AutoWorld: Scaling Multi-Agent Traffic Simulation with Self-Supervised World Models

Traffic simulation aims to generate realistic multi-agent motion, reflecting the inherently multimodal distribution of possible future outcomes. Learning such a distribution with supervised methods typically requires large-scale annotated trajectory data. However, collecting high-quality motion labels at scale is time-consuming and expensive, limiting the diversity and coverage of driving scenarios available for training. In contrast, raw sensor data such as LiDAR can be collected at scale without manual annotation.

The main idea behind our proposed semi-supervised framework is to complement supervised motion generation with a self-supervised predictive world modeling stage that leverages unlabeled data to implicitly learn scene dynamics and, by doing so, provide structured future context that improves downstream motion generation. We instantiate this idea in AutoWorld, a traffic simulation framework that formulates behavior generation as a generative process conditioned on a predictive world model of the environment. In specific, we first learn a motion-aware latent occupancy world model in a fully self-supervised manner from unlabeled LiDAR data, forecasting future latent occupancies. This stage serves as a scalable pretraining phase that implicitly captures motion-driven scene dynamics without requiring semantic or trajectory annotations. From the predicted future latent occupancies, we construct a predictive scene context that summarizes anticipated scene evolution. We then generate agent motions using a conditional diffusion model given both the future latent sequence and the predictive scene context, forming a coarse-to-fine representation of the future scene. Finally, to promote a variety of plausible multi-agent behaviors, we introduce a cascaded latent diversity strategy applied at inference time.