Automated Vehicle Identification Based on Car-Following Data with Machine Learning

Advantages:

  • Identifies autonomous vehicles in real time using very short data windows.
  • Strengthens traffic safety, capacity planning, and infrastructure decisions through accurate analysis.
  • Classifies vehicles using behavioral data instead of visual appearance, revealing specific manufacturers.
  • Deploys flexibly across vehicles, roadside units, or servers using standard sensors.

Business Summary:

Autonomous vehicles now share roads with human drivers, creating unpredictable mixed traffic that regulators struggle to monitor. Since AVs look identical to standard cars, authorities cannot manually track their behavior, leaving traffic planners without the data needed to manage flow, boost capacity, and protect public safety. Without reliable insight into AV behavior, infrastructure strategies and safety evaluations remain critically underinformed.

This system uses standard sensors like cameras and GPS to capture vehicle motion data, then applies machine learning to analyze subtle driving patterns such as speed, spacing, and trajectory. Within just 0.2 to 5 seconds, it classifies vehicles as autonomous or human driven and can even identify the manufacturer. Unlike simulation based or small-scale field approaches, this method works directly in real world mixed traffic, delivering near real time results without specialized infrastructure, making it a scalable and practical tool for traffic management.

 

(a)

 

(b)

 

 Mixed traffic experiment locations. (a) Low-speed experiment. (b) High-speed experiment

 

Desired Partnerships: 

  • License
  • Sponsored Research
  • Co-Development

 

Patent Information:
Title App Type Country Serial No. Patent No. File Date Issued Date Expire Date
Automated Vehicle Identification Based On Car-Following Data With Machine Learning Utility United States 17/884,411 12,639,955 8/9/2022 5/26/2026 8/19/2043