Project · Active

DriveDNA: a naturalistic driving-style dataset and benchmark

PI

What it is

DriveDNA treats driving style as a consistent, driver-specific behavioral pattern in how a vehicle moves under similar conditions — and builds the dataset and benchmark needed to measure it. It spans over 4,100 drives from 465 drivers across 115 vehicle models, with nearly 1,000 hours of recorded driving, and supports tasks including driver re-identification and behavior prediction.

Why it matters

Driving-style research has been held back by confounds: without controlling for vehicle and environment, it is difficult to tell whether an observed behavioral difference reflects the driver or the conditions they were driving in. DriveDNA’s scale across drivers, vehicles, and drives is what makes it possible to isolate driver-specific style from those confounds instead of assuming it away.

Connections

DriveDNA extends the lab’s driver-behavior modeling work alongside TriDrive and DriveMotion, and draws on the same production-vehicle instrumentation lineage as OpenLKA. It sits in the lab’s AI for mobility and Traffic flow theory threads.

Project team

  • Hao Zhou (PI)
  • Yuhang Wang
  • Lingyao Li

Outputs & releases