AnisoLift: Anisotropic Latent Representations for Coarse Particle Liquid Enhancement

2026-06-09

AI Summary

High-resolution particle-based fluid simulation is computationally expensive, and existing coarse-grained enhancement techniques rely on generating additional particles, leading to inefficiency and limited representational capacity. This work proposes AnisoLift, a novel framework that introduces learnable anisotropic ellipsoidal latent variables into each coarse-grained particle, enabling the capture of directional local structures in high-resolution flow fields without introducing extra particles. By jointly optimizing particle dynamics and geometric structure—augmented with residual state correction and a physics-geometry hybrid loss—the method significantly enhances simulation fidelity while maintaining the original particle count. Experimental results demonstrate that AnisoLift outperforms current approaches in both detail preservation and physical plausibility.

Abstract

Particle-based liquid simulation is widely used in graphics and physical modeling, but high-resolution rollouts remain computationally expensive. Consequently, many methods aim to recover fine-scale dynamics and dense transport patterns from coarse particle simulations. However, these methods typically rely on additional particle generation, which still incurs considerable computational overhead and leads to poor representation. To this end, we propose AnisoLift, a structured latent closure framework that augments each coarse particle with learnable anisotropic ellipsoidal components. This allows the model to capture directional local structure from the underlying high-resolution flow without introducing extra particles. Given a coarse simulation, our model predicts residual corrections to particle states to bring the updated state closer to the aligned high-resolution teacher. Our training objective jointly supervises particle dynamics and anisotropic geometric structure, encouraging both physical consistency and structural coherence. Extensive experiments show that our approach enhances coarse liquid simulations through improving fidelity to fully resolved flow behavior.

Problem

• particle-based liquid simulation
• coarse-to-fine enhancement
• anisotropic representation
• latent closure
• high-resolution flow

Innovation

• anisotropic latent representation
• particle-based liquid simulation
• structured latent closure
• coarse-to-fine enhancement
• ellipsoidal components