Digital twins using NVIDIA Omniverse and OpenUSD: A proof-of-concept implementation

Abstract

Operational digital twins of buildings require a common environment in which Building Information Modeling (BIM)-based spatial models, live telemetry from building automation systems (BAS), and institutional records of equipment and spaces are integrated to support monitoring, analysis, and decision-making at facility scale. Existing platforms and research efforts still leave open challenges in integrating heterogeneous source systems, representing building semantics within the integrated environment, and supporting analytical workflows over the composed information at runtime. This paper investigates NVIDIA Omniverse and OpenUSD as a scene-native foundation for addressing these challenges through a shared representation. The paper contributes (i) a modular five-layer reference architecture for digital twin development, (ii) a prototype implementation realized as a custom Omniverse Kit application integrating BIM-derived context, BAS telemetry, and institutional facility records in an academic building, and (iii) an analysis of how native OpenUSD constructs can encode operational building context directly within the scene. The case study demonstrates that the composed scene, continuously updated with live operational data, can support historical trend review and system-aware workflows grounded in current building context. The results show that OpenUSD can host operationally relevant building semantics alongside spatial and operational content within a single scene, positioning Omniverse as an extensible foundation for the analytical, AI-assisted, and lifecycle-spanning workflows expected of next-generation facility operation.