RAVENS Mobility Dataset for MEC Orchestration in Vehicular Scenarios — Turin (TuST)

This dataset supports research on proactive service migration in Multi-access Edge Computing (MEC) environments for vehicular scenarios. It was generated using a simulation setup built on OMNeT++ (6.1), Simu5G (1.2.2), SUMO (1.22), and Veins (5.3.1), using a subsection of the Turin SUMO Traffic (TuST) scenario covering approximately 1.2 × 1.2 km with 10 base stations. The dataset comprises 45 simulation runs, each corresponding to a 3,600 s window during the 7–8 AM peak hour, with approximately 1,300 vehicles per run. Mobility and radio network data were collected at 3 s intervals through RAVENS, a data collection system interfacing with the ETSI MEC Location Service and Radio Network Information Service at each MEC Host. The repository contains: Raw data: Per-run CSV files as exported by the RAVENS Data Collection System in Data Saving mode, containing per-UE location, speed, bearing, serving base station, and Layer 2 uplink/downlink delay measurements. Each run uses a different random seed to ensure variability in vehicle density and mobility patterns, generated using SUMO scripting tools with manually adjusted density scaling factors following a normal distribution (mean = 0.3). Processed data: Feature-engineered CSV files ready for model training, including 5 lagged time steps of position, speed, bearing, and serving cell, Euclidean distances to all 10 base stations, one-hot encoded categorical features (96 features per timestep), and multi-horizon target labels (t+3 s to t+15 s) with an exit class for UEs departing the scenario. The 45 runs were split into 31 training, 7 validation, and 7 test runs.The mobility traces are derived from the Turin SUMO Traffic (TuST) scenario introduced in: M. Rapelli, C. Casetti, and G. Gagliardi, 'Vehicular Traffic Simulation in the City of Turin from Raw Data,' IEEE Trans. Mobile Comput., vol. 21, no. 11, pp. 4189–4204, 2022, doi: 10.1109/TMC.2021.3075985.