GenSUMO: GenAI-Creation of Critical Scenarios for Autonomous Vehicle Testing

Recent advancements in autonomous vehicle research highlight the importance of Machine Learning (ML) models in tasks like motion planning, trajectory prediction, and emergency management. To support AI development, we propose a novel approach for generating on-demand datasets using the Simulator of Urban Mobility (SUMO) and a Generative Adversarial Network (GAN). Our method focuses on capturing critical events such as sudden pedestrian crossings, near-misses, and collisions, providing essential data to improve vehicle models' responses to emergency situations.