Making skin for robots is hard. But success could bring huge benefits for humans, too
Gordon Cheng's lab at TUM is developing artificial skin to provide robots with tactile feedback.
To recreate a human in robot form is no small piece of work. Making a human brain? Researchers have already started modelling synapses and neurons in software and hardware. A robot that can move like a human? Researchers are already building artificial muscles, joints and tendons for the bipedal machines.
But perhaps one of the greatest challenges in building a truly human-like robot is the skin.
Innovation
First, there's the matter of size: at 1.5m to 2m square, skin is the largest organ in the human body. Then there's everything it does. As well as keeping our insides inside and the outside world out, skin does a lot of amazing, complex things. It has separate receptors to detect different sensations: pressure, texture, vibration, cold and heat, plus the ability to pick up diverse sensations from even the slightest touch, across the whole body.
SEE: An IT pro's guide to robotic process automation (free PDF) (TechRepublic)
For anyone seeking to create a robot skin that can pick up the same number of sensations as the human version, adapt like biological skin, and collect and process information from millions of sensors a second, one of the main barriers to overcome is power.
Skin tissues contain millions of receptors that gather information. A robot with a similar density of sensors, sampling that information hundreds or thousands of times a second, would take a lot of energy and processing power.
As such, after covering just one robotic arm in electronic skin and processing the data using traditional computing methods, Gordon Cheng, professor of cognitive systems at the Technical University of Munich (TUM), was convinced that the systems used by the human body would be a far more useful model.
"When we used conventional wisdom and power to make sense of the data, it worked to some extent. But when we tried to scale it up, we needed more and more computers," Cheng says. One of the most clever things about the whole biological system is that it won't send information to the brain until something changes and it's necessary, he says.
That's because the skin is designed to pass on only the information that the brain needs, when it needs it. When you put your socks on this morning, your skin told your brain when they were covering your feet. But your skin knows your brain doesn't need to keep being told that you're still wearing socks throughout the day. So the skin receptors dial up the signal when the sock goes on, and dial them down until it comes back off at the end of the day.
Cheng's lab has created skin 'cells' with sensors for movement, pressure and other sensations, which only relay information when a change happens. The event-based system cuts power consumption by 90%, making their widespread use more feasible.
The lab at TUM has used the cells to cover most of a human-sized robot called H-1, which can use the feedback from the cells to help adjust its movements: the cells on its arms allow it to determine the right pressure to use for giving a hug, while the cells on the soles of its feet help it adjust to walking on different terrains.
At the National University of Singapore (NUS), researchers are also looking to make skin less draining on robotic systems by using biologically inspired computing: NUS' electronic skin uses neuromorphic chips, which are inspired by the way information is processed in the human brain, to keep the system's power needs down.
SEE: Using data science and analytics to fight childhood cancer
The NUS' skin, which uses Intel's Loihi neuromorphic chips, is also event-driven. It's modelled after the 'spikes' of activity that are passed through human nerve fibers, and only passes on information once there's a change in the sensations it's picking up. That not only compresses the amount of data, it needs one hundred times less power.
While humans' skin and nervous systems may be the model for electronic equivalents, the biological model only goes so far. Our skin, brain and nerves don't get updates: we're stuck with pretty much the same processing power and sensor capabilities over our lifetimes. Advancements in software and hardware, however, will eventually mean that the capabilities of robot skin will outstrip that of human skin. NUS' skin senses touch over 1,000 times faster than its human equivalent already, and electronic skin's abilities are only likely to get more advanced over time.
"We've already demonstrated that through our use of our technology, we are able to give not just the sense of touch, but actually superhuman sense of touch," Benjamin CK Tee, deputy head of outreach and innovation in NUS' department of materials science and engineering, says.
Benjamin CK Tee (far left) and his team have developed electronic skin that responds 1,000 times faster than its human equivalent.
Just like human skin, robot skin will need to feel pain, to serve as an early warning system that alerts the robot when it's at risk of damage.
RMIT University has created a prototype robotic skin that can feel pain, realistically recreating the way that skin detects sensations like heat or cold all the time, but pain is only registered once certain thresholds are breached: when heat becomes hot enough to damage the skin, for example.
The sensitivity of electronic skin could be adjusted to recreate other skin conditions, such as sunburn. And recreating these conditions in robot skin could help researchers working on the biological versions to understand more about them, and how to treat them.
That's not the only way that fully realised electronic skin could help humans. Today's prosthetics may be able to look and even move similarly to human joints, but they don't have the same sensing abilities. "A prosthetic arm can significantly enhance people's lives, but it's still not quite close to a human limb. It doesn't have the ability to sense. Imagine having this electronic skin stretched out over your prosthetic arm, it could bring it a little bit closer to a real limb-type experience," says Madhu Bhaskaran, professor and co-leader of the functional materials and microsystems research group at RMIT University.
Sensory signals from the skin are already transmitted to the human brain via electrical signals, so in theory connecting the electronics in a prosthetic to the nervous system shouldn't require a lot of extra engineering know-how.
However, it will need extra materials science know-how. Any electronic skin that connects up with human tissue will need to be biocompatible (that is, the body won't try to reject it), and able to withstand the harsh environment of the human body (the salty, wet environment of human tissue typically isn't a welcoming place for electronics).
SEE: The future of wearables: Why your smartwatch could soon be your doctor's favourite gadget
It will also need to withstand all the stretching and bending that human limbs go through, without cracking or warping, and last a long time.
Human skin also has the rather impressive ability to self-heal: a small cut can disappear in a matter of days, and for bigger wounds it can create a whole new material to cover the gap, in the form of a scar.
Researchers are already working on materials with similar self-healing properties to human skin. Carnegie Mellon University engineers, for example, have created a class of soft, stretchable polymer containing liquid alloys which allow it to self-heal if punctured, for example. Others have suggested that graphene could also be used to create self-healing robot skin. Meanwhile, NUS researchers have come up a foam material, with nerve-like electrodes embedded within it, that can self-heal if it's damaged.
While many engineering and materials challenges – from longevity to biocompatibility and even aesthetics – remain to be solved, the benefits of electronic skin for both robots and humans are clear.
"Skin gives us an entire sense of the world. Not only that, it gives us the context to interact with others, for example, a handshake or a fistbump," says Tee from NUS. "I think the technologies we're developing will allow robots and humans to actually collaborate much more effectively and the social impacts can be very positive."