Technology news and discussion on TalkFuse.
Physicl raises $10M to make the digital world real enough for robots
Physicl has raised $10 million in seed funding from Inovia Capital to build realistic digital environments for training robots and physical AI models. The company aims to address a key limitation in robotics: simulations can provide training data more cheaply than real-world collection, but they must accurately represent objects and physics to be useful.
Founder Alex de Vigan is adapting technology developed at his previous company, nfinite, which created 3D product models for retailers. Physicl’s Studio product offers 10,000 environments and 100,000 objects for generating synthetic training data, including difficult cases such as a cup obscured by clutter or shadow. Early customer SpAItial used Physicl’s data to improve its simulation tools; Physicl ultimately hopes to generate new training scenarios automatically as models need them.
In short
- Physicl raised $10 million from Inovia Capital.
- Its Studio product includes 10,000 environments and 100,000 assets.
- The startup aims to generate realistic synthetic data for robot training.
Explained simply
Robots learn to handle objects and move through spaces using examples, often collected by observing the real world. That can be slow and costly, so developers also use computer simulations to create training data.
For simulated examples to help, they need to reflect how objects look and behave in real life, including when lighting is poor or objects are partly hidden. Physicl makes digital scenes and objects for this purpose, drawing on technology developed by founder Alex de Vigan’s previous company, nfinite, which made 3D product models for retailers.
Physicl’s Studio provides a library of environments and objects that companies can use to generate training examples. One early customer, SpAItial, used its data to improve simulation tools; Physicl says it hopes to create new scenarios automatically as models need them.