Nanite vs. Traditional LOD: Real-Time Performance and User Experience in VR for Participatory Urban Planning

Abstract

Virtual reality (VR) now becomes an emerging tool in urban planning to help people visualize designs, especially in participatory planning. However, showing high-polygons 3D models in VR can slow down computers, particularly in real-time rendering scenarios. Traditional level of details (LODs) techniques addresses this issue through manual mesh simplification and distance-based model switching, but these approaches are labor-intensive and may result in visual artifacts. Nanite, Unreal Engine 5’s virtualized geometry system, promises to overcome these limitations by dynamically streaming and rendering the only visible pixels in real time. Although Epic Games officially promotes Nanite as a superior alternative to conventional LOD workflows, several inconsistent evaluation results have been reported in the community. Several reports have already shown that Nanite incurred substantial GPU overhead and did not outperform traditional overdraw LODs in gaming and filming area, while its applicability in urban planning workflows remains underexplored. Accordingly, this study evaluates the real-time performance of Nanite in comparison to traditional LOD-based workflows, with a focus on computational efficiency and user experience. The University of Waterloo’s new M4 building and its surrounding campus serve as the case study. A 3D model of the building was developed using Autodesk Revit, processed in ArcGIS Pro, and imported into Unreal Engine to develop a VR environment. Two VR versions were created: one using a traditional LOD workflow and the other using Nanite-enabled assets, both derived from the same base model to ensure a fair comparison. User experience data and computational performance metrics were collected from 28 participants who experienced both two VR models. The results show that Nanite significantly reduces polygon counts compared to traditional LOD, confirming its advantage for dense architectural models. However, Nanite also introduces higher GPU overhead and less stable frame rate under VR conditions, challenging its direct applicability in participatory planning. User experience data indicated that participants preferred the traditional LOD approach with 64% favouring it over Nanite. This research addresses a gap in the literature by assessing Nanite’s suitability for participatory planning applications through both system performance evaluation and user experience analysis within architectural VR environments.