As far as I'm concerned, fully autonomous cars will never happen. A human driver can drive off-road and do anything he wants in unmapped places. In order for a machine to be able to do the same, it would require a full AI: not going to happen.
The best I'm hoping for is a car that keeps driving itself on known roads, and asks for assistance from the human driver when it enters or approaches a difficult situation (detects a deadlock with other autonomous cars, needs to go off-road/unmapped-path, or even through a difficult intersection)
This will enable the driver to take a phone call (or whatever) 99% of the time, drive through a tough spot for 5 seconds and resume the autopilot.
The machine is never bored and never falls asleep, and the human driver can take over for short periods of time when needed. But to make sure the autopilot doesn't do anything stupid, you need a robust ADAS in order to garantee the safety of the car, passengers and environment. That's what the ADAS is for. Put it in "active mode" when the machine is driving (actually prevents the machine from doing something stupid like running over an old lady or crashing into a tree) and on "warning mode" when the human takes control (so that he can do something "dangerous" to clear a difficult situation if needed like driving "far too close" to an obstacle or ignoring some rules in order to resolve a deadlock)
@DanielMast: In 30 years, a robot will be driving antique cars built today.
They should organize the first test drives in Sheffield, England.
My 83-year-old mother recently purchased a Smart Car (or something similar). Now she happily tootles around all over the place. You can see when she's recently driven past somewhere by observing the "civilians" climbing down from trees and extracting themselves from bushes and hedgerows.
By the time your robot comes along, the combination of my mother and evolution will have created a city full of people with hair-trigger reflexes who can leap out of the way of a robot-controlled car at a moment's notice LOL
I think it depends on the mindset of the manufacturer.
Cars programmed by Toyota, or any auto manufaucturer for that matter, I would trust lesss than a system programmed by, say, Google.
The auto industry has management that doesn't get this new fangled software aspect. They are too used to their old methods and will be making silly design decisions and staring at their watches expecting the software to be cranked out the same way they expect fenders to be cranked out.
I used to work for a company that was dragged kicking and screaming all the way into the computer age and by God they weren't about to change their methods of project management. Which led to some spectacular failures that were blamed on the programmers.
While I agree that the days of all vehicular traffic being composed of autonomous vehicles is far away, I don't agree that there will need to be uniform standards worldwide on those vehicles. I think the rollout will come in stages. It might start with a highway lane being made available only to autonomous cars, and autonomous operation only allowed on highways. Then once a level of comfort with the vehicle performance is achieved, it might open to city traffic, then suburbs, then rural areas. Or rural first, but in stages regardless.
And if crossing a boundary means that different regulations are in effect, well, the car knows where it is and can be instructed to stop at the border and insist that the human take over because it is not allowed to drive in the place they are entering. No need for uniformity, just the ability to adapt to the different requirements.
We worry a lot about possible failure of our autonomous systems, and in these early days of the technology perhaps those concerns are valid. But I am convinced that the systems will ultimately perform better than the average driver and a whole lot better than some of the yahoos that infest the roadways around my home. Ultimately, I think autonomous vehicles will prove far safer in passenger miles per accident than with human drivers, at which point autonomous will become mandatory rather than special case.
Absolutely agree with Junko that fully autonomous car is 30+ years away. There are so many hurdle ahead even we are technologically ready. Just a couple days ago, Google self driving car has hit 1 million miles. It is an impressive milestone. However, it is said as self driving. It doesn't mean there is attendence. There is a person sitting in the driver seat just in case. Why is there a driver in a driverless car?
Folks in transportation department are typically conservative. There is a good reason for it - better be safe than sorry. To get those folks to agree a fully autonomous car without a driver actively monitor the environment isn't an easy task. Prove of concept and confirm with mileages. Is 1 million miles enough? Probably 10 or 100 million will give the general public more confidence. A billion will surely impress the regulatory agencies.
In addition to regulatory restriction, to sort out the responsibility in case of accident isn't an easy task. Insurance, car makers and owners will all need to agree a term. A term and condition will definitely require attorney expert in the area to build.
Next, the general public, the market need to have enough confidence and demand of the product. Toyota prius takes year to penetrate the market. The change is simply the addition of alternative fuel and the change of drivetrain. Imagine the willing of people adapting the technology. In general, people are excited before the product hit the shelves. Once it is on the shelves, people become indecisive. There are no doubt early adapter. To hit the mass market will nonetheless take time.
In the Siiicon Valley, people talk a lot of self driving car and show excitment about it. To my experience, people in valley are more inclined to new technology. You will see more Tesla in the valley than anywhere else in the nation. Most of the iProducts spread out in the valley pretty rapidly. I can see the first mass market self driving car appear in the valley first. Will it roam in a crowded city or just self drive along highway, only time can tell?
Yes i agree AEC Q100 (at least grade 1, preferably grade 2) would be required. For camera systems typically ISO26262 ASIL B required for functional safety today. For sure this will increase for semi-autonomous driving. Adaptive cruise control systems which are available today based on RADAR technology seem to be assessed up to ASIL C minimum with a preference for ASIL D.
There's a new application area emerging where the information from RADAR and camera sensors are fused to create some kind of 3D environmental mapping of the vehicle surroundings. Scenarios are run based on a long list of potential risk factors and "safe" decisions are made about vehicle dynamics. This requires a LOT of processing power and car makers are trying to figure out how to achieve ASIL safety at very high performance without too much redundancy (since redundancy costs money..).
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...