an incurable degenerative neuropathy, Charcot Marie Tooth (CMT) type 2A, which is slowly paralyzing my legs. Currently, at age 28, I’m completely paralyzed from the knee down, and my thighs are getting weak enough that I have trouble walking up and down stairs.
My dream is simple: I want to be able to run on the beach, give my wife piggy-back rides, and someday maybe go on a hike with a dog. There’s no easy route from here to there, and figuring out how to biologically cure CMT 2A (and then regrow my muscle that’s turned to scar tissue from disuse) is the most direct, but also seems the slowest. My bet in the PhD is that the best odds for accomplishing that dream in the next few years lie in powered exoskeletons. I’ve made some progress, and have some ways to go. This series of posts is my attempt to lay out the logic of the plan I’m following to restore mobility for people with disabilities like mine, and put my existing work in context.
Exoskeletons are a pretty simple idea. To quote a friend of mine who’s a mechanical engineering PhD, “I guess it’s not actually that hard to strap electric motors to a person.” In an ideal world, you’d attach motors to your body (with some sort of clever mechanism), and they would just kind of “meld” into your body, and you’d be able to intuitively perform feats of strength impossible for mere mortals.
There is no shortage of clever ways to attach motors to your body: there are literally thousands of mechanical prototypes around, and dozens of devices you could buy. The exoskeleton dream captures the imagination of many brilliant mechanical engineers, and many devices I’ve come across are extremely clever and have found nice solutions to reduce weight, improve battery life, and increase comfort. Some have even found a niche where they can be commercially useful (like therapy or joint stress reduction).
That being said, every exoskeleton I’ve ever put on (including some $250k+ devices with fancy 3D-printed titanium components, and professional FDA-approved exos designed by teams of engineers over many years) has felt “clunky.” At the end of the day, despite all the clever mechanical engineering and lightweight materials, the motors were always doing a poor job of guessing what I wanted them to do. At worst, it felt like I was running a 3-legged race with another person. At best, it felt like moving through mud.
From my point of view, the hardware in most exoskeletons is already good enough to be usable. Sure, we only get 30 minutes of battery life on powered exoskeletons, and sure, they’re heavy and bulky. But they have the torque to support themselves and you, and if they really melded into your body they’d feel amazing for those 30 minutes. The big glaring problem is that almost nobody wants to wear the exoskeletons for the 30 minutes the battery lasts, even as it is. The bottleneck isn’t weight or battery life, the bottleneck is the users.
While there is plenty of work still to be done improving the mechanical design of exoskeletons, and I’m glad smart people are working on it, I feel like the biggest problem holding back every exoskeleton (past, present, and future) is actually really simple: how does the exo know what to do with its motors so that it feels like a part of your body?
Really, I think this breaks down into two questions:
- How does an exoskeleton know what you want to do?
- If it knows what you want to do, how can the exoskeleton help, without making a “3-legged race”?
Neither question is trivial. Somehow, the exoskeleton needs to sense what you want to do with extremely low latency (<1ms, if the haptics literature is any guide) and high fidelity. Once it knows what you want to do, it has to provide some sort of assistance that your brain feels is “natural,” instead of feeling like there’s another thinking agent strapped to your body trying to push you in a direction you don’t want to go. These are both open research questions, with no obvious answers.
I’ll dive into each of these problems in more depth in follow-up posts, and talk about the landscape of things people have tried, and what I think is eventually going to work.
Getting to a working exoskeleton that feels good to wear is the goal of my life (and by extension, my PhD). I’m in a race against my disability to try to figure this out before I end up in a wheelchair. I’ve got a plan, which I’ll describe in more detail in the coming posts, and I’m making progress, but I could always use more ideas and more help! It’s my hope that this blog series gets more people interested in this problem, and sparks healthy discussions about the best way to solve these problems.
I’m much more interested in solving mobility problems for people with CMT and other disabilities than in being right about my research. I could use your help! There are plenty of opportunities to help with technical research efforts, which I go into in Part 2. I could also use help spreading the word, so please share this with people who might be interested. Finally, if you know anyone who might be interested in supporting our lab’s work, please get in touch!
