A robotic drummer

  • Thread starter Thread starter pinball1970
  • Start date Start date
Click For Summary
SUMMARY

The discussion critiques the paper titled "Robot Drummer," which claims to present a humanoid capable of high-precision drumming across various songs using reinforcement learning. Participants express skepticism regarding the robot's performance, noting issues such as incorrect rhythms, lack of a bass drum, and overall poor timing. The consensus is that the claims of "high-precision drumming" are overstated, with the robot's actual performance failing to meet the standards suggested in the abstract. The need for more transparency in the research and its methodologies is emphasized.

PREREQUISITES
  • Understanding of reinforcement learning principles
  • Familiarity with MIDI transcriptions and their application in music
  • Basic knowledge of humanoid robotics
  • Awareness of drumming techniques and musical timing
NEXT STEPS
  • Research reinforcement learning frameworks for robotics
  • Explore MIDI transcription techniques for music performance
  • Investigate humanoid robotics and their applications in music
  • Study the impact of timing and rhythm in musical performance
USEFUL FOR

Musicians, robotics engineers, and researchers interested in the intersection of AI and music performance, particularly those evaluating the effectiveness of robotic systems in replicating human musicality.

pinball1970
Gold Member
2025 Award
Messages
3,661
Reaction score
5,795
TL;DR
In this paper, we introduce Robot Drummer, a humanoid capable of expressive, high-precision drumming across a diverse repertoire of songs. We formulate humanoid drumming as sequential fulfillment of timed contacts and transform drum scores into a Rhythmic Contact Chain. To handle the long-horizon nature of musical performance, we decompose each piece into fixed-length segments and train a single policy across all segments in parallel using reinforcement learning.
Paper https://arxiv.org/abs/2507.11498

Article https://techxplore.com/news/2025-08-robotic-drummer-gradually-human-behaviors.html

Not my area of expertise at all, Robotics, programming but I do play drums. If you play the clip there is a lot wrong the with basic things the drummer is playing.
In the animation the drummer is stood, there no bass drum, is playing open hand - not usual.
If the animation is purely for illustration and not representative of the music then we can look at that also.
I cannot hear a separate BD, sounds like backing track.
Goes out of time.
Wrong rhythm and also goes "upside" down in Roxanne.

So I disagree with the abstract "high-precision drumming across a diverse repertoire of songs."

Opinion on what they have actually done from our high tech members will be welcome and also musicians who may be a little confused with the claims and results.
 
  • Like
  • Informative
Likes   Reactions: phinds, berkeman and Greg Bernhardt
Computer science news on Phys.org
I think they have some catching up to do... :wink:



1754670917310.webp
 
  • Wow
  • Like
Likes   Reactions: Greg Bernhardt and pinball1970
pinball1970 said:
Opinion on what they have actually done from our high tech members will be welcome and also musicians who may be a little confused with the claims and results.

My opinion is not printable within forum rules.

Whoever has published this should have their funding withdrawn.
 
Last edited:
  • Agree
  • Like
Likes   Reactions: 256bits and pinball1970
pinball1970 said:
TL;DR Summary: In this paper, we introduce Robot Drummer, a humanoid capable of expressive, high-precision drumming across a diverse repertoire of songs. We formulate humanoid drumming as sequential fulfillment of timed contacts and transform drum scores into a Rhythmic Contact Chain. To handle the long-horizon nature of musical performance, we decompose each piece into fixed-length segments and train a single policy across all segments in parallel using reinforcement learning.

So I disagree with the abstract "high-precision drumming across a diverse repertoire of songs."
From the article:
They found that the robot effectively learned complex rhythmic structures and played the songs with high accuracy, often with a rhythmic precision of over 90%.

So if the robo improves, then it becomes what "extreme ultra high precision" gobblylygook.
The authors of this 'study', and pretty much anything associated with AI, or robotics, fell into the present day common theme of overselling inferiority and defective products or research.

"To address this gap, we introduce Robot Drummer, a general RL framework for humanoid drumming that learns to perform drum patterns derived from MIDI transcriptions."
That's what I thought - Midi drumming, as the simulation video sounds so awful x86 16 bit, which is grating to the ears.

And only for the reason of over hype should this be cancelled. If they can't be truthful, then when will they ever. I imagine a lot went into this project to have the simulated robo read 'sheet' music, which is quite an accomplishment. Emphasis on THAT would be more in accordance with Machine Learning and the algorithms that went into it. But the article is sparse of details, and one cannot say if the robo is actually reading music on its own, or just flopping its arms around from directions given by a non AI program.
 
  • Like
Likes   Reactions: berkeman

Similar threads

Replies
2
Views
3K
  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 11 ·
Replies
11
Views
6K
  • · Replies 13 ·
Replies
13
Views
4K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 1 ·
Replies
1
Views
10K