RobotStudio® HyperReality- Industrial Physical AI in Precision Manufacturing

In this talk

1. 0:00

   Vision must handle real-world variation

   Robots need vision as well as movement. In electronics manufacturing, shifts in parts, lighting, materials, and camera views can affect precision.

2. 0:20

   Validate vision systems in simulation

   ABB says vision validation traditionally relied on physical samples and real-world images. RobotStudio HyperReality moves validation earlier using virtual simulation data that ABB describes as 99% accurate to the physical world.

3. 0:37

   Replicate and refine the station

   The digital twin simulates the product, camera, lighting, and robot before production begins. ABB compares a real robot and camera view with a simulated view, tuning materials, lighting, and camera settings to match the station.

4. 1:06

   Generate labeled synthetic data

   Controlled randomization varies light, camera, and material properties to produce thousands of realistic images. Each image is automatically labeled for AI training and validation, without interrupting the production line or waiting for a physical product.

5. 1:32

   Close the real-to-simulation loop

   After deployment, feedback returns to the simulation to improve the AI model and inform the next deployment. ABB describes this feedback cycle as real-to-sim-to-real.