For his sixth birthday, my son received a Lego Ninjago building set featuring a dragon guardian, which we assembled together. Or rather: I handed him the pieces, while he clicked them together at great speed. His eyes were already on the next step while his fingers rotated the blocks at precisely the right angle. I watched with justified pride, while at the same time wondering: why is something we take for granted, such as the human hand, so difficult to put into a robot?
Read the opinion piece in De Tijd ( in Dutch)
The hand is one of the biggest obstacles to the humanoid robot revolution. To move our hand, we use more than 30 muscles. If you want a robotic hand with the same degree of freedom of movement, you basically need the same number of motors. But where do you put them all?
Some manufacturers cram them into the hand itself, which is quite a feat, as there is hardly any room in slender fingers. In humans, many of those muscles are not located in the hand but in the forearm. Quite a number of robot builders copy that trick. Our evolution took millions of years to refine the coordination of all those muscles. Engineers are now trying to replicate that in a handful of design cycles, and at a steep price.
Besides the motors, there is the problem of sensors. In a well-known experimental video, a woman lights a match. She can see perfectly what she is doing, but she feels nothing because researchers have administered an anaesthetic, preventing her fingertips from transmitting tactile information. The result is striking: she suddenly becomes extremely clumsy. Our fingertips contain thousands of tactile sensors, which measure not only touch but also vibration and temperature. Here too, robotics still has a way to go.
"The refined coordination of all those muscles is the result of millions of years of evolution"
All that sensor data also has to be processed. Enter the brain, or in the case of a robot: computing power. Manual dexterity and brain size appear to have gone hand in hand throughout our evolution: free hands capable of using tools required ever finer motor skills, and those motor skills demanded greater brain capacity. Hand and brain thus drove each other forward. The robot brain must evolve as well, and for that, data is currently lacking.
Then there are the materials. Our fingers combine hard components (bones and nails) with soft tissue, and it is precisely this combination that makes us dexterous. A robotic hand therefore also becomes a highly complex Lego construction set in its own right: countless motors, sensors, and a combination of hard and soft materials, all controlled by a powerful brain.
Together with Melexis, our researchers recently developed an open-source construction kit for robotic fingers that can sense both pressure and shear forces through 3D tactile sensors. Several robot manufacturers are testing it. And in the Flemish Fidex project, researchers from Ghent University, VUB, and Imec are developing modular fingertips that combine different sensors. What we hope to see is research from Flemish laboratories ending up, quite literally, in the hands of future robots. Before long, my son will tear open another box of Lego, with the same natural ease as always. I am curious to see the first humanoid robot capable of clicking such a building set together. Small and large developments in the tech world, explained by an insider.