DYMO-Hair: Generalizable Volumetric Dynamics Modeling for Robot Hair Manipulation

Abstract-- Hair care is an essential daily activity, yet it remains inaccessible to individuals with limited mobility and challenging for autonomous robot systems due to the fine-grained physical structure and complex dynamics of hair . We introduce a novel dynamics learning paradigm that is suited for volumetric quantities such as hair, relying on an action-conditioned latent state editing mechanism, coupled with a compact 3D latent space of diverse hairstyles to improve generalizability. This latent space is pre-trained at scale using a novel hair physics simulator, enabling generalization across previously unseen hairstyles. Experiments in simulation demonstrate that DYMO-Hair's dynamics model outperforms baselines on capturing local deformation for diverse, unseen hairstyles. DYMO-Hair further outperforms baselines in closed-loop hair styling tasks on unseen hairstyles, with an average of 22% lower final geometric error and 42% higher success rate than the state-of-the-art system. Real-world experiments exhibit zero-shot transferability of our system to wigs, achieving consistent success on challenging unseen hairstyles where the state-of-the-art system fails. T ogether, these results introduce a foundation for model-based robot hair care, advancing toward more generalizable, flexible, and accessible robot hair styling in unconstrained physical environments. Hair is central to personal identity and self-esteem [1], [2], yet routine care is difficult for individuals with limited mobility due to reduced coordination, strength, and flexibility [3]. To improve accessibility and autonomy, robot hair care systems have been explored [4]-[7], but existing approaches rely on either handcrafted trajectories or rule-based controllers, restricting generalization across diverse hairstyles and goals. To address these limitations, we propose DYMO-Hair, a model-based robot hair care system. Our system is capable of generalizable and flexible visual goal-conditioned hair manipulation, across diverse hairstyles and objectives in unconstrained physical environments. Chengyang Zhao, Uksang Y oo, Jonathan Francis (by courtesy), Jeffrey Ichnowski, and Jean Oh are with Robotics Institute, Carnegie Mellon University, Pittsburgh, Pennsylvania, USA. Arkadeep Narayan Chaudhury is with Epic Games, Inc., Pittsburgh, Pennsylvania, USA. Giljoo Nam is with Meta Codec Avatars Lab, Pittsburgh, Pennsylvania, USA. Jonathan Francis is with Bosch Center for Artificial Intelligence, Pittsburgh, Pennsylvania, USA. Figure 1. We introduce DYMO-Hair, a unified, model-based robot hair care system.