Newly developed robot can play the drums, listen and learn

It seems simple: The robot ZRob, a small machine easily placed on a table, holds a drumstick and beats a drum. In reality, it is anything but simple. ZRob’s arm has a flexible grip, just like the human wrist. It also has an artificial intelligence (AI) that helps it optimize its movements.

If you put several ZRobs together, they can play drum rolls, which requires very fine movements. They can also play rhythms entirely different from what humans can achieve but with a touch of human sensitivity.

“Playing together, the robots will find a pattern unlike everything else,” Mojtaba Karbasi says. He has developed ZRob as part of a doctoral thesis at the RITMO Center for Interdisciplinary Studies in Rhythm, Time and Motion, University of Oslo. Karbasi will defend his dissertation November 29.

A bio-inspired approach

Humans still manage several tasks better than robots. Humans are far more complex. For instance, we can see a glass of water, estimate its weight, and thus lift it without spilling.

“Humans talk, move and touch things. We have several senses and can act on what we sense with adapted movements. Machines cannot do this,” Alexander Refsum Jensenius, professor at RITMO and Karbasi’s supervisor, says.

He believes that machines will have to become multimodal, i.e., have more senses, if they are to interact well with people in the future. Karbasi had this in mind and used a so-called bio-inspired approach when developing ZRob and the AI model that comes with it.

Similar to a human wrist

“We are not trying to copy a human, but rather to take the human body as a starting point. Musical robotics is a controlled way of working with this. It does not have potentially dangerous consequences,” Jensenius says.

One of the things that makes ZRob more human is his arm. Two springs—one on each side—give the arm some of the flexibility of the wrist.

The springs and the AI make the robot adaptable, an important feature when hitting a drumhead in motion.

ZRob has two senses: It listens and can sense where its arm is.

The body of the robot shapes the behavior

Karbasi explains that the springs in the various versions of ZRob are of different sizes. The AI helps each robot to find its behavior.

“If a robot is trying to play a double stroke, it wants to find the right action parameters to execute it. If you change the springs, the behavior will change, too. The robotic body thus helps to shape the behavior,” Karbasi says.

If several robots play simultaneously, you may hear rhythms humans cannot play.

“Different physical characteristics completely shape the output. It is like a piano piece played by forty fingers—unpredictable, and it is hard for humans to imagine how it was produced.”

He does not believe that this kind of music can touch people the same way as humanly played music. Nevertheless, he believes that his robots can inspire musicians.

“A robot that precisely mimics human behavior is not inspiring. Then humans could do the same, even better. However, if we manage to create robots that take some humanity into them but are unique in their way, it inspires to a much greater extent,” Karbasi says.

It also opens up new possibilities; he adds and mentions the newly developed robots of the company Boston Dynamics: The robots resemble humans, but have a different way of standing up and can even turn their upper body 360 degrees.

Jensenius believes that ZRob is interesting for musicians since it can create new types of rhythms and play faster than humans play.

In addition, it can be a practical help.

“If you play the guitar and sing, you can get a drum robot to play with you. If you are a drummer who has lost an arm, you can use a robot as a replacement,” Jensenius says.

The research also has a high transfer value, according to Jensenius.

“If we manage to create complex drum robots that sense and adapt their behavior accordingly, we are closer to creating robots that can perform other types of actions, such as lifting a glass of water or cutting a cucumber without problems.”

Karbasi has also published multiple papers on ZRob, and some of his latest research is published in the journal Frontiers in Robotics and AI.

Provided by
University of Oslo