Slashdot Mirror
MIT Study Shows Stop Lights Won't Be Necessary In The Future (computerworld.com)
Lucas123 writes: An MIT [Senseable City Lab] study based on mathematical modeling demonstrated a likely scenario in which high-tech vehicles, using sensors to remain at a safe distance from each other as they move through a four-way intersection, can eliminate the need for traffic lights in the future. By removing the waits caused by traffic lights, these so-called Slot-based Intersections speed-up traffic flow.The study claims this kind of traffic-light-free transportation design, if it ever arrives, could allow twice as much traffic to use existing roads.
What would be the comparative advantage with respect to a roundabout?
I know they are not very popular in the US, but they can be very efficient, and prevent the frustration of waiting at busy intersection (especially if going in the non-popular direction).
How are they supposed to cross? Without lights, there could just be continuous stream of them walking over the road and cars can't pass since they try to avoid hitting them thus causing even bigger jams in big city centers.
Seems like they still need to play a part in any scenario with or without traffic lights. Dropping traffic lights may be fine for autonomous vehicles, but that also means that people will just walk when they want to, which'll make cars stop to let them pass. Perfectly fine when there's few pedestrians, but when there's a lot, the automobile traffic will totally come to a half if no one is stopping people from walking into the street.
Give a hand, not a hand-out.
I believe that I have seen animations for driverless cars careening through intersections for years, as in going back to the 70s.
What I find much more fascinating is the economic impact of this sort of thing. How much economic activity is generated by traffic lights. Obviously there are the companies making them, maintaining them, their electrical usage, the cost in having people idle at them, and the ticket taxes generated by having police ticket people for not obeying the tax laws.
I would not be surprised that the savings to the taxpayer and the public by removing a single unimportant traffic light could be well in excess of $100,000 per year. For instance there was one major downtown street near my old house where they had the lights perfectly timed so that you pretty much missed all of them. Thus the average speed on that street in low traffic was maybe 15mph tops. With about 8 lights and the street being 1.5 miles the savings in time alone to get that up to 25mph would be astounding, let alone in gas.
Also many busy intersections are pretty much car accident factories. So to remove those would be just another layer of costs removed.
But what is interesting about all the above costs is that they are all very parasitical. Most of the costs in having a traffic light don't really "benefit" society. Obviously a typical traffic light today massively reduces accidents and other problems but when we have 100% SDCs their removal will only be a net benefit to all.
Where this is also going to get interesting is that some traffic lights are political. For instance there is a neighbourhood in my old city where a 3 way stop was replaced with a traffic light. This then encouraged people to take a short cut through a rich influential neighbourhood so within about 10 days the light was removed and went back to a 3 way stop. I can see attempts to prevent self driving cars from "navigating" through rich neighbourhoods but that is going to impinge upon fundamental freedoms and those laws are going to be hard to sustain. But with enough political influence there will be a way to keep the plebs away from the rich. Which will simply be part and parcel of the many many stupid laws that I see coming when politicians don't realize that every stupid traffic law they implement will be diligently followed by computerized cars. I can see every squeaky wheel along rural highways calling for the speed limit to be dropped in front of their house because of "the children" thus the speed limit will be very much an indication of how influential any given household is in rural communities.
...Happening already
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
Humans suck at driving. The problem is that 99.999% of the time, you can be borderline incompetent at driving and you'll still get there safely, because things only go wrong on rare occasions. Most of the time, at city street speeds, you could glance at the road for two seconds out of every ten, and you wouldn't crash, because there just isn't much happening. There are situations, however, in which humans are physically incapable of being good drivers. For example:
And in some cases, each of those situations can result in a crash with a human driver, depending mostly on luck. Computers, by contrast, won't exhibit any of those physical failings, and thus won't crash in any of those situations, typically.
So the key question is whether they will crash more often in other situations where a human wouldn't (e.g. when nothing is going wrong). As long as that answer is no, then they will likely be safer than human drivers.
That's just not true at all. Google's self-driving cars have clocked over a million miles on the roads, with basically no at-fault crashes. That's a far cry from barely being able to work in a pristine environment on a fixed guideway. It has some ability to recognize pedestrian behavior, avoid obstacles in the road, handle traffic lights (as long as it knows to look for them at a particular intersection), etc. It does require a lot of pre-mapping of the terrain so that it knows where to watch for traffic lights, roughly where lanes are, etc., but still, they've gone way beyond a subway system on a fixed track as you imply.
Check out my sci-fi/humor trilogy at PatriotsBooks.
This is the future. Pedestrians go in tubes, duh.
Irony: Agile development has too much intertia to be abandoned now.
You're operating under the mistaken notion that the primary purpose of traffic lights is to increase safety and maximize throughput. In reality, the primary purpose of traffic lights is to make city streets so annoyingly slow that drivers will be forced to get on the freeway as soon as possible, then sit there at 3 MPH. It's called "traffic calming", mostly ironically, as the primary effect appears to be an uptick in road rage.... :-D
Check out my sci-fi/humor trilogy at PatriotsBooks.
Like pretty much everybody, I never look at what's ahead when in a bus. The bus stop and go, and I just don't care. It doesn't "damage" myself. Also, my parents never said "whoa" and "go" to signal me anything when I was a kid.
Sorry, but your comment is idiotic.
The issue right now is basically what you explained for why human drivers suck. Automated cars are GREAT at the 99%. It's the edge cases, in bad weather, with mechanical failures...that will be the true test.
Has google been testing their cars in blizzards? Mostly they seem to be in fairly nice safe weather conditions. This link details Ford's ventures into this issue and it seems like good progress. But when the car simply loses traction on a snow covered street, who does it try to protect? The driver? or the pedestrian walking in the road because the sidewalk isn't plowed? How well does this system cope with a build up of ice across a vehicle? It's easy to cherry pick these questions obviously, but giving full access requires either the ability to deal with everything as well or better than humans. (which might be possible in most cases)
People in cars cause accidents....accidents in cars cause people
People did that when automobiles were first becoming mainstream because the driver was used to talking to their horses. Are you sure your not just old?
I think you underestimate just how much I just dont care.
Automated cars are GREAT at the 99%. It's the edge cases, in bad weather, with mechanical failures...that will be the true test.
Automated cars are better than humans in bad weather. They have many types of sensors, while humans rely on vision. Tesla has recommended that owners engage Autopilot during snowstorms, rather than trying to drive. The sensors can detect other cars in fog or heavy rain, avoiding fatal pileups. As for mechanical failures, those cause a very small percentage of crashes, but I see no reason to assume that humans can handle them better than a computer. At the very least, the computer would have a faster reaction time.
Humans suck at driving.
Actually, they don't. That's why driving hasn't already been replaced, unlike say, computing a FFT by hand or screwing on a million bottle caps. For example, in the US we're down to about 11 deaths and 1850 crashes per billion vehicle miles.