Nvidia Corp is working to make humanoid robots safer around people, arguing that they’ll need to handle split-second decisions before they can be trusted to work closely with humans.
The chipmaker is offering software and semiconductors that will allow humanoids to enter the workplace and truly interact with people – even making physical contact if necessary. Nvidia’s Halos software, developed from systems used for self-driving vehicles, will be the basis of computers that give robots a much better awareness of what’s happening around them, the company said in a statement on June 22.
Nvidia and its Silicon Valley peers are racing to develop technology for robotics, billing it as the next big market for artificial intelligence. The machines will evolve into a market with billions of devices, tech executives predict.
Humanoid robotics – technology designed to resemble people, with heads, torsos and limbs – will generate US$200bil (RM829.8bil) in revenue by 2035, according to Barclays estimates.
But the industry will have to overcome some hurdles. Current safety systems force robots to stop or reduce speed when there’s a chance they might come into contact with a human. That makes them less productive. The safeguards also hinder cooperative activities, such as handing an item to coworkers or helping them lift something heavy.
"If you think about safety in the context of a traditional robot, it is basically you need to put it in a cage or you need to have sensors which can detect there’s an obstacle and the robot comes to a stop,” said Amit Goel, a senior director of product management at Nvidia. "But that is not enough for a humanoid robot.”
Nvidia will provide technology for robots like Agility Robotics’ Digit, allowing the machines to make their own decisions based on analysis of what’s going on around them. That will require hardware that supports safety awareness, according to Nvidia. The company is offering up Halos as an operating system that runs on top of its IGX Thor hardware.
The machines will also have connections to external sensors. An autonomous forklift, for example, will be able to tap cameras in a warehouse to see what’s around a corner and decide whether it can keep going full tilt or slow down to avoid a collision, the company said.
Nvidia, the world’s most valuable company, generates much of its revenue from chips used in AI data centres. But it has gradually been expanding into other fields – often offering technology for free – in an effort to broaden the use of artificial intelligence.
In another step aimed at accelerating robotics adoption, Nvidia created a lab where robot makers and customers can run safety tests before going to regulators to get the necessary certifications. Nvidia’s engineers will do pre-inspection work and help with engineering tweaks as needed.
The constraints for robots are different than they are for autonomous vehicles, which generally only have to avoid making contact with people and other objects. A humanoid has to be more flexible, said Pras Velagapudi, chief technology officer of Agility Robotics. That company’s humanoid has been deployed to customers including at Toyota Motor Corp manufacturing facilities in Canada.
"Safety design has to be much more advanced because you need to actually reason about, based on what you’re seeing in the environment, what things you can and can’t touch, move, or exert force on,” Velagapudi said. "And the scales of these forces can’t be small. You can’t sort of avoid the problem by making your robot so weak that it’s just going to brush up against a person and sort of have no impact, because any robot that’s that weak is also not going to be able to do useful work.”
Though engineers are still confronting the technical challenges of creating artificial humans, the opportunity is big enough that the industry can learn in stages. Machines that take care of people in their homes will come later, built on the back of factory robots, Velagapudi said.
"We’re starting in warehouse and logistics, which is one of the most structured environments, but already has many billions of dollars worth of addressable market,” he said. "We can move into retail, we can move into healthcare, we can move into construction. And each of those is incrementally more difficult, but is on its own a huge industry that can support a huge number of robots.” – Bloomberg
