Imagine a world where robots seamlessly navigate our human-centric environment, effortlessly grasping tools and objects just like us. But here's the catch: creating a robotic hand that matches the dexterity and versatility of the human hand is one of the most daunting challenges in robotics today.
Two decades ago, Rich Walker, now a director at Shadow Robot, fondly recalls crafting the company’s first robotic hand using wood, springs, and rubber bands. "We built it with whatever we had lying around," he says. Fast-forward to today, and I’m at Shadow Robot’s north London headquarters, witnessing their latest innovation. These robotic hands feature cylindrical 'forearms' housing small electric motors called actuators, which pull on metal tendons to move the fingers with remarkable precision. During a hands-on demo, I expected clumsy, erratic movements, but instead, I was amazed by the smooth, controlled manipulation of blocks and cups—a testament to how far the technology has come.
Currently, around 200 of these hands are in use, primarily by researchers at universities and tech firms. Walker describes them as "development kits for dexterity," enabling users to explore what’s possible before scaling up to larger projects. "It’s about understanding what you can achieve in terms of dexterity and then deciding what to build next," he explains. This groundwork is crucial if robots are ever to truly integrate into our human-designed world, where nearly every tool and device is tailored for the human hand.
But here’s where it gets controversial: Bren Pierce, founder of Bristol-based robotics start-up Kinisi, boldly claims, "The hand is the hardest, most complex part of any humanoid robot." His company’s KR1 robots, currently undergoing commercial trials, use interchangeable grippers—strong 'gorilla' pincers for heavy lifting and suction devices for delicate tasks. Yet, like many in the industry, Pierce dreams of a single, all-purpose hand that can do it all. "For 40 years, people have been dreaming of one robot hand to rule the world. Many believe the humanoid hand could be the answer," he says.
Kinisi has developed a three-fingered prototype that Pierce describes as "pretty good." However, the real challenge lies in making it durable, scalable, and affordable. Their current prototype costs £4,000 ($5,400), a staggering 10 times the price of their simple pincer gripper, which retails for just £400. And this is the part most people miss: even tech titan Elon Musk, speaking at the All-In Summit last September, acknowledged that creating a functional robotic hand is one of the three biggest hurdles in humanoid robotics—alongside developing advanced AI and mass production.
Tesla’s upcoming humanoid robot, Optimus, promises to address this with a hand boasting "human-level manual dexterity." But not everyone is convinced. Nathan Lepora, Professor of Robotics and AI at Bristol University, dismisses Musk’s claim as "rubbish." Having dedicated his career to robotic hands, Lepora believes human-level dexterity is still a decade away. "It’s not happening in two years," he asserts.
Lepora’s work, like Shadow Robot’s, focuses on tendon-driven hands, which he believes will eventually become more affordable and capable. However, he’s impressed by Chinese firms like Wuji Technology, which are taking a different approach—using motors embedded in the fingers and joints to drive movement. "They’re creating bespoke motors that fit within the hand’s structure, and it’s looking very promising," he says. Wuji’s latest hand features four independently controlled joints per finger, enabling intricate movements, and co-founder Yunzhe Pan promises increased durability and affordability in future iterations. Currently priced at $12,000 (£8,800), the hand also incorporates piezoelectric sensors for a sense of touch—a potential game-changer for humanoid robots.
But here’s the real question: Can these advancements overcome the durability issue? Pierce points out that while research labs boast impressive sensors, they often fail after just six months. "For industry, we need robots that last 10 years," he emphasizes. Yet, there’s reason for optimism. Lepora notes that tactile sensing, once a distant dream, is now within reach thanks to the billions being invested in humanoid robotics. "Things are changing," he says.
So, what do you think? Is Elon Musk’s vision of a human-like robotic hand achievable in the near future, or is Nathan Lepora’s decade-long timeline more realistic? And which approach—tendon-driven or motor-embedded—will ultimately win out? Let’s spark a debate in the comments!
