Slashdot Mirror
The Intelligent Intersection Could Banish Traffic Lights Forever (arstechnica.com)
Jonathan M. Gitlin reports via Ars Technica: With a degree of coordination -- between vehicles, and with traffic infrastructure -- traffic chaos should theoretically be banished, and less congestion means fewer pollutants. Clemson researcher Ali Reza Fayazi has provided a tantalizing glimpse at that future, a proof-of-concept study showing that a fully autonomous four-way traffic intersection is a hundred times more efficient at letting traffic flow than the intersections you and I currently navigate. Because cars don't sit idling at the lights, Fayazi calculated it would also deliver a 19 percent fuel saving. Fayazi designed an intersection controller for a four-way junction that tracks vehicles and then uses an algorithm to control their speeds such that they can all pass safely through the junction with as few coming to a halt as possible. What makes the study particularly interesting is that Fayazi demonstrated it by interspersing his own physical car among the simulated traffic -- the first use of a vehicle-in-the-loop simulator for this kind of problem. Fayazi drove his real car at the International Transportation Innovation Center in Greenville, South Carolina, where a geofenced area was set up to use as the simulated intersection. Using GPS sensors, his car was just as visible to the intersection controller as the virtual autonomous vehicles that were also populating its memory banks. Ideally, Fayazi says he'd like to have tested it with an autonomous vehicle, but they are hard to come by, particularly in South Carolina. Instead, the intersection controller directly governed his speed in the study (as it did with the simulated vehicles), and this controller sent him a speed to maintain in order to safely cross the junction. Over the course of an hour, the intelligent intersection only required 11 vehicles to come to a complete halt. By contrast, when the simulation was run with a traffic light instead, more than 1,100 vehicles had to stop at the junction over the course of an hour.
Aside from the obvious pedestrian issues (which can be addressed by the Buddhist street crossing technique: just walk at a constant speed through the intersection), this is a pretty obvious solution.
...is way to make drivers drive at the mandated speeds. Which, of course, they won't do; you only have to look at how speed limits are obeyed to know that.
I'm sure it looks good if there are only cars. The minute a pedestrian or bike comes through, its all going to hell in a handbasket. The gall of them, wanting to cross the road!
not CLEM-zun
There will be a random chance that a car in the middle lane will do an illegal left-hand turn in front of the fast lane and turn lane when the light turned green. One of the reasons why I don't like to drive in San Francisco.
No traffic lights == no red light cameras..
The problem here is he's not looking at the bigger picture. In a congested big city you would be simply moving the traffic jam to the next traffic light. In bigger cities the lights all have to be coordinated so that you don't end up with a backlog of red light traffic that extends all the way back to the previous traffic light. And then having idiots that get stuck in the intersection blocking cross traffic.
Maybe it works in Greenville, population 100,000 but not in a city of 2-3 million or more.
Aside from the whole how do pedestrians and other vehicles with either non-functioning or missing telemetry modules, how will this be kept secure and without the possibility of jamming or spoofing the signals?
...and motorcycles, and baby strollers, and delivery people moving boxes, and letter carriers, and dogs, and cats, and children playing.
https://www.youtube.com/c/BrendaEM
I've driven a bunch of times lately in Europe and its amazing how long you can go without ever coming across a traffic light.
This sort of optimization can work well, obviously, for traffic continuing through the intersection, but what about real-world traffic that needs a left-turn at intersections?
Umm... so what? This is like saying don't build roads because inevitably people won't be able to get to the road.
Anyhow, simulations like this have been around for decades, I remember seeing one in the 90s I'm sure.
But for now - as in, for at least the next couple of decades and probably longer - we need to allow for human drivers.
For as long as I can recall, my region has had semi-intelligent lights, with pre-set schedules tweaked by data gathered from sensor loops in the pavement.
Unfortunately, while that tells you if at least one person is waiting at the light, it doesn't tell you how many cars are coming, how far away they are, or how fast they're moving (though you could simply assume the speed limit for that last one).
By adding in some cameras to identify vehicles that aren't yet at the intersection, lights could anticipate and change for optimal flow. I can't count how many times I've sat waiting at a light while the other direction has no cars... only to see the light start changing they come around a corner. A smart light would have seen there was no traffic in the other direction and immediately flipped to green as I approached resulting in a lot less car idling. And you wouldn't have to rip up the intersection to put in the sensor loops, either.
Intelligent Intersections Already Exist, they're called "Roundabouts" or "Traffic Circles" depending on your geographic location.
In most cases they eliminate the need for idling cars, they process much more cars per minute than a traditional traffic light, all cars can go through at an intelligent speed without the need for computers or artificial intelligence calculating algorithms.
Best of all, it works with human controlled vehicles too. There's going to be a lot of them on the road for decades even if we have AI cars available to the public tomorrow.
It's really quite ridiculous we don't have more of them in the US instead of four-way stops and traffic lights.
"That's the way to do it" - Punch
Funny how this pops up and most of the comments are immediately pointing out obvious problems that everyone would have thought about. It's a proof of concept, and as a concept this is a seriously good one. Yes there's issues with pedestrians, and with cars not enrolled, but there's ways to manage and work around those. Intersections like these are the way of the future, as are autonomous vehicles. Driving a car is going to go the same way as riding a horse - a hobby activity not an every-day thing.
Greenville itself may only have 100k (don't know the statistic), but the city with all it's satellite suburb towns is closer to 1 million. That said, Greenville has much better traffic than other cities of similar size. Unlike most places they usually plan ahead and expand roads before they're needed, instead of waiting until 10 years after the expansion is needed.
It's not "Small town America" it's "midsized city America". A lot more people live across the country in cities smaller than 3 million than live in cities of more than 3 million people.
Will the same solution work in New York City? Probably, probably not. It would certainly require a lot more processing.
"That's the way to do it" - Punch
Great if all traffic is going straight, or turning right, but what about those left turns?
This essentially relies on batching up packets of cars and letting them cross as a group. I wonder how well this works during heavy traffic periods.
This is probably a bad idea. We know that eventually the roads will be owned by the corporations. The corporations will want to give preferences to businesses that are willing to pay more. Then we have a road neutrality movement. I have enough aggravation supporting net neutrality. Please don't add road neutrality to this list!
Pretty sure I saw this concept on the Jetsons already.
In your gleeful rush toward a greed-based solution that prioritizes the wealthy, you missed the part where the red lights no longer exist.
By spacing the traffic appropriately, cars simply arrive at the intersection at pass through without stopping, allowing everyone gets to their destination as efficiently as possible. There is no inefficiency for you to monetize unless you artificially introduce it. If you want to see how well that works, just look at the arguments surrounding net neutrality, zero rating, throttling, etc.
They've already done this in Moscow. There are blue-designated center lanes that were apparently originally for Party officials, required a blue light on the top of the car. With the fall of the Soviet Union these essentially got turned into pay-to play lanes, and rich people bought permission to use these lanes and to have the blue lights on their cars.
Do not look into laser with remaining eye.
The city I live in doesn't even have the budget or ability to synchronize the lights with each other. Now someone thinks cities will be able to coordinate lights with traffic?? It won't happen where I live, not ever.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
In most large cities, bicycles are becoming a larger and larger percentage of the traffic flow. Even New York is doing it. The wave of the future isn't smarter cars, it is no cars...
It's not a bug, it's a feature...
Already working, no computers needed. Check out Ho Chi Minh City, old Saigon, a city of 8+ million people, 10 million motorcyclesâ, and about two dozen traffic lights.
If you've never driven a motorbike through a high traffic unsignalized intersection, you should try it. It's said that one achieves a certain sort of enlightenment the first time one drives through Ben Thanh Circle.
Also, isn't this kind of what roundabouts already accomplish in a fairly elegant way?
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
The issue the GP is alluding to is what happens when traffic volume is higher then the max volume the intelligent intersection can handle, what happens when traffic is backed up to the intersection. Roundabouts are great until they reach their capacity then they don't work.
Knowledge = Power
P= W/t
t=Money
Money = Work/Knowledge so the less you know the more you make
Looking forward that's effectively an exaggerated figure, since there will be more hybrids which won't save nearly as much.
Someone had to do it.
I don't think congested city has much to do with the outcome of the solution (if it is correct). If the algorithm covers and controls approach speed of every car to any intersections, it could clear up any congestion.
However, I think the study jumps the conclusion. There are so many assumptions in the study that DO NOT and CANNOT apply to the real world.
I don't like the way the researcher (Ali Reza Fayazi) quantifies his study...
Over the course of an hour, the intelligent intersection only required 11 vehicles to come to a complete halt. By contrast, when the simulation was run with a traffic light instead, more than 1,100 vehicles had to stop at the junction over the course of an hour.
In other words, the quantification comes from his own simulation. What would you think about it? He could simply find a way to work around the flaw in his traffic light algorithm, so that his "intelligent intersection" could result better.
Another thing, because it is a simulation, the implementor can plug in all variables to make its outcome look good. In this case, I have a feeling that the researcher wants grants and/or advertise for his work. This type of study is not as simple and would need a lot more data than this study has. Of course, it may be possible in the future, but it is still far away as long as we include humans in the road.
What's this idling he's talking about? The switch to smart intersections will be slower than the switch to electric vehicles, so stopped cars won't be using nearly that much power. True, they may be running heat/AC and such, so it's not zero, but it's not the big deal that it was. Even hybrids shut off their engines at stop lights, and now some gas-only cars have added that technology.
But that's just a nitpick. Obviously there's a lot we can do to improve intersections, and this idea is nothing new.
Most of the backups come from people coming to a complete stop, and the extremely slow acceleration for all but the front car once traffic is moving. That is, at the front light when there's multiple intersections backed up. You would greatly increase the capacity of those intersections, but probably at the cost of increasing the number of people willing to get on the road in the first place.
I can add one more. Position is based on GPS instead of some sort of radar system once cars are in range - at least in the demo. GPS is somewhat accurate on the open road, but Google Maps gives an impression of much higher accuracy than is really possible in a moving vehicle. They snap you to the nearest road based on the curvature you follow compared to the documented curvature of the road (partly pioneered by this guy).
The problem here is he's not looking at the bigger picture. In a congested big city you would be simply moving the traffic jam to the next traffic light. In bigger cities the lights all have to be coordinated so that you don't end up with a backlog of red light traffic that extends all the way back to the previous traffic light. And then having idiots that get stuck in the intersection blocking cross traffic. Maybe it works in Greenville, population 100,000 but not in a city of 2-3 million or more.
There is a point where the road can't handle the traffic no matter how efficient the traffic controls are, but fully autonomous controls could greatly improve traffic even in those congested cities. At worst, they would minimize the windows of greatest congestion buy keeping traffic flowing as long as possible. There would need to be some algorithms that adjust for multiple intersections. In some cases, the answer might be stop and go.
With very light traffic, traditional lights that turn green when they detect an approaching vehicle are already optimal. With very heavy traffic, traditional lights that let each queue go in turn are already optimal
Not even close.
Roundabouts have significantly higher bandwidth for the same area of ground covered, and also don't require the input and output roads to be as large, due to having a peak traffic flow closer to the average traffic flow.
19% reduction in fuel costs?
Big Gasoline "LOBBY LOBBY LOBBY LOBBY LOBBY!"
I get around 30 miles per gallon. According to this guy, I will get almost 6 miles per gallon more. I'll believe that when I see it.
The shepherds did so well protecting the flock that the sheep no longer believed that wolves existed.
Than the rotaries our not so bright Governor put on Route 1/117 - the rotaries themselves are OK but without signage it gets fun.
All self-driving cars would need to add those LCD windows that can be made opaque on command. The software would block the windows when going through the intersection so the passengers wouldn't freak out.
I'll be long gone before this happens. I'm one of those that ENJOYS driving. I don't want any silly computer to take that enjoyment away.
I cannot remember which one it was, but one part of one comic had a way to change lights that was quite interesting. If more than one car is coming to an intersection and the light was red, you could put in money to make it change. The person who put in the most got it to change. Many times it would only take a penny, but if many rich people were in a hurry it could start a bidding war. I always think of this when I come to stupidly controlled red lights where I am the only car.
On another note, I was just in Norway and Denmark for 2 weeks. I rented a car and in most areas I was at there were no lights or stop signs. Almost all traffic was controlled by roundabouts where needed and places where there would have been a stop sign in the US, there was a yield sign in Norway. I liked it.
With complete autonomy of the vehicles, yes, it should work anywhere. The controllers will all need to be interlinked and work together.
One nice thing about complete autonomy is that the car can communicate it's intended route to the controller network too, so that it can adjust traffic to allow the car to proceed in it's required direction without having to stop.
We've all marvelled at the amazing choreography in this video (https://www.youtube.com/watch?v=nl-3Kx7jr4g), but with full autonomy this could be every junction all the time... well, within reason - we've no autonomous pedestrians. But it would be possible to get somewhat close.
That said, with full autonomy, it would be possible to have a public transport system that would pick you up at point A (your front door) and drop you directly to point B (work, your parents', that restaurant where you are meeting your friends) just by inputting the time you need to be at point B, and the requested time of collection at point A. Having an autonomous bus network that could work through this information, the system would tell you the time the bus will arrive at your door (as close to your requested time as possible), you board (within a specified amount of time or the bus leaves and you have to book another one), and arrive at your designated time. The 'expressness' of your booking would be a factor of the cost of the journey - if you only want to spend 20 minutes on the bus, you'd pay more than if you are happy to spend 30 minutes (thus allowing the bus to pick up x extra passengers) - as well as how much time before your booking and your pickup (if you book a bus for 5 minutes time you pay more than if you book it for a hour, or 2 hours time - more time allows for better calculation of routes). The bus (or central controller) will decide the route to take in order to collect other passengers at their pickup points, and drop people at their destinations. Bus routes would cease to exist, with any bus going to any pickup or dropoff point, depending on how a central controller calculated the most efficient use for every bus over a period of time. You might always get a bus at around 08:25 and arrive at work at 08:55 every day, but could take a complete different route to get there.
Basically, complete autonomy could completely do away with the need for private car ownership by a majority. Even if you want to go from New York to LA by road with your family, a car-bus could pull up outside your house and be exclusively yours to get ye there, or even for the 2 weeks you are spending there. On return to New York, the car goes back for a quick service and the next day is driving another family to Chicago, while you are hopping on a city bus to work, and you spouse on a separate inter-city bus to get to New Jersey, and the kids onto the school bus, or whatever.
Anyway, just a thought. Elon, should you wish to do this, give me a call and we can talk through it!
You don't generally find traffic signals on freeways. This article is about junctions on city streets, exactly where you'd expect human beings to be crossing the road.
You are not alone. This is not normal. None of this is normal.
Or when they PAVE a nice bike route right BESIDE the main road, only for cyclists to use the main road anyways
That depends on how many lanes designed for automotive traffic a cyclist has to cross to reach the lane for turning left.
In practice, one faulty component - hardware or software, though most likely a carbon unit - will bring the whole rotten edifice down.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
We don't come to a stop at traffic lights because other cars need to go through. We stop at red lights at 3am when there isn't another car for miles.
We stop at intersections because that's where absolutely anything is allowed to happen.
"constant speed" doesn't exist with construction, weather, birds, animals, clouds, rain, sun, winter. Oh yeah, and turning.
Intersections aren't intersections, they are cross-roads. An intersection is where a road is no longer a straight-away. We assume that we can drive smoothly on a straight-away. Even in snow and ice and fog we can still drive 140kph on a straight road. We don't dare do so through an intersection -- there are too many things that happen in an intersection.
If you want to have less traffic, you need fewer cars or more roads. I'm really sorry, but more people in a smaller space just isn't solvable. Welcome to density.
And besides, this is all pointless. Even if you could magically get twice as many cars to flow at full speed on a road, those cars need to go somewhere. City planners intentionally put bottlenecks into traffic areas to ensure that congestion is, say, on the highway and not in-front of your house.
be in a community with the extra land near intersections required for the additional infrastructure
issues those outside the scope, of problems a civil engineer can realistically be expected to address, other than by attempting to minimize infrastructure costs.
The real issue his bicycles! Which just need to be strait up banned from use on public roads IMHO.
Providing separate but equal facilities for cyclists alongside those for motorists requires "extra land" and thus is likewise "outside the scope, of problems a civil engineer can realistically be expected to address". What workaround did you have in mind that doesn't require a cyclist to spend upwards of 1000 USD per year on automobile acquisition, maintenance, parking, fuel, and insurance, nor to move his family to a different city with a more practical public transit schedule?
According to TFA :
the target isn't actually human drivers (who as you mention will never actually follow speed instruction anyway),
but semi-autonomous cars (basically the system feeds directly into the cruise control of the car, and at that speed, cars cross while mathe-magically managing to avoid each other).
Yeah. Right. All the cars at an intersection needing to openly agree together at which speed each should drive through a public open protocol.
I can't see what could possibly go wrong/get abused~
(Yeah, the various three letter agencies of the world can finally blame their attempt at assassination by remote car hack to "bugs in the traffic management software")
More seriously :
Luckily most modern car have the *adaptive* cruise control type, coupled with forward collision avoidance, so your (high range) car should be able to notice that (because of buggy / malicious speed instructions) it is on a collision course with another car and apply emergency brakes to reduce risks.
(But sadly, current modern cars don't have a very wide angle into their FCAS, and will consider a car coming sideways only at the very last moment.
So you're still in for some ruined car body, but at least you'll probably still be alive)
Note:
the simpler version shown in the Youtube animation (with groups of car crossing in turns, but never cars coming from both perpendicular streets at the same time, only next group once the perpendicular group has finished), is actually being implemented in some european cities (I've seen it in Switzerland) in the traffic lights : the lights try to intelligently switch green on one perpendicular axis or the other, depending on the incoming traffic trying to minimize the slowing down of group of cars (i.e.: if you travel the correct speed, the traffic will always suddenly turn green once you approach and turn back immediately red once you've crossed ; and conversely on the side ways.
i.e.: the traffic light vary their cycles so it looks like the youtube animation with nobody waiting at the crossing due to red light).
but the implied end goal of lanes of car safely criss-crossing each other simply by keeping the correct speed sound a little bit dangerous, error prone and open to hacking.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Roundabouts (traffic circles for those who use stupid words for them) are the solution. Also having most vulnerable road user priority rules (pedestrian > cyclist > motorcyclists etc.).
This might go well ... until the next solar storm comes that emits enough radio noise for cars to lose GPS satellite lock:
http://gpsworld.com/defensenew...
I wouldn't be surprised if auto insurance companies start putting in exceptions for that one, just like the 'act of god' clauses in home owners insurance.
Build it, and they will come^Hplain.
Foot bridges & underpasses cost alot and underground ones may have issues with the homeless out of work truckers living in them. And you better hope they don't try to change them as sex offenders just for public urination as that will just end up costing all of us big time with court / jail / prison costs vs maybe an fine / a small jail time just for public urination.
Great idea, but there is one technological issue that needs to be addressed before the idea will work.
Who will pay when things go horribly wrong?
Will it be the stoplight manufacturers? Will it be the software design company? Will it be the city in which they are installed , namely city taxpayers?
What will happen when two or more autonomous vehicles crash in the intersection?
I'm guessing it will be the last, so it will behoove citizens to also install virtual taxpayers, who will foot the bill by instantaneously generating bit coins, virtual.police, virtual police, virtual insurance companies, virtual body shops, and virtual lawyers, and virtual judges as a back-up plan just in case all involved fail to obey Murphy's Law.
AI to control traffic has been around for years. It's probably a classic project in introductory AI courses.
Switching most stoplights to one side blinking yellow and the other side blinking red between the hours of 9pm and 6am. Most stoplights seem to be set up to control heavy traffic during the day, and are completely superfluous in light traffic. It sucks to come to a stop and wait in the middle of the night at a large, empty intersection. Set the lights to blinking, so that only traffic from the lighter travelled road has to stop, and save half the electricity of keeping the lights on. Or save even more with a different duty cycle.
This would be a lot easier to implement than getting an AI through the government procurement process, because nothing new would have to be bought.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
On the other hand, the hobos in Seattle don't pay any attention to signals anyway. So lets go ahead with this.
Have gnu, will travel.
As often as everything breaks down do you really want to submit yourself to this?
Caution: Contents under pressure
Roundabouts have a far bigger capacity than signaled intersections. For really busy times you can use part-time signal-controlled roundabouts to handle even more.
Drill baby drill - on Mars
most modern cars shut down at intersections rather than idling
So how do you rev your engine when some poor old lady with a walker is only half way across the intersection when the lights change?
Have gnu, will travel.
Because cars don't sit idling at the lights, Fayazi calculated it would also deliver a 19 percent fuel saving.
Except that by the time this comes anywhere close to being plausible, cars won't be running on petrol any more.
systemd is Roko's Basilisk.
Motorcycles and bicycles would have to be banned. You can build an automated car that coordinates with traffic and maintains exactly 36 mph to reach an intersection at a moment it can pass through. But can you put all of that on a motorcycle without ruining it? Traffic at a right angle would also pass quite close to you and that would unnerve almost any motorcyclist and probably a bunch of car drivers as well. And then we come to the bicycles. Not only would mounting the gear be a disaster on a bicycle but at best the bicycle could be forced to slow down but no way to make it speed up. Obviously eliminating motorcycles and bicycles would cause major social unrest. And even positive things can have unseen issues. Obviously self driving cars would cure the drunk driving issue. But fear of loss of a driver's license may be the only reason some people stay sober and counties depend upon the huge fines levied for driving drunk and so do auto insurance companies. All in all, the idea of an automated intersection works well with robots in a warehouse but with humans on roads the social disruption would be far too great.
I had this idea back in junior high school, right around the time I was getting really excited (what passed for really excited before a late puberty) about how the 8008 would someday change the world.
Plus, I would be turning sixteen soon, and having to navigate all that legacy infrastructure in my dad's fuel-guzzling pickup truck with the sticky clutch pedal, so my mind was specially tuned to the many imminent electronic upheavals sure to enlighten all the tired brick and mortar and asphalt in the blink of an eye.
I can believe that intersections might be more efficient if controlled by computers. But 100 times? That sounds like a cherry-picked scenario. I live in Houston, where there is congestion everywhere. I highly doubt that computer control could achieve 100 times the efficiency.
The article didn't say how many cars per minute / hour were simulated in the test, but I'd bet it was small enough that most cars could get through without stopping. Those kinds of intersections are already not a problem. When you start introducing REAL congestion, every algorithm will eventually break down.
Would it be any worse than what we've got now?
"When you have eliminated the unacceptable, whatever is left, however improbable, must be the truthiness" - Holmes
It could be I am unfamiliar with the algorithm so I can't predict how it would respond when the limits are reached. Before this is hailed as the cure for traffic jams its limitations need to be understood.
Knowledge = Power
P= W/t
t=Money
Money = Work/Knowledge so the less you know the more you make
Yet, how will local police make money if there are no lights to run or speed traps?!
Self-importance and self-indulgence is the root of ALL evil.