Datasets

Traffic Datasets

This page provides access to traffic datasets collected and developed as part of my research in traffic flow theory, heterogeneous traffic modeling, and intelligent transportation systems. These datasets are intended to support research in traffic operations, machine learning, computer vision, and transportation engineering.

Chennai Traffic Dataset

The Chennai Traffic Dataset is a large-scale drone-based trajectory dataset capturing heterogeneous urban traffic conditions. The dataset includes high-resolution vehicle trajectories extracted from aerial videos and covers multiple traffic environments.

Dataset Highlights

• Approximately 59 hours of drone recordings
• Vehicle trajectories with spatial coordinates
• Mixed traffic conditions
• Suitable for traffic flow modeling and machine learning applications

View Dataset Details

Applications

• Traffic simulation calibration
• Roundabout capacity analysis
• Vehicle trajectory prediction
• Behavioral modeling

Dataset Download

Vehicle-Specific Dataset

The Vehicle-Specific Dataset contains trajectory data for different vehicle classes operating under heterogeneous traffic conditions. The dataset enables the analysis of vehicle movement characteristics, lane usage, lateral position preferences, overtaking behavior, and traffic interactions in lane-free traffic streams.

The figures below illustrate the lateral position preference distributions for different vehicle categories. Darker regions indicate a higher probability of vehicle presence at a given lateral position across the roadway width.

Motorized Two-Wheelers (MTW)

Auto-Rickshaws

Cars

Commercial Vehicles

Key Observations

• Motorized two-wheelers predominantly utilize the left-most portion of the carriageway and exhibit high lateral flexibility.
• Auto-rickshaws show a moderate concentration near the left side while occupying a wider lateral range than commercial vehicles.
• Cars primarily operate within the central region of the roadway and display greater lateral dispersion under heterogeneous traffic conditions.
• Commercial vehicles exhibit strong lane-like behavior with concentrated lateral occupancy due to their larger dimensions and limited maneuverability.

Vehicle Categories

• Motorcycles
• Cars
• Auto-rickshaws
• Buses
• Trucks

Data Collection Methodology

Traffic videos were collected using aerial drone platforms. Vehicle trajectories were extracted through computer vision techniques and post-processed to obtain high-quality trajectory data suitable for traffic flow analysis and model development.

Citation

If you use any dataset from this repository, please cite the corresponding publication and acknowledge the dataset source.