Fair-VARO SUMO Simulation Data and Experimental Results

Abstract

This dataset accompanies the manuscript “Fair-VARO: Balancing Reliability and Participation Fairness in Vehicular Federated Learning.” The study investigates the trade-off between update completion reliability and successful-participation fairness in vehicular federated learning using SUMO-based traffic simulations and a GRU traffic-state classifier. Fair-VARO combines vehicle-state reliability scoring with compensation based on successful aggregation histories. The experimental evaluation covers different data heterogeneity levels, component ablations, controlled state–data permutations, explicit wireless completion models, an independent simulated road network, parameter sensitivity, and controlled state-report falsification. Update failures, successful-participation fairness, aggregation-weighted label representation, and predictive performance are evaluated separately. The data are generated through simulation and do not represent measurements collected from an operational vehicle fleet. Third-party data and licensing: The primary road-network materials include OpenStreetMap data (© OpenStreetMap contributors). The source file SUMO/primary/map.osm and applicable OpenStreetMap-derived road-network data remain subject to the Open Database License (ODbL); see https://www.openstreetmap.org/copyright. The selected CC BY 4.0 license applies only to materials for which the authors hold the relevant rights and does not replace third-party licenses.