Case Study: Pickin’ Chicken With Synthetic Data

Massachusetts-based Soft Robotics is using AI to help food manufacturers cope with labor shortages. The start-up augmented its soft-grip technology with 3D vision and synthetic data1 to train robots in the handling of chicken parts as they move from processing to packaging.

A case study2 published by microchip maker Nvidia, whose Issac Sim platform helped Soft Robotics to model processes, examines the company’s response to the challenges posed by handling varied items in unstructured environments. It details how mGripAI applies Nvidia’s Ominverse Replicator3 to generate thousands of images and address challenges such as occlusion and variation in lighting.

Soft Robotics utilizes ray tracing4 technology applied in the gaming industry and physics-based accuracy provided by Omniverse to replicate the complex scenarios. The application-specific models enable Soft Robotics’ customers to increase production efficiency, in part through cutting onboarding times from months to days, Nvidia says.

This synthetic data approach boosts production line pick accuracy, enabling the soft-grip robots to handle up to 100 picks per minute. Robots equipped with the mGripAI tech mitigate staffing challenges in an industry rife with health risks for workers and consumers, ensuring continued productivity and safety.

Smart Robotics mGripAI builds on tech used by Canadian vendor JPM Solutions for clients in high-volume production of pre-made sandwiches5.