Skip to content

← ArchivePaper2023

Towards Realtime: A Hybrid Physics-based Method for Hair Animation on GPU

Li Huang, Fan Yang, Chendi Wei, Yu Ju (Edwin) Chen, Chun Yuan, Ming Gao

SCA9 citesCFX

This hybrid physics-based method targets real-time hair simulation for games, avatar streaming, and metaverse applications.

Abstract

This hybrid physics-based method targets real-time hair simulation for games, avatar streaming, and metaverse applications. It treats inter-strand collisions as a means to preserve overall hair volume, resolving them with an explicit Material Point Method while simulating individual strands with a semi-implicit Discrete Elastic Rods model. The GPU pipeline reaches up to 260 frames per second with more than two thousand simulated strands on an Nvidia GeForce RTX 3080.

How to read this

Category
Method: real-time hair simulation on GPU (hybrid physics)
Contributions
  • A hybrid scheme that simulates individual strands with a semi-implicit Discrete Elastic Rods model.
  • Treats inter-strand collisions as a volume-preserving process resolved with an explicit Material Point Method.
  • A GPU pipeline reported to reach up to 260 FPS with more than two thousand simulated strands on an RTX 3080.
Context
Builds on continuum/MPM hair simulation (cf. McAdams et al., 2009 on detail-preserving continuum simulation of straight hair) combined with Discrete Elastic Rods, targeting games, avatar streaming and metaverse use.Builds on: Detail-Preserving Continuum Simulation of Straight Hair
Correctness
Performance is demonstrated on specific hardware (RTX 3080) at a given strand count; the volume-preservation framing for collisions is an approximation, and behavior outside the tested regime (denser hair, other GPUs) is not claimed.
Clarity
Moderately technical; a first pass conveys the MPM-plus-DER split, a second pass is needed for the coupling and the semi-implicit integration.
How to read it
Read for the architecture (which solver handles strands vs collisions) first; do a second pass on the GPU pipeline and time-stepping if real-time performance is your goal.

Builds on

Built upon by

Nothing yet.

Related work

Keywords

This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →