NVIDIA Brings AI to the Supply Chain

NVIDIA cuOpt AI software optimizes route planning, warehouse picking, fleet management and more logistics challenges to control skyrocketing delivery costs from factories to stores and homes.
by Alex Fender

Editor’s note: The name of the NVIDIA ReOpt AI software was changed to NVIDIA cuOpt in March 2022. All references to the name have been updated in this blog.

The global pandemic has created a supply chain crisis affecting just about every person and business on the planet.

For consumers, supply-and-demand imbalances might be written off as one less present under the Christmas tree. Businesses, however, tally the toll as millions of dollars in profits.

To help the $9 trillion logistics industry efficiently route goods from factories to stores and homes, NVIDIA today debuted NVIDIA cuOpt AI software.

Combining NVIDIA RAPIDS software with local search heuristics algorithms and metaheuristics such as Tabu search, NVIDIA cuOpt is an accelerated solver for big data that optimizes vehicle route planning and logistics in real time.

NVIDIA cuOpt delivers new tools for dynamic logistics and supply chain management to a wide range of industries, including transportation, warehousing, manufacturing, retail and quick-service restaurants.

A demonstration of cuOpt at NVIDIA GTC used NVIDIA Isaac Sim, powered by the NVIDIA Omniverse platform for simulating and collaborating within virtual worlds and 3D workflows, to first create a scalable robotics simulation application showing dynamic re-optimization in a photorealistic environment.

At a time when the global supply chain faces massive disruption, NVIDIA cuOpt provides the AI software required for everything from vehicle routing for last-mile delivery to efficiently picking and packing of warehoused goods bound for homes and offices.

From Oven to Door

Leveraging big data quickly is crucial to Domino’s Pizza. The company says there are 34 million ways to build a pizza — based on the potential combinations of toppings, crust, cheese, sauce and size. Delivering all those possible pizza combinations as efficiently as possible is important to hungry customers. Working with the NVIDIA cuOpt team, the company implemented a real-time predictive system that helps it meet important delivery standards for customers eager for dinner.

“The results we achieved working with NVIDIA showed great promise in delivering real-time and actionable insights that help us better serve our customers each day,” said Andrew Birch, vice president of data science at Domino’s.

Domino’s pizza delivery route optimization using NVIDIA cuOpt

Planning for a Post-Pandemic Economy

The logistics industry is a highly integrated supply chain network that links producers and consumers through multiple transportation modes, including air and express delivery services, freight rail, maritime transport and truck transport.

Delivering goods directly to a customer’s door, called last-mile delivery, has been a costly challenge even before the pandemic disrupted logistics just as millions more people switched to shopping online.

It’s estimated that over half of total shipping costs result from last-mile deliveries, impacting profitability, according to ABI Research. Companies typically eat about 25 percent of that cost themselves — a number that is increasing because of supply chain inefficiencies, according to logistics company Onfleet.

cuOpt, now available for early access, offers software-driven process changes, including:

  • Route optimization. Dynamic route optimization uses advanced algorithms that provide dispatchers and drivers with routing that can dramatically reduce miles, fuel cost, CO2 emissions and idle time. Companies gain real-time feedback on road conditions, traffic data and route suggestions to create more accurate arrival windows.
  • Advanced constraints. Modeling movements of vehicles that have finite capacities and different costs. That includes factoring in such items as fresh produce that must be carried by refrigerated trucks, or that customers are only available during specific windows in time.
  • Warehouse picking. Companies can create better human-machine interaction for automated storage and retrieval in picking and packing that dynamically routes and re-routes robots for efficient truck loading as new orders arrive.
  • Fleet mix optimization. Tracks routing across the supply chain with locations and fleet statistics for airplanes, trains, trucking and container fleets. The software can take into account the number of available pilots, drivers or ships to operate these vehicles on a given day and factor in maintenance costs.

NVIDIA Accelerated Computing Powers World-Leading Accuracy

NVIDIA GPUs offer the computational power needed to fuel the most ambitious heuristics while supporting the most challenging constraints. cuOpt takes advantage of NVIDIA’s massively parallel architecture to generate thousands of solution candidates and refine them to select only the best one at the end.

As a result, cuOpt can scale to the largest problems in seconds with world-class accuracy. The software showed a 2.96 percent accuracy average gap, compared to the world’s best known solution for the number of vehicles, in early tests on the Gehring & Homberger benchmark.

Try out NVIDIA CuOpt AI Software and CuOpt Launchpad.

Watch NVIDIA CEO Jensen Huang’s GTC keynote address below.