Velodyne says its sensors, which once cost $75,000, are declining in cost enough that the new generation will ultimately cost about $500. And the sensors, once comparable in size to a fried chicken bucket, are now roughly the size of two stacked hockey pucks.
Graphics chip maker Nvidia Corp. introduced a lunchbox-size computer for autonomous driving called the Drive PX 2, packing the power of 150 MacBook Pro laptops. Nvidia said it can be quickly trained to identify objects more effectively than computer algorithms coded by human programmers over the course of years.
This is a huge development, said Glen De Vos, vice president of engineering at Delphi, because "a few years ago, you wouldn't have been able to fit all the computers in the back of the car."
Delphi, which is working to make Nvidia's computers production-ready as the brains of Audi's self-driving cars, presented its own vision last week of how self-driving cars could make people feel at ease with their robotic chauffeurs and better their lives.
In such a car, drivers would have personalized profiles stored, complete with their names and faces, as if on a social media site.
"Safety check," the car says as it prepares to leave, showing through a graphical interface that all sensors are functional. And then: "Ready to drive."
Later, when it needs to hand control back to a human driver, the car addresses that person by name. "Dave," it might say, "prepare to take over."
While Delphi's car is driving, passengers can be prompted to pick up a friend or a potential car-sharing customer. "Do you have time to pick up Thomas?" the car asks. And when a friend is driving in the vicinity, Delphi's concept car points out that "Mike's Cadillac" is nearby.
Large suppliers like Delphi are determined to build software for self-driving cars that seduces humans. It's a matter of survival. These suppliers have historically made money selling discrete hardware modules, and much of the value of a self-driving car will come from software instead.
Helmut Matschi, head of the interior division at Continental AG, gave the example of a technology under development that would grant secure access to the trunk of a car, so a delivery person can open the trunk and leave a package inside.
It requires little hardware, but Continental sees huge benefits. It's not that difficult to imagine that a self-driving car could someday drive to the grocery store, open its trunk to accept the preordered groceries, and then drive home.
"You can imagine how our lifestyle could start to change after that," Matschi said.
The first wave of cars with autopilot may not have these features. But as autopilot becomes more commoditized, standardized and regulated, these will be the features that allow companies to set themselves apart.
In other words, De Vos said, they are "the next wow factor."