AV/ADAS Incident Heatmap

11,900+

560,839 incidents · 3,521 locations

Recording Sites

Downloads

Map .osm · 4.2MB Trajectory .csv · 118MB Paired Bundle .zip · 122MB

Ozone: A Unified Platform for Transportation Research

Bridging multi-sensor data to 1:1 digital twin mapping for autonomous driving research.

Explore on GitHub

Apply for Access

10+ Datasets

1:1 Real-world
Mapping

0+ Applicants

0+ Page Views

Scroll to Explore

Why Ozone

What We Solve

Autonomous driving research is held back by fragmented, incompatible data sources. Traditional approaches fail to bridge the gap between the physical and digital worlds.

ISSUE 01

Heterogeneous Data Slices

Drone footage, RSU feeds, vehicle trajectories, and simulation outputs exist in incompatible formats with no unified schema.

ISSUE 02

No Real-to-Digital Mapping

Existing benchmarks lack precise 1:1 correspondence between real-world locations and their digital twin counterparts, limiting research validity.

ISSUE 03

Fragmented Toolchains

Researchers juggle disconnected tools for simulation, sensor modeling, and scenario testing — wasting months on integration instead of innovation.

The Solution

One cohesive digital twin
ontology system.

Ozone maps real-world traffic to high-fidelity digital twins, unifying multi-sensor data into a single research framework.

System Architecture

From heterogeneous data sources to digital twin basemaps, CARLA co-simulation, and platform capabilities.

TRAJECTORY

High-resolution Vehicle
Trajectory Datasets

High-resolution sensor recordings and vehicle trajectories from real-world driving scenarios, powering the latest of AD research.

Aerial Trajectory Extraction

Bird's-eye view captures of real traffic flow, enabling precise trajectory extraction and scene understanding at scale.

Autonomous Driving Data

Large-scale, multi-modal datasets that capture the full complexity of real-world driving — from high-resolution LiDAR point clouds and camera imagery to lane-level vehicle trajectories and object interactions. Designed to accelerate research in perception, prediction, and motion planning for autonomous vehicles.

Basemaps

Specialized Digital Twin Basemaps

Purpose-built basemaps for autonomous vehicle testing, sensor simulation, and research environments. More maps are continuously added.

Foundation

Foundation · Module 01

Digital Twin Base Map

High-fidelity 1:1 mapping from real-world to simulation. Capture road networks, terrain, and infrastructure with centimeter-level accuracy for precise AV testing and validation.

HD MapsLiDARRTK1:1 Scale

Simulation · Module 02

SUMO-CARLA Co-Simulation

Seamless integration between SUMO microscopic traffic and CARLA driving simulators. Synchronize vehicle dynamics, sensor data, and traffic flow in real time for mixed-reality testing.

SUMOCARLAReal-time SyncMulti-agent

Simulation

Safety

Safety · Module 03

Corner Cases Testing

Systematic generation and replay of edge-case scenarios — sudden cut-ins, pedestrian occlusions, adverse weather — to ensure autonomous systems handle rare and critical situations safely.

Edge CasesScenario GenSafety Metrics

Behavioral Science · Module 04

Human Factor Study

A complete driving simulator platform for studying human-AV interaction. Measure response times, trust calibration, and situational awareness with HUD overlays and eye-tracking integration.

HUDEye TrackingDriver BehaviorSimulator

Behavioral Science

Perception

Perception · Module 05

Sensor Simulation

Physics-accurate simulation of LiDAR, camera, and radar sensors within the digital twin. Test perception algorithms under controlled noise, occlusion, and multi-modal sensor fusion conditions.

LiDARCameraRadarSensor Fusion

Full-stack AV · Module 06

Autonomous Vehicle Simulation

End-to-end AV algorithm development and benchmarking. Deploy planning, control, and prediction modules within the Ozone digital twin for reproducible, standardized evaluation.

PlanningControlPredictionBenchmarking

Full-stack AV

Research Team

The Minds Behind Ozone

An international team of researchers from MIT, Tongji, Southeast University, and beyond — united by a shared vision for intelligent transportation.

Project Team

Ou Zheng Ph.D.

Zhiling Research / Peking University

Ruyi Feng Ph.D.

Southeast University / The Hong Kong Polytechnic University

Yufeng Yang

Zhiling Research

Shengxuan Ding Ph.D.

Zhiling Research

Lishengsa Yue Ph.D.

Tongji University

Ye Li Ph.D.

Central South University

Yunhan Zheng Ph.D.

Peking University

Minwei Kong

Massachusetts Institute of Technology

Zhibin Li Ph.D.

Southeast University

Ao Qu Ph.D.

Massachusetts Institute of Technology

Dingyi Zhuang Ph.D.

Massachusetts Institute of Technology

Meng Li Ph.D.

Korea Advanced Institute of Science and Technology

Dongjie Wang Ph.D.

University of Kansas

Wangyang Ying Ph.D.

Arizona State University

Data Contributors

Ziyun Jiao

Sichuan University

Guo Xiang

Zhiling Research

Request Access

Apply for Access

For researchers, students, and industry practitioners working on transportation and autonomous driving research.

Personal Information

Full Name *

Email Address *

Please enter a valid email

Institution *

Country *

Role *

Student Faculty Industry

Research Purpose *

0 / 200

I agree to the Terms of Use. Data will be used for non-commercial research only, not redistributed, and I will cite Ozone in any publications. For commercial licensing, please contact us.

Please complete all required fields

Citation

Our paper introducing the platform and the unified framework is available on arXiv.org here. To reference the platform, please use:

@article{zheng2026ozone,
        title={Ozone: A Unified Platform for Transportation Research},
        author={Zheng, Ou and Feng, Ruyi and Yang, Yufeng and Ding, Shengxuan and Yue, Lishengsa and Li, Ye and Zheng, Yunhan and Kong, Minwei and Zhuang, Dingyi and Qu, Ao and Li, Zhibin and Wang, Dongjie and Ying, Wangyang},
        journal={arXiv preprint arXiv:2604.10959},
        year={2026},
        doi={10.48550/arXiv.2604.10959}
}