Giant warehouses can often be self-contained ecosystems that require the same infrastructure one might find in their local neighborhoods. Water, electricity and HVAC are essential, of course, but an intentional and safely designed traffic pattern for vehicles, equipment and employees is just as important in the day-to-day operations of these large facilities.

So where does a company start in planning the internal layout of these buildings? Sometimes with a supercomputer.

Researchers at the Robotics Institute at Carnegie Mellon University used allocations on Bridges-2 at the Pittsburgh Supercomputing Center (PSC) to optimize both warehouse layouts and robot coordination. The supercomputing time, acquired through the U.S. National Science Foundation program ACCESS, led to models for greater efficiency and safer working conditions for robots and humans alike.

“[We] have a server in the lab, but [it] can only run very small setup experiments,” said Jiaoyang Li, an assistant professor at the Robotics Institute. “Because of this parallelization requirement, I feel like for this line of research, it would be super challenging if we didn’t have a large computing center server [like Bridges-2 available].”

Li’s team used the artificial intelligence and machine learning capabilities of Bridges-2 to run hundreds of simulations simultaneously, training the models to account for many types of problems much quicker than traditional methods. AI helped determine efficient shelf layouts, inventory placement and robot routes while keeping the warehouse understandable for human workers. The team’s methods help thousands of robots avoid traffic jams and collisions, improving package handling and reducing delays.

You can read more about this story here: Robo-Ready Warehouses Key to Faster Shipping

Resource Provider Institution(s): Pittsburgh Supercomputing Center (PSC)
Resources Used: Bridges-2
Affiliations: Carnegie Mellon University
Funding Agency: NSF
Grant or Allocation Number(s): CIS220115

The science story featured here was enabled by the U.S. National Science Foundation’s ACCESS program, which is supported by National Science Foundation grants #2138259, #2138286, #2138307, #2137603, and #2138296.