Physicl raises $10 million to build realistic 3D worlds for training robots

[Tilly14]

Robots need to learn about the physical world, and one start up wants to give them better practice environments. TechCrunch reports that Physicl has raised a $10 million seed round led by Inovia Capital

Founder and chief executive Alex de Vigan previously spent a decade building nfinite, a company that made 3D product models, and Physicl reuses that 3D pipeline. That experience should help the company move quickly

Its product, Studio, launches with 10,000 3D environments and 100,000 objects. Users build scenes to generate synthetic training data for tricky situations, such as a cup on a cluttered counter or partly in shadow

An early customer is SpAItial, which builds world models, and its chief executive Matthias Niessner said that otherwise you have to curate the data by hand, which is a very long cycle. De Vigan said something was missing in AI, namely its ability to understand, interact with and manipulate space and objects

Training robots in simulation is far cheaper and safer than in real kitchens and warehouses. Edge cases like clutter and poor lighting are exactly where robots fail today, so generating lots of them makes sense. Realism will be the key test, because robots trained on unrealistic scenes struggle in the real world

Will simulation be enough to train useful home robots? And what task would you most like a robot to handle?

Physicl raises $10M to make the digital world real enough for robots | TechCrunchtechcrunch.com