ControlMap: Controllable High-Definition Map Generation for Traffic Scenario Simulation

The paper introduces ControlMap, a data-driven pipeline for generating controllable high-definition (HD) maps using latent diffusion and ControlNet for spatial conditioning. This approach allows for adjustable conditioning strength through classifier-free guidance and supports city-level style transfer, enhancing the diversity and realism of generated maps. The model's ability to accurately reflect specific road topologies and city details addresses the limitations of existing generative models, making it a significant advancement for practitioners in autonomous driving simulation who require high-fidelity map data.