Research videos aren’t known for being highly narrative. However, sometimes you find something that just keeps getting better the longer you look at it. “Beyond Manual Dexterity” is one such video. The extremely dexterous robot gripper will amaze you in the first few seconds, but then things start to go off track.
The video was first shown alongside a research paper of the same name at the IEEE International Conference on Robotics and Automation in Rotterdam this week. The research, carried out by Switzerland’s EPFL, explores how robotic hands can grasp a wider range of objects beyond their existing limits.
“Deep learning models have significant advances in dexterous manipulation techniques for multi-finger hand grasping,” the researchers said. “However, contact information-based understanding in cluttered environments remains largely unexplored.”
For decades, roboticists have looked to the natural world for inspiration. Today’s robots are built on these foundations. Look no further than the quadrupeds and humanoid robots that currently flood the market. Living things are a rich and seemingly endless source of inspiration, but they can also drive designers into a corner.
But why should robotic hands be constrained by the limitations of their natural counterparts? A common answer may be a lack of imagination. But the research on display here shows what these systems look like when people intentionally break their bonds.
The results may be a little spooky. The human-inspired hand design, which can bend over and grab a wide range of objects, or break away and crawl onto another, is firmly planted within the uncanny valley. Perhaps there would be less anxiety if it didn’t look so much like a human hand to begin with.
Image credit: EPFL (4x speed)
But in order to function in this world, we have to push past that instinctive reaction. Beyond that, we find a fascinating way forward for robotic grasping. Whether something like this scales well is another question. So far, this research has undoubtedly generated a lot of interest in the field.
“You often see it in scary movies, but I think we are the first to bring this idea to robotics,” researcher Xiao Gao cheekily told IEEE.
Biologically inspired robots for a long time. Hello, this is Horrorbot.
