What AV@B Research works on
Division
TL;DR
AV@B Research explores the frontier of autonomous systems and embodied intelligence. Projects aim at technical contributions other teams can use, not demos.
We plan to publish, open-source our work, and collaborate with Berkeley researchers. Students can join a project in their first semester.
Topics
- Vision-language-action models
- World models
- Autonomous driving
- Perception and 3D vision
- Multi-agent prediction
- Motion planning
- Reinforcement and imitation learning
- Localization and sensor fusion
- Model predictive control
- Efficient onboard inference
What a project produces
Every project should end in something another team can cite or run: a paper, a trained model, a dataset, or an open-source system. Work happens inside Berkeley Driver, so results can be tested on real vehicles.
Joining a project
Pick a topic above, ideally one the vehicles need solved, and start on a real project.