2026 · Nokia Bell Labs · Cambridge
Agentic 3D Modelling for Physically-Grounded 6G Simulation
An end-to-end agentic pipeline where LLM agents use Blender and MCP to generate physically grounded, city-scale 3D twins for high-fidelity ray-tracing and neural physical-layer research.
Problem. Training neural physical-layer models for future 6G networks needs city-scale geometry that is both detailed and physically credible for ray-tracing — expensive to author by hand.
System. An agentic pipeline built on Blender’s Python API and the Model Context Protocol, enabling LLM agents to generate and refine digital twins grounded for simulation use.
Outcome. End-to-end deployment of agents that produce twins suitable for high-fidelity ray-tracing workflows underpinning neural PHY research.