Scaling physics-to-protocol digital twins for 5G/6G networks

Digital Twins (DTs) are central to the 6G vision, serving as the definitive proving ground for pervasive network intelligence. The ultimate value of a DT hinges on closing the sim to-real gap, which requires faithfully modeling dynamic, dense multi-gNB (next-generation Node B)/multi-UE (User Equipment) behaviors. However, current platforms can emulate high-fidelity but isolated slices of the protocol stack; scaling this fidelity horizontally across dense environments encounters prohibitive computational bottlenecks. This creates a binary trade-off: scale broadly using abstracted simulators that sacrifice protocol realism, or achieve genuine protocol execution within the narrow bounds of costly hardware testbeds. This paper introduces a Physics-Based Digital Twin (PBDT) architecture that breaks this deadlock. We deploy a centralized proxy that interconnects fully 3GPP-compliant OpenAirInterface (OAI) gNB and UE stack instances, replacing per-node baseband processing with slot-level message routing. The proxy conditions each gNB UE link with channel metrics (SINR, RSS, RSRP) derived from site-specific ray tracing, so every node perceives a radio environment shaped by 3-D urban geometry, enabling direct MAC-to-MAC communication at deployment scale on commod ity infrastructure. Experimental validation confirms the physics fidelity of the ray-traced channel conditioning along a vehicular mobility trace and demonstrates a 3.4× per-slot cost reduction over baseband-coupled baselines, extending the feasible real-time operating range by 3× on commodity hardware. The platform provides a native environment for training, validating, and stress testing AI-driven RAN optimization against real 3GPP protocol behavior, closing the loop between algorithm development and deployment-grade network dynamics.