@Bert, I hear you. I am very congnizant of your argument and I do agree. There are a lot of hurdles. And yes, indeed, this could be the automotive industry's 3DTV disaster.
But what I was trying to get at in this story is the following.
There is a self-driving car, and there is a semi-autonomous car.
The semi-autonomous car requires, as you pointed out, a driver's assistant.
In my mind, those two have always been two separate things.
But as I talked to more people, I realized many in the automotive industry view the semi-autonomus car (or ADAS) as a stepping stone for the ultimate self-driving car.
What I wanted to point out in the story is that the architecture decisions carmakers are making in 2014 are important because eventually that will become the basis for the self-driving cars. It almost doesn't matter when self-driving cars eventually come to streets. What matters is that carmakers need to get this semi-autonomous cars right first. And for that, there is still a lot of work to be done.
Larry, the truth is, today's autonomous cars either run on prescribed roads, or they require constant attention by the driver. The latter being, of course, more of a driver assist feature, than real autonomy.
I'm aware of these. I've read the articles. The cars cannot detect all threats, the driver has to intervene, and/or the roads have been cleared, are known to be free of potholes and construction zones, or tricky intersections, or blind corners.
So we can't extrapolate. Not yet.
I have also heard arguments about "why can't these autonomous cars read road signs?" And my response is, DO THEY? I've never argued that they can never do so, but do they today? Until they can, they won't be ready for prime time. The human behind the wheel won't be able to take a snooze or read a book.
All I'm really arguing is, unless one is in the employ of an auto company, there's no reason to downplay the obvious obstacles ahead still. Even if on-board car sensors are good at detecting small objects, even if vehicle-vehicle comms have been thoroughly worked out, that political quagmire of vehicle to infrastructure is still needed for unapologetic autonomous driving. Until someone can clearly explain why not, I can always come back with examples that will require driver attention.
Bert, I might be tempted towards your argument except for a few facts. Self-driving cars are in fact a reality today, albeit mostly still in the form of science projects. Google cars are not only logging serious miles, but they are also doing useful things in the process. This is being done on existing open roads, not on special closed tracks. This can be done, it just has to be made less weird.
Secondly, there are real tasks that can be done. 3DTV is a gee-whiz technology that to date has not provided enough real utility to make it worth the accomodations that it requires. Why do we still have to wear glasses? If it was easier it would be more likely to be the default decision.
Finally, it is not an all-or-nothing proposition. Cruise control is a start. Adaptive cruise control is a measurable improvement. Lane assist is another, as is self-parking. It doen't take too many more steps before we are handing over the keys.
Junko, do you remember all the nagging articles about 3D TV coming? It's coming! it's coming! And yet, no one was asking! I think it is plainly obvious now, if not back then (to some people).
You've heard it here first: the same applies to self-driving cars. The auto manufacturers seem to be the only ones trumpeting this recurring fanfare. And in order to make the idea seem plausible, the auto manufacturers HAVE to pretend that no third party, slow moving, bureaucratic infrastructure effort is needed. Because if the auto manufacturers said that it will takes bureaucratic/political involvement to make these dreams come true, everyone's eyes would glaze over.
So I ask readers once again. Go through the long list of road signs and signals, and pavement markings, and ask yourself these simple questions: (1) Would an autnomous vehicle also need that information? (2) If not, why not? (3) If yes, how would they get the info?
Road signs and signals are used when the driver needs more than his own eyes and ears, to make the right decision. Traffic signals at intersections. Warnings about what is ahead of a blind turn. Merging lanes. Yield signs. Road crews at work. And on ad nauseam.
If the infrastructure needs to communicate these things to a human driver, I don't see how the on-board sensors of an autonomous vehicle would be able to get by without it.
Just as 3D TV hype was used to create more demand for the flat TV market, my bet is that the autonmous vehicle hype is used to combat the flat auto market, which you reported on recently.
Every time when we write about self-driving cars, we find our reader community is divide in two: Those who believe it ain't happening; Others who think that it will come, perhaps over time, and only in limited places (i.e. highways), but it will take "a long time."
I think our readers' instincts are correct; and yet, it is clear that nobody in the automotive industry is twiddling their thumbs waiting for the right moment to arrive. They find a big opportunity in rolling out semi-autonomous cars, sooner than later.
That means setting up a big goal like the launch of self-driving cars in 2020 for the commercial market. In turn, it means, self-driving cars that need to be tested on the road must get done by 2017.
2017? Now suddenly that feels so soon to me.
How ready carmakers really are (and chip vendors who supply some of the key technologies to OEMs and tier ones) is unclear. But clearly, the amazing race is already on, and the second half of 2014 seems like the first mile stone for silicon vendors waiting for car OEMs' platform decisions.
While there are a number of ways in which driving a car is more dangerous and problematic than flying an airplane, it should be noted that flying an airplane is the more complicated effort. Even if pilots are professionals, the fact that autopilot has existed for as long as it has--even in commercial flights which involve hundreds of peoples lives--shows that the notion of a self-driving car is not so far-fetched. If a plane can fly itself, a car should be able to--at some point--drive itself.
NASA's Orion Flight Software Production Systems Manager Darrel G. Raines joins Planet Analog Editor Steve Taranovich and Embedded.com Editor Max Maxfield to talk about embedded flight software used on the Mars on EE Times Radio. Live radio show and live chat. Get your questions ready.