AI and the World's Hard Problems · Theoretical computer science · arXiv · 2026-10-21
CorteXforge: Scenario-Based RF Dataset Generation on Real Testbeds
Deep learning for RF signal analysis relies on large annotated datasets, but synthetic datasets trade physical realism for controllability, while uncontrolled over-the-air measurements offer realism with limited reproducibility. CorteXforge is a scenario-based framework that automatically translates high-level RF scenarios into synchronized experiments on programmable SDR testbeds via the SLICES-RI/CorteXlab platform, producing fully annotated SIGMF datasets that capture hardware impairments and propagation effects. Validation through automatic modulation classification using ModForge, an OTA dataset generated with the framework, revealed a synthetic-to-real performance gap and showed that synthetic pretraining remains beneficial for adapting to OTA data.