Google's Self-Driving Car, a Presentation

November 8, 2011 6:48 PM

Here's a really cool presentation on how Google's self-driving car works: Part 1, History. Part 2, Implementation. Part 3, Use Case

Some highlights from Part 2:

• Modeling the environment
• Isolating moving objects from the background
• Negotiating an intersection

posted by odinsdream (55 comments total)

I've never understood how this thing isn't wildly illegal.
posted by PostIronyIsNotaMyth at 6:53 PM on November 8, 2011 [1 favorite]

PostIronyIsNotaMyth, I was at the most recent Google Faculty Summit, where someone asked the legal implications of such an autonomous car. The response from the researchers was that this system was being treated the same as cruise control, and that the person behind the wheel was still responsible for all of the operations of the car.

Of course, things will get hairy once some parent sends their kid of non-driving age to grandma's house, but we're not at that point yet.
posted by jasonhong at 7:21 PM on November 8, 2011

Thanks to Google Plus, it only drives in a Circle.
posted by Apropos of Something at 7:25 PM on November 8, 2011 [6 favorites]

But before it could do that they had to fix the odd Buzz.
posted by PostIronyIsNotaMyth at 7:27 PM on November 8, 2011

Stuff like this definitely makes me feel like I'm living in the future.

My dream is to work on the usability and human factors issues on self driving cars like these when I graduate. I can already imagine a million.
posted by Defenestrator at 7:29 PM on November 8, 2011

Ok, my jaw dropped when I saw them driving Lombard street full of pedestrians.
posted by loquacious at 7:34 PM on November 8, 2011

My next car will drive itself (probably delivering location aware ads as it does so, but if it's a free car, I'm all about it).
posted by el io at 7:53 PM on November 8, 2011

Seriously, folks. Get over the dream of the future where we've all got our own automobiles (self-driving or just good old monkey-driven). That's a part of the current mass-fantasy that ain't gonna survive. It's not good for us. It's not even as convenient as we think. And there won't be rocket-packs either.
posted by philip-random at 7:59 PM on November 8, 2011 [3 favorites]

Automated commuter buses that customize their routes to deliver riders to doorsteps would be fine too.

I'm not greedy.
posted by el io at 8:08 PM on November 8, 2011 [4 favorites]

Seriously, folks. Get over the dream of the future where we've all got our own automobiles

It's not so much "folks" dream, as it is the car industries dream. The car industry will fight self driving cars, they don't want them. It means fewer cars sold and a basically a radical change to their business.
posted by stbalbach at 8:10 PM on November 8, 2011 [1 favorite]

How funny, I totally passed that car on my way home from work today near Stanford. I was thinking to myself I bet that's the google self driving car and lo and behold I was right.
posted by wilky at 8:38 PM on November 8, 2011

> Get over the dream of the future where we've all got our own automobiles...

I once thought as you did, then recently I started looking at the numbers. In 2009, transit motor buses moved 18 billion passenger miles, using 421 million gallons of diesel, 138 million gallons of natural gas, 5 million gallons of other. Fpr the sake of oversimplification, we'll assume that's all diesel. That's 31.9 passenger miles per gallon.

However, that's mostly diesel, so the carbon load is higher than gasoline. I could go dig up comparison numbers, but I'm feeling lazy, so let's just compare to a diesel automobile. My wife's 5 passenger station wagon is diesel, though, so we can do a straight across comparison: With 1 passenger, she gets 38.5 around town, 42 highway.

Buses can be very efficient in high density situations, and okay on reasonably travelled commute paths, but pretending that buses are useful for suburbs is hogwash (and largely a myth which is really driven by income redistribution), and wasting a hell of a lot of fuel and burning a sh*load of carbon that we don't need to.

(Don't get me started on trains...)

If you're going to make that argument, you're going to have to convince everyone to abandon their small towns and suburbs and move to Manhattan density cities. Frankly, I don't think that's going to happen any time soon, so we'd be far more realistic to abandon public transit except where it actually makes sense and pursue alternate energy systems that can get us clear of fossil fuels (and that may just be biofuels or other carbon neutral mechanisms).

But in the mean-time, we can also look to telecommuting and carpooling as things that actually do work, and self-driving cars as a way to reduce accidents and allow for behaviors that can dramatically increase the throughput of existing roads and mileage of the vehicles on those roads.
posted by straw at 9:06 PM on November 8, 2011 [1 favorite]

Seriously, folks. Get over the dream of the future where we've all got our own automobiles (self-driving or just good old monkey-driven). That's a part of the current mass-fantasy that ain't gonna survive.

Maybe. But people love their cars and the autonomy. And no smelly homeless guy snoring in the seat behind you.

The thing is, as all the cars become automated real efficiencies can be realized. For example, on a highway, most of the energy is consumed in wind resistance, but that can be mitigated by drafting off the car ahead. That's exceedingly dangerous for humans, but trivial for a computer.

Plus, if you think of the cars as packets, 30 years of data traffic management insight is now useful in another context.

I know you love your supertrain, with it's cool music and great coffee. But people love their cars. The answer is no.
posted by Pogo_Fuzzybutt at 9:08 PM on November 8, 2011

Oh god, the audio quality!
posted by Theta States at 9:37 PM on November 8, 2011

What are the odds of me getting a self-navigating jetpack then?
posted by Samizdata at 9:38 PM on November 8, 2011

Self-driven vehicles will probably disrupt vehicle usage.

If you could call a car on your phone that would take you where you wanted to go at any time and would arrive within 5 minutes wouldn't you take it?

What if it cost less than owning a car? I would ditch my car in a second (burbs dweller) if I could do this. Taxis currently cost more. But if that changed due to self-driven vehicles I'd happily ditch mine.

Perhaps some people would want their own vehicle that they would allow others to use when they didn't require it.

Perhaps the taxi-minibus could take a few passengers. You could say you wanted it just to take you or you could say you were willing to share for a lower price. You could make the trade off depending on how urgently you needed to go somewhere.

Self-driven vehicles could also cut the cost of deliveries as well. Cutting out having to pay a driver could lower those costs significantly. Perhaps golf-cart sized self-driven short range vehicles could deliver mail and more.

The technology is just awesome though. Sebastian Thrun is a great academic and engineer. It is just so great to be able to see things like this happen during our lifetime.
posted by sien at 9:59 PM on November 8, 2011

When I was at camp as a kid, the counselors once drove us to a nearby park for a several day hike. One of them was a prankster and he did this thing while driving on these narrow country roads that I found interesting (stupid, but interesting).

When there was an oncoming car, he'd move his hands over the steering wheel like he was turning left, into the approaching car. He wouldn't actually be turning the steering wheel--he'd just pretend to. In almost every case, the oncoming car would turn right into the ditch.

Now it never occurred to me we notice the hand location of other drivers. It is completely unconscious, but our brain brings it up to our conscious when it decides it is important enough. There are probably quite a lot of things like this that our brain does without our knowing it.

The question is, do these self-driving cars do this? Are they using neural network type computers that learn things that don't need to be programmed?

Another thing, that might shock the Bay Area Google employees. People don't drive the same in other places. There are lots of unwritten rules in every state and every country that the local drivers know, and newcomers struggle with.

For example, the Google car stops for someone crossing the street in Palo Alto. In some places in New England, no car would ever do that.

In Italy, I found that drivers do this trick where they look the other way, then enter the intersection. They know that you'll notice them looking the other way, so you have to let them in, because they are certainly not going to stop for you! They can't even see if you are stopping or not! How would you even program a car to handle this type of driving?

It may be that now days, with drivers increasingly taking their eyes off of the road to text or press menus on their computerized electronic dashboard, that the Google car ends up as the best driving car on the road.

But I'm worried that it is missing some basic things (involving human behavior) that may be very difficult if not impossible to program for, which could lead to disastrous consequences if left out.
posted by eye of newt at 10:21 PM on November 8, 2011 [1 favorite]

I've never understood how this thing isn't wildly illegal.--PostIronyIsNotaMyth

Because every 'driverless' car actually has a driver, ready to grab the wheel and take over at any moment.

There was actually an accident once with one of these cars, and Google claims the driver was actually in control in that situation.
posted by eye of newt at 10:36 PM on November 8, 2011

What? He was causing major accidents as a prank? I don't think I can even see the hands of other drivers usually, because they're far away and because of glare on the windshield.
posted by !Jim at 10:39 PM on November 8, 2011 [1 favorite]

What? He was causing major accidents as a prank?

Apparently. Maybe not so major.
posted by dhartung at 10:54 PM on November 8, 2011 [1 favorite]

eye of newt, at least in the case of the asshat playing his "prank" that you describe, I would think that a self-driving car would actually be at an advantage, since it would look at the actual physical situation and not at the driver's little trick. So no ending up in ditches. The stunt you describe Italians doing would end up in an accident, of course, with the human driver at fault.

Personally, the whole automated driver thing makes me a little nervous because of the number of variables involved, but then driving on today's roads with so many distracted and unskilled drivers also makes me a little nervous.
posted by moonbiter at 11:07 PM on November 8, 2011 [1 favorite]

While I think this is awesome, the biggest losers with something like this are the taxi and bus drivers. More contribution to those unemployment numbers!
posted by Defenestrator at 11:56 PM on November 8, 2011

The stunt you describe Italians doing would end up in an accident

I doubt it. Automated cars have good reflexes.
posted by ryanrs at 12:11 AM on November 9, 2011

I saw a talk by Prof Thrun in which he said that one of the frustrating things about being in a self-driving car is how slowly it typically drives...unlike human drivers who tend to impatiently charge ahead it'll wait around for as long as it takes to get a handle on the safety of the situation. So, it might drive you crazy to be stuck on a one-lane road behind a self-driving car...but it's not going to be making snap decisions about ambiguous situations using tenuous cultural cues as a human driver often does, it'll just wait for the situation to clarify itself.

I very distinctly remember the first time I saw video of the self-driving car making a left off University Ave in Palo Alto. I used to live in that neighborhood and would go to crazy lengths to try not to make those lefts because the street was always such a morass of pedestrians, cyclists, lost tourists, and aggressive local drivers that it was just not worth the pain. It made that left with aplomb, and my mouth dropped open. Super amazing.
posted by troublesome at 12:37 AM on November 9, 2011 [1 favorite]

Seriously, folks. Get over the dream of the future where we've all got our own automobiles (self-driving or just good old monkey-driven). That's a part of the current mass-fantasy that ain't gonna survive.

The dream? If you're talking about the U.S things have been that way for decades. If you don't think it's going to continue, you have to explain why. It's ridiculous to say "oh, cars aren't sustainable, therefore we won't have them anymore". First of all, when has unsustainability ever stopped society as a whole from doing something? Second of all, you haven't actually done anything to show why they're unsustainable.

Why would, for example, an electric car powered mostly by wind, solar or nuclear not happen not 'happen'?

And you have to look at timeframes. These self driving cars are only a few years out. Already you can buy cars that will manage themselves on the highway, simply requiring you keep your hands on the wheel for liability sake.

This 'carless' future you're imagining wouldn't happen in the next 50 years. So why not work on driverless cars in the short term.

Plus, as other people have mentioned there are a TON of things you can do with driverless cars that you can't do with normal cars or normal public transportation. You could simply punch your destination in with your phone, have a car, or a bus come pick you up and drop you off while you just dial in your speed, privacy, expense settings and so on. You wouldn't even need to have your own car.

In Italy, I found that drivers do this trick where they look the other way, then enter the intersection. They know that you'll notice them looking the other way, so you have to let them in, because they are certainly not going to stop for you! They can't even see if you are stopping or not! How would you even program a car to handle this type of driving?

Well, it won't be able to prevent idiots who try to psych out other drivers from ramming into it. At the same time these other drivers who are trying to play mind games with each other would certainly recognize a car with no driver and adapt, either giving it a wide berth or perhaps driving aggressively hoping the car won't try anything.
posted by delmoi at 12:56 AM on November 9, 2011 [1 favorite]

eye of newt: In Italy, I found that drivers do this trick where they look the other way, then enter the intersection. They know that you'll notice them looking the other way, so you have to let them in, because they are certainly not going to stop for you! They can't even see if you are stopping or not! How would you even program a car to handle this type of driving?

I think there may be a more rational explanation for this behavior than that it's a brazen trick. I've done some pretty considerable driving in Italy and several other European countries, and I think the actual explanation is as follows:

In most continental European countries, drivers approaching an intersection must yield to other drivers coming from the right. That means, when you approach an intersection, you only have to look to the right, since any drivers approaching from the left have to yield to you.

This is, in fact, a rational behavior that could easily be programmed into self-driving cars, and could do away with a number of 4-way stop sign intersections in the States (as could traffic circles, which are really pretty awesome).
posted by syzygy at 1:24 AM on November 9, 2011

I've never understood how this thing isn't wildly illegal.

Because a well-designed automated system is about a bajillion times more reliable than the average human? How many airplane crashes are the fault of the autopilot? How many of the pilots?

I doubt the Google car would run red lights through pedestrians crossing the road, which puts it well ahead of the petrol tanker this week and every other cyclist in Wellington, including the one who clipped me a couple of weeks ago.
posted by rodgerd at 1:28 AM on November 9, 2011

Plus, as other people have mentioned there are a TON of things you can do with driverless cars that you can't do with normal cars or normal public transportation. You could simply punch your destination in with your phone, have a car, or a bus come pick you up and drop you off while you just dial in your speed, privacy, expense settings and so on. You wouldn't even need to have your own car.

Okay, if you watch the third video google already has this on their campus
posted by delmoi at 1:32 AM on November 9, 2011

Looking through the related videos I came across this one which shows some seemingly random people going for a ride in the google car as it tries around a pretty small motocross track. Their reactions are pretty entertaining, and they get up to about 41MPH in a pretty small area.
posted by delmoi at 2:02 AM on November 9, 2011

Delmoi, that link is broke and I can't find which one you're talking about.
posted by loquacious at 2:31 AM on November 9, 2011

The end of drunken teenagers killing themselves and others on twisty dark suburban roads can't come too soon.
posted by tempythethird at 2:42 AM on November 9, 2011

If you're going to make that argument, you're going to have to convince everyone to abandon their small towns and suburbs and move to Manhattan density cities.

Don't need to be nearly that dense. Many old-timey American small towns were mostly walkable until we screwed them up and turned them into freeway exits. I blame things on a combination of the love affair with the car and around the 1930s or so where we decided everyone should have a house leading to the subdivisons of tickey-tacky.

But the self-driving car can be good medicine. Cars are expensive, gas is relatively cheap. Once you have a car it makes a lot of sense to pretty much use it for everything. When we get cheap taxis that can come by and pick us up with no driver inside, lots of people don't need to own their own cars anymore. You switch to paying by the hour. The nearly $0 marginal cost of operating a vehicle goes away, and people will likely respond by driving less. Maybe they stop tricking themselves into thinking they save money by living in the sub/exurb.
posted by floam at 3:54 AM on November 9, 2011

Also, same can be said of car-sharing programs like Zipcar.
posted by floam at 3:57 AM on November 9, 2011

I also have suspicions that personally owning an entirely self-driving car might end up being a bit of a problem. Once the user isn't operating it, the car is, the automaker is going to be on the hook for a lot more liability. They are going to want to ensure software fixes are deployed across the board, and that the things are well-maintained, stuff like that. I have a feeling we're going to see these in fleets, and companies that deploy them for personal rapid transit, long before you can pick up your own. When and if you can, maybe it's going to be some kind of managed deal where you're not really allowed to toy with it. Or maybe they'll successfully lobby to get laws written that let them off the hook.
posted by floam at 4:09 AM on November 9, 2011

Also, they probably won't want their first-generation self-driving cars on the road at all after a decade goes by and they're fifty times safer and perhaps do some kind of fancy new car-to-car signaling or something.
posted by floam at 4:13 AM on November 9, 2011

I can't help but thinking: what if the Yellow Pages went from "helping you find companies you want to contact" to "controlling how your car drives" in under 15 years. Kinda creepy, no?
posted by ShutterBun at 4:30 AM on November 9, 2011

Umm, your 31.9 passenger miles per gallon comes from American statistics, straw, where only hippies and poor people take public transit. Amtrack gets 50 passenger miles per gallon, ala a 25mpg car with two people.

There are places where carpooling has raised the passenger mpg for cars of course, but almost all other countries should get vastly higher ratings for busses and trains, due to both affordability and habits, including eco-consciousness. A French TGV consumes considerable electricity hurdling down the tracks at 300km/h, but good luck finding a seat with nobody sitting in the adjacent seat, excluding first class.

You might rephrase philip-random's comment as saying American's standard of living must decline until the passenger mpg improves, but all the pain in the transition comes from our transportation policies.

We should halt all the various old subsidies immediately, start increasing the taxes on gas, and start building more electric bus lines, light rail, etc., transitioning people towards transport options that aren't capped like cars.
posted by jeffburdges at 5:04 AM on November 9, 2011

Now it never occurred to me we notice the hand location of other drivers. It is completely unconscious, but our brain brings it up to our conscious when it decides it is important enough. There are probably quite a lot of things like this that our brain does without our knowing it.

The question is, do these self-driving cars do this?

It could probably be easily programmed in. Facial recognition is already pretty good. This is what fascinates me about this concept, because adding another piece of sensory information, and another, and another, and another is totally conceivable with the computer system. Humans are already at or beyond their physical limitations (i.e., speed of vehicles and typical density is beyond what neurological reaction time can safely work with).

I mean, consider coupling this system with deep information knowledge. An oncoming car is careening towards a human driver. The human is going to react instinctively, and with a limited amount of time. They're going to jerk away in a random direction and side-swipe their neighbor.

In microseconds, a computer driver could look up the license plate for the oncoming car, pore through the driver's social media history, research the make and model of the car, what tires it comes with, analyze the coefficient of friction, perform thousands of physics simulations with the known information, note every single other moving car in the scene within sight, and then choose among a myriad of best-case options for avoiding collision not only with the oncoming car, but every other single object around.

Humans are awful at driving. Really.
posted by odinsdream at 5:08 AM on November 9, 2011

In Italy, I found that drivers do this trick where they look the other way, then enter the intersection. They know that you'll notice them looking the other way, so you have to let them in, because they are certainly not going to stop for you! They can't even see if you are stopping or not! How would you even program a car to handle this type of driving?

Funny you should ask. This is specifically mentioned in the video.Yes, it's solved.
posted by odinsdream at 5:14 AM on November 9, 2011 [1 favorite]

I liked how on the highway section the Google car was keeping a safe distance and the cars behind it were all RUM RUM HURRY HURRY SCREECH HURRY SCREECH HURRY SWERVE SCREECH FUCKER WHY ARENT YOU TAILGATING HURRY RUM SCREECH
posted by seanmpuckett at 5:15 AM on November 9, 2011

Funny you should ask. This is specifically mentioned in the video. Yes, it's solved.

Woah, you actually watched the whole thing and paid attention? Nice going,

A. part of the problem
B. Big Brother's bitch
C. Person who honestly wants to learn more about future technology and hasn't been burned too many times in the recent past
D. ?
posted by ShutterBun at 6:04 AM on November 9, 2011

In microseconds, a computer driver could look up the license plate for the oncoming car, pore through the driver's social media history, research the make and model of the car, what tires it comes with, analyze the coefficient of friction, perform thousands of physics simulations with the known information, note every single other moving car in the scene within sight, and then choose among a myriad of best-case options for avoiding collision not only with the oncoming car, but every other single object around.

Seems like "keep to the right (of-way) side of the road" would be a pretty efficient default there.
posted by ShutterBun at 6:10 AM on November 9, 2011

At some point in most of our lifetimes computer driven cars will become many times safer than human driven ones and insurance companies are going to notice. Once that happens, how long before the insurance companies stop letting humans drive at all?
posted by octothorpe at 6:17 AM on November 9, 2011 [1 favorite]

Before any of this happens, there will have to be a certain amount of time for "actual driving conditions" to occur, in order for insurance companies to have enough data to come up with something approaching "actuarial" data, which they're so fond of. Before that happens, a pretty good handful of people are gonna have to bite the bullet and get some cutting edge insurance for this new technology. (how 'bout it, Google? Put your money where your mouth is, so to speak?)

Anyone want to give me odds against this never, ever happening?
posted by ShutterBun at 6:33 AM on November 9, 2011

Woah, you actually watched the whole thing and paid attention? Nice going,

I, uh, posted this thread. So... yeah.
posted by odinsdream at 7:08 AM on November 9, 2011 [2 favorites]

For those who think that this means the end of privately owned cars, have you ever used a taxi? If you're lucky, it's been cleaned recently, but if there isn't a human driver keeping it clean, the lessons arising from the Tragedy of the Commons show it'll become a fetid dump in very short order. Some people, especially if they think they're not being watched and won't face any repercussions, seem to take delight in graffiti, vandalism, and just trashing anything they can get away with.

In a city, or anyplace where parking is at a premium, sure, it may work out well to have a system like the car sharing service Zipcar -- if the vehicle isn't clean, there's a clear track record of who had it last and enforcement of cleanliness and vandalism rules. For a taxi-like service, though, where there's plausible deniability (it was like that when I got it, I just was in too much of a hurry to report it . . . ), calling a car may well resemble rolling the dice -- and if it turns up snake eyes too many times, many may just buy a car so they have a known good ride when they need it.

Also, may professionals work out of their cars -- they have their computer, files, tools, and so on in there, again encouraging ownership. Repairmen, contractors, vets, and so on will need their own vehicles.

And, of course, there will be those who want to own a vehicle for the prestige. We've already seen that the beginnings of automation -- the auto-braking, parking, and similar features -- show up first in high-end luxury vehicles, and even though Google has made some very promising deveopment, I would expect auto-drive to be first offered on high-end Mercedes or similar long before it'll be available to the rest of us proles.
posted by Blackanvil at 7:25 AM on November 9, 2011 [1 favorite]

Yeah, this is pretty OT, but I do tie it back to automated vehicles in the third paragraph...

jeffburdges, dang it, I can't come up with the stat right now, but I read something that suggested that as a percent of miles traveled per year, European public transit use was like 20% vs a U.S. number of circa 10%. Table 5 in the Victoria Transport Institute Online TDM Encyclopedia has a "transit split" (no idea if that's trips, miles, what) that says Germans use transit for 8%, Austria for 17%, U.S. for 2%.

The big difference is that Germans are traveling 23 miles a day (and spending 80(!) minutes doing it), while this report (all the usual caveats about "information is worth what you paid for it") suggests that in 1995 US residents were traveling an average of 29.3 miles per day. I'm sure that number is much higher today.

So, yeah, if you want to change how we use energy for transportation, raise fossil fuel taxes to the point where it actually makes a difference in operating costs (and begins to offset some of the real externalities of fossil fuel use, it's nowhere near close now, and starts to dwarf some of the ancillary fixed costs that make public transit more expensive in the U.S. than driving), and wait 20 years for suburban development patterns to change such that we start to live in places that encourage pedestrian behavior (which ties in to speculation that "transit-oriented development" is really "pedestrian-oriented development").

But in the mean-time, expect that another trend will be rather than paying drivers to move big heavy buses around empty routes at midnight, we'll start to have automated smaller vehicles running demand-optimized routes at lane densities that are double what they are today, with parking happening automatically (dramatically reducing urban miles of driving).

To Blackanvil's speculation about how these technologies will be deployed, lane deviation warnings have been deployed for a decade in commercial trucking, we're already seeing lane following behaviors in high end cars, I think an option on the European Ford Focus of this year is reading road signs (to tell you when you're exceeding the speed limit). A ballpark I've heard is that a Google car (or the equivalent from the Italian and German teams who are doing very similar things) costs roughly half a million bucks. Moore's Law suggests that this will be a thousand dollar premium in a decade and a half.

(For the record, I live in a town of about 60k people, deliberately live walking distance from downtown, drive less than 4k miles a year on my vehicle, work from my home office and walk to client meetings, but my wife works for a school system and puts closer to 18k a year on hers. I hope that I can maintain a lifestyle such that I don't drive, and we can continue to subsidize the school system, because when you count how much she drives to work with the kids I'm not sure she breaks even.)
posted by straw at 9:11 AM on November 9, 2011 [1 favorite]

Seriously, folks. Get over the dream of the future where we've all got our own automobiles (self-driving or just good old monkey-driven). That's a part of the current mass-fantasy that ain't gonna survive. It's not good for us. It's not even as convenient as we think.

Spoken like someone who imagines we all want to live in a densely-populated metropolitan area.

One of the few recent developments that have made me think I might be able to live without a car someday is Ithaca Carshare (I'm sure there are similar services springing up in other progressive isolated communities). But they're still in the early stages and it's going to take a while for people to adopt them en masse, and like many shared services, their success will be vulnerable to those who deface, damage, or neglect the shared resource out of assholishness or laziness.
posted by aught at 9:14 AM on November 9, 2011

To straw, concerning buses:

The mistake you're making is in assuming that increased bus coverage and density won't result in less car use and more bus use, and that this won't increase the passenger-miles-per-gallon ratio for buses. Many areas buses are not reliable transportation, or their use is looked down on, so people drive most of the time. Also, most gas-powered cars aren't as fuel efficient as your wife's car; those buses might be less efficient than that car.

Also, what do you have against trains?
posted by JHarris at 9:41 AM on November 9, 2011

JHarris, I'm not convinced on increased bus density. Cars are generally enough faster than buses (because of multiple stops) that unless you get heavy traffic (which buses are subject to as well), you have to be making less than your average Starbucks barista to make a bus worth your time. Or, the bus has to be useable time. I think that latter aspect is actually the thing that could switch people to public transit, but you'd actually have to have transit agencies thinking that way. I used to commute into San Francisco on Golden Gate Transit: When they got it right it was nice motor coaches, but the space was just a little tight to use a laptop, and about 1/3 of the time you'd get a bus or a driver that bounced all over the place and was too rough to read.

And GGT still isn't publishing their schedules in GTF, so 511.org, Google Maps transit directions, and the Golden Gate Transit PDFs can all give different schedule information.

I think this is a general problem with the fact that the customers for transit (federal and state funding agencies) are not the users of transit (the people who would ride it, if they could figure out the schedules and if it made sense). My wife's job for the school district right now involves teaching developmentally disabled kids life skills. One of those life skills is using the bus system for transportation. We've spent quite a while driving around with a GPS tracker and a camera trying to build usable transit maps for her students, because the materials provided by the transit agency suck. This won't change until the bus system is actually making money on riders, rather than losing money when they accumulate additional riders. But if they raise the direct costs to the point where fares actually make a substantial contribution, they'll drive more people back to cars.

Buses make sense in population densities where parking is the limiting factor.

And I actually don't think my wife's car is all that efficient. If you normalize for CO₂ output, 40MPG diesel is probably about 25MPG gasoline. Which also puts into perspective some of the numbers for rail...

On trains: Electrified light rail can be efficient, but you have to wait a decade or three for development to catch up with the rail. However passenger rail on shared freight tracks can't really ever be efficient: You're stopping and starting that train a lot, the FRA requires that rolling
stock on shared freight tracks be unreasonably heavy, you just end up burning a hell of a lot of fuel for moving around the equipment, and the overall efficiency numbers don't look all that good. Per passenger mile you're doing a hell of a lot better flying than training (also because pushing around air at sea level vs 36k feet).
posted by straw at 10:08 AM on November 9, 2011

JHarris, I'm not convinced on increased bus density. Cars are generally enough faster than buses (because of multiple stops) that unless you get heavy traffic (which buses are subject to as well), you have to be making less than your average Starbucks barista to make a bus worth your time. Or, the bus has to be useable time. I think that latter aspect is actually the thing that could switch people to public transit, but you'd actually have to have transit agencies thinking that way.

More bus use should reduce traffic, which would make both buses and remaining cars faster and reduce emissions. I think you're on to something with the "useable time" thing though, that could really be wonderful if funded well and kept up-to-date.

We've spent quite a while driving around with a GPS tracker and a camera trying to build usable transit maps for her students, because the materials provided by the transit agency suck. This won't change until the bus system is actually making money on riders, rather than losing money when they accumulate additional riders.

I don't think it is necessarily the case that you must build in a profit motive to make a federal agency responsive, but serious work must be done in listening to patrons' wishes and not giving in to administrative inertia. This is a process problem, it has to do with the way federal agencies have evolved. They don't have to remain that way.

Will have to take your word for it planes vs trains.
posted by JHarris at 1:43 PM on November 9, 2011

In cities where buses have transitways and bus-dedicated lanes, it is often faster to bus from the suburbs.
posted by Theta States at 6:51 AM on November 10, 2011

There isn't any question that trains and busses achieve vastly more passenger miles per gallon than cars, straw. As I observed, virtually every other country in the world prove that point, including both rich and poor countries. In particular, passenger aircraft average 1.4 MJ/passenger-km while Japanese railways consume about 0.35 MJ/passenger-km (cite).
posted by jeffburdges at 9:02 AM on November 10, 2011 [1 favorite]

jeffburdges, clearly I need to dig deeper, but the reference for the .35MJ/passenger-km, "What Was Accomplished in Fiscal 2004 in Relation to the 2005 Environmental Goals? (PDF)" doesn't in fact mention passengers, and looks like the interpretation which resulted in that number on the Wiki page may be making the usual obfuscation (which seems to be deliberate with many transit agencies) of confusing seat miles with passenger miles.

Not that it isn't possible to run buses and trains at high efficiencies (as your pictures show, and as dense urban cores and commute services show), just that a lot of claims for both are extremely inflated.

This PDF, which I haven't dug into, says that Germany's ICE is using 1.17MJ/passenger-km, better than air, but not nearly as better as the Wikipedia claim for Japan.
posted by straw at 5:22 PM on November 12, 2011

I adore train travel in Germany, no seat or train reservations, hop an ICE at 11pm, find an empty row, stretch out and sleep all night. All that comfort ain't great for the passenger-km rating though. Aircraft are much more fully occupied.

French TGVs won't even sell you tickets without fixed seat and train reservations. Also, TGVs don't run all night, you'll hardly ever find an unoccupied row, etc. I've never witnessed two unrelated Americans sitting side-by-side on the Amtrack NE corridor line, the busiest line in the country, although presumably that happens during rush hour.

A TGV Sud-Est seats 2760 and weighs 424 short tons. An average car seats 4 and weighs 2 tons, even a even a SmartCar weighs 0.8 tons. Ergo, a TGV with one person per 3.2 seats achieves the same occupancy per tonnage as crowded cars, i.e. every train passengers gets an extra two seats if we ignore that the TGV travels 2.5x times as fast.

An empty moving train or bus burns infinitely more energy than an empty non-moving car of course, but trains and busses obliterate cars once anybody really uses them. Americans would definitely use a 300 km/h alternative to the NE corridor. Around 1h30 for Boston-NYC or NYC-Washington? Hell yes!

And never mind all the human hours put towards activities more worthwhile than driving!
posted by jeffburdges at 12:10 AM on November 13, 2011