Helping Robots Get Hands-On Makes All The Difference
Computing power has made it possible to train robots for seemingly endless tasks. As a society, we are questioning how far we should take artificial intelligence now that the possibility for doing so appears limitless.
However, even as “brains” we incorporate into robots grow faster and more complex, there’s a human capability that’s still holding them back: dexterity. Robots still can’t quite compete with our ability to grasp, feel, touch, manipulate, and identify objects.
But, there are researchers and tech developers working on ways to solve that problem right now, and it could change the way we employ robots in the future.
Image Source: IEEE Spectrum
The Challenge Of Mimicking Human Motion
Many researchers, like roboticist Amir Shapiro of Ben-Gurion University of the Negev in Israel, fully understand the challenge of mimicking the human hand and all its capabilities in robot form.
As Shapiro points out, “we have more than 20 degrees of freedom in our hands” and attempting to reproduce that would require more than 20 different motors, programed and automated separately.
In addition to sounding complex, these components would seriously weigh down and inhibit a robot’s arms. 3D printing, laser cutting, and other technologies—like those used to create a remarkably complex experimental robot hand at the University of Washington last year—can help lighten the load and allow for more fluid, anthropomorphic movement.
Simplify The Way We Grip
Other robotic developers have gone for a more simplified approach. Rather than attempting to recreate a human hand, developers create robots with grippers or claws comprised of two or three digits.
This structure requires fewer motors and other components but it still allows for a range of different gripping and handling capabilities. This may be the structure we see in robots used to perform various hands-on tasks in warehouse, assembly lines, and manufacturing facilities.
Image Source: MIT Technology Review
Combining Machine Learning With The Right Touch
A notable example comes from RightHand Robotics, Inc., a startup based in Somerville, MA. They have developed a new robotic picking platform that’s comprised of a robotic arm with gripper.
Using cloud-based, machine learning capabilities, the robotic platform can pick up, identify, and sort a range of different objects.
RightPick, as it’s called, shows how a higher level of the robotic dexterity and machine learning would be needed to successfully perform industrial sorting and packing tasks without much human intervention—particularly when the objects vary considerably.
It’s going to take more than just a smart artificial brain to get human jobs done, but as developers are uncovering new ways to get robots moving like humans, they are advancing into our lives even faster than most of us expected. What are your thoughts on improvements in robotic dexterity and their use in the workplace?