Robots have since evolved and now exhibit far more human-like qualities, with movements that have become softer and subtler. However, that has been true mostly for humanoid robots, which are a tiny minority compared to the industrial robots that have helped manufacture our goods — such as cars — for decades.
Around 3 million robots work in factories around the world, with about a third of those in the automotive industry, according to an industry body. Now, a company called Micropsi Industries is looking to make even industrial robots closer to humans. “We make a control system that allows industrial robots to do things that without our software they couldn’t do,” says Ronnie Vuine, Micropsi’s founder, “which is essentially having hand-eye coordination and adapting to changing conditions in the environment as they do their work in a factory.”
The company’s first product, called MIRAI, uses artificial intelligence (AI) and cameras to train robots to perform tasks that would be impossible via traditional, pre-programmed movements.
Vuine became interested in AI while a student at Berlin’s Humboldt University in the 2000s. “There was a working group that was interested in how machines learn in the real world when there’s no engineer around to tell them what to do, but they just need to sort out and find out what to do to survive. How would you do that? So that’s been our research interest.”
Vuine says that AI was distinctly unfashionable at the time, but when Google purchased AI company Deep Mind in 2014, it showed the team how AI had become more mainstream and was the motivation they needed to push forward. Micropsi was founded in the same year.
The company is now developing its products for various brands of manufacturing robots. “By far the most advanced industry when it comes to deploying robots at scale is automotive,” Vuine says. “Cars are the most complex artifact we make at scale as humans. We also make planes, and they’re more complex, but we don’t make as many of them. Cars are just the most advanced automation game we play.”
Demonstrating the technology to CNN, MIRAI allowed a robot arm to pick up a thin computer cable dangling from a person’s hand and plug it into a switch — a delicate task that is considered too hard to manually engineer a robot to do. Vuine says MIRAI can teach the robot how to do this in about an hour, with a human tutor. After that, using cameras and lights to see what it’s doing, the robot can perform the task by itself. “The cable, of course, jiggles about, so you can’t fully know and predict where that’s going to be, but the robot will reliably pick it and then insert it,” he says.
This opens up options for automation to carry out tasks previously handled by humans, which could prove especially useful in producing electric cars. “Automotive is moving to electric. There are much more cables to be plugged in,” says Vuine. “Of course, it’s important in electronics, where you have ribbon cables (to connect to circuit boards). All of these applications couldn’t be done with robots (previously). You would have to use a human, or you couldn’t do it at all, and would need to redesign your product for manufacturability.”
Having recently moved its headquarters from Berlin to San Francisco, the company is now looking to expand from cars to other products, like power tools and white goods, as well as other fields altogether, like logistics. In the future, the system could power humanoid robots, too. “The software that drives the robot would be very much applicable outside a factory, in a service robot that does your dishes,” Vuine says. “We sometimes do playful demos that show these capabilities.”
The hurdle to that expansion is not the software, he adds, but the robots themselves. “Robots are not made of soft material like humans. They’re made of metal, so it hurts if they hit you. You need to go very slowly, and you need to put lots of safety around, and lo and behold, you’ve created a machine that’s too expensive and too cumbersome to live in your home. We just haven’t solved that yet.”
About The Author
