VTrackIt

A SYNTHETIC SELF-DRIVING DATASET WITH INTELLIGENT INFRASTRUCTURE & POOLED VEHICLE INFORMATION FOR THE DEVELOPMENT OF SAFE & ROBUST AUTONOMOUS VEHICLES

Abstract: VTrackIt is the first comprehensive synthetic driving dataset that encourages integrating intelligent infrastructure for safer and more robust Autonomous Vehicles (AVs). Many AVs experience risky scenarios due to a wrong understanding of the surrounding environment. Vehicle-to-Infrastructure data transfer will support reducing hazardous edge cases. For example, learning that an AV is on a left-turn or exit lane helps regulate its actions to achieve safer lane change, trajectory projections, etc. Our experiments show that integrating surrounding vehicles and infrastructure information when training AV modules significantly reduces the number of hazardous edge cases.

Infrastructure Data

Lane Marking

Lane Width

Traffic Signs

Intersection Data

Traffic Regulations

Others

Ego Vehicle Data

Physical Attributes

Global Location

Vehicle Kinematics

Birds-Eye-View

360 Perception Coverage

Others

Neighboring Vehicles Data

Physical Attributes

Global Location

Vehicle Kinematics

Lane Off-Center

Relative Location to Ego

Others

DirectSim v1.0

User-friendly Platform

Automated Scenario Generator

Manual Generator

20+ Scenario Customizations

ROS Communications

Others

Terms of Use

We released VTrackIt to support advancements in Artificial Intelligence (AI) and Machine Learning (ML) for computer vision and perception tasks including motion trajectory forecasting. VTrackIt and any associated documentation are available under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 License (“CC BY-NC-SA 4.0”). The full text of the license is accessible at Creative Commons.

Copyright & Non-commercial Use

We license vTrackIt to you for non-commercial use only. Under the terms of the CC BY-NC-SA 4.0. The term “non-commercial” means “not primarily intended for or directed towards commercial use, advantage or monetary compensation.”

Citation

VTrackIt was released on July 2022 and it is available on arXiv as "VTrackIt: A Synthetic Self-Driving Dataset with Infrastructure and Pooled Vehicle Information". When using the VTrackIt dataset, please use the following citation:

@misc{2207.11146,
Author = {Mayuresh Savargaonkar and Abdallah Chehade},
Title = {VTrackIt: A Synthetic Self-Driving Dataset with Infrastructure and Pooled Vehicle Information},
Year = {2022},
Eprint = {arXiv:2207.11146},
}

Team

Abdallah A. Chehade

Associate Professor

Mayuresh Savargaonkar

Ph.D. Student

VTrackIt Dataset

VTrackIt

A SYNTHETIC SELF-DRIVING DATASET WITH INTELLIGENT INFRASTRUCTURE & POOLED VEHICLE INFORMATION FOR THE DEVELOPMENT OF SAFE & ROBUST AUTONOMOUS VEHICLES

Infrastructure Data

Ego Vehicle Data

Neighboring Vehicles Data

DirectSim v1.0

Terms of Use

Terms of Use

Copyright & Non-commercial Use

Citation

Team

Abdallah A. Chehade

Mayuresh Savargaonkar

Sample Scenario

Download

Training Dataset

Validation Dataset

Testing Dataset

Get in touch

University of Michigan-Dearborn

Department of Industrial and Manufacturing Systems Engineering