Unified GPU-Based Simulation of Porous Flow, Absorption, and Diffusion in Cloth–Liquid Coupling

Cloth–liquid interaction involves complex physical phenomena such as porous flow, absorption, emission, and diffusion, which are difficult to model in a unified and stable manner, especially in real-time or high-resolution simulations. Existing approaches often address these effects in isolation or rely on unstable pressure estimates near boundaries, leading to numerical instability and limited scalability.In this paper, we propose a unified GPU-based framework for stable and efficient simulation of porous cloth–liquid interactions using Smoothed Particle Hydrodynamics (SPH). The proposed method reformulates porous flow as a directional process inspired by Darcy’s law, preserving physical intuition while avoiding unstable SPH pressure magnitudes through a virtual-pressure formulation. Absorption and emission are governed by a saturation-centered mass management scheme, and diffusion is handled as a sequentially decoupled stage, together ensuring mass conservation, directional consistency, and numerical robustness. To achieve high performance, all stages of the cloth–fluid coupling pipeline are executed entirely on the GPU, supported by a Bitonic sort–based hashing structure for efficient neighbor search.We evaluate the proposed framework through both qualitative and quantitative experiments, including frame-time analysis, scalability with increasing particle counts, mass conservation error measurement, and module-wise performance breakdown. The results demonstrate that the method achieves stable wetting behavior, reduced non-physical artifacts, and significant GPU-based speed-ups compared with baseline implementations. Overall, the proposed approach provides a stable, interpretable, and high-performance solution for large-scale wet-cloth simulation.