TUM-AVS/FM-AD-Survey

Organizes 348+ peer-reviewed papers across six categories—scenario generation, scenario analysis, datasets, simulators, benchmark challenges, and foundation model implementations—enabling systematic study of how large language models, vision-language models, and diffusion models address autonomous driving validation. The survey maps foundation model architectures (LLMs, VLMs, MLLMs, world models) to their specific applications in testing and safety-critical scenario understanding. Maintained by TUM's Autonomous Vehicle Systems group with monthly updates tracking emerging research in generative and multimodal approaches for autonomous vehicle testing.

185 stars.