The first self-checkout system was installed in a Kroger grocery store outside Atlanta in 1986. It took decades, but the technology has finally spread across the U.S. And with grocery stores moving toward automation, robotic bagging may not be far off.
This week, MIT's CSAIL division is showcasing RoboGrocery, which combines computer vision and soft robotic grippers to bag a variety of items. To test the system, researchers placed 10 objects that were unknown to the robot on a grocery conveyor belt.
Products range from delicate items such as grapes, bread, kale, muffins and crackers to more rigid items such as soup cans, meal boxes, ice cream containers, etc. A vision system comes into play first, detecting objects on the belt before determining their size and orientation.
When the picker touches a grape, pressure sensors in the finger determine that the grape is too delicate to fall to the bottom of the bag — something many of us have no doubt learned the hard way — and then the picker recognizes that the soup can is a more rigid structure and pushes it to the bottom of the bag.
“This is an important first step towards robots packing groceries and other items in the real world,” said Annan Zhang, one of the study's lead authors. “While it's not yet ready for commercial deployment, our work demonstrates the power of integrating multiple sensing modalities into a soft robotic system.”
The team notes there's still plenty of room for improvement, including upgrading the gripping device and imaging system to better determine how and in what order items are packed. Once the system is more robust, it could be deployed not just in grocery stores, but in industrial spaces like recycling plants.