Data-Tesla-Traffic-Light-and-Stop-Sign-Control

Overview

This repository hosts the Tesla Traffic Light and Stop Sign Control – Vehicle (TLSSC-V) Dataset, introduced in the TRB 2026 Annual Meeting paper:

Li, Z., Zhang, P., Liang, S., Zhou, H., Ma, C., Li, Q., Yao, H., & Li, X.
Benchmarking Tesla’s Traffic Light and Stop Sign Control: Field Dataset and Behavior Insights
TRBAM-26-05481, 2025.

The dataset provides high-resolution, synchronized vehicle trajectory and driver-perspective video data of Tesla’s Traffic Light and Stop Sign Control (TLSSC) feature interacting with Traffic Control Devices (TCDs), including both traffic lights and stop signs.

It is the first publicly available dataset focusing on ADAS–TCD interactions, supporting research in traffic operations, driving behavior modeling, safety evaluation, and simulation.

---

Key Features

• Vehicle platform: Tesla Model Y with TLSSC enabled.
• Data type: GPS-based vehicle trajectories (10 Hz) and driver-perspective videos.
• Geographical coverage: Madison, Middleton, Fitchburg, and Verona, Wisconsin (USA).
• Behavior taxonomy:
  1. Stopping behaviors:
     • Stop before a red/yellow light
     • Stop before a green light (permission required)
     • Stop before a stop sign
  2. Accelerating behaviors:
     • Accelerate after permission at a green light (before/after stop)
     • Accelerate after permission at a stop sign
  3. Car-following behaviors:
     • Standard car-following (gap levels 2, 4, 7)
     • Car-following through intersections (gap levels 2, 4, 7)
• Novel finding: Empirical car-following detection threshold ≈ 90 m for TLSSC-V.

---

Data Structure

Trajectory data are stored in CSV files with the following key variables:

Column Name	Description
Time	Timestamp in ISO 8601 format
Longitude, Latitude	Raw GPS coordinates of TLSSC-V
Longitude_smoothed, Latitude_smoothed	Smoothed GPS coordinates
Speed, Speed_smoothed	Raw & smoothed vehicle speed (m/s)
Longitude_lead, Latitude_lead	Raw GPS coordinates of lead vehicle (if present)
Speed_lead, Speed_lead_smoothed	Lead vehicle speed (m/s)
Additional fields	Elevation, heading, accuracy metrics (PDOP, HDOP, VDOP)

---

Data Quality

• Sampling rate: 10 Hz
• Smoothing: Moving average (1-second window)
• Quality metrics:
  • Anomaly Acceleration (%)
  • Anomaly Jerk (%)
• Post-processing: Reduced acceleration & jerk anomalies to near zero.

---

Example Use Cases

• Calibrating car-following models (e.g., Full Velocity Difference Model – FVDM)
• Investigating ADAS behavior at intersections
• Evaluating safety and comfort of ADAS–TCD interactions
• Integrating empirical ADAS behavior into microscopic traffic simulations

---