MiTra: A Drone-Based Trajectory Data for an All-Traffic-State Inclusive Freeway with Ramps

This dataset contains original videos, tracking logs to validate the videos, and extracted trajectory data of naturalistic traffic collected using unmanned aerial vehicles (drones) over a 900 m section of the A50 urban freeway in Milan, Italy. Nine flight campaigns, totaling 135 min, were conducted using six drones flying in a line to capture comprehensive coverage across all traffic states, from free flow to congested conditions.
The dataset offers detailed trajectory data extracted from single drone videos (54 datasets from nine flight campaigns of six drones) and 9 datasets of stitched footage from all six drones. With a granularity of 30 frames per second, we extracted 124,641 vehicle trajectories from single drone videos and 24,161 trajectories from stitched footage. This enables complete vehicle tracking across five distinct categories: Cars (73.0%), Medium Vehicles (13.4%), Heavy Vehicles (11.3%), Motorcycles (2.1%), and Buses (0.2%).
Of the total vehicles, 76.9% traveled straight on the freeway, while 13% merged and 10.1% diverged using on-ramps and off-ramps, respectively, in both directions. Regarding lane changes, 51.9% of the vehicles in the dataset executed at least one. Among these, 24.8% performed a single lane change, while 27.1% changed lanes multiple times.
In addition to trajectory data, this dataset includes the original videos and tracking files, which show the recorded traffic scenes, provide visual context, and enhance the usability and interpretability of the trajectory data. The tracking files can be used to map vehicle IDs in the video, enabling various analyses as detailed in the user guide.
This dataset facilitates the analysis of driving behavior, traffic dynamics, and vehicle interactions, offering valuable insights for research, planning, and policymaking in transportation and urban mobility.