Research organization
Google DeepMind
World models, robot learning, VLA systems, embodied reasoning
Official research siteDirect answer
What does Google DeepMind research?
World models, robot learning, VLA systems, embodied reasoning
Physical AI work
Google DeepMind contributes to the Physical AI ecosystem through world models, robot learning, vla systems, embodied reasoning. This profile connects its public research projects with the roadmap topics needed to understand them.
Notable projects
- RT-2
- Open X-Embodiment
- Genie
- Gemini Robotics
- ASIMOV
Related paper explainers
RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
RT-2 co-trains vision-language models on web and robot data so semantic knowledge can influence actions.
Open X-Embodiment and RT-X
Open X-Embodiment combines robot datasets across institutions and trains policies that transfer across embodiments.
Genie: Generative Interactive Environments
Genie learns controllable interactive environments from unlabeled internet video.
Genie 2: A Large-Scale Foundation World Model
Genie 2 generates action-controllable 3D environments from a single prompt image.
Genie 3: A General-Purpose World Model
Genie 3 generates interactive environments that can be explored in real time from text descriptions.
DreamerV3: Mastering Diverse Domains through World Models
DreamerV3 uses robust normalization and objectives to learn across more than 150 tasks with one configuration.
Gemini Robotics 1.5
Gemini Robotics 1.5 turns visual observations and instructions into motor commands while supporting multi-step physical tasks.
Gemini Robotics-ER 1.6
Gemini Robotics-ER 1.6 improves spatial reasoning, multi-view understanding, tool use, and safety-oriented robot reasoning.
ASIMOV Benchmark for Robot Semantic Safety
ASIMOV evaluates whether robot-brain foundation models understand unsafe physical situations and safety rules.