Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Incorporating Ephemeral Traffic Waves in A Data-Driven Framework for Microsimulation in CARLA

Alex Richardson, Azhar Hasan, Gabor Karsai, Jonathan Sprinkle

TL;DR

The paper tackles the challenge of evaluating autonomous vehicle performance under realistic traffic waves by creating a data-driven cosimulation in CARLA that reproduces real-world stop-and-go dynamics around an ego vehicle. It uses high-resolution I-24 MOTION data to drive a boundary-controlled framework with visible and ghost vehicle regions surrounding the ego, enabling time-space diagram fidelity. Key contributions include a 1-mile I-24-based environment, a ghost-cell boundary mechanism, and an open cosimulation pipeline for testing wave mitigation and perception-driven autonomy, demonstrated through two qualitative scenarios. While achieving perceptual fidelity to real traffic waves, the work also identifies sim-to-real gaps in CARLA's vehicle behavior and discusses future directions such as IDM-based agents and smoothing of ghost inputs to improve realism and scalability.

Abstract

This paper introduces a data-driven traffic microsimulation framework in CARLA that reconstructs real-world wave dynamics using high-fidelity time-space data from the I-24 MOTION testbed. Calibration of road networks in microsimulators to reproduce ephemeral phenomena such as traffic waves for large-scale simulation is a process that is fraught with challenges. This work reconsiders the existence of the traffic state data as boundary conditions on an ego vehicle moving through previously recorded traffic data, rather than reproducing those traffic phenomena in a calibrated microsim. Our approach is to autogenerate a 1 mile highway segment corresponding to I-24, and use the I-24 data to power a cosimulation module that injects traffic information into the simulation. The CARLA and cosimulation simulations are centered around an ego vehicle sampled from the empirical data, with autogeneration of "visible" traffic within the longitudinal range of the ego vehicle. Boundary control beyond these visible ranges is achieved using ghost cells behind (upstream) and ahead (downstream) of the ego vehicle. Unlike prior simulation work that focuses on local car-following behavior or abstract geometries, our framework targets full time-space diagram fidelity as the validation objective. Leveraging CARLA's rich sensor suite and configurable vehicle dynamics, we simulate wave formation and dissipation in both low-congestion and high-congestion scenarios for qualitative analysis. The resulting emergent behavior closely mirrors that of real traffic, providing a novel cosimulation framework for evaluating traffic control strategies, perception-driven autonomy, and future deployment of wave mitigation solutions. Our work bridges microscopic modeling with physical experimental data, enabling the first perceptually realistic, boundary-driven simulation of empirical traffic wave phenomena in CARLA.

Incorporating Ephemeral Traffic Waves in A Data-Driven Framework for Microsimulation in CARLA

TL;DR

The paper tackles the challenge of evaluating autonomous vehicle performance under realistic traffic waves by creating a data-driven cosimulation in CARLA that reproduces real-world stop-and-go dynamics around an ego vehicle. It uses high-resolution I-24 MOTION data to drive a boundary-controlled framework with visible and ghost vehicle regions surrounding the ego, enabling time-space diagram fidelity. Key contributions include a 1-mile I-24-based environment, a ghost-cell boundary mechanism, and an open cosimulation pipeline for testing wave mitigation and perception-driven autonomy, demonstrated through two qualitative scenarios. While achieving perceptual fidelity to real traffic waves, the work also identifies sim-to-real gaps in CARLA's vehicle behavior and discusses future directions such as IDM-based agents and smoothing of ghost inputs to improve realism and scalability.

Abstract

This paper introduces a data-driven traffic microsimulation framework in CARLA that reconstructs real-world wave dynamics using high-fidelity time-space data from the I-24 MOTION testbed. Calibration of road networks in microsimulators to reproduce ephemeral phenomena such as traffic waves for large-scale simulation is a process that is fraught with challenges. This work reconsiders the existence of the traffic state data as boundary conditions on an ego vehicle moving through previously recorded traffic data, rather than reproducing those traffic phenomena in a calibrated microsim. Our approach is to autogenerate a 1 mile highway segment corresponding to I-24, and use the I-24 data to power a cosimulation module that injects traffic information into the simulation. The CARLA and cosimulation simulations are centered around an ego vehicle sampled from the empirical data, with autogeneration of "visible" traffic within the longitudinal range of the ego vehicle. Boundary control beyond these visible ranges is achieved using ghost cells behind (upstream) and ahead (downstream) of the ego vehicle. Unlike prior simulation work that focuses on local car-following behavior or abstract geometries, our framework targets full time-space diagram fidelity as the validation objective. Leveraging CARLA's rich sensor suite and configurable vehicle dynamics, we simulate wave formation and dissipation in both low-congestion and high-congestion scenarios for qualitative analysis. The resulting emergent behavior closely mirrors that of real traffic, providing a novel cosimulation framework for evaluating traffic control strategies, perception-driven autonomy, and future deployment of wave mitigation solutions. Our work bridges microscopic modeling with physical experimental data, enabling the first perceptually realistic, boundary-driven simulation of empirical traffic wave phenomena in CARLA.

Paper Structure

This paper contains 22 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Stop-and-go Traffic Wave Data and Theory
  4. Control of Stop-and-Go Waves
  5. Real World Traffic Flow Data Collection
  6. Bird's Eye View Traffic Simulations
  7. CARLA Cosimulation Framework
  8. I-24 Environment
  9. Road and Environment Generation
  10. Cosimulator Design
  11. Vehicle State
  12. Configuration
  13. Simulation State
  14. CARLA Vehicles
  15. Simulation Control Flow
  16. ...and 7 more sections

Figures (4)

  • Figure 1: The I-24 60.6 mile to 61.6 mile region shown as a simplified diagram, with each segment at 0.1 mile in length demarcated. Cameras are shown at approximate spacing, with the green segments showing the beginning of each road, and the red segments showing the end of each road.
  • Figure 2: Two time-space diagrams with overlay of the ego vehicle and its surroundings at two time, $t_0$ and $t_1$. The direction of travel is "up" in the figure, and time advances to the right. (Left) at time $T=t_0$. Downstream cars are placed in their exact positions and given their velocity directly based on the representative traffic data. The simulated cars in CARLA are in the the immediate surroundings of the ego vehicle, and managed by its traffic control framework. Upstream cars (lower in the figure) are positioned directly by data. (Right) at time $T=t_1$ the ego car and its surroundings have moved downstream. The car has now encountered a traffic wave, which means that additional cars will begin to appear in the local CARLA framework as the surroundings pick those cars up directly from the positional information gained by the measured data.
  • Figure 3: Scenario A simulation results - a Bird's Eye View of the cosimulation state at (a), a screenshot of CARLA at (b) and a time-space diagram showing the simulated ego vehicle's trajectory vs the I24 empirical trajectories at (c).
  • Figure 4: Scenario B simulation results - a Bird's Eye View of the cosimulation state at (a), a screenshot of CARLA at (b) and a time-space diagram showing the simulated ego vehicle's trajectory vs the I24 empirical trajectories at (c).