Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SynTraC: A Synthetic Dataset for Traffic Signal Control from Traffic Monitoring Cameras

Tiejin Chen, Prithvi Shirke, Bharatesh Chakravarthi, Arpitsinh Vaghela, Longchao Da, Duo Lu, Yezhou Yang, Hua Wei

TL;DR

SynTraC introduces the first public image-based traffic signal control dataset generated with the CARLA simulator, pairing real-style intersection imagery with traffic signal states and multi-objective RL rewards. The dataset encompasses diverse weather and time-of-day conditions, four synchronized RGB cameras, and ground-truth vehicle bounding boxes and rewards, enabling offline RL experimentation. Extensive experiments reveal a pronounced gap between image-based RL and feature-based control, affected by detection quality and camera viewpoints, while real-world generalization is demonstrated via fine-tuned detectors and Tempe intersection data. The work provides a reusable data-generation pipeline and codebase to drive future advancements in image-based TSC, highlighting the need for robust perception-to-control algorithms under perception uncertainty.

Abstract

This paper introduces SynTraC, the first public image-based traffic signal control dataset, aimed at bridging the gap between simulated environments and real-world traffic management challenges. Unlike traditional datasets for traffic signal control which aim to provide simplified feature vectors like vehicle counts from traffic simulators, SynTraC provides real-style images from the CARLA simulator with annotated features, along with traffic signal states. This image-based dataset comes with diverse real-world scenarios, including varying weather and times of day. Additionally, SynTraC also provides different reward values for advanced traffic signal control algorithms like reinforcement learning. Experiments with SynTraC demonstrate that it is still an open challenge to image-based traffic signal control methods compared with feature-based control methods, indicating our dataset can further guide the development of future algorithms. The code for this paper can be found in \url{https://github.com/DaRL-LibSignal/SynTraC}.SynTraC

SynTraC: A Synthetic Dataset for Traffic Signal Control from Traffic Monitoring Cameras

TL;DR

SynTraC introduces the first public image-based traffic signal control dataset generated with the CARLA simulator, pairing real-style intersection imagery with traffic signal states and multi-objective RL rewards. The dataset encompasses diverse weather and time-of-day conditions, four synchronized RGB cameras, and ground-truth vehicle bounding boxes and rewards, enabling offline RL experimentation. Extensive experiments reveal a pronounced gap between image-based RL and feature-based control, affected by detection quality and camera viewpoints, while real-world generalization is demonstrated via fine-tuned detectors and Tempe intersection data. The work provides a reusable data-generation pipeline and codebase to drive future advancements in image-based TSC, highlighting the need for robust perception-to-control algorithms under perception uncertainty.

Abstract

This paper introduces SynTraC, the first public image-based traffic signal control dataset, aimed at bridging the gap between simulated environments and real-world traffic management challenges. Unlike traditional datasets for traffic signal control which aim to provide simplified feature vectors like vehicle counts from traffic simulators, SynTraC provides real-style images from the CARLA simulator with annotated features, along with traffic signal states. This image-based dataset comes with diverse real-world scenarios, including varying weather and times of day. Additionally, SynTraC also provides different reward values for advanced traffic signal control algorithms like reinforcement learning. Experiments with SynTraC demonstrate that it is still an open challenge to image-based traffic signal control methods compared with feature-based control methods, indicating our dataset can further guide the development of future algorithms. The code for this paper can be found in \url{https://github.com/DaRL-LibSignal/SynTraC}.SynTraC
Paper Structure (12 sections, 3 equations, 4 figures, 4 tables)

This paper contains 12 sections, 3 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. SynTraC Dataset Generation Pipeline
  3. Traffic Scene Configuration
  4. Simulation and Data Generation
  5. Reward Calculation
  6. EXPERIMENTAL EVALUATION
  7. Evaluation Setting
  8. Evaluation Results
  9. Influence of Different Camera Angles
  10. Generalization on Real-world Data
  11. Discussion
  12. Conclusion and Future Work

Figures (4)

  • Figure 1: Overview of SynTraC, showing the images from different cameras, weather conditions, times, and comparison with other traffic signal datasets.
  • Figure 2: Overview of SynTraC data generation pipeline.
  • Figure 3: Example of lane detection computation. The direction of cross-product $V_1 \times V_2$ points upward (yellow dash), indicating the vehicle is in the right lane.
  • Figure 4: Example images for our chosen scenario. In the bird-view image, vehicles wait north and south while no east or west car coming. The traffic signal is red in this scene.