Research

We are interested in building intelligent and useful robotic systems that can be deployed reliably in challenging environments. Our research broadly focuses on the intersection of machine learning and robotics, with an emphasis on large-scale robot learning (foundation models of & for robotics), reinforcement learning, and multi-agent and human-robot interaction.

We take a full-stack approach to robotics by focusing on all aspects of the problem, ranging from algorithmic innovation to system design, and value both the analytical and empirical nature of science. We draw inspiration from cognitive science and psychology to build physical AI systems at the interface of perception, learning, and control.

Robot Foundation Models

Foundations of scaling up robot learning: architectures (vision-language-action models, world models, etc.), systematic and scalable evaluation, understanding data quality, learning from diverse data sources, cross-embodiment learning.

Relevant Publications

1.
Zha, L. et al. LAP: Language-Action Pre-Training Enables Zero-shot Cross-Embodiment Transfer. Robotics: Science and Systems (RSS) (2026).
1.
Hirose, N., Glossop, C., Shah, D. & Levine, S. OmniVLA: An Omni-Modal Vision-Language-Action Model for Robot Navigation. in IEEE International Conference on Robotics and Automation (ICRA) (2026).
1.
Deshpande, A., Hendrix, R. & Shah, D. MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulation. arXiv preprint (2026).

Interaction and Autonomous Improvement

Algorithmic frameworks for modeling multi-agent and human robot interaction, learning from feedback, continual learning, autonomous improvement at scale.

Relevant Publications

1.
Zhang, M. & Shah, D. Visual Verification Enables Inference-time Steering and Autonomous Policy Improvement. Robotics: Science and Systems (RSS) (2026).
1.
Hirose, N., Shah, D., Stachowicz, K., Sridhar, A. & Levine, S. SELFI: Autonomous Self-Improvement with Reinforcement Learning for Social Navigation. in Annual Conference on Robot Learning (CoRL) (2024).
1.
Stachowicz, K., Shah, D., Bhorkar, A., Kostrikov, I. & Levine, S. FastRLAP: A System for Learning High-Speed Driving via Deep RL and Autonomous Practicing. in 7th Annual Conference on Robot Learning (CoRL) (2023).

Open-World Systems

Robust and reliable systems for long-term deployment of robots in the real world.

Relevant Publications

1.
Hirose, N. et al. Learning to Drive Anywhere with Model-Based Reannotation. IEEE Robotics and Automation Letters (RA-L) (2025).
1.
Team, G. R. & . Gemini Robotics: Bringing AI into the Physical World. Technical Report (2025).