The Uber vehicle, which uses Nvidia’s technology, struck and killed a woman crossing a street in Arizona. Video of the accident suggests the vehicle did not “see” or attempt to avoid the pedestrian, raising questions about the technology in the Uber vehicle and concerns about autonomous driving technology in general.
Following the accident, Uber suspended North American tests of its autonomous vehicles. Up until now, Nvidia had been conducting testing of self-driving cars worldwide – including in California, New Jersey, Japan, and Germany. The company decided to suspend testing as a cautionary measure until more is known about the causes behind accident.
“We suspended testing not because we know that we’re doing something wrong,” Nvidia CEO Jensen Huang said during his keynote address at the company’s annual GPU Technology Conference in San Jose. “We are already using all of the best techniques we know how to keep things safe. [W]e stopped for a second … for a simple reason – good engineering. Somebody has just learned something. There is an incident of great importance. We should stop and see what we can learn from them.”
Nvidia’s DRIVE platform is used by over 370 companies developing self-driving technology, including automakers and robo-taxi companies, and makers of self-driving hardware. According to Reuters, Nvidia CEO Jensen Huang expects investments in self-driving technology to rise – not fall – in the wake of the Uber fatality.
The latest reports on the accident suggest that the Uber vehicle involved – a Volvo SUV – was equipped with fewer sensors than had been available on the ride-sharing company’s previous self-driving fleet of Ford Fusion cars. According to reports, when the company switched over to the new vehicles it also decided to scale back on the safety sensors used to detect objects in the road.
The new Uber driverless vehicle, say reports, uses only one roof-mounted LiDAR sensor compared with seven LiDAR units on the older Ford Fusion models that Uber had been using. According to interviews with former Uber employees, this reduction in sensor units resulted in a blind zone around the perimeter of the SUV that cannot fully detect pedestrians.
Velodyne, the maker of the LiDAR sensor used, has acknowledged that with only the rooftop LiDAR unit, there is a roughly three meter blind spot around a vehicle. Autonomous vehicles operated by Waymo, for example, have six LiDAR sensors, while General Motors’ vehicle contains five, according to information from the companies.
In January, Uber and Nvidia announced that the ride-sharing company had selected Nvidia technology for the AI computing system in its fleet of self-driving vehicles. At that time, Uber had said that its AI self-driving cars and trucks had already driven more than 2 million autonomous miles, and had completed over 50,000 rides.
Uber self-driving fleet to use Nvidia technology
Volvo to build tens of thousands of self-driving taxis for Uber
Nvidia invests in autonomous trucking
Uber buys autonomous truck startup, partners with Volvo in self-driving car effort