The weird, counter-intuitive science of traffic jams
washingtonpost.com
washingtonpost.com
If you don't think in static terms, this is obvious. Across any segment of time, some fraction of commuters will change their habits based on their expectations of time/cost/comfort. Some commuters can respond daily if, for example, they have a car and affordable parking at their destination as well as access to public transit. Other commuters may not have a car at all but will eventually buy one if it begins to offer substantial advantages over its cost. The point is that the system is dynamic.
And if, magically, the City of Boston could wave a wand and redirect 1% of its commuters to the subway, they'd soon discover that the 18% improvement in traffic flow has tempted many other commuters back onto the roads, and the benefit vanishes.
I think that there should be a law that a car's horn should be as loud inside the car as it is outside. That's how loud it is for nearby cyclists & pedestrians -- drivers would be much more likely to use their horn only when it's really needed if it was as loud for them as it is for others.
What's more I think that the manner in which many people drive is also an expression of their personality. Sure, some people buy a toyota tercel and drive around in a machine-like state, purely for the purpose of transportation ... but I think the majority of people drive in a more "personalized" manner, for lack of a better term.
I'm sure it's the case that google-self-driving-cars will be more efficient (and safer) transportation devices ... but the challenge will be that the masses will not easily migrate away from their personality-extensions (mustang, bmw, cadillac, convertible, truck, etc) into a pod-like utilitarian mode of transportation.
No doubt, one reason why traffic is so inefficient is precisely because of these "personality" effects ... but there is much in life that is inefficient. In an age of google (and others) identifying the "inefficiencies" in our daily lives, we need to have a discussion about what features of our behaviour, and our daily lives, define us as individuals and make us human.
Of course if part of your "personality" involves driving recklessly or putting people in danger, this is not a good state of affairs. But you get the gist of what I'm saying...
Look how quickly we switched from horses to cars. Horses are living animals which people probably had emotional bonds with in an owner-pet like way. Plus a lot of people, even to this day, love to ride horses. There was a huge infrastructure and layers of jobs to support that lifestyle, between horse-breeders, knackers, horse feed, etc. But once cars became practicable and affordable, they replaced horses in the time of a generation.
Imagine the 80s exec on his office phone behind the leather desk.
The power player in the early 90s in a "CAR phone".
The suburbanite with the really long cord so he/she can use the phone anywhere in the living room/kitchen vicinity.
etc.
All of these were replaced by mobile phones which also become diversified as personality brands. The same will no doubt happen with self driven cars.
So let's look for what the equivalent of the pickup, mustang, or prius will be in the self-driven era.
My prediction: - Mustang will be fast looking, but self-driven. With a manual option (for driving on a track some place) - Pickup will be strong looking, but self-driven and with a manual mode expected to be used a lot - Prius will be small and utilitarian looking, without much expectation to be manually driven
cool
First, when most vehicles are self-driving, you no longer need traffic lights or lanes. An urban street that is currently 2+2 lanes could be used as 4+1 in the morning and 1+4 in the evening if that is what traffic requires.
Second, self-driving vehicles remove a current disincentive to commuter microcars. A single person could easily get around in a car a third the size of a current subcompact if they didn't have to worry about being run over by land yachts. Beyond the lower cost, a half-lane vehicle could have a speed advantage in urban traffic.
However, there's one further requirement for removing traffic lights and lanes and such: no pedestrians (try to imagine crossing the street in the above simulation). If you believe that cities should be designed for human beings, then this is a real problem.
So let's say you want to remove the traffic lights and create a high-speed continuous-flow system comprised of lightweight energy-efficient micro-cars. How could you make this compatible with human habitation? By removing the roads entirely, and running the vehicles on cheap segregated infrastructure, either below or above ground.
Sound fanciful? This technology already exists, and is much more mature than self-driving cars. It's called "Personal Rapid Transit" (http://en.wikipedia.org/wiki/Personal_rapid_transit), and has been prototyped successfully for decades. The first commercial system began operating in 2010, and there are now three PRT networks in public operation (The Masdar Institute in Abu Dhabi, Terminal 5 at London Heathrow Airport, and a national park in Suncheon South Korea).
When it comes to self-driving cars, freeways / motorways will be relatively easy to automate, because they are already pedestrian-free continuous-flow environments. I expect to see vehicles available which can do that by the end of this decade, or possibly early next decade. Mixed-mode surface streets will be far more difficult, however, and while they are likely to become automated in the longer term, I suspect that this will only happen at very low speeds -- possibly as low as 10 km/hr, which is the speed at which humans and vehicles can interoperate in a truly safe manner.
Or, bridges.
Just start walking, the cars will steer around you! More seriously, there are already separate express and local routes for many major roads across the US, I don't think it has to be a radical departure from what is already there. Best part is, the way that self-driving cars are progressing to drive on regular roads, the infrastructure can grow more specialized for self-driving cars organically over time.
I'd say the easiest way to have streets with self driving cars and without traffic lights would be to build pedestrian overpasses. Just a nice small bridge to cross the roads.
My problem with that approach is that there seems to be very little room for error. And while we might say that AI drivers wouldn't make errors, the rest of the equipment is not necessarily so reliable. Seems like such an intersection would be one seized brake line or one engine sputter away from massive fiery death.
Imagine being able to ride your bike in complete safety because the car behind you is literally incapable of running you over.
That's one of the trickier aspects of robo-cars that we'll have to figure out how to deal with when they start making up a significant amount of traffic. When humans are in control, there will always be a few people who are inclined to act like jerks as long as they can get away with it. On the roads today, the imperfect drivers around us and the prospect of a nasty crash tends to keep us all honest. What would the troublemakers do if the roads were 95%+ robo-cars that will drive such as to never hit them, no matter what they do?
Lots of cities move at bicycle speed or lower, and the only reason cyclists are not out in traffic lanes away from where they can be "doored" is aggressive drivers.
You might even decide to bike if it got safer.
The difference is that the attention span, view range and response time of an AI is much superior than that of us.
Of course it won't be 100% perfect. But just enough to make traffic some orders of magnitude safer.
I'd love one of those, you could even have a pool of larger internal combustion engined ones if you know your trip is going to exceed the range of the golf cart.
No need for self driving cars to do this…
100% self-driving cars may reduce the marginal infrastructure costs of enabling this conversion, thus making it economical to do on more of the roads where it might be useful; its done now in some places, but its usually limited to the very worst bottlenecks because the cost of doing it safely (usually involving retractable dividers and lots of signage and other signalling systems.)
Learn people how to drive properly and keep their distance and things will be better. A good indication is on how much fuel you used in the traffic jam. (Electric car excluded).
An auto-distance holder also helps to reduce the oscillation by 1 car, however correctly adjusting by hand is better.
I suspect it applies to any large multi-lane highway / motorway / autobahn etc
It was really interesting to see this as it demonstrates behaviour I've been experimenting with and have long wondered what difference it made. It's great to be vindicated!
It won't fix it then, but it should help.
From the Minnesota DOT: "Minnesota DOT engineers developed what they call a "zipper," which meshes cars quickly. Signs advise drivers of the upcoming lane closure, tell them to use both lanes up to a point and then direct them to take turns merging. When traffic is flowing, drivers merge early to avoid unsafe maneuvers. But when traffic is congested, motorists make full use of both lanes. The data revealed that the change reduced traffic lines by 35 percent and also brought down "lane changing conflicts," says Craig Mittelstadt, Minnesota DOT's work zone safety specialist."
From the Virginia DOT: "The late merge should be considered for 2-to-1 lane closure configurations to improve throughput when large numbers of heavy vehicles are present (>20 percent) for the majority of the time and congestion and queuing are often present."
The Virginia site recommends late merges for when >20% of the vehicles are heavy vehicles, which is due to their merging complications (maybe something to do with those pesky "lane change conflicts"). It is obviously not a general recommendation.
There is no clear data for either position on throughput...because ultimately throughput is limited by the bottleneck, which is the lane closure, not the merge behavior. People who spend too much time in their cars tend to be religious about the appropriateness of certain driving behaviors, and like any faith-based conviction, will tend to over-extrapolate empirical evidence to their context of choice.
You mean, besides empirical evidence? Nope.
I see it almost every day on my commute, which is riddled with construction. On the days that everyone merges well before the last possible moment, traffic moves at a quicker pace, because there is no one to cut you off, and my commute is shorter. On days that people decide to cut others off at the last moment, my commute is measurably longer.
If you merge ahead of time, you are more likely to find an existing gap in traffic, which has little to no effect on the traffic's speed. Whereas, if you are merging at the last possible moment, you are necessitating a gap be created specifically for you, forcing others to slow down.
If cars stay in the eliminated lane until the last moment, and then the speed of traffic increases after the bottle neck, it was the merging at the last possible moment that caused the slow down in the first place.
The behavior you are observing is not zipper merging -- it's people being assholes and trying to skip traffic.
If everyone merged at the last second it would move better.
I didn't say "this one time back in 2001". This is everyday for years. These are consistently observable data. Cars merge early, my commute is under 30 minutes. Cars merge late, and not just one or two (the north shore of chicago is chock full of entitled assholes), my commute is over 40 minutes. Sounds like a falsifiable hypothesis to me.
One of the main reasons airplanes and trains are much safer than automobiles is that planes and trains have outside coordination of traffic, a third party who can see the bigger picture. Automobile traffic is thousands of individual actors who focus mainly on their own interests, and consider others around them only to avoid near term problems. Hoping everyone will magically act in concert, without government enforced obligation, is a pipe dream. Which is what the article points to, and proposes to solve with self driving cars. Self driving cars would allow the outside coordination that could facilitate zipper merging.
- The traffic jam is longer because less road is used and the traffic jam often reaches so far back into intersections that other roads get blocked that are unrelated to the actual blocked lane.
- Several merge points slow down the traffic more often which create several waves of breaking and accelerating, instead of just one breaking instance.
By merging at last possible moment, more cars can be fitted in the same length of the queue. obviously traffic wont have higher throughput through this particular cause of queue, but the overall effect on the city is substantial.
There are two things I do not trust: I do not trust that our technology can successfully and intelligently respond in the event of an traffic or roadway anomaly, the kind that usually causes car crashes i.e. damaged roads, debris falling off of cars, erratic drivers, etc. Although I do trust safety features, which are pretty nice lately, I would rather have not have a wreck.
Finally, I do not trust other drivers and I do not know how they will react/interact with a self driving car. It only takes one time for an idiot to flip out on the highway and cause a wreck.
When I feel that the self driving can adequately handle anomalies, wrecks, and suicidally bad drivers, then I will have no qualms with the self driving car. Till then, I'm not willing to bet my life that it works well enough.
Anecdote: I was driving on a 5 lane interstate when a driver to my right started to merge into my lane. I merged left and at the same time looked. There was a car in the left lane, so I stopped before I merge all the way into the lane. The driver in the left lane, being on her phone and not paying attention, panicked and way over corrected to her left. This sent her across the inside lane and into the median barrier then over corrected again and flew across all five lanes coming to a rest unscathed on the shoulder. It could have been a lot worse, and I doubt a self driving car would have done a maneuver that aggressive putting all five lanes at risk.
This was before I started thinking about the unknown variables that cause accidents: deer, a tire exploding, other debris (nails, glass, etc), some manufacturer or software bug.
It's all too easy to imagine the software doing quite unexpected things when it gets to an unknown. Even if you could take control of the car I imagine most people wouldn't be paying attention to actually do it.
This is just my panicky train of thought though. The tech is cool.
Not saying we'll get there anytime soon, I don't know well enough to say. But I find that people in these discussions tend to overestimate the skill of most human drivers on the road.
The fact that we'd be distanced from the stress of driving, as you say, is also a great benefit, we're not really made for coping with the irritating nature of congested driving.
As to the article, in Atlanta one major change that helped congestion was to implement throttled access via on ramps governed by lights. This tends to keep people from merging in groups which has a rubber band effect on existing traffic.
Still more can be done by restricting areas of the highway to limit the lane changes. In Chicago I saw this done with bypass areas separated by cement walls. A simple solution but a bit cumbersome. Still separating some outside lanes would help.
* Far fewer cars needed — as more cars could be shared
* Far fewer parked cars — as cars needn't stay with the 'driver'
The two cases are linked: a car that doesn't have to wait around where the passengers are dropped off can return to base or can be away performing more lifts while the original passengers are busy.
Families would have need for fewer vehicles. Sharing a vehicle would make more sense.
If cars are busy during a drop-off-pick-up cycle there is no need for them to be clogging up the streets.
You'd get off a plane in a foreign city hit Web2point0nameforcarservice.com and a car would drive up (sending you a picture of the car and it's license to you and your details to the site (so we know who to bill if you throw up in the car for example)).
This way you spread the cost of owning a car but it's available to you if you want to go away for the weekend (since you just mark it as out of service).
It's going to be a wild ride (pun intended).
AutoTaxi.
1. Most outward trips would be chained: drop 1, drop 2, drop 3, return to base; rather than: drop 1, return to base, drop 2, return to base, drop 3, return to base. As would most return journeys.
2. The roads would be clearer in cities due to fewer parked cars. In some cases there would be an extra lane now available.
3. Cars could still have multiple bases, especially if they are shared. Car parks could be re-utilised as a stop-off points where empty cars could 'swap out' and remove themselves from the system briefly.
4. Chains could be linked: do the early run, do the late run. Do the Smiths; do the Joneses.
5. Empty cars with no urgent pick-up could wait for the roads to clear before embarking on a return journey.
6. Would anyone care if a car returns to base slowly if there wasn't an urgent passenger waiting?
Cars could be more efficiently used and that would certainly imply that the fewer number of cars in existence would be being used for more trips throughout the day.
I admit that it will be hard to remove the rush-hour problem entirely without a significant amount of car-sharing. However with a fully self-driving system cars could do clever passenger shares (and/or mid-route swaps). Passengers would be free to only car-share within their sex / family / workplace / social networks etc.
Some small pilot studies have been carried out in Europe, often with successful results. Basically, the idea is that we become more alert when a situation requires us to be. An intersection that appears riskier forces us to give it more conscious and subconscious attention.
http://knowledge.allianz.com/mobility/transportation_safety/...
Anecdotally, I can certainly testify to the effects of traffic lights on traffic. I grew up and worked for a lot of my life in LA, a city as famous for its traffic as for its movies and weather. There are certain parts of the city in which every damned block has a traffic light, and blocks are very short in the denser areas. These areas are traffic hellholes. Beverly Hills in rush hour would have made a cameo appearance in Dante's Inferno, had he been around to experience it.
This is a weird adaptation of the Peltzman effect[1], which found that people drove more recklessly when wearing seatbelts.
[1] http://gregmankiw.blogspot.com/2006/07/peltzman-effect.html
level crossings have been very common here in the uk for a long time... open crossings are the exception rather than the rule, and you have to go far out of your way to find them.
We know.