Date of Award

Spring 6-1-2026

Document Type

Thesis (Undergraduate)

Department

Computer Science

First Advisor

James Mahoney

Second Advisor

Lorie Loeb

Abstract

Presented in this paper is a GPU-native approach to interactive fluid simulation within Unreal Engine 5. The system, BioFluidSim, implements an incompressible Navier-Stokes solver using Unreal’s Niagara Grid2D compute shader pipeline, with a modular biological emission output stage parameterized from experimentally measured Lingulodinium polyedrum bioluminescence behavior. The system is evaluated against FluidNinja Live, a commercially available fragment shader fluid implementation, as a performance baseline. Beyond performance, BioFluidSim offers greater physical fidelity than the fragment shader baseline. Helmholtz–Hodge pressure projection enforces a divergence-free velocity field at runtime, a physical constraint approximated but not enforced by fragment shader approaches. The biological emission stage is parameterized directly from experimentally validated Lingulodinium polyedrum literature, grounding flash intensity, decay rate, and emission threshold in measured dinoflagellate response rather than artistic approximation. Together these properties distinguish BioFluidSim from visually motivated fluid implementations that prioritize appearance over physical correctness. Performance profiling via Unreal Insights and stat unit reveals a consistent architectural advantage on the CPU side. BioFluidSim reduces game thread cost to 1.98 ms in the water case study and 1.87 ms in the snow case study, compared to 2.77 ms and 2.45 ms respectively for the commercial baseline. The isolated Niagara solver cost totals 0.372 ms per frame across 15 compute stages, with the Pressure Jacobi pass accounting for the majority of that budget. GPU steadystate cost is competitive across both case studies, with BioFluidSim achieving a lower GPU cost than the commercial baseline in the snow configuration. The solver is integrated with a Blueprint detection and tracking architecture evaluated across water and snow environments, projecting 2D simulation output as surface effects via parallax occlusion mapping. A snow case study demonstrates solver generality: transitioning from bioluminescent water to granular snow deformation requires only runtime scalar adjustments on the Manager actor, with no modifications to the underlying architecture. Results confirm real-time viability at 60 fps on the test hardware, and support the case for Niagara Grid2D compute shaders as a performant, physically grounded, and extensible foundation for interactive fluid simulation in game engines.

Download

Share

COinS