BMW’s virtual factory uses AI to hone the assembly line

German car manufacturer BMW plans to start making electric vehicle powertrains at a massive plant in Regensburg, Bavaria, later in 2021. Well before new parts roll off the production line, the entire production process will be performed in stunningly realistic detail in a virtual version of the factory.

The simulation allows managers to plan the production process in greater detail than previously possible, says Markus Grüeneisl, who leads the production strategy at BMW. “We now have a perfect digital twin to our real-time production,” he says.

The simulation is part of BMW’s plan to use more artificial intelligence in production. Grüeneisl says machine learning algorithms can simulate robots that perform complex maneuvers to find the most efficient process. BMW wants to use the simulation over time to teach robots how to perform increasingly complex tasks.

BMW used a software platform called Omniverse, developed by chip maker Nvidia, to rebuild the production line in Regensburg. Last year, BMW said it was using an Nvidia AI platform called Isaac to train robots for certain new tasks.

“I’m very sure that in the future we can just put a new robot in this facility and say, ‘Okay, talk to the other robots and find the best way to produce this body,’” says Grüeneisl.

Manufacturers have long used computer simulations to hone their assembly lines. But Omniverse allows to simulate the entire production process with photorealistic details, and with physical properties such as gravity and different materials. It is possible to explain the production process from start to finish and see how changes to one part can have knock-on effects on another. It is easier to build a more complex virtual environment because different 3D models can be imported into the system. Omniverse uses an open file standard that is compatible with many computer-aided design packages.

The software also simulates human worker avatars grabbing parts and tools and assembling components to find the best procedure and minimize ergonomic issues. It could also ensure that fewer employees can do a particular job, says Grüeneisl.

“We’re doing AI simulation of how people move around the factory,” said Richard Kerris, general manager for Omniverse at Nvidia. He calls the project “one of the most complex simulations ever done”.

There is a growing interest in using AI to control robots and other industrial machines. Encouraged by recent advancements in AI, some startups are focused on letting robots learn in simulation how to perform devilishly difficult tasks such as grabbing irregular objects, technology that could eventually help automate a lot of ecommerce and logistics work. This often uses an AI approach called reinforcement learning, where an algorithm experiments and learns, from positive feedback, how a specific goal can be achieved.

“This is definitely the way to go,” said Ding Zhao, a professor at Carnegie Mellon University who focuses on AI and digital simulations. Zhao says simulations are critical to using AI for industrial applications, in part because it’s impossible to run machines through millions of cycles to collect training data. In addition, he says, it’s important for machine learning models to learn by experimenting with unsafe situations, such as two colliding robots, which is not possible with real hardware. “Machine learning requires a lot of data, and collecting it in the real world is expensive and risky,” he says.

Willy Shih, a professor at Harvard Business School who specializes in manufacturing technology, says the sophistication of simulation is steadily increasing, and he says simulation mainly saves time and money by preventing future manufacturing problems.

Shih says there is a lot of hype around AI for manufacturing, but adds, “There are a lot of uses, too” for the technology.

Nvidia CEO Jensen Huang discussed BMW’s use of Omniverse during his keynote at the company’s annual GTC conference, held virtually on Monday. Nvidia initially made graphics chips for gaming, but broadened its focus as these chips proved adept at training AI programs. The company has since jumped into several other industries where AI is important, including automotive and medical imaging.

Share this:

Related