Vehicles that virtually drive themselves are no longer the stuff of science fiction, with Google and other companies working to develop self-driving cars. These automated vehicles not only offer improved safety and fewer traffic jams, but real environmental benefits as well.
Imagine standing on a street corner in Manhattan, in need of a taxi to take you from Union Square up to Columbia University. You punch your destination into an app on your phone, which tells you the cab should be there in 30 seconds. A car pulls up as promised, and you get in and relax on your way uptown, just as with any cab ride. The only difference? No one is driving this cab.
This is one scenario that impressive advances in self-driving vehicle technology have turned into a legitimate possibility in the not-so-distant future. Since the early days of cars, we have been looking for ways to automate them, including the cruise control and parking assist systems available today. But as high-profile projects by Google and others have begun to show, we are getting closer to much more thorough automation, where cars may actually start to drive themselves with only minimal assistance from the person at the wheel. From drastic safety improvements to huge reductions in fuel use and accompanying emissions, the potential of self-driving cars is impressive. And they may be coming sooner than you think.
These self-driving cars will undoubtedly be using some COTS GPS system. They will follow each other like lemmings through every mapping error and local lane closure.
I have a GPS built into my phone that shows a local street going through the center of a church 2 blocks from me. It’s been like that for 6 years and has never been changed. Good luck with self driving cars and road construction.
It could work just have the vehicle follow an embedded wire in the roadway. This would be great on the highways think of the fuel savings. The cars could follow each other very closely in effect drafting one another. This could double every vehicles fuel mileage.
And overheat all but the lead.
Not only does this require a massive capital investment by government, but also by individuals. For some politician’s idea of utopia.
Overheating shouldn’t be a problem. This isn’t NASCAR. My son followed me on the highway quite close with my Lincoln and the car completed the trip with an average of 60mpg. The cars normal would be about 32mpg on the highway. The Lincoln didn’t overheat even with the AC running.
Cooling requires airflow. Improved gas mileage comes from reduced airflow. YMMV
Matt, not necessary. The AI can be trained to follow the road signs. In fact, for simple highway driving, there were demonstration models available in the 80s and 90s.
Things got much more complicated in situations other than limited access highways, but with increased computing power, it is possible given current teechnology.
(http://en.wikipedia.org/wiki/Driverless_car)
Guys, stop being silly. The self-driving car was perfected for highways long before the invention of GPS. The issue will be dealing with the 1 in 10,000 idiot. The situation where you NEED to not abide by the rules in order to avoid getting killed. Teaching a computer to follow the rules almost all the time never works. You will go for months without problems, and then the
AI will decide that it needs to hit your garage with full force repeatedly for no aparent reason, or go endlessly through the McDonalds drivethrough, or go from La Porte to the Woodlands with a quick stop by the old Sugarland Sugar mill (20 miles out of the way).
Automatics would already be on the road if a 99% solution was acceptable. However, it’s not. You need a 100% solution, and until we can develop full sentience, I’m not seeing that happening.
And that’s not counting the simple fact that the AI won’t consider the possibility of a car going the wrong way on a road, and won’t be able to tell if a driver three car lengths away is drunk and that its needs to keep its distance. The AI won’t be able to adapt to a deer or dog on the road, and won’t be able to make the judgement call of hitting the dog, driving off onto the shoulder (or is it a ditch), or driving onto oncoming traffic (which might be safe or might not depending on what sort of road it is). Then, will it do the same for an armadillo? Roadkill? A Cardboard box? A Semi Tire? A person standing in the road? A person laying down? Some of these should be avoided at all costs (up to and including crashing the car if necessary), and others should be ignored.
“The situation where you NEED to not abide by the rules in order to avoid getting killed. Teaching a computer to follow the rules almost all the time never works.”
Infamously, Airbus lost two planes to this, and the Russians lost the Kursk.
Airbus’ first public demo, the pilot did a flyby. The computer, having been put into landing sequence, would not release the plane from landing sequence.
So it landed. In the trees.
A pitot tube on a trans-Atlantic Airbus flight became plugged with ice, telling the computer that the airspeed was 0 mph. It went nuts; the pilots could not wrest control of the plane from the computer, and it crashed into the ocean with great loss of life.
The nuclear sub Kursk had a reactor incident. The computer had been programmed to require a 7-minute delay before restart. The sub sank to crush depth in under 7-minutes.
In my 30 plus years as a computer jock, I saw many egregious errors, but none that killed people. We had a saying, “A new software release is replacing the old bugs with new bugs.”
This could completely change ‘the debate’. Some might still insist that people kill people, but it will become obvious that cars kill people.
I am not suggesting AI and definitely not full sentience I have a little tickle in the back of my mind saying that full sentience in computers might not be a good idea. The science fiction book Colossus by DF Jones comes to mind. It would only work with limited access highways and no manual operated vehicles would be allowed. No complicated computer programs just follow the wire and stay at a set distance from the car ahead. Switch to manual control on the off ramp.
Matt, that would require specialized equipment that is quite expensive, on a specialized roadway that it quite expensive. One would be useless without the other, and there would be no possible transition period. Plus, it would require automatic override of the controls (because people will take it out of auto for many reasons, from a “what does this button do” moment up to and including a massive wreck right in front of them), and GC detailed exactly what the problems are with that.
Remember, the Stanley Steamer and other early cars ran on Kerosene (lamp oil) before the proliferation of gas stations, and ran on carraige roads before modern highways were built. Your idea, going to the Minority Report-style automatic highway might work eventually. However, you need a functional intermediary.
Finally, I was being silly with the sentient computer quip. Besides, I share Asimov’s view on the Golem/Frankenstein-style sentient creation tale.
We could spend a trillion dollars to build this thing, and it would be as empty as the HOV lanes.
Bakersfield to Modesto might be a good location.
Other problem with this is when there is an accident of any sort, the insurer will immediately sue the maker of the AI system.
At the moment because everyone can make a mistake and cause an accident, it suits everyone to accept the possibility and arrange for it. So we have no-fault insurance and reinsurance, because we all know we risk having a bad day.
But as soon as you have someone you can blame for the accident the game theory defection option will have big payoffs. Hence lawsuits will be inevitable and whoever makes these things is going to go bankrupt. No technology can be perfect, especially on initial rollout.