Physicl raises a $10 million seed to simulate hard cases for robots

TechCrunch reports that Physicl, a startup founded by Alex de Vigan, raised $10 million in seed funding led by Inovia Capital. The company generates realistic 3D scenes used to train robots and world models, focusing on the kind of tricky, cluttered situations that are hard to capture with general video training alone. Its product, called Studio, ships with 10,000 ready-made 3D environments and 100,000 object assets, which can be used to produce synthetic training data for edge cases such as a cup sitting on a cluttered kitchen counter or an object partly hidden in shadow. World-model company SpAItial is an early customer and used Physicl's tools while building its Echo-3 model. Physicl's pitch is aimed specifically at post-training on difficult manipulation scenarios, rather than competing on broad general-purpose video training.

Why it matters

Robots learning to handle messy, real-world clutter rather than clean lab setups is one of the harder unsolved problems in robotics, and synthetic data is a cheaper way to get there than filming it all by hand.

Source: TechCrunch