A digital twin sounds a lot like a long-lost sibling you found on Ancestry, but it has nothing to do with that. Rather, it’s a virtual replica of a real space, such as a warehouse or a factory, that include technology, people and processes. Digital twins have the potential to forever change how these spaces function while impacting millions of jobs in the manufacturing and logistics sectors.
Manufactured and shipped goods are the lifeblood of our society, and companies around the world are filling warehouses and factories with robots. For example, Walmart in April started rolling out autonomous forklifts in distribution centers. Just this month, the semiconductor maker Lam Research released a collaborative robot (dubbed a “cobot”) to conduct high-precision maintenance on product material.
But to introduce robots without a hitch requires serious planning, and digital twins make that possible—all without having to subject workers and the space they inhabit to potentially dangerous environments.
“You can simulate sensors, you can simulate robots, you can simulate people, you can simulate machines, you can train your A.I. agents, all of that in the simulated world,” Amit Goel, director of product management for autonomous machines at Nvidia (NVDA), told Observer.
Digital twins are particularly useful in tasks like deploying new robots, training and certifying operators, and introducing new manufacturing technologies. “Those tests are cheaper, faster and, most importantly, safer,” Leo Moran, a senior manager at Kalyspo, an automation consulting company, told Observer. “That’s really crucial in a world where the application of robotics is accelerating at an unbelievable pace.”
Kalypso offers services like supply chain optimization, autonomous manufacturing and digital twin creation. The company runs 3D models on Nvidia’s Omniverse 3D development platform and the Isaac Sim robotics simulation platform.
On Omniverse, for example, Kalypso tests new features like ray tracing, which simulates how light behaves in the real world. (This feature can be used in digital twins of warehouses for a more realistic simulation experience.) “The renderings look amazing. You can look at how light moves, how objects interact. It looks much more real so you get out of that uncanny valley you often have in this space,” Moran said.
More than the beauty they evoke, these digital replicas allow companies to ask questions such as “What are the safety implications of having 50 robots in a room instead of 20?” and “Will people still be able to safely move through there?”—all without subjecting workers to uncertain environments, Moran added.
While digital twins are arguably one of the most influential behind-the-scene elements working to upend the industrial sector as we know it, they’re not all roses. Building digital twins can be really expensive because of the computing power they require, said Nvidia’s Goel. In November, Nvidia released a suite of generative A.I. blueprint models to help propel clients to create digital twins at scales as large as entire warehouses at the highest fidelity.
Moran noted the biggest challenge isn’t necessarily the technology itself, but the people building and operating it. “It’s a very multi-disciplinary job,” he said. “You need to have an understanding of robotics, automation, computer science, controls engineering, mechanical engineering. The supply for talent does not meet the demand today.”
Sometimes called mechatronics engineering by institutions like Johns Hopkins Whiting School of Engineering and Rochester Institute of Technology, this sort of multidisciplinary education is still catching up with the technology industry’s needs.
