Why driverless cars may mean jams tomorrow
economist.com
economist.com
However they won't solve the fundamental problem of roads being a scarce resource that lack a price signal so the demand is infinite. If you set the price too low you get congestion. If hamburgers were free, everyone would want one and a queue will develop, so instead of paying money you pay with time. It's just the same as "free" parking where you pay time instead of money. Obligatory mention of Donald Shoup and "The High Price of Free Parking".
Something is "free" if, at a price of zero, there is enough supply to meet demand.
It follows that nothing is universally free, it can only be free within spatial (and probably temporal) constraints. The usual candidates (water, air) are certainly not free everywhere.
It also implies people are willing to move from owning a car to time sharing a car. That's not likely, because it'll be businesses that will offer it, and they need to make a profit. As it stands, Uber is more expensive than taxis (and they undercut taxis using dank investment monies); take away the driver and replace it with a more expensive car, and I don't think they'd be able to operate cheaper. Plus, traffic comes in peaks, so that fleet would have to be optimized to handle said peaks.
And of course, a large percentage will need (and want) to use it. It's possible if the transport system becomes government owned or it becomes competitive with car ownership, but I don't see that happening. Sci-fi utopias usually don't explain why everything is going as smoothly as it seems; one explanation is that companies are just that awesome and managed to get a cooperative traffic system up and running whilst making automagic cars cheap enough for everyone, another is that the government or some other singular but benevolent organization is in control of all transport and is making sure it's standardized.
That's what Google's marketing says. Independent research indicates that self-driving cars may need to be driven many hundreds of millions, or even billions, of miles before their safety is actually proven [1].
This makes sense. For self-driving cars to be safer than human drivers, they don't just need faster reaction times, they also need an ability to understand and reason about their environment that is as good as humans'. And that's a long way off.
_____________
[1] Driving to Safety: How many miles of driving would it take to demonstrate autonomouse vehicle reliability?
https://www.rand.org/content/dam/rand/pubs/research_reports/...
Autonomous vehicles would have to be driven hundreds of millions of miles and sometimes hundreds of
billions of miles to demonstrate their reliability in terms of fatalities and injuries.US airport landing slots work like that now. Commercial flights don't take off for congested destinations until they've booked a landing slot. This eliminated circling the destination waiting for a landing slot.
IMO it's big employers that should push towards setting up large bus / train hubs in cities surrounding them, allowing commuters to only need to get to the bus station themselves. There's some companies that generate multiple percents of commuter traffic pretty much everywhere. They're the ones that should also do their part, by offering more efficient transportation and by opening satellite offices. There's no benefit in having 50.000 people work all in the same campus.
I guess it depends where. Here in London buses are everywhere.
It's an interesting take on who's responsibility it is to provide transport. My personal view (neither more not less valid than yours) is that it should be the job of the local authorities to provide transport (council or state). It will help companies of all sizes to grow rather than leaving the responsibility on a few big corporations, and it would give access to transport to every employee, not just googlers. That's only possible if the tax system is well thought and functioning, including taxing said big corporations.
It takes me about 45 minutes to get to work by walking + train, and there's about one train every hour (less often in the morning), so it's inflexible.
If I go by bike, it's about 55 minutes (but then I need a shower, which can take a bit more if your company doesn't provide the right facilities). I'd say that the time saving for a train is ridiculous if we factor in the price for the railway system.
If I go by car it's 25 minutes, but then I need to find a parking spot, and usually pay for it.
If I could rideshare to work and just be dropped in front of my office, and forget about it, with a price on par (even slightly higher) than my car ride, I'd pick that solution most of the times (I'd still ride my bike for training that's it).
Also I might have errands to run, team sports (or I may not, depending on the weather). OR how about you having to leave early, because you have to pick up a sick child from school? Current ride sharing leaves you stranded or unable deal with unplanned events. That's why people drive.
With driverless cars, that can pick you up on demand, some of these issues go away, maybe, but how do you coordinate sharing a random ordered ride with someone? How long are you suppose to wait, before the car decides that the other person is now late, and takes off without him?
Not to be too snarky, but as a non-american it has been amazing to watch the internet debate on the future applications of automated cars these last few years. Slowly but surely people are starting to figure out that public transport is a good thing.
Sure, I don't live in a big city, so maybe the public transport is not as effective as it could be; but many people I spoke with even in London (and other places like Berlin, Paris, Milan) seem to share the same suffering: constantly using public transport ONLY is annoying. Unless you plan your life now & forever (e.g. you buy an house on the right subway line to your job, and never change jobs) you'll find a moment when it's hard to do what you'd like to do. Want to go to the gym? Ouch, the last subway ride ends before you'd go out. Need to collect your kids before going home? No luck, their school is in a badly-served zone.
That's the issue with public transit.
Self-driving cars available on demand might solve a lot of that. However, like most "utopias," I don't think we will ever achive the ideal efficiencies that self-driving car advocates imagine. The real world is too complicated.
Is that really a BUS problem in your area? This happens in Italy as well. But the root cause is cars - cars that happen to be "momentarily parked" in bus stop areas, that force bus drivers to stop in the middle of the road.
> They are slow.
Very often, that's true, they're painfully slow.
But trains don't work at all well for serving smaller volumes of people moving in stochastic patterns at arbitrary times of day. Neither do buses. Cars serve this need very well -- they also generate massive negative externalities in the process, but they do serve this need.
Why should we care about "people moving in stochastic patterns at arbitrary times of day"? Transport planners call this "incidental travel," implying that it isn't really significant. But this is because transport planners don't have a good means of addressing it. In fact, incidental travel actually the majority of trips that people take: going shopping, visiting friends, going to the doctor's, eating out, etc. The balance is about 60/40 for incidental travel / commuting.
People own cars to serve this need, because nothing else will (unless you're in a mega-city that can afford to run high-frequency mass transport 24/7). Then, since car ownership is a sunk cost, people use their cars for commuting as well, where they're really sub-optimal.
Travel demand is an incredibly heterogeneous problem, which can only be addressed with a hierarchy of modes. That includes mass transport at fixed schedules, individual transport that is on-demand, and intermediate collective transport modes that currently doesn't really exist. Autonomous vehicles with 4-8 person capacities and demand-responsive routing would solve a broad class of transport problems that conventional public transport cannot -- and could do so without many of the negative externalities of private cars.
I've written much more on the subject, if you're interested: http://archive.podcar.org/blogs/nathan-koren/article/news/th...
The original claim was that automated cars will reduce traffic by being more efficient (somehow) and people sharing them for trips. I pointed out that we have the sharing aspect of the issue largely figured out already. The point I did not make, but implied, is that automating cars, buses, trains, boats and aeroplanes will not by itself have much impact on traffic. Instead only a significant shift to more shared (instead of individual) use of these transport modes would. I feel that the image of the trafficless automated transport utopia that is so often conjured on HN, reddit etc. is a massive red herring. Because the benefits people claim stem from automation actually stem from public transport.
Whether the shift to shared transport will happen or not and if it happens how big the shift will be, are completely open questions as far as I can see. I believe that, trips for commuting have lots of scope for being bundled into shared/public transport. Meanwhile the bulk of other trips will continue to be made largely with personal transport. (And I make no claim as to whether those vehicle are owned by the end user or by a fleet operator, that's an entirely different debate.) Not coincidentally, this is exactly what happens in many big european cities that have both good public transport and the road infrastructure to allow for lots of cars to travel at the same time. Many people over here commute daily by public transport, bycicle, rideshare etc., but they also have at least one car for those other trips.
What about more efficient use of roads because automated cars have a shorter reaction time and thus do not need as much safety distance between them?
What's interesting is that automated vehicles, coordinating their interactions, should be able to accommodate additional traffic much more smoothly than human-driven vehicles. They will accommodate it by slowing down. Counter-intuitively, capacity is maximised at around 10 mph (16 km/h).
Here's a spreadsheet for doing those capacity calculations: https://docs.google.com/spreadsheets/d/1F3w5-hcopm2GBGz3UswN...
For humans, the reaction time is ~1 second. If you take that down to 0.1 seconds, you'll notice that the 10 mph capacity increases by about 1/3rd. That's not too shabby! But at a more acceptable 50 mph, the capacity increase is only 15% -- and that's if humans drove safely, which they don't. Actual capacity increase versus actual human driving behaviours, at highway speeds, can be expected to be nil.
Conclusion: automation will only bring capacity gains if roads get slower -- probably much slower.
Where I think we disagree is here:
> we have the sharing aspect of the issue largely figured out already.
No, we really haven't. With non-automated vehicles, sharing only works in situations where you can amortise the cost of a driver. This requires maintaining an average occupancy of > 25 people, which in turn limits you to fixed routes at fixed schedules. Hence, bad for serving incidental demand. Automated vehicles open up the possibility of having a range of smaller vehicles that are demand-responsive yet shared. This would bridge the gap between current private and public transport options, serving a substantial travel regime where current technologies are inefficient.
Have you ever been to Switzerland?
> incidental travel [is] actually the majority of trips that people take: going shopping, visiting friends, going to the doctor's, eating out, etc. [...] People own cars to serve this need, because nothing else will
I use public transit (or just plain walking) for all of these tasks. Admittedly, this only works because I live in a city rather than a village in the countryside: When I visit my brother who lives in a small village, I arrange that he picks me up with his car from the nearest train station.
But it's nothing like grandparent's very American perspective of "public transit is only good for commuting".
Er, yeah. I reside in the UK, and have lived and worked extensively in Europe. As a public transport planner. So I kinda know what I'm talking about here.
I use public transport for all of those tasks too -- because I live in zone 2 of London. If I lived in a much smaller city, or even in zone 5, it would be much more difficult for me to do so. The ability of public transport to serve incidental travel depends greatly on the size and shape of the city, and on your place within it. So yes, it can work for some people -- but certainly not for all people, nor even for most people. Not even in Switzerland.
America is egregious for how badly its cities are laid out; with the exception of New York, they are almost wholly ill-served by public transport. Fun fact: with an average 9% occupancy, buses in America produce more emissions per passenger-mile than single-occupancy SUVs. This is not a problem that cannot be solved with more mass transport: it requires entirely new forms of transport, or spatial reorganisation of the city, or (most likely) both.
I mentioned Switzerland specifically for a reason, which is that it has an unusually high number of small fairly remote towns where cars are actually banned and nearly everyone reaches them by train on relatively low capacity lines, and even the tiniest rural villages are served by well synchronized bus services, directly contradicting your statement.
There really aren't any areas of the country where car ownership is absolutely essential, or even all that desirable. Sure lots of people have cars, but almost nobody has to.
Yes, of course as you point out it's a multifaceted problem that requires land use decisions instead of just blindly building train lines to nowhere. But you present our current car dependence as inevitability. It absolutely is not, it's a public policy choice we make.
Uhm, I guess then 600.000 inhabitant Copenhagen must be this mega-city where high-frequency mass transport runs 24/7 (a driverless metro) and people use bikes to get around in stochastic patterns at arbitrary times of day (because there are bike lanes everywhere).
In the future I imagine, every train or bus station would have a small fleet of self-driving cars that people travelling nearby could use on-demand. I think this arrangement has a lot of advantages over using cars point-to-point.
The cars themselves can have a lower maximum speed, and lower range because they don't need to go very far, and lightweight, because they're consequently unlikely to get into high speed crashes. This also means your access road network can take up less space, and be more pedestrianised.
Additionally, because people use these cars for last mile transport, the effective catchment area of a station is much higher allowing them to be more viable even at lower densities, and also allowing less frequent stops to cover an area effectively.
Also, once people are used to using transport-as-a-service interfaces, it's much easier to make them aware of new transport options, and convince them to switch. If someone is opening an app anyway to order a ride to the shops, you can offer them a slightly cheaper journey involving a bus interchange. This in turn allows the networks to be more flexible, as you don't need to worry as much about providing a consistent service pattern for people to learn. So if you notice a lot of people go to the shops on Saturday afternoon, you can add a few buses at that time and people will start using it immediately.
And nope, I have no professional experience with this, I just like trains a bit too much and want more of them. :D
Even more than that, people who like transport often fail to consider it in a holistic context which includes land-use patterns, development typologies, etc. Appreciating the inter-relatedness of these things is not common, even amongst professionals. Maybe especially amongst professionals.
Anyhow, if you happen to be a coder and would like some professional experience with this, give me a shout. I run www.podaris.com, and we're trying to make it easier to build cities the way you like 'em.
Good public transport is really good for both the rider and the city.
You get exercise. Most people slack off in bad weather, so there is a bonus (I'm also in Norway. I walk and ride the bus 95% of the time. You get used to it and dress appropriately.) Sure, most public transport doesn't provide door to door service - cabs do, though, and those generally have more options. And though there are limitations, you don't have to own a car nor pay for all the expenses with it.
Some of these are bad or incomplete public transport issues, which means they can be improved.
Its like a psychological formula for making most/many people uncomfortable.
When I see a heavy truck merge a mile ahead on the interstate, I will go ahead and get in the left lane. Heavy trucks cannot accelerate like a passenger car. If I can detect the age of the driver in front of me, I can anticipate an older person with slower response time and tendency to drive slow, so I accommodate them.
In a traffic jam, I will leave 50'-100' buffer zone between me and bumper of car in front. Unfortunately, some drivers will quickly pass me and take up this zone, because people like the idea of "getting ahead". This only clogs visibility and increases braking. There are certain situations where this is not appropriate, because that buffer zone can mean other drivers getting caught at red lights. There are some tight situations where it is ideal to drive bumper-to-bumper. Interstate jams are ideal for the buffer.
If the light turns red and I can see it a mile away going 55mph, I will begin my coast with minimal braking so as to conserve momentum. Most times I never get below 30 and can always catch green lights this way. I feel even better if people stay behind me while I do this, because I am essentially saving their energy by setting our pace. Sometimes slower is faster.
I'd like to think that just a handful of mindful drivers in a jam can improve traffic flow.
Also, I would argue that maintaining a buffer zone in heavy traffic, though it may feel like you are breaking less/going faster, is not actually helpful. The flow rate of a section of highway (cars/hour) is the product of speed (miles per hour) and density (cars/mile). Leaving a space in front of you lowers the density, decreasing overall throughput. Ideally with self driving cars we could have them going at high speeds very close together (or even mechanically coupled like a train).
The idea is that if the buffer were used more frequently, then the average speed could increase. Multiplied by a lower density number, it could potentially be a higher flow rate. It requires more energy and time to move from a complete stop than it does to increase speed while already moving. The buffer could eliminate jams completely, if we define "jam" as standstill traffic.
Merging would be smoother because there wouldn't be stop and go. Think of all the thousands of people in a 5 mile jam stretch that slow to a halt and wave their hand while the other person takes time to look at them and make sure they really do have clearance. Every time that happens, dozens of cars behind must also come to a complete halt. With the buffer, speeds may temporarily slow but never halt when introducing new vehicles into the line. If the buffer was a standard implemented in drivers ed, then all drivers would have a protocol for jams.
And conversely there are days when the stars seem to align and your commute is unusually short and free of significant time going 0-5 mph. Sure you still crawl, but by comparison it feels great.
Again, this is almost entirely an artifact of human-caused preventable road accidents. I fail to see how, if you accept the premise that driverless cars will dramatically reduce or outright eliminate accidents, traffic will not be permanently improved.
I also see a lot of talk (in other threads) about the eventual banning of drivers in favor of a totally automated driving world. Can no one see the potential problems with this? Even eliminating the potential for government control, what kind of perverse incentives might arrive for companies? Taking you on routes with more stores as a form of advertising is something I can think of just off the top of my head.
Not to mention all of that kind of talk is very city-centric. Some people do have legitimate need to go places that don't have roads, or at least mapped roads. Rural places still exist, and it annoys me to see how no one seems to think about these places when talking about driverless vehicles, or banning drivers.
However, the self driving car solves many other problems. For instance, in some of our cities most of the traffic is people driving around looking for some place to park. That is it, that is what they are doing rather than going somewhere important.
At some stage the existing dumb cars we have parked on our streets now will get sent off to the recycling facility, if, at the same time, more people are getting glorified driverless cabs and not buying their own cars, then there really can be a sizeable bit of highway real estate to recover. Hopefully because the self driving cars go in convoy this extra road space can be for people, e.g. on bicycles, and not for yet more automobile infrastructure.
Then there is the school run, no need for the journey there and back, plus the ride sharing can get smarter to know who needs to be where and when, i.e. self aware of all the meetings in town.
https://www.motorists.org/blog/speed-limits-slower-safer/
Additionally, self driving cars appear to have a higher accident rate per mile driven (5x). This appears to be at least partly the result of the self driving cars performing unorthodox maneuvers in some situations.
If everyone strictly obeyed the posted speed limit and treated it as both a minimum and a maximum speed, as driverless cars would likely aim to do, roads would be safer. The interim state of having both driverless and human-driven cars on the road at the same time is where problems may arise.
We can get AI reaction time down to the speed of light since all vehicles can communicate their speed, direction, and location simultaneously.
It is a totally new epoch of transportation and the old rules no longer apply.
Still I am confused by articles logic - Why would driveless car be any better than the current ownership model? The logic presented is - AIs that can pilot cars more closely together will boost road capacity. But using article's logic on people movement - it might entice people who actually don't drive due to various reason to actually buy a car. The increased number of cars might lead to the said jam.
One of my controversial opinion is that going forward governments might be tempted to label ride sharing companies as basic necessities, just like internet, to try and improve transportation.
Agreed, I expected an article about the intricacies of the fruit preserve industry.
The Economist is famous for using puns in its headlines, and this is a play on a well-known expression (that is itself a pun):
And accidents I believe will also be minimized as soon as we get to full-automation. Even the less than ideal automatic Teslas seem to be faring slightly better than human drivers. Imagine how far we can get in 10-20 years. I'd guess maybe 1% the accidents we have now.
It just runs counter to reality to think that every empty space you create is occupied immediately. After all there are less congested cities and more congested cities. Mostly because the less congested cities have a higher road/road width/road quality to commuter ratio. That's why Seattle and SF have huge issues. With the water and mountains it's impossible to get a good infrastructure to population ratio. And no, cities with better traffic don't start going on the road to occupy space just because it's there. There's always a need for people to commute paired with inadequate infrastructure.
By utilising the infrastructure better, self-driving cars can basically increase one side of the fraction. Higher number of drivers will probably happen (seniors and children especially who couldn't drive before), but they also usually I doubt it's even more than 20-30%, seniors won't start commuting to jobs they don't have, children have more vacation, they're already only a smaller part of the population.