Imagine, if you will, the home robot of the future. It picks clutter off the floor, sweeps, and does the dishes. And it has to do so perfectly: If the robot has an error rate of just 1 percent, it will drop one dish out of a hundred. Totally unacceptable. In no time, your floor would be covered in shards and the robot would get stuck in a sad, vicious feedback loop, dropping dishes and sweeping them up and dropping more dishes, ad infinitum.

To avoid this domestic nightmare, engineers will have to give robots a keen sense of touch. And for that, the machines will need fingertips, perhaps like the one recently described in the journal Science Robotics. It feels in a decidedly nonhuman way, by sensing the subtle changes in the finger’s own magnetic field, and it could one day make for ultra-sensitive prosthetic hands and robots that don’t maim tableware (or people) because they can’t control their grasp.

You, a human, can feel pressures and textures with your fingertips, thanks to specialized sensory cells in the skin called mechanoreceptors. These, along with the nervous system at large, translate mechanical information from the environment into signals your brain can comprehend as the perception of “touch.” Combined with thermoreceptors (which sense temperature) and nociceptors (which sense pain), you’re able to manipulate the world around you without hurting yourself.

We need robots to do the same, only we want them to avoid hurting themselves and us. A robot might, for instance, be put to work assisting the elderly, lifting them in and out of bed. “With tactile feedback, it would be much safer for the domestic robot to interact with people,” says computer scientist Youcan Yan of the University of Hong Kong and the City University of Hong Kong, lead author on the new paper describing the system. “And the robot can accomplish much more challenging tasks that they couldn’t achieve before, like dexterous grasping and manipulation.”

It’s just not feasible, though, to try to replicate the wildly complicated human system of touch—so Yan’s team kinda replicated it. The “skin” of their robotic fingertip is made of a flexible magnetized film, which generates a magnetic field within the device. The supporting “bone” of the finger is a circuit board, which is studded with sensors that monitor the magnetic field. If you jab, say, a pencil eraser into the fingertip, the magnetic skin depresses and the finger’s magnetic field subtly changes, which the sensors analyze to determine where the eraser is making contact. If you slapped these fingertips on a humanoid robot, the machine would be able to localize where its fingers were making contact with an object, improving its grip.

Video: Yan et al., Sci. Robot. 6, eabc8801 (2021)

In the GIF above, you can see just how sensitive the team’s design is: As the object is dragged across the fingertip, the system senses not only where it’s touching, but the shape of the rounded object.

The robotic fingertip also picks up the slippage—known to nerds as shearing force—of the object that it’s holding. Imagine you’re lightly gripping a water bottle while filling it up. As the bottle gets heavier, gravity tries to pry it out of your hand. When you feel that happening, you tighten your grip, obviously. But it’s not so obvious to robots that they need to do the same.

Below, watch as the shearing force (represented in the graph by the blue line) climbs as liquid is added to the bottle. At bottom, the gripper with the sensor activated can detect this and tighten its hold, while the gripper at top without the sensor eventually drops the bottle. That’s the difference a magnetic field makes.

Video: Yan et al., Sci. Robot. 6, eabc8801 (2021)

“This work is part of a very exciting new wave in tactile sensing, where folks are building sensors that give incredibly rich, raw data that is not necessarily intended for human consumption,” says Columbia University roboticist Matei Ciocarlie, who wasn’t involved in this new research. Interestingly enough, Ciocarlie’s group at Columbia has developed a similar robotic finger that uses light instead of magnetism. Their robotic finger’s “bone” is studded with LEDs and photodiodes, which detect how light changes within the structure when the skin is deformed.

With either method, the idea is to help robots fix their embarrassing lack of dexterity. We humans effortlessly manipulate the world around us—because we built the world around us. If there’s any hope of making these machines truly useful in our world, they’ll have to be more sensitive to it. A sense of touch will go a long way, whether a robot is packing boxes in a warehouse or helping tidy up a household.