A Reproducible UAV-Assisted VANET Dataset Generator for Fragmentation Risk Analysis in Intelligent Transportation Systems
For researchers working on connectivity management in intelligent transportation systems, this provides a modular and reproducible dataset generation framework to support fragmentation risk analysis, though it is an incremental contribution.
This work proposes a reproducible ns-3-based dataset generator for short-term fragmentation risk prediction in UAV-assisted VANETs, simulating highway scenarios with various traffic conditions and providing labeled data for supervised learning. The generated dataset is characterized in terms of balance and feature ranges, offering a practical basis for future studies.
Vehicular Ad Hoc Networks (VANETs) are a key component of Intelligent Transportation Systems, enabling cooperative communication among vehicles and between vehicles and roadside infrastructure. However, their highly dynamic topology makes them vulnerable to network fragmentation, particularly in highway scenarios, low-density traffic conditions, localized accident zones, and communication-stressed environments. Although Unmanned Aerial Vehicles (UAVs) have been increasingly investigated as temporary aerial relays for improving VANET connectivity, reusable, future-labeled, and reproducible datasets designed to support short-term fragmentation risk analysis remain limited. This paper proposes a reproducible UAV-assisted VANET dataset generator for short-term fragmentation risk prediction. The proposed framework simulates a two-lane highway scenario in which vehicles move in opposite directions while UAVs operate as aerial support nodes. It incorporates multiple data collection profiles, including free-flow traffic, localized accidents, sparse extended topologies, dense bursty traffic, and mixed stress conditions. During each simulation episode, the generator periodically extracts mobility, topology, UAV coverage, and communication-window features, then assigns each sample a future fragmentation label based on the network state observed after a configurable prediction horizon. An illustrative generated dataset is descriptively characterized in terms of scenario balance, UAV policy balance, future-label distribution, scenario-specific label behavior, and representative feature ranges. By providing a modular, extensible, and reproducible ns-3-based data-generation framework, this work offers a practical basis for future supervised learning studies and connectivity management strategies in UAV-assisted VANETs.