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

TL;DR

DYMO-Hair tackles autonomous, generalizable robot hair styling by introducing the first 3D volumetric hair dynamics model trained with large-scale synthetic data and a ControlNet-inspired latent-space editing paradigm. The system combines a compact, pre-trained 3D latent space with an action-conditioned dynamics model and a MPPI-based planner to perform visual goal-conditioned styling in 3D space. A novel Genesis-based PBD hair simulator enables large-scale synthetic data generation for strand-level, contact-rich dynamics, supporting robust closed-loop manipulation. In simulation and real-world wig tests, DYMO-Hair outperforms baselines in local deformation modeling and goal attainment, achieving zero-shot transfer to unseen hairstyles and demonstrating the potential for generalizable, accessible robot hair care in unconstrained environments.

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. In this work, we present DYMO-Hair, a model-based robot hair care system. 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. Using the dynamics model with a Model Predictive Path Integral (MPPI) planner, DYMO-Hair is able to perform visual goal-conditioned hair styling. 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. Together, 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. More details are available on our project page: https://chengyzhao.github.io/DYMOHair-web/.

Figures (11)

• Figure 1: DYMO-Hair Overview. We introduce DYMO-Hair, a unified, model-based robot hair care system. We propose the first 3D volumetric hair-combing dynamics model, featuring a novel learning paradigm. It uses an action-conditioned latent state editing mechanism, coupled with a compact 3D latent space of diverse hairstyles, enabled by our novel hair-combing simulator, for generalizable dynamics modeling. Building on this model, we develop DYMO-Hair with a MPPI-based planner for closed-loop visual goal-conditioned hair styling.
• Figure 2: Comparison of Hair State Representations. (a) Colors distinguish individual hair strands. (b) We show a resolution of 2K points, the maximum used for point cloud–based methods in our experiments (see \ref{['sec:exp-dynamics']} for more details). (c) We show $64 \times 64 \times 128$ grids with a voxel size of about 5 mm. Colors denote local strand orientations. Red dashed box: a zoomed-in region. Brown dashed box: the corresponding local strand segments.
• Figure 3: DYMO-Hair's Dynamics Model Overview.Left: State latent space pre-training. A 3D volumetric hierarchical model with vector quantization enables compact compression while preserving detailed representation capability. Right: Dynamics learning. The pre-trained model is adapted to capture hair dynamics in a ControlNet-style framework, formulating dynamics as action-conditioned editing in the pre-trained state latent space. zero: zero-convolution; copy: weight copying for initialization; $\oplus$: element-wise addition; $\otimes$: 3D attention-based feature fusion. In this phase, only the motion encoding path is trainable, with all pre-trained components frozen.
• Figure 4: Constraints for PBD-based Strand-level, Contact-rich Hair Combing Simulation.Top: Each constraint’s formulation and intended effect. Bottom: Simulation results under progressive constraint addition. Starting from the initial state (far left), a combing motion is applied along the white dashed arrow. The red dashed box marks the contact-rich region, with its simulation results shown on the right. With all three constraints, the hair maintains a realistic shape; the twist constraint, in particular, preserves curvature and prevents gravity-induced oversmoothing.
• Figure 5: Experiment Setup.Left: Synthetic hair used for training and evaluation; simulation environment. Right: Real-world setup for evaluation.