MUNICH – Jensen Huang isn’t bothered by the differences between the five levels of autonomy that the US-based Society of Automotive Engineers (SAE) has defined for driverless cars. “As long as it works,” the Nvidia CEO told reporters here Thursday.
Huang, who was in town for the chipmaker’s GTC Europe developer conference, said he expects vehicles with level 2 autonomy – where the driver remains in control – to be around “for a long time.”
Robo-taxis with level 4 or 5 autonomy – where human drivers no longer are in charge – are likely to start trials next year and “10s of thousands” can be expected to be on the road by 2020.
Current level 2 technology is not particularly good, Huang said. Hence efforts by automakers to develop so-called level 2+ functionality, which should improve the workings of automatic controls.
Volvo Cars announced Wednesday it was adopting Nvidia technology for its next-generation models, which will have level 2+ automation. (See story at: https://tinyurl.com/y9cc2lt6)
Huang said it will be much easier to develop robo-taxis than autonomous passenger vehicles, because the former will drive on pre-defined routes while the latter will be expected to go anywhere. “Robo-taxis only have to serve in a constrained environment,” he said.
In tomorrow’s cities, Huang predicted, there will only be a relatively small number of passenger-operated cars. “I think most cities will end up having mostly taxis and public transport,” he said. “Automated transport will help cities reduce conguestion, no question.”
The transformation of personal mobility now underway provides “gigantic opportunities” for the global automotive industry, Huang said. That is because, with new business models, carmakers will be able to focus on the 10 trillion miles driven each year, rather than on the 100 million cars sold.
Nvidia’s “accelerated computing” offering, which addresses the entire software stack rather than just the chip, is designed to provide the car industry with the vastly higher requirements that come with increasing automation. “Safety is computing power,” Huang said. And that computing power also is essential to make cars more intelligent, he added.
A software-and-sensor solution should come to the rescue in situations where level 2 or level 3 technology needs to return control to a human driver who may not be paying attention anymore. In such situations, Huang said, artificial intelligence can help. “Driver monitoring is really important,” he said. “AI will tell you if you fall asleep.”
Is the auto industry up to the task when it comes to implementing such radical changes to a tried and tested approach to mobility? “The biggest challenge for the auto industry is developing the software capability,” Huang said. Continuous over-the-air software upgrades and improvements to the vehicle over its lifetime will become standard, he predicted.
That’s a big change that will alter the relationship between automaker and car user in a positive way. Said Huang: “The benefit is that you’re connected to your customers.”
-By Arjen Bongard