The December 2022 issue of IEEE Spectrum is here!

Close bar

Totally Autonomous Cars May Not Evolve From Semi-Autonomous Ones

Reason: nobody wants to pay for systems that you still have to supervise

2 min read
Totally Autonomous Cars May Not Evolve From Semi-Autonomous Ones
Illustration: Getty Images

Robocar theorists fall into two camps: those who think driver-assistance features will evolve to perfection and those who think robocars will emerge in one fell swoop. The car makers favor evolution, Google favors revolution.

Now comes an interesting new argument for the revolutionary side, from Brad Templeton, a dot-com veteran who writes about autonomous cars. He bases his position on two observations: first, that driver-assist programs require the driver to remain vigilant, which people are bad at doing. And second, that such programs just haven’t been compelling enough to get many people to pay for them. Reason: you still need to supervise them.

Templeton cites the example of adaptive cruise control (ACC), which uses sensors and software to maintain a set distance between your car and the one in front: 

Uptake on it is quite low — as an individual add-on, usually costing $1,000 to $2,000, only 1-2 percent of car buyers get it. It’s much more commonly purchased as part of a ‘technology package’ for more money, and it’s not sure what the driving force behind the purchase is.

Legal and other considerations suggest that slow-and-steady evolution will never get these systems over the hump to full autonomy, he argues. Instead, a clean break is needed, one that doesn’t count the cost (at least on the first few cars that are sold). “The first car is the car of greatest risk, you will do all you can reasonably do to make it as safe as you can,” Templeton argues.

That will mean loading it to the gills with laser ranging systems (lidar) as well as radar, sonar, cameras, GPS, minutely mapped roads, and on and on. As for those driver-assist programs, bits and pieces of them may make it into the final product as self-contained modules. That way, there’d be a fully tested backup system—one that can at least pull the car over to the side of the road.

The first true robocar would thus be the first of its kind, and not the last in a series. And though this is, of course, the Google approach, there is no particular reason to expect it to bear fruit in five years’ time, as Google has asserted.

The Conversation (0)

Chinese Joint Venture Will Begin Mass-Producing an Autonomous Electric Car

With the Robo-01, Baidu and Chinese carmaker Geely aim for a fully self-driving car

4 min read
A black car sits against a white backdrop decorated with Chinese writing. The car’s doors are open, like a butterfly’s wings. Two charging stations are on the car’s left; two men stand on the right.

The Robo-01 autonomous electric car shows off its butterfly doors at a reveal to the media in Beijing, in June 2022.

Tingshu Wang/Reuters/Alamy

In October, a startup called Jidu Automotive, backed by Chinese AI giant Baidu and Chinese carmaker Geely, officially released an autonomous electric car, the Robo-01 Lunar Edition. In 2023, the car will go on sale.

At roughly US $55,000, the Robo-01 Lunar Edition is a limited edition, cobranded with China’s Lunar Exploration Project. It has two lidars, a 5-millimeter-wave radars, 12 ultrasonic sensors, and 12 high-definition cameras. It is the first vehicle to offer on-board, AI-assisted voice recognition, with voice response speeds within 700 milliseconds, thanks to the Qualcomm Snapdragon 8295 chip.

Keep Reading ↓Show less