Nvidia is Making Safer Robots for the Factory Floor

Nvidia’s interest could help bring AI onto the factory floor for more than just maintenance purposes.

Photo by Simon Kadula on Unsplash.

Sign up to uncover the latest in emerging technology.

Nvidia wants its manufacturing robots to lend a helping hand – or get out of the way.

The chip giant is seeking to patent a system for “reactive interactions” for robotic applications and automated systems. Nvidia’s system uses machine learning to allow for “predictive control of a robot” while it’s performing certain tasks, allowing it to react to unexpected movements to avoid hurting the person working with it. 

First, Nvidia’s robotics system picks up “current state information” about its surrounding environment to determine what tasks it can perform and how it can perform them. For example, if the robot is meant to repeatedly take an object in the manufacturing process from a human hand, it will come up with a “set of potential grasp options,” or ways that it can perform this handoff without incident. 

Those grasp options or other task options, as well as environmental state information and potential constraints or obstacles in performing the action, are then fed to what Nvidia calls a  “path planning and optimization process.” Using a model, this process plans out the most optimal path of motion while avoiding collisions or obstacles, and minimizing “cost factors or functions” such as additional unnecessary movements, thereby saving energy. 

When going through the motions, this system also modifies or optimizes its actions as it performs them, using computer vision to keep a constant watch on environmental factors and “contact detection” with the human it’s working with. This system also is aware of the entirety of the machine in its movement, not just the robot hand performing the task.

Nvidia noted that existing systems generally aren’t “smooth, intuitive, or reliable, which may result in a human making a quick or unexpected movement,” which, along with potentially causing injury, can cause a user to block the camera being used to control the robot’s movement.

Photo via the U.S. Patent and Trademark Office.

Most of the time in industrial settings, machines are fenced off from human workers to avoid collisions or accidents, and cannot operate if a person is within a certain radius, said Rhonda Dibachi, CEO of Manufacturing as a Service company HeyScottie. This is because these machines can’t sense the world around them, and “don’t know who’s there and who’s not,” she said. 

“That limits their effectiveness,” said Dibachi. “Because if you can’t interact with a robot, they’re just another machine, and you can’t interact with that machine in real-time.” 

Nvidia’s patent provides just one example of companies attempting to work around this problem. Manufacturing robotics firm Fanuc and electronics company Panasonic both have patents in the works for collision avoidance technology, Dibachi noted. Where Nvidia’s invention differs, however, is that it’s essentially a “traffic cop,” she said: Rather than attempting to predict human movement, or giving the robot a set path in which it can move, Nvidia’s tech simply watches the person it’s working with and moves accordingly. 

“A lot of the other patents say the problem with collision avoidance is that there’s a myriad number of calculations that you have to do … and that slows it down” she said. “You have to balance collision avoidance with task completion. This is not a timid robot.” 

Nvidia isn’t currently a leader in the industrial machines market, but its strength in AI could change that. The company is currently the darling of the AI industry, with its data center revenue (which encompasses its AI work) reaching a record of $10.32 billion in the recent quarter, up 171% year over year. That kind of industry grip gives the company room to play around with slipping AI into different industries, whether that be industrial machines or medical x-rays. Plus, with a predicted market size of over $1 trillion by 2032, factory gadgetry could be a lucrative bet.

“AI currently has a foothold in manufacturing with predictive machine maintenance,” said Dibachi. “But there are so many other areas that you can play with. This could allow Nvidia to not just be the maintenance guy.”