The auto industry is going all-in on robotics. The automotive sector has become the number one adopter of industrial robots, making up 33% of all installations in the US last year, according to a 2024 study by the International Federation of Robotics. Key reasons include transitioning to more electric vehicles as well as labor shortages.

Automakers employ a variety of robots that range from collaborative robots (or “cobots”) to six-axis robotic arms. But the latest—and buzziest—tech is the humanoid robots that are walking their way onto the factory floor.

Recently California robotics company Figure, which raised $675 million earlier this year, released a video demonstrating how its Figure 01 humanoid operates in a BMW factory; in August they released Figure 02, which can have natural speech conversations thanks to an OpenAI integration and is being tested by BMW now. And this summer, Tesla showcased its Optimus Gen 2 at the World AI Conference in Shanghai and says it plans to use it in Tesla factories soon; Elon Musk has suggested that Optimus has the potential to become more valuable than Tesla’s other products combined. Mercedes-Benz and China’s Nio are also experimenting with humanoids in their factories.

Previously, Hyundai purchased Boston Dynamics in 2021, although there’s no word on BD’s Atlas humanoid being used in Hyundai factories just yet—more on them in a minute.

How robots are evolving in car manufacturing

“Robots have been around for many years,” says Jose Favilla, global leader for Industry 4.0 at IBM. “The key going forward is evolving them from performing repetitive tasks to be more ‘intelligent’—having the ability to perform a wider variety of tasks according to the different conditions they will face in a plant environment.” Their interactions with plant systems would be more efficient because they won’t have to access systems through screens like people do, he adds. “They’ll get direct digital messages on potential problems, be able to look up previous maintenance records, and use their own or access other intelligent systems to fix a machine.”

“Humanoid hands are a major topic of research and a breakthrough may come in the next few years,” says Dr. Yue Hu, head of the Active & Interactive Robotics Lab at the University of Waterloo. “Since [humanoids] are also equipped with legs and are generally shaped like humans, there is no need for factories to reshape their environment.” Auto plants currently take up a lot of real estate because they need to house large machines that are dedicated to only one task, Hu notes. “This may lead to a more sustainable future and smaller factories where general-purpose humanoids could perform multiple tasks in a highly-reduced space.”

Legs vs. wheels

“Legged systems have the advantage of being able to move in a more versatile way through complex environments, such as climbing stairs and [stepping] over obstacles, unlike wheeled robots,” Hu says. But they are still prone to instability, she adds, which is where wheeled robots have an advantage.

Robotics companies themselves have strong opinions on the matter. “Two-legged walking robots are essentially useless given current actuator technology,” says Geordie Rose, CEO and co-founder of Sanctuary AI, which manufactures the wheeled humanoid Phoenix. In April, the company announced a partnership with Canada’s Magna, a worldwide car parts supplier with 343 factories.

“Any company building two-legged robots will be limited to niche use cases, such as moving empty totes, that are better performed with special-purpose robots,” Rose continues. “This is because two-legged walking is inherently unstable, power inefficient and highly restricts the quality of the motors in the upper body because of the need to make the upper body as light as possible. Mobile dexterous manipulation robots for work require wheeled bases, which is how our systems are architected. Wheels aren’t just a good idea—they are required for a mobile manipulation product.”

Agility Robotics, which manufactures the Digit humanoid, sees well-designed legs as essential. “Digit has been designed to take on a diverse set of tasks and its legs are a big differentiator,” says Daniel Diez, Agility’s chief strategy officer. Digit is currently being used in warehouses for Amazon and Spanx; Agility is working with car companies right now, but they aren’t saying which ones yet.

Agility studied the biometrics of how animals walk and run and concluded that bird-like legs with giant ankles and toes were the way to go (watch Digit in action here). The avian-styled legs mean the robot can squat straight down and fold up for storage. “The design of Digit’s legs allows it to move around much in the same way a person does,” Diez adds. “This enables it to perform a more diverse set of complex tasks and is a big reason we are able to commercially deploy our robots today.”

Two legs or four?

Back to Boston Dynamics: This spring the robotics company retired its decade-old Atlas humanoid and debuted a new version soon afterward. A viral video of the updated Atlas has amassed six million views, and the robot’s very un-humanlike motions in the clip prompted Elon Musk to suggest that it was possessed.

Atlas is still in the experimental stages. In the automotive realm, the company’s real work dog is Spot, the quadruped robot that has been roaming car factories for a while now. In fact, IBM collaborated with Boston Dynamics to develop AI-based solutions for Spot in 2021.

Spot’s latest joint venture is with BMW. In a BMW plant outside of Birmingham, England, Spot autonomously conducts industrial inspections and feeds back data so that teams can do predictive maintenance. One example is its acoustic sensor, which can catch a compressed-air leak before it becomes a pricey repair (a single quarter-inch leak can cost up to $8,000 per year, a rep for Boston Dynamics tells us). The robots can be followed on a facility map using specially-designed fleet management software that lets operators plan routes, schedule missions, monitor progress and control robots remotely.

Do animal-based robots have an advantage over the humanoids? “Each one has pros and cons,” Favilla says. “Humanoid-based robots can be more versatile and easier to integrate into existing plant environments. The animal-based robots can be better for specific tasks, such as crawling within small places, which would be difficult for humanoids to do.”

The human + humanoid factor

Because humanoid robots look, well, human, is there any worry that their living, breathing coworkers might start anthropomorphizing them?

“I think machines introduced in factory settings where they have repetitive, dedicated tasks will be perceived like other technologies that have been introduced in similar settings,” Hu says. “Maybe at first there will be a transition period, but it would be normalized faster [compared] to other environments that require more social interactions.”

Hu’s research examines human-robot interaction (HRI) and looks for ways robots can perform their tasks while appearing socially acceptable. She notes that the way people perceive robots socially is a crucial factor in caregiving or education settings, but “in a factory or automation setting this would be largely mitigated.”

Will humanoids take human jobs?

“Robots often become a lightning rod for automation and job insecurities,” says Diez. “It’s understandable, especially when the technology takes the form of a humanoid. When thinking about humanoid robots or automation in general, it’s important to remember … that technology advances do not decrease the number of available jobs. It has simply changed the type of jobs available.”

“There are currently 8.2 million job openings in the U.S., but only 6.5 million unemployed workers,” says Rose. “The world needs more workers, especially in the kinds of jobs that are too dull, dirty or dangerous for people to do.”

“Companies need to automate their way out of this problem,” Diez adds, “or the success of the business itself could be at stake.”