In a significant step for robotics, engineers from the University of Washington and UCLA have developed a flexible sensor-filled membrane or ‘skin’ that opens the door to critical advancements in how robots grip and manipulate objects.
For all the apparent precision and dexterity of modern robotics technology, there’s scope for vast improvements when it comes to robots sensing what they are handling and making necessary adjustments.
While we’ve previously reported on how humans are reassured by the fallibility of awkward robots, if one happens to be performing surgery, or even simply handling breakable objects, there’s a strong incentive for a more human level of care.
Read more: Number of service robots to reach 264 million by 2026
Drawing on bio-inspiration
Scientists often turn to the natural world when problem solving – to draw upon solutions that evolution has had millions of years to refine. From camouflage to locomotion, nature offers a rich vein of inspiration. This latest research produced a sensor skin that mimics the way our own fingers experience tension and compression.
“It’s really following the cues of human biology,” said lead author Jianzhu Yin. “Our electronic skin bulges to one side just like the human finger does.”
The flexible sensor layer is formed by embedding serpentine channels, half the width of a human hair, into silicone rubber. These pathways contain conductive liquid metal that can be manipulated in a way that would break solid wires. As the channel geometry changes, so too does the amount of electricity that can flow through them.
Read more: CSAIL team pairs robots with VR for smart manufacturing
Entering uncanny valley
The sensor skin can be stretched over any part of a robot’s body or prosthetic to relay data about shear forces and vibration that is vital to handling objects. The new technology uses what’s known as a ‘multimodal’ approach to enable dexterous manipulation. Unlike past tactile sensor designs, the skin incorporates normal forces, shear forces and vibration.
This wealth of extra data creates new possibilities for machine-learning – ultimately leading to robots that are more aware and therefore more capable. Once robots can sense like we do, they need to be able to similarly react and adapt.
“If a robot is going to dismantle an improvised explosive device, it needs to know whether its hand is sliding along a wire or pulling on it,” says senior author and UW professor Jonathan Posner. “To hold on to a medical instrument, it needs to know if the object is slipping. This all requires the ability to sense shear force, which no other sensor skin has been able to do well.”
We may be some way off the technology seen in science fiction films such as Ex Machina, but as social cognizance advances in robots, alongside physical detection and response, we continue along the path to uncanny valley.