MIT Introduces RoboGrocery

The first self-checkout system was installed in 1986 at a Kroger store outside Atlanta. It took several decades for the technology to spread across the US. Given the growing automation in grocery stores, robotic bagging appears to be a natural next step.

This week, MIT’s CSAIL department is showcasing RoboGrocery, a system that uses computer vision and a soft robotic gripper to bag various items. In a test, researchers placed 10 unknown objects on a grocery conveyor belt.

The items ranged from fragile products like grapes, bread, kale, muffins, and crackers to heavier items like soup cans, meal boxes, and ice cream containers. The vision system detects the objects first and then determines their size and position on the belt.

When the robotic gripper makes contact with the grapes, sensors in its fingers detect their delicate nature, recognizing that they require careful handling to avoid damage — a lesson many of us have learned through experience. The robot then identifies the soup can as a sturdy object and places it at the bottom of the bag.

Annan Zhang, one of the study’s lead authors said, “This is a significant first step towards having robots pack groceries and other items in real-world settings. Although we’re not quite ready for commercial deployment, our research demonstrates the power of integrating multiple sensing modalities in soft robotic systems.”

The researchers acknowledge that there is still significant potential for enhancement, particularly in refining the gripper and imaging systems to optimize packing decisions. As the technology advances and becomes more reliable, its applications could expand beyond grocery stores to industrial settings like recycling facilities.

Learn more about how AI agents work from here: