A Federated Learning Dataset for Driving Behavior Prediction in a V2X-Satellite Integrated Network

This dataset offers a comprehensive, multi-scale simulation environment for research in Federated Learning (FL), V2X communications, and autonomous driving. It bundles four distinct scenarios with systematically varying vehicle densities: K=100, 200, 300, and 400. For each scale, the dataset provides per-vehicle private data files for FL, a global validation set for model evaluation, and second-by-second snapshots of traffic and communication (V2R\/S2R) states. This collection is specifically designed for researchers to evaluate the scalability and performance of algorithms in dynamic vehicular networks.