# Could red lights be eliminated if all vehicles were automated? [duplicate]

At a typical intersection that uses red lights to control traffic, vehicles traveling on one roadway have a green light and can travel through the intersection, often at full speed, while vehicles traveling on the intersecting roadway have a red light and must come to a full stop until the signal changes.

This is a good system for human-controlled vehicles, but I'm wondering if, in a world where all vehicles are automated, it might be possible to eliminate red lights completely and instead use precision timing to permit traffic from both roadways to pass through the intersection in roughly an A-B-A-B pattern. Kind of like how some marching bands can have two line of marchers pass through each other orthogonally.

What would be the limiting factors to such a system? Let's assume that we'd never want two vehicles to come any closer to each other orthogonally than they would driving parallel in adjacent lanes. At what kind of traffic density could traffic in both roadways go full-speed through the intersection?

• Welcome to Worldbuilding! Have a +1 just for finding a way to properly use orthogonality in a question. Jun 1, 2015 at 18:52
• The limiting factor would be . . . Pedestrians. If you could eliminate them from the equation (via dedicated pedestrian bridges/underpasses, etc.), red lights would no longer be necessary. Jun 1, 2015 at 20:17
• Instead of a red light, whenever pedestrians step near the road -- perhaps a designated "Wait her for cars to stop" spot, the pedestrians can just walk up to the intersection, and the cars will automatically stop for them -- no red light required. Perhaps cars will have some kind of beacon that indicates that they are stopping for a pedestrian, or maybe even a laser that tags pedestrians that they see so the pedestrian knows whether or not the approaching car sees them. Jun 1, 2015 at 20:23
• @Johnny And now, the annoying kids can play a much more annoying game than "Press the button at the pedestrian crossing and run away." Jun 1, 2015 at 23:47
• It's already possible like this without automation. If you don't believe me just take a trip to India. :-) Jun 2, 2015 at 2:28

## Yes, for some cases.

Assuming the conditions you've given, mainly this is "a world where all vehicles are automated".

Interleaved single vehicles
For intersecting traffic travelling the same speed, vehicles need to stay 7.4 meters apart. Assuming you have perfect information and control of the cars, the spacing is just a car with enough space in front or behind to allow another car to pass as you described. If the car is 4.5 meters long and you only need one meter diagonally, (since the closest the cars come is at the corner for an orthogonal approach) which means $1\over \sqrt 2$ or $\approx 0.707$ meters in front and back for buffer. Since the car approaching the passing car is approaching the other side, the width of about 1.5 meters also should be accounted for. So, 4.5 m length, plus $\sqrt 2$ m for the buffers, plus the 1.5 m width is about 7.4 meters total space required per car.

If a car is traveling at 100 km/h, it travels its own length and buffer space in just over a quarter of a second. This means if that car is travelling perpendicular to a stream of traffic, there needs to be at least that much time gap between two of the cars (with all buffers accounted for). If they're travelling at the same speed, then that time in terms of distance is that same as the intersecting car length and its buffer.

The density would then be about 135 cars per kilometer. Around "jam density", which is ten times higher than today's "stable" density for traffic flow.

I'm not accounting for the wind turbulence between vehicles passing each other at high speed, which would be rather confounding to this whole exercise.

Interleaved trains of vehicles
As mentioned by JDługosz in the comments the cars could also be in tightly packed groups, with the same average density. So the density won't change, but vehicle efficiency can be increased. Higher efficiency can be gained by grouping vehicles close together because the drag on rear vehicles is decreased; they ride in the draft of the first. The efficiency won't increase dramatically beyond the already significant boost from the sub-10 meter following distance, but would likely be worth the trade-off. Turbulence between passing would be minimized as it only greatly affects the lead cars.

Hybrid interleaving

Also pointed out in the comments, the high density closely packed traffic could also be used up to the intersections where the cars actually accelerate to gain the proper spacing. This will still have the turbulence problems as individual vehicles.

The reason this will not happen anytime soon
The human occupants will freak-the-eff out. Grandpa will instinctively lurch the wheel to avoid hitting one car only to slam into another (this is a joke, clearly manual control would be disabled during such a maneuver). You can test this freak out on yourself, just watch a few "near miss" videos and tell me you don't clench a little even watching a video for the second time when you know what will happen. Now imagine being in the vehicle and not knowing.

More likely, for safety and passenger sanity, high speed roundabouts will be used. This would:

• Minimize car-to-car relative velocities.
• The turbulence problems associated with interleaving high-speed cars will be reduced
• Allow vehicles to do something other than go straight through an intersection.
• Require more space.

But this would only be for intersections that:

• Don't also need to allow pedestrians to cross, people will still be walking around in the future, presumably.
• Don't have any other unautomated vehicles, like bikes, segways, or McFly-hoverboards.
• Don't have space restrictions like those inside cities.

One other very cool idea about automated cars is having a form of Uber, where people let their car go drive people around in full auto mode.

• Assuming that Grandpa even has a steering wheel to jerk, of course. Why include one if the entire thing is automated? Jun 1, 2015 at 19:15
• Some answer or comment on this site once said that self-driving cars of the future would probably have opaque windows, so that its passengers would never have to see the terror of the journey. That would solve the problem of old people like us who aren't used to such efficiency. Jun 1, 2015 at 19:19
• When approaching an intersection, the vehicles would be able to speed up to allow for a larger buffer zone between vehicles, while maintaining the overall throughput of the roadway. The larger buffers would allow for the cross traffic (which is also moving faster through the intersection) to pass. imagine, if you will, a slow conveyor belt feeding a much faster one. The bags would still pass at X bags per minute, but the velocity and spacing would both be much greater. Jun 1, 2015 at 20:24
• @JDługosz Yeah, I suspect that would improve the overall efficiency by allowing the rear cars to draft the lead car. Jun 1, 2015 at 20:36
• In this world, Google already has a prototype car with no manual controls. Jun 2, 2015 at 3:30

No, because not all vehicles will be automated. You also need to consider bicycles, and to some extent pedestrians. So for the non-automated users of roadways, some sort of red/green light system will remain.

So lights will stay, but automated cars will automatically time themselves to hit lights instead of doing stop/go traffic. You'll get something similar to your marching band pattern, but "A" might be a block of 10 cars, followed by a large gap, then "B" as another large group of cars. And so on. This leaves room for pedestrians and reduces the possibility that people in the cars will freak out and think they're almost dying.

• Comments are not for extended discussion; this conversation has been moved to chat. Jun 3, 2015 at 8:38

Yes, this is completely possible and the basic logic for it has already been mapped out -- http://www.citylab.com/tech/2012/03/what-intersections-would-look-world-driverless-cars/1377/

Scroll down to the video near the end of the page to see a simulation of a busy 6-lane intersection with fully automated vehicles.

Note that the simulation includes human-driven vehilcles as well -- "The yellow cars pausing at the intersection in this simulation are old-timey human-driven vehicles.... Those human-driven cars would have to wait for a signal that would be optimized based on what everyone else is doing. And the same would be true of pedestrians and bike riders."

• That looks exactly like the intersections in Cairo, Egypt. Jun 1, 2015 at 23:10

Not today, but we certainly could, but you would still have to stop and many times. This is just based on me being a transportation planner for several years in the recent past.

Marching-band style - I think we would have to convert all of the junctions into roundabouts. The problem is that with unbalanced intersections, you don't want a roundabout: in rush-hour where everyone is making a 3/4 left turn, some poor suckers are going to have to wait. Also, there will be times when you have to stop for pedestrian & cyclist movements, but our driverless cars are getting really good at handling that.

Alternative - We have very sophisticated (and really cool!) models, which have come a long way. In Florida, we had cameras at every light as well, for inserting data into the model, for turning a signal if traffic didn't behave as expected, and for communicating with the other signals downstream.

The combination of the model, the cameras, etc., could become sophisticated enough to communicate with the cars, so that lights aren't needed. But your car would most likely still have to "stop" at some critical junctions, and to allow for pedestrian movements.

Note - By the time we are fully driverless, there will presumably be a lot more cars on the road from (1) population (2) from the ability for people to move out of town, and (3) ability for some disabled & visually impaired to 'drive'.

Further, because of these smart cars, we'll have a lot more capacity so you won't be seeing as much rapid expansion as you do now. So it will be the "same" roads, but with a lot more vehicles happily crammed into them.

• +1 Good point about increased percentage of the population being able to drive. Jun 1, 2015 at 22:48
• Yeah, I'm torn on the effects of driverless cars. On one hand it helps the transportation-disadvantaged, including children and the elderly, while reducing crashes. On the other hand, people will have no problem living a 3 hour commute in the awful suburbs since they can read, sleep, etc. Sprawl: ick. Oh, and I'm in Portland. so looking out the window and imagining the multi-modal chaos those poor cars will have to negotiate. Jun 2, 2015 at 4:22
• I'm in those burbs (Tualatin) bike commuting to another burb (Lake O.), but I'm currently buying another house off west Burnside and doing the reverse commute. I would love a self-driving car for that. As long as cities have limited urban sprawl, like Portland (mightiest of cities), things shouldn't be too bad. I think all around it's a positive thing for the world. Bar hopping will be easier at least. Jun 2, 2015 at 5:11

No because it is not fault tolerant

This would work perfectly until the first car suffers a blowout or a sensor malfunction or until a child runs out into the road. Automated vehicles will be designed to move in patterns that are fault resistant or otherwise they will result in spectacular pile ups.

Intercrossing the cars at speed means that no car is in a position to safely take emergency action to avoid impact. This is not the case with, for example, close-following car "trains" since automated cars have no reaction time (or trivial reaction time, more accurately) and thus the whole "train" can be brought to a halt as a single unit and avoid accident.

Yes because signal lights are designed to be human understandable

Answering just the first part of the question and ignoring the second part, it seems likely that red lights themselves will be removed and the city will instead communicate directly with the cars to interrogate them about their destinations and plan traffic flows accordingly. Cars would still stop at intersections but advanced traffic flow management would minimise the time that they would do so and avoid needless waits at uncontested intersections. The only lights needed would be those that signal to pedestrians when they are safe to cross the road.

• Some places have removed traffic lights altogether and crash numbers reduce- I saw it on a tv show. Probably because traffic lights are inefficient at managing people and cars- self-managed is better. Jun 2, 2015 at 11:33
• You seem to ignore the potential of car-car communication. There's no reason that the vehicles can't communicate with the vehicle(s) around them and behave much like the 'trains' you mentioned. If an obstacle were to be detected, all lanes might 'compress' away from the hazard, as well as simultaneously slowing the traffic speeds. This could be localized as much as traffic load and buffer space allowed. Jun 2, 2015 at 15:35
• @MrWonderful: No, that doesn't work. Because the cars are coming from different directions, they cannot deal with the cars in the orthogonal direction even with car-car communication. If a car going North needs to emergency stop to avoid a hazard, no amount of car-car communication will allow the car going East to brake in time to avoid it in an orthogonally crossing arrangement. The car train approach only works if the cars need to slow or take avoiding action in the same direction Jun 2, 2015 at 17:24
• @JackAidley - Perhaps, but vehicles don't usually stop instantaneously. Even in the event of a catastrophic failure (tie-rod breaks, wheels fall off, etc.) the vehicle with the malfunction will take some distance to slide to a stop. A vehicle with tire/road contact should have better braking capability than the vehicle with the steel/road variety. It is conceivable that if an accident were avoidable at all (by a human driver, for instance,) an automated vehicle could do the same. If one can avoid the accident, then they all could. If it is unavoidable, losses could still be minimized. Jun 2, 2015 at 18:10
• @MrWonderful It's because cars don't instantly stop that there's an issue. The car with an issue begins to slow immediately. That means it spends longer in each lane than it's supposed to. There is a car going the other way that's about to occupy this car's newly-freed space. High-speed interleaved traffic without pretty good margins between cars makes accidents in that case unavoidable by anyone; that's one reason we don't do it right now. Jun 2, 2015 at 19:29

This is rather unlikely, from an engineering perspective. A marching band can pull off a trick like that because they're moving relatively slowly, and because a person is generally about as wide (distance from right to left) as they are long (distance from front to back). Cars, on the other hand, are considerably longer than they are wide. This means, as Samuel pointed out in his answer, that you need very wide buffers between cars in order to make it work.

But there are problems with that. First, the system might work well, if timed perfectly, for intersecting flows of cars, but as soon as the first emergency vehicle, bus, or semi truck shows up at an intersection, the entire scheme blows up. (Hopefully not literally!)

Second, the required spacing makes for very low vehicle density, when one of the main selling points for the idea of ubiquitous automated vehicles is that such a system could safely handle high vehicle density at high speeds, while human drivers can't. To make this work, then, you'd have to either stay at low density the whole time, or be constantly speeding up and slowing down, with the line of cars expanding and contracting like an accordion, as you move through the city. Acceleration and deceleration waste energy; the most efficient way to drive is to get up to a stable speed and stay there.

A more likely scenario would be to have your AI actively control the traffic lights to improve efficiency. Have you ever been stuck for what feels like five minutes at a red light, with long lines of cars stacked up in both directions, while maybe one car every ten seconds goes by in the cross street? Or had to wait through multiple red lights to make a left turn because the green left arrow only lets two or three cars through at a time for whatever insane reason? (Why yes, I have lived in LA. Why do you ask?) Fix problems like these and you'll do a lot more good than you would by eliminating stoplights entirely. (Not to mention it works a lot better during the long transitional phase when most cars are still driven by a human being.)

Though you might not have red lights, you might still end up with behavior similar to modern intersections, albeit with much more efficiency. For instance, while cars constantly intersecting at full speed introduces way too many variables and unnecessary dangers, I can see cars positioning themselves into bursts; you either speed up or slow down until cars on your road are clumped together, and then passage of the intersection alternates between these clumps. This should only provide a minimal delay at each light, rather than the possible minutes-long waits we have today. After you get into a burst, it may seem like you're not waiting at all. Plus, this allows for a more scalable design than to simply throw a car at an intersection and try to figure out who needs to slow down to avoid a wreck; this way, cars can be handled in groups of known size and speed, which makes adding a hundred feet or so for burst switching not all that bad.

However, at the end of the day the roadways will probably still get congested; the efficiencies of automated transportation will eliminate most traffic, but I'm sure there will still be areas in which you slow to a crawl or even stop before going through the light. These areas will operate much like modern intersections, but without anyone having to worry about bad drivers who didn't see the light change.

The only exception to this is when someone wants to cross the street. For that, you'd have to stop everyone who's perpendicular to the crosswalk. You might just let people cross when the parallel burst is going, but I'd rather be on the safe side (otherwise people will complain, and we'll all go back to human-driven cars). Another alternative is just to build sidewalks over the roads, so humans don't have to worry about intersecting with cars.

I spent some time overseas in a country that uses roundabouts far more frequently than stoplights. Notably, drivers very, very rarely stopped when entering a roundabout to yield to cars already in it. Instead, drivers rolled into roundabouts at nearly full speed and simply continued the flow of traffic once inside. It was a little hectic, but I never saw an accident or even a near-miss!

These roundabouts were used by manual drivers. I'm sure that automatic cars would be able to use them even more efficiently.

It depends on if the automated vehicle requires visual input to make judgements.

Many automated vehicles that are in use at the moment, use a mixture of camera and depth data (sonar, lidar etc).

If the car in front is perhaps travelling on a motorway, quite abit further ahead, they may be out of range of many depth sensing equipment, so an on-board camera may want to keep track of the car from afar using visual signels (red light might be used in this)

• Why not WiFi / BluTooth / Some-other-RF-technology? Jun 2, 2015 at 18:24
• Can get alot of interference with all of those, plus, it would be difficult to pinpoint your car in relation to another car without line of sight as the only coordinates you could transmit across the RF tech would probably be GPS, which even the best GPS are only accurate to a few feet, which when travelling at 70mph makes all the difference. I mean its possible, but I think would be too unsafe to be used on it's own, when you could easily add camera/depth data to increase safety Jun 3, 2015 at 7:33
• I never intended to replace any other technology, but to augment it instead. You mentioned an issue with out-of-range sensors, and it seems that, coupled with GPS tracking to within a few feet, some RF technology would be sufficient to keep track of the vehicle that was out-of-range. Jun 4, 2015 at 7:36

If you take the question literally: If all cars are automated, then sure, you don't need traffic lights, because traffic lights are designed to convey information visually to humans. With automated cars it would be far more reliable and efficient to have radio signals sent to a receiver in the car.

Several posters here have rejected the premise that all vehicles on the road are automated, saying that there would be bicycles and pedestrians. But who says? There aren't supposed to be bicycles and pedestrians on interstate highways. Maybe we have overpasses or tunnels for pedestrians.

But assuming that what you mean is not traffic lights per se, but rather the idea that all traffic in one direction must stop for a period of time while traffic in the other direction flows; and that one way or another we manage to not have pedestrians crossing the intersection:

Yes, with an automated system you could presumably have vehicles travelling closer to each other without creating danger. But, as Jack Aidley points out, you surely still want to build some safety factor into the system. Exactly how reliable are we assuming that the automated cars are? What if a car can't QUITE accelerate at the intended speed, or if a car has a mechanical failure, or hits a pothole or some object in the road? If cars are whizzing past each other at 100 miles per hour with inches to spare, than even the slightest disruption of the system could result in disaster.

Ultimately, cars going one way have to slow down or stop to let cars go the other way. Presumably an automated system could analyze the flow going in each direction and time this better than our simple traffic light systems. To take the obvious case: If there are no cars at the intersection right now who want to go north or south, then there's no reason why cars going east or west should have to stop. My first thought if I was trying to program such a system would be that we would time it so that the direction with the most traffic gets the most "go" time. Like, if there are twice as many cars going east/west as going north/south, then we should give twice as much time to east/west cars as to north/south cars until the proportion changes.

I suspect that having cars slow down as they approach the intersection to lets cars going the other way pass, and then speeding up when it's his turn, would be more efficient than having them drive up to the intersection at normal speed and then stop. Accelerating and decelerating take energy.

Would it be more efficient, either in terms of energy use or of the time drivers have to wait, to have cars go one at a time in each direction and interweave like you describe, rather than letting ten pass in one direction and then ten in the other? I'm not quite sure. I'd have to model that mathematically, or build a simulation.

Simple answer: Who pays when there's a failure i.e. crash Assuming there are humans in the car, the cost is high if death or injury occurs.

Assuming there are just goods, there are still costs on failure.

Insurance is a mechanism society has devised to offset the fail rate in cars. Will insurance companies back a "no-red-light" car operation service?

Governments support the red light mechanism because it's considered part of road services, and people are willing to pay taxes for such road services...i.e. the overall costs, behavior and reliability are considered worth it.

I suspect you just need a couple of high profile deaths (people burn to death) before your no-red-light solution gets replaced with red-light solution.

It's not just a question of whether you can get cars to interleave 90% of the time.

It's the 0.01% problem. You have to say what happens then.

Roundabouts (Traffic circles) would be much better both for automatic cars and manually driven vehicles. It allows for a relatively continual flow of traffic, and during the transition period, both types of cars can share the road and the roundabouts. Pedestrians and cyclists should be given bridges or tunnels to bypass the roundabout completely.

OTOH, roundabouts do need to be engineered properly. My home city of London, Ontario has a city council which has an unfortunate habit of falling for fads with taxpayer dollars and has been building roundabouts that are far too small. Roundabouts in the UK and Europe are generally twice the diameter of the ones that are being built here, so in effect what we now have is a rather large 4 way stop, rather than a means of allowing traffic to flow smoothly.

This is the one huge disadvantage of roundabouts (or cloverleaf intersections like we see on expressways and highways): they take up a lot of land area. Retrofitting every intersection to be a roundabout would involve demolishing a lot of real estate to get a large enough diameter circle. Quite frankly, it might be more cost effective in terms of real estate to go for an alternative like Personal Rapid transit (PRT), which also avoids issues like stop lights and having pedestrians and cyclists sharing the same roadway.

• We don't have many roundabouts here in the US, but when I've encountered them, I've found them awkward and dangerous. Instead of just stopping and waiting your turn, you have to merge in to a lane of moving traffic, and often switch lanes between an inner and outer circle. Sure, getting out of the roundabout is easy enough if you're in the outer lane, but you had to get on it first.
– Jay
Jun 2, 2015 at 14:09
• @Jay: That's because in the US, roundabouts are often used as a means of impeding the flow of traffic. Jun 2, 2015 at 18:13

No, because there are a lot of cities that still look like this:

http://www.mapmoose.com/maps/London.jpg (8MB image).

This picture doesn't do it justice and London is hardly the worst example of this either. There just aren't that many intersections and you can't deal with every possible way that two (or more) roads might intersect. You could make the argument that it would be better if the road layout was well designed but the cost of redesigning and rebuilding a city is prohibitive.

Traffic flow is also not going to be consistent even with driverless vehicles. It might work in an idealised highway/motorway environment but on regular roads you have to consider cars pulling into the flow such as out of driveways or from street parking. If you have a consistent traffic flow you can't add or remove any cars since the flow behind where the car was added or removed would either need to slow down to allow space for another car or speed up to fill the gap that was left. This would then mean that the smooth flow through an intersection with large gaps between cars as described in @Samuel's answer would break as the distance between each car and each car's speed has been changed from the norm.

• Would it be posible to time it so one car moves in every direction without hitting anything ? Jun 2, 2015 at 10:46
• I've no numbers to back it up but I doubt it, will add a bit more to my answer about traffic flow in general. Jun 2, 2015 at 11:01
• There are also a lot of places in the world that aren't cities. What's the cost of putting over/underpasses on every road like this one? en.wikipedia.org/wiki/Nevada_State_Route_722 And how about deer, cows, wild horses, bears, &c? cache4.asset-cache.net/xt/… Jun 2, 2015 at 23:36

YES

Swarm Technology.

I have actually been thinking about this a while ago but some odd reason, but I do generally think this is completely possible, but not with today's technology.

If we we're able to have each car as a node where they relayed information about themselves - such as telemetry, speed, length, sensor information (Sonar, RFID, proximity) etc - back to the central server.

With all this information being relayed back to the central server (Has to be a quantum server) the server can then relay individual commands back to the car, wherever that would be to slow down or to speed up, stop encase of an emergency etc.

If we were able to tick off all of these requirements, then one could assume that this quantum server will always know the position of every car connected to it, and thus know exactly what is happening with every node connected to it and react with what is being presented to it by each cars data.

• Use a distributed algorithm instead of a centralized one. We've been using distributed scheduling algorithms in mobile ad-hoc networking for a couple of decades now. They would need to be tweaked but a lot of the foundation is already available. Jun 2, 2015 at 19:39

There are already some great answers, but I wanted to take a somewhat different wording on my approach.

Traffic lights are there to regulate the traffic, but essentially, for a logic point of view, they try to turn synchronous something which, by essence, is asynchronous. The idea is to group the cars in bunches. Those bunches are moved together in each iteration. This helps regulating the traffic.

Why would you want to suppress them? If you actually make a world where everything IS synchronous. If you make an analogy to an FPGA, if you have everything running in the same clock domain, no need to insert flip-flops.

Now, it is doubtful that you will get to that point. Indeed, pedestrians, bicycles, animals, children running around, accident and emergencies are asynchronous. Accidents or seemingly rare and/or discrete events could probably be handled by a short-interruption of the system, and adjusting regulation measures. However, continuous, asynchronous events are a pain. All those pedestrians wanted to cross that street. You need to tell them to wait until you (here the AI) set the measures to let them pass. And probably a traffic light at a crossing point is the easiest. You need to synchronise them: use a flip-flop.

You should note that the cars are synchronous only as long as you run on the same clock domain. If your AI is limited to a city, all the cars coming to it have the same problem. And if your AI is worldwide, you'll have some fun in synchronising the whole clock thing on that scale!

Similarly, how are your cars starting to get in-line?

Now if you had some more items, like a separate levels or ways for cars and pedestrians, or communication devices directly implanted into pedestrians, or confined all animals inside and no-one ever gets out without its car, etc.

TL;DR

Probably not. But maybe you can reduce their amount.

For vehicles you can just use roundabouts to remove some traffic-lights, but not for pedestrian crossings, (unless you build foot-bridges or pedestrian tunnels to replace them.) Even some roundabouts need red-lights where they intersect with trains/trams, but those could, (where there is enough space) be solved with bridges/flying-intersection over the top of the tram-line. (Some of the parts of the unfinished London Box have such raised roads.)

If you take the Napoleonic method, (tear our large swaths of your city) then you can have the space needed to do this... or invent flying cars and then you have moved the problem into a higher dimension, with additional problems and possibilities.)