Secure Cloud‑Streaming Digital Twin (SCSDT): A Cloud‑First Metaverse Architecture for Industrial Training

Traditional extended reality (XR) training faces a fundamental trade-off: high-fidelity, in-situ VR solutions offer superior pedagogical outcomes but are limited by expensive, resource-intensive hardware, while web-based alternatives sacrifice immersion for scalability. This research addresses this gap by presenting a novel architecture that delivers immersive XR training with operational efficiency. We introduce SCSDT, a threelayer architecture that fuses cloudGPU streamed metaverse delivery, zerotrust provisioning, and adaptive digitaltwin orchestration. Demonstrated on aviationmaintenance and industrialcyberrange pilots, SCSDT cuts VM provisioning time by an order of magnitude and reduces perlearner power draw by ≈60 %. Repeated-measures ANOVA on a cohort of six technicians over five instructional iterations shows a very large competencegap reduction (η²ₚ = 0.91, p < 0.0001). Our empirical evaluation showed a provisioning speed-up of 12.8× and a 62% energy reduction per learner, with statistically robust learning gains evident across iterations. The system operates over a single HTTPS port and streams at just 200–500 kB/s, enabling widespread access even on LTE links. By inheriting the pedagogical richness of in-situ VR while matching the operational efficiency of web-based labs, SCSDT delivers a superior cost–benefit ratio for enterprise training and successfully answers our research question.