One Robot
World models for robots — realistic simulation for faster robot training
One Robot builds simulation environments (world models) that are physically and visually realistic, trained on a robot team's own data. The platform lets robotics teams discover policy failures in simulation, generate targeted synthetic training data, and ship more reliable manipulation policies without being bottlenecked by physical robot availability. It is focused on contact-rich manipulation, deformable objects, and precise assembly tasks.
A world model is trained on the team's robot data to learn physics and visuals, then used to roll out policies in simulation, discover failures, and generate synthetic trajectories to fix them.
Robotics teams building action models for manipulators
Background.
- Status
- launched
- Business model
- unknown
- Company
- One Robot
Founders
- Hemanth Sarabu
- Elton Shon
Similar projects.
Editorial take on the space this project sits in — momentum signals, adjacent moves, our call on whether the wedge is real. Get pinged when we publish a new read or when the landscape shifts.
Have a take on this space?
Tell us what you’d build differently, where you think the incumbents miss, or what we’ve gotten wrong about this project. Comments + reactions are coming soon.