Kroger Builds an AI-based Warehouse that is Sweeter than Honey

Artificial intelligence and machine learning enable robotic shoppers to fulfill online grocery orders in less than 10 minutes. Instead of walking through the aisles, the cloud-based system guides these robots as they make a beeline for each item. AI controls and manages everything from the ordering process to the delivery.


Article Key

Image credit: Kroger

The growth in online grocery orders skyrocketed due to the pandemic, and while it has declined slightly in recent months, the US online grocery market is still huge, posting $8 billion in sales in February 2021 alone. Grocery stores have embraced the shift. Many have invested in building or bolstering online ordering apps and shopper services, but other stores have revamped their ecommerce sales and customer fulfillment from the ground up.

Kroger Co., the parent of more than a dozen retail chains with 2,800 stores across the US, falls into the latter category. Kroger has built two customer fulfillment centers (CFCs)—and is planning to open six more—to meet the increase in online shoppers. Located in Ohio and Florida, Kroger’s facilities are powered by technology from Ocado Group, an online grocery store in the UK that has partnered with stores in Europe, Japan, and now the US.

The Ocado Smart Platform (OSP) technology encompasses ecommerce, order fulfillment, and logistics and relies on artificial intelligence from customer order to delivery. Ocado uses digital twins of its CFC to determine optimal performance. In 2020, the company ran more than 620,000 simulations.

Customers place orders through an AI-based app that remembers favorite items and preferred brands. The app can even suggest items the user might be running out of, based on prior purchases. When the customer submits the order, the technology kicks into high gear at the CFC.

Abuzz with Activity

The CFC is a huge warehouse and uses artificial intelligence, advanced robotics, and automation to fill thousands of orders per day. Dubbed “The Hive,” it is a giant grid that sits above multiple layers of product-filled totes. On the grid is a swarm of square robots that buzz over the totes collecting items for packaging. Kroger’s facilities use about 1,000 robots, but Ocado’s UK location uses twice as many.

The OSP uses artificial intelligence to determine the best route for grocery fulfillment. Each bot can travel up to nine miles per hour and communicates with the cloud 10 times per second to establish its location on the grid. The bots zoom past each other with only 5 millimeters separating them, and they have retractable wheels that allow them to change axes on a dime. A vacuum winch enables the bot to collect the totes from beneath the grid.

The technology is fast—it can pick 50 items in 10 minutes—and it boasts a 99 percent accuracy rate. The bots work together to retrieve items from the Hive and bring them to the packing station, where they are bagged and prepped for delivery.

Busy as a Bee

The packing stations are manned by human workers, but AI controls the bagging process. The totes are sent to the packers, who look at a screen to see where to bag each item. AI determines how to pack groceries using the fewest number of bags, placing the heaviest items in the bottom, and balancing the weight of each bag.

Most items are hand packed at the Kroger facilities, but some of Ocado’s international locations have robotic packers. In those locations, items are moved from the totes to the bags using a robotic arm with a flexible suction tip. A camera, computer vision, and machine learning on the robotic arm determine how to pick up the item and into which bag each item should be placed. The arm can’t pick up delicate or perishable items, so humans do that.

When the groceries are ready for delivery, totes are scanned and loaded onto trucks based on location and route.  The AI algorithms make 14 million calculations per second to determine the best delivery route, based on time of day, construction, and traffic.

In addition to logistical efficiencies, the OSP brings operational efficiencies, too. Machine learning predicts purchases and helps manage inventory across the store’s 50,000 products. Because the system can predict demand, it limits food waste. The Ocado system has an average of .04 percent food waste, compared to the industry average of 2-3 percent of sales. It also allows the bots to self-report problems, so operators can schedule maintenance.

Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries.

Find Intel® IoT Market Ready Solutions

Search the Intel® IoT partner ecosystem and discover 130+ end-to-end IoT solutions.

Become a Member of the Intel® Partner Alliance

If you are delivering repeatable and scalable IoT offerings using Intel technology, the Intel® Partner Alliance program gives you more opportunities to build your business. Get started.