Optimizing Real-Time Multi-layer Garment Simulation in UE Chaos Cloth

With the rapid growth of virtual idols and the metaverse fashion industry, the demand for scalable, real-time simulation of multi-layer garments has increased substantially.However, the default Chaos Cloth workflow in Unreal Engine frequently encounters instability and garment interpenetration, particularly when handling complex USD-based assets.This study proposes an optimized Chaos Cloth workflow to enhance the multi-layer garment simulation's fidelity and stability of multi-layer garment simulations.The key improvements include restructuring data flow nodes, integrating kinematic collision bodies, and utilizing GPT-4o to assist in generating weight-replicating Marvelous Designer fabric parameters.Additional enhancements, such as mesh simplification, UVbased selection, and hierarchical collision handling, improve simulation performance and reduce manual workload when using three garments: a wool-blend suit, a silk Hanbok, and a five-layer medieval costume.Controlled experiments were conducted under combat and walking animations.The results demonstrated consistent frame rate gains ranging from 7 to 9 FPS and reduced simulation volatility on an RTX 4060 GPU.This methodology enhances visual realism and workflow efficiency, supporting the scalable deployment of assets across various platforms.While further automation and user validation are still needed, the proposed solution offers a practical and reproducible pipeline for VTuber production, virtual concerts, and broader real-time digital fashion applications.