The car company’s expertise in AI could help it design a working prototype, but delivering a reliable product on schedule will be challenging
30 December 2021
In August 2021, Elon Musk announced that Tesla would build a humanoid robot designed to “eliminate dangerous, repetitive, boring tasks” and respond to voice commands, promising to show off a prototype in 2022. Can the company deliver on Musk’s goal?
Tesla has achieved a great deal since Musk founded the electric car firm in 2003: building a valuation of $1 trillion, selling in excess of half a million cars and installing a global network of more than 2000 charging stations for them. But there have also been failures and delays.
Musk promised to have a million self-driving taxis on the road by 2020. He has long touted the imminent arrival of full autonomy for his cars; scheduled a Tesla lorry for production in 2020 and a Cybertruck soon after in 2021. All of those deadlines have been or are due to be missed. Musk himself has admitted that he lacks punctuality but insists that most of his predictions come to pass eventually.
The robot, referred to as Optimus inside the company, will be 173 centimetres tall and weigh 57 kilograms, and it will be able to carry a cargo of up to 20 kilograms, according to Musk’s presentation in August.
He said much of the technology in Tesla’s self-driving cars is applicable to humanoid robots and should give them a head start. “Tesla is arguably the world’s biggest robotics company because our cars are like semi-sentient robots on wheels,” he said. “It kind of makes sense to put that onto a humanoid form.”
Tetsuya Ogata at Waseda University in Tokyo, Japan, believes that engineering of the robot must be progressing well, or the company wouldn’t make such bold claims. But he expects that it will not only run into AI problems, where Tesla certainly has a lot of experience, but hardware problems, where it doesn’t, because humanoid robots are much more complex than cars.
“It’s very difficult to develop robot hands that can perform the same tasks as a human,” he says. “How to reproduce senses that allow tactile feedback is also a big problem.”
Zhongyu Li at the University of California, Berkeley, says he admires the vision, but thinks the deadline is “very ambitious”. He expects Tesla to hit its target of demonstrating a prototype of some kind, but perhaps encounter problems bringing it to market.
“Getting a prototype to walk for some short demos is not that challenging for their clever engineers, but getting humanoid robots to reliably operate in daily life is another story. It needs reliable hardware, a robust control algorithm that can prevent the robot falling, recover from a fall, and detect and avoid obstacles, and these may take years,” he says.
Others believe that the technology is possible, but not in the slender form that Musk promises. Florian Richter at the University of California, San Diego, points to the Atlas robot from Boston Dynamics which can run, jump and perform a range of tasks, but which also has a bulky body and a large backpack-style battery pack.
“They have a lot of work to do. I think their goal of a hardware prototype within a year is totally feasible, but with probably half of their desired power and some sort of weight compromise,” says Richter. “They also should be able to get it walking around on flat surfaces pretty quickly, but other human-level tasks like grasping will take a few years of research and a lot of innovation.”
Neither Tesla nor Elon Musk responded to a request for interview.
More on these topics: