FEDORA has published Deliverable D5.2 – Modular Simulation Platform and Sim2Real Techniques, bringing together new tools designed to make the project’s simulation environment more flexible, reusable and better able to support the transition from virtual experiments to real-world applications.
A central achievement is a modular simulation platform that acts as a common orchestration layer between different simulation environments and the logic modules developed within FEDORA, including control, optimisation and behavioural models. Instead of building separate integrations for each module and simulator, the platform enables them to communicate through a shared interface, making it easier to connect the same logic to different simulation environments or, in the future, real pilot sites. A demonstrator for traffic signal control illustrates this approach, showing how different control strategies can be tested within the same framework and evaluated through a common pipeline.
The deliverable also explores Sim2Real techniques—methods for transferring solutions developed in one simulation environment to another, and ultimately towards real-world applications. In a proof-of-concept experiment, an AI-based traffic control strategy was trained in one simulator and tested in another. The results showed that simply transferring the trained strategy is not enough: using the proposed Grounded Action Transformation method, developed and evaluated within FEDORA improved performance in almost all of the tested configurations.
A third strand of work focuses on foresight and scenario testing. The new platform can create reproducible scenarios combining traffic demand, traveller behaviour, network conditions, control policies and even emerging services such as UAV-based passenger transport. This allows different mobility scenarios to be compared and evaluated using common performance indicators. The approach is currently being tested in the Aglantzia network in Cyprus.
Together, the three components demonstrate how FEDORA’s Simulation Space can bring different models, technologies and experiments together through common interfaces. The work provides an important step towards a flexible simulation environment where new mobility solutions can be tested, compared and refined before moving closer to real-world deployment.