Industrial robots are very good at doing the same thing over and over again. But give one something delicate to pick up, an object with an unusual shape or something that starts slipping between its fingers, and things get a little more complicated. The robot needs to know what's happening when it touches something, not just where its arm is moving. 

That's where robotic tactile sensing comes in. Sensors built into robotic hands and grippers help them detect physical contact. But getting more detailed information about that contact often means adding more sensors, wiring and electronics. New research published in November 2026 suggests there could be another way. 

em360tech image

Researchers have developed an electronic skin that can work out where it's being touched using just four electrical sensing channels. It's a promising development for robotic touch, although there's still quite a bit of work to do before it could help industrial robots handle objects more precisely.

Why Is Tactile Sensing So Difficult for Industrial Robots?

Think about what happens when you pick up a glass. Your fingers can feel where they're touching it, how tightly you're holding it and whether it's starting to slip. You don't need to watch every little movement to adjust your grip because your sense of touch is already giving you much of the information you need. 

Robots don't have that natural sense of touch. Instead, they need sensors to tell them what's happening when their grippers come into contact with something. Detecting pressure is one thing, but working out exactly where it's being applied and how that contact is changing is a lot more difficult. Many tactile sensors use groups of tiny sensing elements called taxels. 

They're a bit like the pixels in a digital image, except they detect physical contact instead of light. The more taxels a sensor has, the more individual points it can measure. The problem is that adding more sensors can also mean adding more wiring, more electronics to read their signals and more information for the system to deal with. 

And when you're trying to fit all of that into something as small as a robotic finger, space becomes another problem. There's also more to touch than pressure alone. A robot might need to know whether an object is moving sideways, starting to slip or changing shape as it's being held. 

A February 2026 study published in npj Robotics looked at how different types of sensors and the physical properties of electronic skin affect a robot's ability to recognise soft objects. The researchers found that combining different ways of sensing touch worked better than relying on just one. So while adding more sensing elements can help robots collect more information, it's not necessarily the best way to improve what they can feel.

How Does Geometric Encoding Improve Robotic Touch Sensing?

One way around the problem could be to design sensors that can collect more information without needing a separate sensing element for every point of contact. That's the idea behind geometric encoding. Instead of simply adding more sensors, researchers change how the sensor is built so that different touches produce different patterns of electrical signals. 

In a study published in npj Flexible Electronics on 19 November 2026, Shaoyu Cai and colleagues used this approach to create a thin, flexible electronic skin that could identify where it was being pressed.

Encoding touch through sensor geometry

The electronic skin contains tiny channels filled with liquid metal. Because the metal carries electricity, the researchers can measure its electrical resistance, which is simply how much it slows down the flow of electricity. When someone presses against the skin, the soft material changes shape. 

This also changes the shape of the liquid-metal channels, which affects their electrical resistance. But the clever part is how those channels are arranged. Instead of keeping them separate, the researchers designed paths that weave around and overlap with one another. So when pressure is applied somewhere on the skin, it affects several channels in different ways. 

Think of it like pressing down on a soft mat with several winding wires running through it. Depending on where you press, some wires will change shape more than others. If you can measure those changes, you can start working out where the pressure came from. The researchers used a similar idea, but with a much more precise design. 

Different contact points create different combinations of resistance changes across just four sensing channels. That means the shape and arrangement of the channels are already helping to identify where the sensor has been touched.

Decoding more information from fewer measurements

Of course, the sensor still needs a way to turn those electrical changes into useful information. The researchers developed software that learns which signal patterns match different positions and amounts of pressure. 

When someone touches the sensor, the software looks at the changes across the four channels and works out where the contact happened and how much force was applied. They also developed a faster way to test different channel designs through computer simulations. This allowed them to compare layouts and find ones that produced clearer differences between touch positions before making the sensors. 

And the results were encouraging. Across three separately manufactured devices, the researchers reported an average touch-location error of just 0.230 millimetres. One of the devices also achieved a force-measurement error of 0.043 newtons. 

In other words, the electronic skin could work out where it was being touched with a high level of accuracy, even though it was only collecting measurements from four electrical channels. The researchers also demonstrated the technology through pressure-sensitive handwriting and a wearable keyboard. 

But they haven't yet shown it working on an industrial robot. So while the results tell us quite a lot about what this sensor design can do, they don't prove that it would make a robot better at picking up or handling objects.

Is Smarter Sensor Design Better Than Adding More Sensors?

Geometric encoding is an interesting way to get more information from fewer electrical channels. But it's not the only approach researchers are looking at, and using fewer sensors isn't always the goal. An October 2026 study in Nature Machine Intelligence, for example, described an electronic skin with more than 10,000 sensing units. 

Instead of trying to reduce the number of sensors, the researchers developed a way to process and compress the information they collected. The system could read touch information at 100 frames per second while reducing the amount of data it needed to send by hundreds of times. 

They also tested it on robots carrying out tasks that needed very fine control. So we've got two different ways of dealing with a similar problem. Geometric encoding tries to get more detail from a small number of measurements. The larger electronic skin collects information from thousands of points, then makes that information easier to handle. 

And there's another part of the problem to consider: actually making the sensors. In January 2026, researchers writing in npj Flexible Electronics developed a way to make flexible electronic skin using laser-patterned circuits and methods that could avoid some of the more complicated manufacturing steps. 

They also demonstrated sensors fitted to robotic grippers. These studies show that researchers aren't all trying to solve robotic touch in the same way. Some are working on the sensors themselves, while others are finding better ways to handle their signals or make the technology easier to produce. 

Which approach works best will depend on what the robot needs to feel and what it needs to do with that information.

What Would Better Tactile Sensing Mean for Industrial Robot Dexterity?

Better tactile sensing could help industrial robots handle objects that need a little more care than simply picking something up and putting it down somewhere else. Imagine a robotic gripper picking up a soft piece of fruit. It needs to hold the fruit tightly enough that it doesn't fall, but not so tightly that it gets crushed. 

If the gripper can feel exactly where it's touching the fruit and how much pressure it's applying, its control system could use that information to adjust its grip. Researchers are already testing electronic skin for tasks like this. A May 2026 study published in Nature Communications described a flexible sensor fitted to a robotic hand. 

It could respond to pressure changes in 47 milliseconds, and the researchers demonstrated the hand grasping different fruits under controlled conditions. Similar technology could eventually help robots with precision assembly, where even a small amount of unexpected pressure might mean a component isn't lined up properly. 

It could also help them handle objects with unusual shapes, where the points of contact aren't always the same. But there's something important to remember here. Knowing where an object is touching a robot doesn't automatically mean the robot knows how to handle it. 

It still needs the right gripper, suitable control software and enough information to understand what's happening as the object moves. For example, knowing how hard something is being pressed won't necessarily tell the robot whether it's starting to slip sideways. 

And just because a sensor can locate a touch within a fraction of a millimetre doesn't mean the robot itself can move or position objects with that same accuracy. The real test will be whether better touch information actually helps robots handle difficult objects more successfully.

What's Still Preventing Advanced Tactile Sensors From Reaching Factories?

Are you enjoying the content so far?

Getting good results in a laboratory is an important first step. But there's a big difference between testing a sensor under controlled conditions and using it on a robot that's working all day in a factory. A sensor might be very good at detecting where someone presses against it. 

But what happens when it's gripping several parts of an object at once, moving quickly or being used thousands of times? Those are some of the questions researchers still need to answer.

Performance under real working conditions

One of the first things to test is whether the sensor can deal with several points of contact at the same time. A robotic gripper might touch an object in three or four places, with the pressure changing at each point as the object moves. The Cai research showed that its sensor could identify touch positions and measure pressure applied directly against it. 

But it didn't establish how well the design would handle several simultaneous touches or sideways movement. Speed is another concern. If something starts slipping, the robot needs to detect it and react quickly enough to stop it from falling. So it's not just about how fast the sensor collects information. 

The software also needs time to work out what's happening, and the robot needs time to respond. Then there's the question of how well the sensor holds up over time. Repeated pressure, wear and changes in temperature could affect the material and the electrical signals it produces. 

Since geometric encoding relies on recognising patterns in those signals, changes to the sensor could affect how accurately it works. The Cai researchers also found that changes to things like material stiffness or channel size would mean recalibrating their simulation model. How often that might be needed in a working robot still needs to be tested.

Manufacturing and system integration

Even if the sensor works well, manufacturers still need to be able to make it reliably. The Cai researchers tested three separately made devices, which is a useful start. But making three sensors that work well isn't the same as producing thousands that all perform consistently. The tiny liquid-metal channels would need to be made accurately and protected from damage. 

The flexible skin would also need to fit onto robotic grippers without affecting their movement, and it would have to withstand repeated use. Then there's the rest of the system to think about. Having fewer electrical channels could make some parts of the hardware simpler. 

But the robot would still need electronics to read the signals and software to work out what they mean. Those parts could bring their own costs, maintenance needs and technical challenges. For manufacturers, the most useful evidence would come from testing the complete system on real handling tasks.

  • How often does the robot successfully pick something up?
  • Can it avoid damaging delicate objects?
  • Does it react when something slips?
  • And can it keep doing those things reliably after thousands of movements?

Until those questions have been answered, geometric encoding is a promising new way to design tactile sensors rather than a technology that's ready to make industrial robots more dexterous.

Final Thoughts: Better Robotic Touch Starts With Better Information

Making industrial robots better at handling objects isn't necessarily about giving them more sensors. It's about helping them get the information they need to understand what's happening when they touch something. Geometric encoding offers an interesting way to do that. 

By changing how a sensor is designed, researchers have shown that it's possible to detect touch very accurately without needing lots of separate electrical channels. But detecting touch is only part of the job. Robots still need to use that information to handle objects safely and reliably, and there's more testing to do before we know how well this approach could work in factories. 

As electronic skin improves, we may find that the number of sensors a robot has is less important than how much useful information it can get from them. And with developments like these changing what's possible, EM360Tech continues to look at the emerging technologies that could shape how businesses work in the years ahead.