Physicl raises $10 million to build more realistic robot simulations

In short: Physicl has raised $10 million in seed funding to help robot developers train robots in digital simulations that behave more like the real world.

What happened

Physicl, a new startup founded by Alex de Vigan, raised $10 million from Inovia Capital. The company is taking technology originally built for online shopping, where brands create lifelike 3D images of products, and using it to build detailed 3D environments for robots.

The problem Physicl is aiming at is the “sim-to-real gap.” That is the common issue where a robot does well in a computer simulation, but makes mistakes in the real world because the simulation did not include enough real-life details (like glare, shadows, clutter, and slippery surfaces).

Physicl is also releasing a product called Studio. Studio includes 10,000 3D environments and 100,000 “assets,” meaning digital objects like cups, counters, and other items you might see in a room. Customers can arrange these to create many training scenes, like teaching a robot to spot a cup on a messy countertop, even when the cup is partly in shadow.

Physicl says it wants to support ongoing synthetic data generation, meaning it can create new training examples on demand (like a photo studio that can quickly restage a scene in thousands of different ways).

Why it matters

Robots are hard to train because collecting real-world practice data is slow and expensive. If simulations become closer to real life, robot makers could test more situations before robots ever touch a real kitchen, warehouse, or factory, which may help reduce errors when robots are deployed.

Source: TechCrunch AI