Do Diffusion Models Dream of Electric Planes? Discrete and Continuous Simulation-Based Inference for Aircraft Design
arXiv cs.LG / 3/17/2026
📰 NewsIdeas & Deep AnalysisModels & Research
Key Points
- The paper presents a simulation-based inference approach to design eVTOL aircraft by learning a posterior over the full design space using two diffusion models.
- It samples over discrete topologies and their corresponding continuous parameters, forming a hierarchical probabilistic model for design Generation.
- The first diffusion model leverages RDLM and UWMs to sample topologies from a discrete-continuous space, while the second uses a masked diffusion method to sample parameters conditioned on topology.
- The approach rediscovers known design trends and physical laws in aircraft design while significantly accelerating the design-generation process.




