When looking for the perfect soundtrack to accompany your holiday celebrations, consider putting away the traditional albums and opting for this technologically savvy rendition of Jingle Bells instead. The performer, the latest work from Josie Hughes and colleagues at the University of Cambridge, is none other than a robot hand designed to play the piano without moving its fingers. While it may not be a virtuoso, the robot exhibits how complex movement can be achieved through clever design.
Typically, robot hands are equipped with controllers that move each finger individually, resulting in a complex network of circuits and motors mimicking muscles. Instead, Hughes’ group designed their robot to operate through “passive dynamics”—only the motorized wrist moves while the fingers remain stationary. They used 3D printing technology to give the robot hand humanlike features, namely bones, ligaments, and joints with specific levels of stiffness. Recreating the mechanical properties of the human hand made the robot hand dexterous enough to play the piano in a variety of styles without moving its fingers. For example, it was able to play both short, detached notes (staccato) and continuous, sliding notes (glissando), depending on how it was placed on the piano.
Getting robot hands to perform complex tasks has been a big challenge—even small children are able to outperform robots on jobs that require manual dexterity. This work demonstrates that humanlike mechanical properties can provide robots with a layer of built-in intelligence and outweigh the need for lots of complex motors. Future robot designs incorporating passive dynamics will be able to perform natural movements with less energy and may be useful in industries like medical prosthetics and manufacturing.
The technology has a long way to go, but in the meantime, let’s give the robot hand a break on the Jingle Bells mistakes—the scientists just have some more practicing to do.
Managing Correspondent: Benjamin Andreone
News Article: Robot hand that plays Jingle Bells could help us make better limbs. New Scientist
Original Article: An anthropomorphic soft skeleton hand exploiting conditional models for piano playing. Science Robotics
Image Credit: Josie Hughes
thenkes best article -_-.
In Facebook’s Intro to Best Deep Learning course, you can learn the foundations of deep learning and build your deep neural networks using PyTorch.
You will gain real-world competence in developing and training deep neural networks through coding exercises and projects.
Yeah, I saw a Mexican robot do better in 1993. https://www.youtube.com/watch?v=PWRd4L_QzX4
I get a bit tired of people saying they’re going to improve prosthetic arms. The problem is not the mechanical actuator – we can already make ridiculously complicated hands. The problem is how to control them in an intuitive and practical way. And don’t start going on about myoelectric or brain gate systems – they are next to useless. The only proven way (the reason most amputees still prefer a Bowen cable body-controlled split hook) is extended physiological proprioception – a concept that is now all but forgotten in the rush to more and more high-tech silliness.