Live data from Hacker News

Why driverless cars may mean jams tomorrow

economist.com

51–60 of 77 posts

Re: Why driverless cars may mean jams tomorrow

#51
post #7

Earlier quoted context omitted.

I also don't follow that argument. IMHO accidents and bad swarm behavior (people following too closely, switching lanes too often, accelerating and braking too much) lead to traffic jams. If we can get that improved our roads could take twice the cars.

More, even. Driving into SF at rush hour is an exercise is mostly about being stationary with occasional jerks forward. You could improve that throughput a hundredfold if things could keep moving.

Yeah, the non-accident, jerky kind of traffic is the one that automatic cars will almost eliminate completely. If we get to fully networked automatic cars we can use swarm intelligence to make them all accelerate at the same time and slowly increase the speed. Traffic is gone.

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.

Re: Why driverless cars may mean jams tomorrow

#52
post #30
post #17

Earlier quoted context omitted.

You mean sharing a ride like say a big, long vehicle with lots of seats that visits a number of stops on a route? Yeah, that would be a great idea. We could improve this even further by giving the ride-share vehicle it's own lane, then turn both the wheels and the road surface into metal to reduce friction. BRB, going to the patent office... Not to be too snarky, but as a non-american it has been amazing to watch the…

No, not like that. Trains work well for arterial routes that need to carry large volumes of people along predictable corridors at predictable times of day -- eg., for commuting. And yes America certainly needs to get better at investing in them. 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 w…

> 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).

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).

Re: Why driverless cars may mean jams tomorrow

#53
post #47

Earlier quoted context omitted.

Or to Europe, for that matter. > 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 vi…

> Or to Europe, for that matter. 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…

> certainly not for all people, nor even for most people. Not even in Switzerland.

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.

Re: Why driverless cars may mean jams tomorrow

#54

Driverless cars will bring safer roads in terms of exposure measures like deaths/injuries per 1 Million/km. They will also lower the monetary cost of transporting a person per km due to reduction in fleet size and optimised driving and resource allocation. This will benefit nations and economies in providing more economic growth, employment etc by moving more people and goods over the same road network and a smaller…

You say reduction in fleet size and whatnot, but I believe that's assuming population remains the same; as it stands, it grows with about 1% per year. That fleet size reduction would need to exceed 1% a year; I don't think that's very likely.

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.

Re: Why driverless cars may mean jams tomorrow

#55
post #35

Earlier quoted context omitted.

I don't disagree. In fact I work on software for cars, so my career depends on personal transport staying relevant. However, that doesn't adress the question of whether automated cars will increase or decrease traffic. 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 fi…

> Because the benefits people claim stem from automation actually stem from public transport. 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?

Reaction time is a relatively small factor in safety distance. Kinetic energy is a significantly larger factor. Currently, humans violate this as a matter of course, which is why we're so bloody unsafe. Automated vehicles will be far more punctilious, which at higher speeds means that the gains from reaction time will be more than offset by the respect for kinetic energy.

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.

Re: Why driverless cars may mean jams tomorrow

#56

This article seems to offer a somewhat limited vision. Personal, driverless cars may mean as many jams as today. What about "ride-sharing" driverless systems, where many people share the same car/minivan in an automated fashion, reducing the number of vehicles on the road and the need for parking? That would make, IMHO, the best compromise between mass transit (effective, but inflexible to people needs) and personal…

Nobody really likes ride sharing, though. That's why people drive in the first place.

Ride sharing doesn't really work, because it assumes that people have very fixed patterns. I get into work at 8 in the morning, but I leave somewhere between 15:50 - 17:00, depending on what happens to land on my desk that day. Finding someone who's going my way, and willing to wait an hour for me to finish up is going to be hard.

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?

Re: Why driverless cars may mean jams tomorrow

#57
post #16

Earlier quoted context omitted.

Nobody really likes ride sharing, though. That's why people drive in the first place.

I assume one interpretation of 'ride sharing' is sharing the same pool of cars, but not the same car at the same time.

Sure, but then you'd still need a large number of cars to handle peak hours, like 7 - 9 in the morning and 15 - 18 in the evening. You're not really freeing up the road if you're not reducing the number of cars.

Re: Why driverless cars may mean jams tomorrow

#58
Driverless is great for quicker reaction times, but they cannot anticipate larger-scale situations. Perhaps if all cars were autonomous and connected it could work, but that will not happen for a very long time.

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.

Re: Why driverless cars may mean jams tomorrow

#59
post #55

Earlier quoted context omitted.

> Because the benefits people claim stem from automation actually stem from public transport. 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?

Reaction time is a relatively small factor in safety distance. Kinetic energy is a significantly larger factor. Currently, humans violate this as a matter of course, which is why we're so bloody unsafe. Automated vehicles will be far more punctilious, which at higher speeds means that the gains from reaction time will be more than offset by the respect for kinetic energy. What's interesting is that automated vehicles…

Actually, here's a more digestible illustration of the capacity gains from automation: https://docs.google.com/spreadsheets/d/e/2PACX-1vQ6ga2atFS69...

Re: Why driverless cars may mean jams tomorrow

#60
post #5

No, the way it will work is that roads will have "surge pricing". You'll bid on road space before you leave, and don't get to leave until there's road space for you. 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.

Then there's hope as it's probably cheaper to have buses full of people than individual cars

It already is, but based on the availability and popularity of buses, it's not likely.

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.

Post reply on HN