World model for Robotics and Vehicle control
Advancing Physical AI through Vision‑Language‑Action Research
We build general-purpose foundation models combining language, vision, and action — enabling embodied agents that learn from and interact with the physical world.
The future is embodied
WoRV is a research team operating with the agility of a startup. We believe the next frontier of AI lies not in models trained solely on internet data, but in Embodied Agents — systems that learn from and interact with the physical world.
We're building a general-purpose Foundation Model unifying language, vision, and action — favoring rapid experimentation and open communication to create products with real-world impact.
Where we apply our research
Autonomous Driving
End-to-end learning for vehicle control, replacing specialized pipelines with unified foundation models.
Robotics
Embodied agents that learn from and interact with the physical world through vision, language, and action.
Agriculture
Commercializing Physical AI for precision agriculture and autonomous farming systems.
Defense
Developing robust autonomous systems for defense applications with real-world reliability.
Recent highlights
SketchDrive (CANVAS)
Demonstrating autonomous navigation through contextual understanding — our flagship published work.
Read more BenchmarkCostNav
A navigation benchmark for evaluating embodied agents in realistic environments.
Read more ResearchMapAnything
Bridging the gap between traditional mapping methods and real-world applications.
Read moreBuild the future of Physical AI
We're looking for researchers and engineers passionate about robotics, autonomous driving, and embodied intelligence.
View Open Positions