Workshop on Data-Driven Autonomous Driving Simulation (DDADS)

This workshop will be held at the CVPR 2025. Details will follow.

Abstract

On-road testing of autonomous vehicles presents significant challenges in terms of cost and safety, underscoring the importance of simulation as a crucial tool to accelerate the development of safe autonomous driving (AD), a technology with enormous real-world impact. To extract useful information from simulation, minimizing the sim-to-real gap by developing good agent behavior models, and ensuring faithful perception simulation is essential. While recent years have witnessed a surge of publications in this rapidly evolving field, several fundamental questions remain unanswered, with the research often scattered across different robotics and machine learning venues and research fields.

This workshop aims to unite leading researchers from various specializations, including perception simulation, behavior modeling, planner development, and safety research, to foster interdisciplinary conversations and collaboration.

Call for Papers

Submission details will follow. Stay tuned!

We encourage submissions from a broad range of topics connected to simulation, including but not limited on how to:

Improve fidelity and diversity of generative models for behavior and sensor simulation,
Develop new metrics for the quality and coverage of generated scenarios,
Combine behavior modeling, perception simulation, and safety requirements,
Create trust and assess the credibility of the simulation toolchain, and
Effectively use simulation to evaluate autonomous vehicles,
Effectively use simulation to train autonomous vehicles in closed loop,
Incorporate recent advances from other fields, such as RL or foundation models.