[Demo] Real2Sim - Sim2Real

We haven been testing and scaling to different traffic environments in research and commercial contexts.

You may wondering (as we have been), beyond nice visuals, what can this do for ADAS system development. Here we show three engineering use case examples developing by our colleagues:

• 3D labelling: Unlike 2D frame-by-frame annotation, this enables working directly in 3D, just imagine you can zoom in, measure, inspect - just easier than 2D.

• Closed-loop virtual testing inside logged data: The DriveTwin shadow testing platform enables running a complete virtual digital twin vehicle - such as high fidelity Simcenter Amesim model - in parallel with real world driving. The 3D model generates novel view traffic for closed-loop testing. It benefits analysis of algorithms and vehicle performance metrics such as autonomous driving safety & comfort, battery efficiency, vehicle suspension dynamics…

• Simulation integration: Combining with traffic scenario simulation like Simcenter Prescan, it could overcome the limitation of data distribution (typical in data-driven AI method) by generating diverse traffic scenarios, including edge and critical scenarios.

Always interesting to see new technologies in AI and Generative AI continue pushing the (already) advanced autonomous driving engineering.

With multiple Siemens engineering Digital Twin x AI colleagues: João Malheiro Silva, Jean Pierre Allamaa, Ajinkya Bhave, Piero Brigida, Sarah Barendswaard, Theo Geluk…