I invented that 20 years ago. I called the vision system "eye-liner and the overall system, "The Autotrain". Oh god, I have been sucked into the comments...
2) I assume you're from the US - you're really freaked out by roundabouts :) It's just a type of intersection like any another, in Europe or the Middle East no one cares and they're everywhere.
Left-turns on classic intersections (without separate light for turning) are the most dangerous, passing a roundabout is safe. Even if there's a collision on a roundabout, there are no fatalities due to the relatively slow speed and no significant difference of speed between vehicles, while a left-turn collision can be deadly or at least most of them are very destructive.
great idea. But I can't help to think that as soon as this gets implemented, some guy will get on the road train and then jump into the back seat and take a nap.
What happens when the truck driver guns it on a yellow and has 6 cars behind it? I'm guessing 2 or 3 T-bones and about 4 hours clearing up an intersection.
For this system to work, all traffic lights will need to be reliably broadcasting their status and how long they're going to be remaining in it so that either the fleet can stop, or so that some of the cars can drop out to avoid running a red.
It would also need to communicate with other vehicles. I can turn right (or left in some occasions here in Canada) on a red if I've got the room, but how do I tell if my fleet has the room? If the truck driver becomes liable for risking my vehicle then he's only ever going to move on his light and when he knows he has the time, which will only cripple intersections with congestion.
It's a great idea and I can't wait for the day, but our road system is overly complex as it is and these systems will need to be equally as complex to guarantee safety.
Edit: Also the driver of the fleet vehicles would need to know the destination and what exit the lead is taking. If I'm going to Montreal, I don't want to end up in downtown Toronto and spend forever trying to get back on the highway. Similarly, it will be especially dangerous if I'm just 'dropped' from the fleet due to a communication problem or if my vehicle doesn't follow because the lead is exiting before my destination is set to.
IMO this system would do best as an advancement of an auto-driver system that can read lights and signals, etc. itself without endangering the driver.
This system is for highway traffic, there's no traffic lights. The maximum number of vehicles in a road train is 5. Obviously it has all kinds of failure recovery systems and in fact some top-end BMWs can already safely pull over to the side of the road if the driver suddenly has a heart attack or is otherwise incapacitated.
Anyone else disturbed/amused to see a Windows screensaver in the bottom right on the dash computer of the guy in a "following" vehicle who's reading the paper?
Thought this notable: SARTRE researchers say that around 80% of accidents on the road are due to human error.
I bet if we extend the definition of "human error" to "something that a human couldn't react to, but a computer could have" that number is a lot higher.
And I've seen demonstrations of sensor platforms that can detect deer far more readily than humans. ISTR that at least at one point you could buy cars with IR HUDs, though I don't recall reading anything about that in a while; I don't know if that's because they don't really work or because they're just boringly common by now on cars I can't afford.
Also, assuming a very good computer driver, a computer driver will have a much better sense of when they are in trouble. They might refuse to drive fast in a low-visibility situation. I see a world in which computers "want" to drive 55 in areas currently labelled 25 (though of course they will honor the speed limit) and in some cases will insist on driving 25 through areas labelled 55 that turn out to be something like forest too dense to see through that may be harboring deer or something, or other situations in which a human would blithely speed through but the computer is "scared". Arguably correctly.
And per my other message, when the computers are tweaked to ignore that and then hit a deer, lawsuits.
We can already build self-driving cars. The obstacles now are societal and legal. In the next 5-10 years they will become common, which solves both the traffic problems of safety and traffic jams.
The USA has very strict liability laws. Any mistake is likely to mean massive lawsuits in the USA even if the system is significantly safer than human drivers. All these tests are happening in the EU, and the EU is already getting ready to implement these systems.
For what it's worth, it flows the other way too. If computer drivers are safer under some circumstances, lawsuits will be filed about how the computer wasn't driving the car.
My father works in the car industry. When I was young (though I heard about this later so the memory is fresher than that), he was called as an industry witness in a lawsuit in which his company was being sued for not putting antilock brakes on this particular car, "resulting" (for some value thereof) in a fatal accident. At the time of the accident, antilock brakes were actually an option on the car... an $X,000 option in mid-1980s money. The idea of putting antilock brakes on all the cars was an economic absurdity.
The car company lost.
I know the cynical answer is to say computer will face major legal barriers but I'm not sure it's that simple. One can look at the lawsuit-happy USA and readily see a world in which licensing restrictions are actually significantly tightened and it becomes illegal to drive a car under most circumstances using the exact same logic. Conclusion? Beats me.
The research is part of the European Union-backed SARTRE project (Safe Road Trains for the Environment) and it's the first time that "vehicle platooning" technology has been demonstrated outside of a simulator.
Uh, the PATH Project at Berkeley demonstrated platooning over a decade ago:
The expression "Road Train" is used in Australia and elsewhere to refer to very long trucks. In the Northern Territory (where I am), a truck can pull as many as 4 carriages.
Passing one on the highways is a test of endurance.
Having one go past at 130 kmh (80 mph) on the other side of the road is an exercise in gripping the daylights out of your steering wheel.
25 comments
[ 1.5 ms ] story [ 78.4 ms ] thread1) Traffic lights
2) Roundabouts
3) Yield/stop signs
Left-turns on classic intersections (without separate light for turning) are the most dangerous, passing a roundabout is safe. Even if there's a collision on a roundabout, there are no fatalities due to the relatively slow speed and no significant difference of speed between vehicles, while a left-turn collision can be deadly or at least most of them are very destructive.
For this system to work, all traffic lights will need to be reliably broadcasting their status and how long they're going to be remaining in it so that either the fleet can stop, or so that some of the cars can drop out to avoid running a red.
It would also need to communicate with other vehicles. I can turn right (or left in some occasions here in Canada) on a red if I've got the room, but how do I tell if my fleet has the room? If the truck driver becomes liable for risking my vehicle then he's only ever going to move on his light and when he knows he has the time, which will only cripple intersections with congestion.
It's a great idea and I can't wait for the day, but our road system is overly complex as it is and these systems will need to be equally as complex to guarantee safety.
Edit: Also the driver of the fleet vehicles would need to know the destination and what exit the lead is taking. If I'm going to Montreal, I don't want to end up in downtown Toronto and spend forever trying to get back on the highway. Similarly, it will be especially dangerous if I'm just 'dropped' from the fleet due to a communication problem or if my vehicle doesn't follow because the lead is exiting before my destination is set to.
IMO this system would do best as an advancement of an auto-driver system that can read lights and signals, etc. itself without endangering the driver.
What the fuck.
I bet if we extend the definition of "human error" to "something that a human couldn't react to, but a computer could have" that number is a lot higher.
Around 100%, probably.
Also, assuming a very good computer driver, a computer driver will have a much better sense of when they are in trouble. They might refuse to drive fast in a low-visibility situation. I see a world in which computers "want" to drive 55 in areas currently labelled 25 (though of course they will honor the speed limit) and in some cases will insist on driving 25 through areas labelled 55 that turn out to be something like forest too dense to see through that may be harboring deer or something, or other situations in which a human would blithely speed through but the computer is "scared". Arguably correctly.
And per my other message, when the computers are tweaked to ignore that and then hit a deer, lawsuits.
It's gonna be interesting times.
My father works in the car industry. When I was young (though I heard about this later so the memory is fresher than that), he was called as an industry witness in a lawsuit in which his company was being sued for not putting antilock brakes on this particular car, "resulting" (for some value thereof) in a fatal accident. At the time of the accident, antilock brakes were actually an option on the car... an $X,000 option in mid-1980s money. The idea of putting antilock brakes on all the cars was an economic absurdity.
The car company lost.
I know the cynical answer is to say computer will face major legal barriers but I'm not sure it's that simple. One can look at the lawsuit-happy USA and readily see a world in which licensing restrictions are actually significantly tightened and it becomes illegal to drive a car under most circumstances using the exact same logic. Conclusion? Beats me.
Uh, the PATH Project at Berkeley demonstrated platooning over a decade ago:
http://www.path.berkeley.edu/nahsc/
Passing one on the highways is a test of endurance.
Having one go past at 130 kmh (80 mph) on the other side of the road is an exercise in gripping the daylights out of your steering wheel.
Are they looking for "No Exit" jokes?