Robot Drummer: Learning Rhythmic Skills for Humanoid Drumming

2025-07-15

AI Summary

Humanoid robots still face significant challenges in high-dynamic, long-duration musical performance—such as drumming—including rhythmic precision, multi-limb coordination, and real-time adaptability. To address these, we propose the “Rhythmic Contact Chain” modeling framework, which encodes drum scores as spatiotemporally constrained striking sequences and trains a unified, cross-phrase policy via reinforcement learning. This enables fine-grained motion decomposition, parallel policy optimization, and adaptive drumstick allocation. Our approach achieves stable, minute-long performances across over 30 rock, metal, and jazz pieces—attaining high F1 scores—and discovers human-like strategies including cross-limb striking and genre-adaptive phrasing. This work overcomes critical bottlenecks in temporal modeling and long-horizon coordination for expressive robotic tasks, establishing a novel paradigm for creative embodied intelligence.

Technology Category

• Intelligent Robots: Manipulation
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Research challenges in human and human-AI computation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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 pieces lasting minutes. In this paper, we introduce Robot Drummer, a humanoid system 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 in to 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 sticks assignments, demonstrating the potential of reinforcement learning to bring humanoid robots into the domain of creative musical performance. Project page: href{https://robot-drummer.github.io}{robot-drummer.github.io}

Problem

• Develop humanoid robot for expressive drumming performance
• Address split-second timing and multi-limb coordination challenges
• Enable long-horizon musical tasks via reinforcement learning

Innovation

• Formulates drumming as sequential timed-contact fulfillment
• Uses Rhythmic Contact Chain for drum score transformation
• Trains single policy via parallel reinforcement learning segments