Robot Drummer: Learning Rhythmic Skills for Humanoid Drumming

TL;DR

This paper presents Robot Drummer, a humanoid robot capable of expressive, high-precision drumming across various genres by formulating drumming as sequential contact fulfillment and training a reinforcement learning policy for long-horizon musical performances.

Contribution

It introduces a novel reinforcement learning approach to enable humanoid robots to perform complex, expressive drumming with human-like strategies across diverse musical pieces.

Findings

Robot Drummer achieves high F1 scores on over thirty tracks.

Emergent human-like drumming behaviors are observed.

The approach demonstrates adaptive and expressive musical performance.

Abstract

Humanoid robots have seen remarkable advances in dexterity, balance, and locomotion, yet their role in expressive domains such as music performance remains largely unexplored. Musical tasks, like drumming, present unique challenges, including split-second timing, rapid contacts, and multi-limb coordination over performances lasting minutes. 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. Through extensive experiments on over thirty popular rock, metal, and jazz tracks, our results demonstrate that Robot Drummer consistently achieves high F1 scores. The learned behaviors exhibit emergent human-like drumming strategies, such as cross-arm strikes, and adaptive stick assignments, demonstrating the potential of reinforcement learning to bring humanoid robots into the domain of creative musical performance. Project page: robotdrummer.github.io