Felipe Codevilla

Universitat Autònoma de Barcelona, Universidade Federal do Rio Grande

Papers

Total Citations

1,174

H-Index

About

Felipe Codevilla is a researcher whose work spans autonomous driving, robot navigation, and multi-agent motion prediction. He is best known for his landmark 2015 paper "End-to-end Driving via Conditional Imitation Learning," which has amassed over 1,000 citations and addressed a critical limitation in imitation learning-based driving systems — the inability to control a trained vehicle's behavior at test time. By conditioning learned driving policies on high-level commands, Codevilla enabled end-to-end neural networks to respond to navigational instructions, a foundational contribution to the autonomous driving field. Earlier in his career, Codevilla contributed to underwater robotics, co-developing DolphinSLAM, a bio-inspired solution for 3D underwater localization and mapping that extended the RatSLAM framework to marine environments, earning 63 citations. More recently, his research has shifted toward multi-agent trajectory prediction, with his work on Latent Variable Sequential Set Transformers (AutoBots) proposing elegant architectures for modeling joint future behaviors of multiple agents simultaneously — a critical challenge for safe autonomous systems. Across these diverse domains, Codevilla consistently bridges biological inspiration and deep learning to solve real-world robotics challenges.