Each clip is a self-contained 20 s multimodal record centered on an automation disengagement: front-view video plus 13 synchronized CAN / perception / IMU signal files.

Why this dataset

Most takeover research comes from simulators or small scripted road experiments. ADAS-TO instead captures how drivers actually resume control during everyday production-ADAS use — across heterogeneous traffic, roadway, weather and vehicle platforms — with the road scene, automation state, driver inputs, vehicle dynamics and post-intervention safety margins aligned on the same control-transfer event.

Clip structure

t = 0 is the reconstructed ADAS disengagement and sits at the midpoint of every clip:

|<--------- 10 s ADAS engaged --------->|<--------- 10 s manual driving --------->|
                                     t = 0

Directory layout

ADAS-TO/
├── <CAR_MODEL>/                    e.g. TOYOTA_PRIUS, TESLA_AP3_MODEL_3
│   └── <driver_NNN>/               🔒 anonymized driver ID
│       └── <route_MMM>/            🔒 anonymized route ID
│           └── <clip_id>/          integer, 0-indexed within the route
│               ├── takeover.mp4          20 s front-view video (20 fps)
│               ├── meta.json             metadata & timing
│               ├── carState.csv          speed, accel, steering, pedals, cruise state
│               ├── carControl.csv        lateral/longitudinal commands
│               ├── carOutput.csv         actuator outputs
│               ├── controlsState.csv     ADAS controller state & alerts
│               ├── drivingModelData.csv  lane-line estimates, desired curvature
│               ├── longitudinalPlan.csv  planner targets, FCW
│               ├── radarState.csv        lead-vehicle radar tracks
│               ├── accelerometer.csv     IMU acceleration
│               ├── VehicleIMU.csv        body-frame IMU / yaw rate
│               ├── Gyroscope.csv         angular rates
│               ├── CameraOdometry.csv    visual odometry
│               ├── LiveCalibration.csv   device → vehicle frame calibration
│               └── LiveParameters.csv    online vehicle-parameter estimates
└── annotations/

15 files per clip (1 video + 1 metadata + 13 CSVs). Vehicle logs are sampled at up to 100 Hz where an rlog was available, otherwise at the qlog rate; video is 20 fps.

Annotations

File	Rows	Contents
`annotations/clip_final_labels.csv`	16,446	scenario label per clip + `lc_direction`, `human_labeled`
`annotations/urgency_stability.parquet`	4,971	take-over urgency `U_main`, post-hand-back stability `S`, settling time
`annotations/selection_metrics.parquet`	6,223	nine safety-selection indicator values per eligible clip
`annotations/selection_flags.parquet`	6,223	boolean flag per indicator
`annotations/safety_critical_749.parquet`	749	manually reviewed safety-critical case set + context labels
`annotations/safety_critical_labels.jsonl`	749	`{clip_path, labels}` multi-label context taxonomy
`annotations/selection_thresholds.csv`	54	every candidate threshold considered, with the chosen one flagged

Scenario partition (`clip_final_labels.csv`)

Every clip was screened with synchronized signals and then manually reviewed (11,697 clips received human review; all cover clips were human-confirmed):

Label	n	Meaning
`cover`	6,346	Maneuver-Filtered ADAS — ordinary lane-keeping / car-following, no maneuver confound
`turn`	4,191	intersection turn or departure from the through path
`stop`	3,082	stopping / decelerating for a traffic control
`lane_change`	2,573	lane transition, merge or fork (`lc_direction` gives left/right)
`Unknown`	254	complex geometry, non-routine scene, or insufficient evidence

Most analyses should start from cover, which removes disengagements that are explained by a planned maneuver rather than by the automation reaching its limits.

Related datasets

Repo	Contents
ADAS-TO-Sample	small sampled subset for quick inspection
ADAS-TO-Critical	491 "difficult hand-backs": high intervention demand and unstable post-hand-back motion

Privacy & licensing

Driver and route identifiers are anonymized (driver_NNN / route_MMM); the mapping to the original device IDs is not released.
Clip signal files contain no GPS coordinates.
Video is raw forward-facing road scene and may contain other road users; use accordingly.
Released under CC BY-NC 4.0 for non-commercial research. Access is granted manually.

Please cite the ADAS-TO paper when using this dataset.

Downloads last month: 35

Total file size:

43.6 GB

You need to agree to share your contact information to access this dataset

🚗 ADAS-TO