Learning Robot Manipulation from Audio World Models

Fan Zhang, Michael Gienger

Keywords: Robot Manipulation, Audio World Model, Flow Matching

TL;DR: We propose a flow matching–based audio world model that facilitates robot policy learning, capturing intrinsic physical dynamics, such as pitch and rhythmic patterns.

Abstract: World models have demonstrated impressive performance on robotic learning tasks. Many such tasks inherently demand multimodal reasoning; for example, filling a bottle with water can make visual information alone ambiguous or incomplete, thereby requiring reasoning about the temporal evolution of audio, accounting for its underlying physical properties and pitch patterns. In this paper, we propose a generative latent flow matching model to anticipate future audio observations, enabling the system to reason about long-term consequences when integrated into a robot policy. We demonstrate the superior capabilities of our system through two manipulation tasks that require perceiving in-the-wild audio or music signals, compared to methods without future lookahead. We further emphasize that successful robot action learning for these tasks relies not merely on multi-modal input, but critically on the accurate prediction of future audio states that embody intrinsic rhythmic patterns.

Submission Number: 47