Slashdot Mirror
Volvo Testing Autonomous Cars On Public Roads
cartechboy writes: "Multiple automakers have already committed to selling autonomous vehicles by 2020, but only a handful of them have actually started testing and developing them. Now Volvo is putting self-driving test cars on real public roads in Sweden among other, non-autonomous traffic. 'The test cars are now able to handle lane following, speed adaption and merging traffic all by themselves, Volvo engineer Erik Coelingh said in a statement. 'This is an important step towards our aim that the final Drive Me cars will be able to drive the whole test route in highly autonomous mode.' The goal for the Drive Me project is to deliver 100 autonomous cars to customers in Gothenburg by 2017."
It's the mode created by lawyers to maintain that you are responsible for the vehicle's operation. because if the software fails, it's Volvo's fault. Imagine you drive through a oil slick, the vehicle takes a corner too fast before becoming aware of the reduced traction. In automatic mode, it's the car's fault, in "highly autonomous mode" it's still yours.
Slashdot's rate-of-post filter: Preventing you from posting too many great ideas at once.
A self-driving car probably won't be as good at anticipating stupid drivers as a really good human driver.
Why should SDCs have to be better than a "really good human driver" rather than an "average human driver"? As long as SDCs are better than an average human, they are a net win. Also, bad drivers will likely be the first adopters, both because many bad drivers don't enjoy driving and because they pay higher insurance premiums.
And how many cars has Google ever sold? With $150k of extra equipment bolted on, they neither know nor care about the realities of car production. Volvo is intent on actually putting these things in the hands of real paying customers. To Google it's just another cool toy to play with. Hire some good AI people and have fun - who cares if it has nothing to do with their business.
More importantly, there are too many unknowns about the Google cars (despite them having been around for years), and about the Volvo cars. How do the cars perform in the rain? Snow (I'm betting Volvo thinks about that more)? How often does the autonomous feature kick out and require the human driver to take over? How much time does the human have to respond? In testing, how often do the human drivers take over control because they anticipate that the car is entering a situation where the autonomous control will have problems? Etc., etc., etc. Without that information we have no way of knowing if it will soon be a viable technology, or it's just a cool demo. However, Volvo is planning to sell these things in 2017, so they're probably working hard on a realistic approach. What year has Google claimed they'll ship product? Sergey's "5 years"? Sure, and I predict we'll have fusion power in 10. Honest. I hired some cool physicists who say they'll have it done by then.