The Skin That Repairs Itself: How Robots Are Learning to Heal Without Us

Jun 4

Written By Michele Edington (formerly Michele Gargiulo)

Out there, in a laboratory in Nebraska, something wild is happening that I haven’t seen enough coverage on. Well, except some overly-exaggerated social media posts for some clickbait.

Just materials, circuits, and a problem engineers have been trying to solve for a long time: how do you build machines that don’t fail the moment something goes wrong?

The team working to solve that problem is led by Dr. Eric Markvicka, the Robert F. and Myrna L. Krohn Assistant Professor of Biomedical Engineering in the Department of Mechanical & Materials Engineering. Working alongside him are graduate students like Ethan Krings and Patrick McManigal, and their work has already been recognized at the IEEE International Conference on Robotics and Automation as a finalist for some of the field’s top awards.

Their research isn’t just for show or some fancy tech convention, it’s measured in layers of soft materials embedded with liquid metal microdroplets, pressure sensors that detect when something is wrong, and circuits that can read those signals and drive a response.

It doesn’t heal like living tissue does, and it doesn’t feel pain (I know you’ve been watching too many sci-fi movies if you’re thinking that way like me). I’m not talking about something “regenerating” in the biological sense, but when this artificial skin is cut, punctured, or stretched beyond its tolerance, the system notices that. Then, through its own built-in chemistry and circuitry, it restores continuity and function without a real person stepping in.

It was engineered to detect damage and recover from it. That simple shift from machines that break to machines that can repair themselves is already changing how we think about durability, deployment, and machine autonomy.

When Repair Becomes the Point

For most of technological history, intelligence was the goal. Especially when everything right now is just all about AI and AGI to the point where I just don’t care to hear more about computers.

Can a machine calculate faster or recognize patterns or learn faster and faster and faster? Intelligence doesn’t keep something alive though, resilience does. And that’s coming from a trauma-survivor. A system that can repair itself functions better and survives longer. This sort of self-fixing machine would fail less catastrophically and operates farther from our careful oversight.

That’s the real breakthrough here that I want to focus on. This robotic skin isn’t “smart” in the way we usually talk about AI, it doesn’t reason or make decisions based on whatever logic has been inputted into it, but it does respond. When the material is punctured or torn, embedded sensing pathways detect the disruption and electrical conductivity drops. Signals change and that local change triggers a physical response in the material itself, allowing damaged connections to re-form.

There’s no central processor weighing the decision and no external command is required. The repair just goes on and happens where the damage occurs.

This strange and slightly magical material combines soft, stretchable polymers with conductive pathways, often using liquid metal alloys, embedded throughout the skin. These pathways act as sensors and as functional circuitry. Now, when damage interrupts them, the system detects the break immediately because the electrical signal changes. In some configurations, the material can physically reconnect when pressed back together. In others, heat or chemical processes help restore conductivity across small gaps.

That might be going down the wormhole a little too much, and I’m not a scientist, so that part really isn’t as deep as I’m trying to go here, what matters isn’t the specific chemistry for me, it’s the architecture.

Each section of skin monitors itself. That decentralization is the key to this whole operation. There’s no single point of failure where damage requires diagnosis from a distant “brain.” The material knows, locally, when something is wrong. That’s independence from constant supervision we often have to give.

Nature Didn’t Invent This, We Just Studied It

Self-repair isn’t a new idea, I mean, nature has been doing it forever.

Plants seal wounds, skin knits itself back together with or without scars, and some animals out there regenerate entire limbs. If you’ve ever bitten on the inside of your lip while chewing food you know what I’m talking about. For like two days it’s swollen and you bite it again and again until it finally heals itself.

What engineers are doing now isn’t copying biology so much as borrowing its logic: don’t prevent damage at all costs, just learn to recover from it instead. That’s a very different design philosophy. Most machines are brittle by design (mostly because I think companies want to sell you replacement parts and also have you replace the entire thing every few years, but whatever). They work perfectly until they don’t. One crack, one tear, one broken connection, and the entire system fails.

Self-healing materials replace that brittleness with tolerance and assume damage will happen and plan for it. And yes, it does worry me that this sort of technology will never see the light of day, as maintenance on machines also generate hella revenue for tech companies (ah yes, “hella”, the most technical term there is).

Machines don’t feel pain because they aren’t live. Let me say that again for the people in the back who are marrying their AI friends. Machines aren’t alive. So the real challenge here is how do you teach one to respond to harm?

Well, this team suggests you give it thresholds. When conductivity drops below a certain level, something is wrong. When strain exceeds a limit, something has failed. When continuity breaks, the system reacts. That’s not awareness and the machine isn’t suddenly alive, but it is self-monitoring. Self-monitoring is the foundation of autonomy in any system, biological or artificial.

This Isn’t Just About Robots

As much fun as robots are, and the center of most of our talks these days, the implications for this tech stretchs far beyond robotics.

Self-healing materials are already being tested or deployed in concrete that seals its own cracks using embedded bacteria, and also coatings that close micro-fractures before corrosion spreads. Polymers are currently being experimented on that recover after repeated stress, and even some battery components that repair internal damage to extend lifespan have made their way into some markets.

This is about durability in unstable environments in my opinion. As systems grow larger, more remote, and more complex, constant maintenance becomes unrealistic. Teaching materials to take care of themselves would vastly help in this future we’re already building.

Of course, there’s more to this whole thing that we’ve glazed over up until now, but me, as a trauma survivor, has honed in on. Our machines are learning how to recover from damage…we often aren’t.

We push through our exhaustion, normalize burnout, and then go ahead and treat breakdowns as a personal failure instead of a structural one. Meanwhile, the material in a lab pauses, reconnects, and continues functioning, because it was designed with recovery in mind.

That contrast might seem overly philosophical right now, but it’s actually practical.

We build machines expecting failure, but we build people pretending it won’t happen. The hypocrisy in this is bothersome to me. We should be allowing ourselves more grace and time and being more kind to ourselves, but here we are.

As environments become more unpredictable with extreme temperatures, storms, isolation, infrastructure stress, etc etc, resilience matters more than optimization. A robot in the Arctic can’t rely on a repair crew to get to it on time. A rover on Mars absolutely can’t wait for spare parts before it kicks out. Infrastructure under constant strain can’t be rebuilt every time it cracks.

Self-healing materials might seem like a luxury, but they’re a necessary response to instability. They turn failure from an endpoint into a phase.

No Myths Required

This isn’t the beginning of sentient machines, I’m not saying the machines are taking over (although, if they are, How AI Would Really Take Over If It Ever Became Sentient).

Nothing here wants anything or chooses survival in the basic way that we do, nothing “knows” it was hurt. These materials just operate under a new assumption: damage is temporary. That little assumption changes how long systems last, how far they can go, and how little intervention they need from us.

That’s not any sort of AI rebellion we should worry about, it’s just our design.

Zoom out far enough and this is about shifting away from disposable systems. For decades, we accepted that things break and just get replaced. Planned obsolescence wasn’t a flaw, it was the business model. Growing up, whenever anything broke at home, dad would just buy a new one. When my washing machine broke a few months ago he told me to go online and find a new one. My husband (who grew up in a single-mother household with 3 other siblings) told me we could buy a $30 part and fix it instead. Oddly enough, it worked. I’d honestly never thought about it before, and was frantically trying to figure out how to make the $1500 for a new washer.

Self-repair breaks this old logic. Anything that can heal becomes harder to discard. Building things to recover invites longevity.

So while yes, we’re changing materials, we’re also working to change expectations.

Survival isn’t about never breaking, it’s about knowing how to put yourself back together, without needing permission.

The Skin That Repairs Itself: How Robots Are Learning to Heal Without Us

When Repair Becomes the Point

Nature Didn’t Invent This, We Just Studied It

This Isn’t Just About Robots

No Myths Required

Related Reads You Might Enjoy

The Fungus in the Backpack: A Quiet Arrest, a Toxic Threat, and the Strange Future of Biosecurity

The Whitest Paint Ever Could Cool Cities and Fight Climate Change