Join 3,553 readers in helping fund MetaFilter (Hide)


Could advances in AI make up for a lack of advances in generating and manipulating energy?
August 20, 2012 5:08 PM   Subscribe

How self-driving technology could make electric cars commercially viable

"Electric cars have been competing with the internal combustion engine for more than a century, and they have never won. Batteries are more expensive, have less range, and require more time to recharge than it takes to fill a gas tank. In late 2010, U.S. Energy Secretary Steven Chu himself articulated the challenge, stating that battery companies have to develop units that last 15 years, improve energy storage capacity by a factor of five to seven, and cut costs by about a factor of three in order for electric cars to be comparable to cars that run on gasoline and diesel. U.S. energy policy has tried to address these challenges. The Energy Department and other agencies have supported the development of battery and recharging technology. In addition, the U.S. government has provided financial support to Nissan, General Motors, Tesla, and Fisker to develop and manufacture commercial electric vehicles (EVs). The government hopes such investment can spur economies of scale, thereby reducing unit costs and making new technologies viable. So far, though, these efforts have failed to produce any game-changing breakthroughs... Is the electric car then history? Will the Leaf and the Volt go the way of the ill-fated EV1, General Motors' electric car from the 1990s? If the status quo persists, they very well might. There are, however, reasons to believe that electric cars might find a viable niche after all -- if we use them in the right way. For the last several years, Google has been testing self-driving cars, primarily in California and Nevada. Its vehicles use lasers, radars, and other sensors to establish their position and identify objects around them. This data is interpreted by artificial intelligence software that enables the vehicle to drive itself. Google's vehicles have now proved themselves in hundreds of thousands of miles on the road. And Google's not the only game in town. Bosch is also developing the technology, and Cadillac has promised to have a car capable of driving autonomously on the highway by 2015. Self-driving technology is gradually moving to commercialization, and when it does, it will liberate the car from its driver, enabling a vehicle to serve more users. According to the Transportation Department, the average U.S. vehicle is used less than one hour per day -- a utilization rate of about 5 percent. Many Americans only drive their cars to work, park, and leave them until they drive home at night, making them essentially unavailable for use by others for most of the day. But if the car could drive itself, it could return home to take the children to school, members of the family shopping, and seniors to visit friends or keep appointments. If the vehicle served even one additional passenger, its utilization rate would double, and its capital cost per user would fall by half."
posted by bookman117 (113 comments total) 17 users marked this as a favorite

 
Stylistically, I agree with 23. But I still read every word on the FP.

As for the concept of self-driving cars, it could either be pure awesome or (due to programming oversight or outright hacking) it could turn into Maximum Overdrive. Make it safe and make it informationally secure, and it'll work like magic.
posted by Strange Interlude at 5:16 PM on August 20, 2012


What an odd argument - electric cars have limited range so lets utilize them as a fleet of constantly available autonomous cabs. When would they be charged - you'd need three or four of them in order to replace one ic engined vehicle negating the lower operating costs. Daft.
posted by zeoslap at 5:30 PM on August 20, 2012 [3 favorites]


FTA: If the vehicle served even one additional passenger, its utilization rate would double, and its capital cost per user would fall by half.

That argument holds equally well for gasoline engines, or diesel engines, or coal-fired steam engines for all it matters.

One and only one advancement in technology will make electric cars practical (and make no mistake, I would sincerely love them to enter the realm of fiscally sound choices, rather than just another form of "bling") - Cheaper, higher capacity batteries that can take a near-full charge in under five minutes.
posted by pla at 5:31 PM on August 20, 2012 [2 favorites]


I don't know too much about electromagnetism, but what if roads were retrofitted with cables that would power vehicles via induction, with the battery serving as a backup? Is that idea simply to inefficient to pursue?
posted by triceryclops at 5:36 PM on August 20, 2012 [1 favorite]


I don't think human beings have a good enough track record to put human lives at the autonomous mercy of our own engineering skills.
posted by bleep at 5:37 PM on August 20, 2012


Charging stations are already swapping batteries so they can charge in the shop overnight. The time-to-fill may not be quite so important.
posted by LogicalDash at 5:37 PM on August 20, 2012 [10 favorites]


triceryclops, that is how the trolleys in Portland and San Francisco work. It's perfectly lovely if you can physically attach the vehicle to the rail. I'm thinking that might be a disadvantage.
posted by LogicalDash at 5:39 PM on August 20, 2012


I don't know too much about electromagnetism, but what if roads were retrofitted with cables that would power vehicles via induction, with the battery serving as a backup? Is that idea simply to inefficient to pursue?

Like a third rail, but on roads?
posted by Pope Guilty at 5:39 PM on August 20, 2012


I don't think human beings have a good enough track record to put human lives at the autonomous mercy of our own engineering skills.

Actually, self-driving cars are thus far safer than human drivers. Human drivers are actually really really bad at driving safely. The bar for "safer than humans" is extraordinarily low.
posted by Pope Guilty at 5:40 PM on August 20, 2012 [49 favorites]


Cities can barely afford to fill the potholes in existing roads never mind retrofit them to provide induction charging (which is super inefficient)
posted by zeoslap at 5:42 PM on August 20, 2012 [2 favorites]


LogicalDash, Pope guilty: he did say "induction" as in, like the wireless charging stations for your phone. I'm too lazy to find a source, but I have seen in engineering magazines that they're working on inductively-coupled roads. Pretty much, to work efficiently, they'd *have* to be self-driving cars, to maintain proper charging speed.
posted by notsnot at 5:43 PM on August 20, 2012 [2 favorites]


If I'm getting this correct, a third rail operates through conduction, direct contact with the energy source. Induction moves electrons using a magnetic field, like a transformer or a wireless phone charger. I know the current would have to be pretty high in order for that to be plausible, and would almost certainly have to rely on the cars to operate autonomously.
posted by triceryclops at 5:47 PM on August 20, 2012


Or what notsnot said. Good to know people are looking into this.
posted by triceryclops at 5:48 PM on August 20, 2012


I already take a self-driving diesel electric hybrid to work every morning. It's paid for by my employers, has huge picture windows, piles of legroom, and a bathroom, no less. I can read a book, listen to David Rakoff on the mp3 player, or write endless and occasionally relevant comments in reply to neat things on metafilter, all in the comfort of my seat.

I want to lean towards the gee-whiz, Dan Dare side of things, but I will be ding-damned before I sit in the cockpit of some ridiculous little computer-guided crunchmobile. Build me a computer that doesn't crash, even rarely, and I'll trust one to carry me around.
posted by sonascope at 5:50 PM on August 20, 2012 [3 favorites]


Wait til they learn how to hoon.
posted by Flashman at 5:52 PM on August 20, 2012 [2 favorites]


I want to lean towards the gee-whiz, Dan Dare side of things, but I will be ding-damned before I sit in the cockpit of some ridiculous little computer-guided crunchmobile. Build me a computer that doesn't crash, even rarely yt , and I'll trust one to carry me around.
According to the National Highway Traffic Administration, car accidents happen every minute of the day. Motor vehicle accidents occur in any part of the world every 60 seconds. And if it’s all summed up in a yearly basis,there are 5.25 million driving accidents that take place per year. Statistics show that each year,43,000 or more of the United States’ population die due to vehicular accidents and around 2.9 million people end up suffering light or severe injuries. In a certain five year period, there had been recorded a 25% of the driving population who encountered or were involved in car accidents. It is also affirmed that car accidents kill a child every 3 minutes.Statistics on the number of car accidents taking place in every state or country is normally based on medical or insurance records filed.

Based on a research conducted years back, an estimated number of 1.2 million car accident deaths occurred last 2004 and 50 million people injured worldwide. In the Global Status Report on Road Safety 2009 made by the World Health Organization, more than 90% of the world’s road casualties happen in low and middle income countries which comprise only 48% of the world’s registered vehicles. The escalating death rates pertaining to driving accidents over the years have already become one of the most serious global issues. It is estimated that by 2020, road accident casualties will exceed HIV/AIDS mortality and disability rate.
posted by Pope Guilty at 6:00 PM on August 20, 2012 [13 favorites]


Holy Wall of Text, Batman!
posted by ChurchHatesTucker at 6:12 PM on August 20, 2012


wikipedia: According to a study by the National Highway Traffic Safety Administration, vehicular communication systems could help avoid up to 81 percent of all traffic accidents.[44] These systems provide communication between nearby vehicles, intersections, etc.

Wouldn't liability litigation be a difficult problem for a self-driving vehicle? Maybe it could have the equivalent of black box recordings that would be admissible in court.
posted by Golden Eternity at 6:14 PM on August 20, 2012


What an odd argument - electric cars have limited range so lets utilize them as a fleet of constantly available autonomous cabs. When would they be charged - you'd need three or four of them in order to replace one ic engined vehicle negating the lower operating costs. Daft.

I can't imagine self-driving cars on the freeway safely anytime soon without a dedicated left lane. What this lane would allow is bumper to bumper driving, and perhaps charging on the go from a charged railing aside or under the car. An electric lane left of the freeway could charge batteries and cram so many cars together as to make the freeway triple its capacity. I think the final touch on all of this would be to allow electric autonomous vehicles to exit left on the freeway using a special lightweight ramp-to-bridge that would be constructed for this purpose, since exiting would be controlled by a computer and ideally suited for this otherwise dangerous exit when conditions are crowded. I also note that so many cars added to the freeway would need new exit ramps anyway, so they might as well go left rather than crossover and cause accidents. We could do it by saving all that money on light rail, which costs 50 million per mile and often loses money.
posted by Brian B. at 6:17 PM on August 20, 2012 [1 favorite]


I don't think human beings have a good enough track record to put human lives at the autonomous mercy of our own engineering skills.

Or our own regulatory skills. How will we regulate autonomous vehicles? Will each car be required to run the same standardized AI that behaves in roughly the same way in every circumstance, or will BMW be allowed to install a performance AI an AI that drives 40% more douchy to satisfy their target audience.
posted by RonButNotStupid at 6:24 PM on August 20, 2012 [2 favorites]


Self driving electric cars are so so early 21st century. Be where the puck is going people.How would you like to never have to leave the house ever again? We need near lightspeed telepresence by wiring into the brainstem, transmit to and from a drone via quantum entanglement. Hate shopping? send the drone. Boss sick of you turning up late for work? It will never happen again if you leave a spare drone at the office.

It is 2012, why do we still have to go places?
posted by Ad hominem at 6:28 PM on August 20, 2012 [6 favorites]


I can't imagine self-driving cars on the freeway safely anytime soon without a dedicated left lane

Well they already are in Nevada...

Self driving cars will have some failures, sure. But current evidence shows them better than human drivers, and it's not surprising. Humans are VERY EASILY distracted. Driving requires paying attention to a lot of things at once, easy for a computer (which never gets bored) but tough for people.

They don't need special infrastructure. The old ideas were all about turning every car into self-driving and then having them all drive 100mph with 1 in separation or some nonsense. They don't need to talk directly to each other, although it would be a good additional signal for them. But you could never rely on it since bicycles, people, animals, trash, etc would not have the signals and have to be navigated as well.
posted by wildcrdj at 6:29 PM on August 20, 2012 [2 favorites]


I don't think human beings have a good enough track record to put human lives at the autonomous mercy of our own engineering skills.

You may want to check out the live videos of Google's car, if you haven't already. It's easily a better driver than most humans.
posted by odinsdream at 6:30 PM on August 20, 2012 [3 favorites]


I'm not sure about everyone else, but if I had a self driving car I'd take a whole lot of all night road trips to every city within a thousand miles of my house. Which would be awesome, but not good for the planet.
posted by miyabo at 6:35 PM on August 20, 2012 [2 favorites]


I'm sure that one model that's had a ton of resources dumped into it by Google is going to perform great. Yes, NASA managed to land several rovers on the moon already - yeah, tons of resources being dumped into a single effort. When it scales up will it have all of those resources all the time? Even in the politically unimportant neighborhoods? I just don't trust human beings with any kind of essential technology anymore.
posted by bleep at 6:44 PM on August 20, 2012 [1 favorite]


And of course when I say "the moon" I mean Mars. See I can't even type right.
posted by bleep at 6:45 PM on August 20, 2012


If they are so great at driving, let's see one in the NASCAR race at Bristol this weekend?
posted by I'm Doing the Dishes at 6:48 PM on August 20, 2012


I am really really really excited about electric self-driving cars!

Imagine if you could commute one hour every day or night, no matter how close you were to a train station, with a small carbon footprint (if you power it with wind/nukes), and just sleep or read or whatever. In the bay area, you could live someplace truly beautiful out in the wild, pay much less rent, and still bring in big city bucks. The effect on the real estate market is something that I think people are underestimating here. From San Francisco, you could spend Saturday in Las Vegas, Sunday in Salt Lake City, and be back to work monday morning.

Also this is something that is truly amazing about Google's R&D. I feel like I'm looking at the new bell labs. Sure it's churning out stupid advertising search algorithms but it's also doing really really cool shit that will actually change people's lives with engineering and technology. In today's business world where businesses don't know where to put their cash because they believe that consumers won't spend and they're afraid of a double-dip, seeing an organization forging a new market through the power of technology (versus apple's "innovative design" or microsoft's "new" version of Winblows) is AWESOME.
posted by The Ted at 6:48 PM on August 20, 2012 [17 favorites]


I'm sure that one model that's had a ton of resources dumped into it by Google is going to perform great

Well there are several of them, but they are prototypes. Prototypes cost more than production models, yes. But I'm not sure why you think it couldn't be produced for a reasonable price.

I don't even know what you mean by "all of those resources all the time". It is a car. It has whatever technology is in the car all of the time, yes...
posted by wildcrdj at 6:49 PM on August 20, 2012 [1 favorite]


"When would they be charged..." Folks trip on the idea of building infrastructure to support electric cars mainly because we have become so used to the petrol-based infrastructure that we completely ignore how mind-bogglingly enormous, convoluted, costly and polluting it is. Supporting electric cars will be a lot simpler and cleaner.

"I can't imagine self-driving cars on the freeway safely anytime soon..." This, I think, is one of the main hurdles for the future of the electric autonomous vehicle. But it's moving in the right direction. I'm really excited for the day that the manually operated, combustion engine driven automobile is history.

Also, everything The Ted just said!
posted by fartknocker at 6:51 PM on August 20, 2012 [3 favorites]


Or another way to put it --- the car requires nothing external to itself. It is not using cloud computing, external sensors, etc. It uses GPS to navigate but a GPS failure would at worst result in getting lost, although it shouldn't since map + local sensor data should be sufficient.

So the neighborhood a car is in would have no bearing on its safety.

It's not clear how much extra cost the tech adds since it's still in the prototype stage, the actual electronics should not be staggering though since we're talking about a computer and some sensors. Software costs scale very well, obviously.
posted by wildcrdj at 6:54 PM on August 20, 2012


And I know others have already pointed this out above, but it bears repeating with added emphasis: WE SUCK AT DRIVING, PEOPLE!

Think about it... stop signs, turn signals, break lights, stop lights... all things we invented to keep from killing ourselves at an even more egregious rate than we currently do behind the wheel. Autonomous cars deal with these things better than we do, and they don't even need them.
posted by fartknocker at 6:58 PM on August 20, 2012 [2 favorites]


Gosh, a company pushing self-driving electric cars thinks that the solution to transportation problems are self-driving electric cars.

Here's some of the flawed logic in their arguments (ignoring the pie-in-the-sky nature of magic cars from the future):
But if the car could drive itself, it could return home to take the children to school, members of the family shopping, and seniors to visit friends or keep appointments. If the vehicle served even one additional passenger, its utilization rate would double, and its capital cost per user would fall by half.

So instead of driving to work, parking, and then driving home, the car drives me to work, then drives back home, then to school, then back to my work, then drives me home, then to school, then back home. Maybe the "utilization" will double, but the number of trips on the roads, i..e the traffic congestion, quadrupled. The energy efficiency was just cut in half, and the demands on the battery quadrupled.

The actual transportation problems we have in general are that the roads are full (and expensive to expand), and that transportation is energy-intensive. The problems that electric cars have in particular are that they have limited range, daytime charging creates additional peak demand on the power grid, and that they are mostly burning coal (since that's where electricity comes from today). So we've managed to make none of the problems better, and all of them worse.

The viability of electric vehicles would be further enhanced if they were used as a service, rather than purchased as assets.

Here's transportation as a service: Daimler has been operating car2go, a "transportation as service" for over a year in several urban markets in North America. It's a carsharing program, where the premise is that hundreds of cars are made available in a central "home" area (which is much larger than the downtown, for instance in Washington, the home area is essentially the entire District); you can use one anytime, and drop it off anywhere in the home area. It's incredibly convenient, and has been a big success - they recently rolled it out here in Calgary. On a typical weekday, we have about 500 trips being made in this central area. If we had the usage that Vancouver (the most successful market) had, we would have almost 1000. So it exists, and it is a success. As it happens, I know roughly how many trips are actually being made in this home area: 750,000.

Nor are buses a realistic alternative for daily commuting for most people. A recent USA Today article recounts the story of a university employee in Arizona who lost her car to an accident and decided to take the bus instead. Her daily commute increased from 20 minutes to one hour each way. Do the math, and her monthly commuting time increased by the equivalent of three working days. Increasing the working month by three days is not social progress -- it is a social and economic catastrophe.

Well, if it doesn't work for one person in Arizona (surely the gold standard for effective public transit), it can't work anywhere. American families spend 15.64% of their budgets on cars; that's 3.4 working days a month.

Transportation problems are basic, and people get hypnotized into thinking that a shiny technology will solve our problems. Once, it was monorails. There are still Personal Rapid Transit advocates around. Self-driving cars have been coming soon, any day now, for decades. So has nuclear fusion. The point I'm trying to make is that it's tenuous to rely for our salvation on technologies that haven't been invented, or that have been invented but not made practical, or that are practical but not at industrial scale. Some make it out of development hell, some never do. North American society hasn't done a very good job of solving our transportation problems using technologies like high speed rail, urban public transit and mixed use development, and we've had these technologies for 50, 100 and 2000 years respectively.

Our main transportation problems are psychological. The problems are the ideas that cars are the only effective solution, or that suburbia is what everybody should live in, or that transit won't work, or that we can't own a small electric car to solve 99.5% of our trips and then rent a big car for the annual road trip. I'm always a little leery of technological solutions to psychological problems.
posted by Homeboy Trouble at 7:00 PM on August 20, 2012 [8 favorites]


Well, they'll still need the stop lights to brake for pedestrians.
posted by Kevin Street at 7:01 PM on August 20, 2012


The thing that I think would be an awful downside of The Ted's vision of being able to live out in "the wild" and work in the city is that that would be the end of "the wild." If there's no barrier of inconvenience to living and working over long distances, eventually everyone moves to those cute little towns and mountainside idylls and builds and builds and builds and before long, it's just tract housing everywhere. In Maryland, a very good example of this, albeit without the magical robot cars, is the former small city/farm town of Frederick, which was pretty much loved to death by DC-working folks right about the time I-270 was widened into a fast, zillion lane highway. Easy routes to everywhere will make everywhere exactly the same—soulless tract mush.
posted by sonascope at 7:03 PM on August 20, 2012 [6 favorites]


According to the National Highway Traffic Administration, car accidents happen every minute of the day. Motor vehicle accidents occur in any part of the world every 60 seconds. And if it’s all summed up in a yearly basis,there are 5.25 million driving accidents that take place per year. Statistics show that each year,43,000 or more of the United States’ population die due to vehicular accidents...

Yikes. Humans suck at driving. I'm sure any of us here could write an algorithm that if employed, would be at least marginally safer than the status quo. On our lunch break. In LOGO.
posted by sourwookie at 7:04 PM on August 20, 2012 [3 favorites]


At last technology catches up with my liver.
posted by lumpenprole at 7:11 PM on August 20, 2012 [3 favorites]


Roughly speaking, drunk driving accounts for one third of all fatal accidents in the US. Even if the AI is "only" as good as the average driver, road fatalities would still fall significantly just from stopping those accidents.
posted by fings at 7:22 PM on August 20, 2012 [5 favorites]


judging from the code I fixed today I'm sure most driving algorithms would probably just take every single left turn till it got where it was going and periodically decide to take itself and any other car around it out simply because the driver did something unanticipated like roll down the windows and turn on the wipers at the same time.
posted by Ad hominem at 7:24 PM on August 20, 2012 [2 favorites]


Well there are several of them, but they are prototypes. Prototypes cost more than production models, yes. But I'm not sure why you think it couldn't be produced for a reasonable price.

Well, the hardware itself isn't cheap. For instance, that futuristic-looking LIDAR sensor on top of the car: the original version, designed specifically for the DARPA Grand Challenge, cost about $75,000; there's now a smaller version with half as many lasers that will only set you back about $30k. Maybe at some point the software will be sophisticated enough to infer the same amount of knowledge about the world from cheaper sensors, like cameras, but we're nowhere near there yet.

Add in the cost of GPS, radar, cameras, actuators, and the computers themselves, and you can see why these things are so pricy. For now, at least.
posted by teraflop at 7:28 PM on August 20, 2012 [1 favorite]


GPS, radar, cameras, and actuators are all super cheap. Well, maybe not the actuators, but since every car already comes with them, there is no price increase.
posted by ryanrs at 7:31 PM on August 20, 2012 [3 favorites]


I'm sure that one model that's had a ton of resources dumped into it by Google is going to perform great. Yes, NASA managed to land several rovers on the moon already - yeah, tons of resources being dumped into a single effort. When it scales up will it have all of those resources all the time? Even in the politically unimportant neighborhoods? I just don't trust human beings with any kind of essential technology anymore.

What do you imagine the resources being here? Once you do the initial research, the actual manufacturing is not the hard part. The hard part is the software, which is kinda free to copy.

So instead of driving to work, parking, and then driving home, the car drives me to work, then drives back home, then to school, then back to my work, then drives me home, then to school, then back home. Maybe the "utilization" will double, but the number of trips on the roads, i..e the traffic congestion, quadrupled. The energy efficiency was just cut in half, and the demands on the battery quadrupled.

Er, no. Utilization can double without traffic doubling, because instead of 2 people having 2 cars, 2 people have 1 car that drives (a little more than*) twice as much.

*because it needs to get from where person 1 dropped it off to where person 2 needs it.

Add in the cost of GPS, radar, cameras, actuators, and the computers themselves, and you can see why these things are so pricy. For now, at least.


The rest of that stuff is a drop in the bucket.
posted by Jpfed at 7:33 PM on August 20, 2012


Er, no. Utilization can double without traffic doubling, because instead of 2 people having 2 cars, 2 people have 1 car that drives (a little more than*) twice as much.

*because it needs to get from where person 1 dropped it off to where person 2 needs it.

Except that person 1 probably dropped it off at their job downtown in the morning. And nobody downtown wants to go anywhere; they just got to their jobs. The main group of people who want to go somewhere are people at home in the suburbs, where person 1 came from, wanting to go downtown.
posted by Homeboy Trouble at 7:45 PM on August 20, 2012


GPS, radar, cameras, and actuators are all super cheap. Well, maybe not the actuators, but since every car already comes with them, there is no price increase.

The LIDAR is certainly the most expensive single component, but don't underestimate the rest of the equipment. The GPS in one of these cars is a lot more accurate than a consumer module, and includes a very precise IMU which is also not cheap at all. And AFAIK, most cars don't have any kind of electronic control for steering, braking or shifting gears, so that stuff has to be added too. And it has to be very reliable.

Don't get me wrong: having worked on this stuff for a while, I think the technology is awesome and has a huge amount of potential. I just think we still have a while to wait before that potential becomes an economically feasible reality.
posted by teraflop at 7:47 PM on August 20, 2012


I should add that the paper describing Stanford's "Junior" (the predecessor to Google's car) is available online, and has tons of juicy information for those who are interested.
posted by teraflop at 7:52 PM on August 20, 2012 [1 favorite]


The ballpark for the self-driving thing is about half a million bucks per car right now, which means that if Moore's Law holds then in 12 to 14 years it'll be a two thousand dollar premium. Price and technology aren't the problem. And differential GPS isn't a concern: Roads move. A lot. As does construction. Global positioning isn't useful, local positioning is.

What I love are the people who say "but most cars spend 22 hours a day parked!": Yeah, that's why we have rush hour. That's why the freakin' bus didn't have room for my bicycle today so I had to wait for the next one. And as others have pointed out, we have commute and anti-commute directions, re-using the vehicle doesn't save that much.

The capital cost of the vehicle is what we pay for convenience. Trying to pretend that that desire for convenience is suddenly going away is silly. If energy costs rise enough to make us take transportation costs seriously we'll react with buses and telecommuting and housing and work development patterns, not by sharing the falling costs of capital, especially since those costs would have to be matched by energy costs for us to shift our use patterns.
posted by straw at 7:53 PM on August 20, 2012 [2 favorites]


Self driving cars will have some failures, sure. But current evidence shows them better than human drivers, and it's not surprising. Humans are VERY EASILY distracted. Driving requires paying attention to a lot of things at once, easy for a computer (which never gets bored) but tough for people.

They don't need special infrastructure. The old ideas were all about turning every car into self-driving and then having them all drive 100mph with 1 in separation or some nonsense. They don't need to talk directly to each other, although it would be a good additional signal for them. But you could never rely on it since bicycles, people, animals, trash, etc would not have the signals and have to be navigated as well.


The real issue here isn't possibilities, but to go from a luxury to necessity, so that it wasn't just a whim to decide to let a machine drive your car, but instead afforded ultra safety + ultra efficiency. To me this means a new cram lane, in order to avoid rush hour at very high speeds while decongesting the freeway at the same time. It's like putting a train on a freeway, because we can, and it reaches a potential in the technology that most people would logically choose. Simply putting more expensive autonomous cars out there under same conditions is merely proof of sameness, and very little changes, except maybe for current non-drivers, and they currently take the bus for far less money.
posted by Brian B. at 7:53 PM on August 20, 2012


If you live in a big city, electric cars are totally viable. How often do you need to drive over a hundred miles in a single day? That's the range of my Nissan Leaf. And spending 20 bucks a month to drive it about a thousand miles is pretty damned awesome. This self-driving nonsense seems a bit bogus however.
posted by Windopaene at 7:57 PM on August 20, 2012 [1 favorite]


I'm glad the Stonecutters relented on this one. Now if they'd just stop rigging every Oscar night...
posted by MoonOrb at 7:58 PM on August 20, 2012


imply putting more expensive autonomous cars out there under same conditions is merely proof of sameness,

No, the claim is that safety will increase significantly even with existing roads and non-driverless cars. If they have fewer accidents than humans (currently true, and there are many reasons to think it will continue to be true) that is all the benefit that is needed.

Efficiency would be nice but is hardly required for self-driving cars to be worth it.
posted by wildcrdj at 7:58 PM on August 20, 2012 [1 favorite]


most cars don't have any kind of electronic control for steering, braking or shifting gears

Sure they do.
posted by ryanrs at 8:20 PM on August 20, 2012


It always amazes me that people are more willing to entertain the idea of robot cars than of good public transit infrastructure.

Anyway, I just took my first car2go trip today. Those smart cars have the lurch-iest transmission of anything I've ever driven (admittedly, that's not saying much).
posted by sevenyearlurk at 8:27 PM on August 20, 2012 [1 favorite]


I worked on this stuff for a little bit too (autonomous vehicle startup that never went anywhere).

I'm not at all worried about the technology itself. Even with just a couple of months of funding, we got a pretty decent little autonomous bot running with lidar, GPS, an IMU, and a camera that could navigate obstructions. The core software and hardware components are off the shelf at this point.

I do think it's unlikely the US will be the first place to get self-driving cars. Any manufacturer would be terrified of huge lawsuits here. Our government treats one of the critical components -- the IMU -- as a weapon, and they're basically impossible to import or export from our borders. And the weather in large parts of the US is far from ideal, I'd like to see any kind of sensor that could hold up in a Midwest blizzard. I'd bet on Australia or Israel or some other sunny country with a flexible government to adopt the tech first.

And I really don't see any reason to think this will be good for the environment. People will simply travel much more when they don't have to drive. Congestion and pollution will be much, much worse. Maybe the government can pass a law that all autonomous cars have to be small and electric.
posted by miyabo at 8:27 PM on August 20, 2012 [5 favorites]


This self-driving nonsense seems a bit bogus however.

The nonsense that Google's been testing and operating successfully? The nonsense that works and has a better safety record than human drivers? That nonsense?
posted by Pope Guilty at 8:32 PM on August 20, 2012 [2 favorites]


What I wonder is-- suppose automated cars take off. And suppose they are "a bit" better safety-wise than current drivers (drunk and stupid though they may be), but that they are not perfect, and that pedestrians continue to be run over. Someone in government is sure to make the point that now that everybody, including the blind and children and so on, can ride in an automated car, the easiest solution to the problem of people getting run over is to simply ban people from the streets entirely, or at least remove whatever legal protection they now enjoy-- after all, why protect someone who's not concerned about their own safety (if they were concerned with their safety, after all, they'd be in a car). So much cheaper and easier than building sidewalks and crosswalks and making everyone slow down instead of traveling everywhere at 60 mph under computer control.

And if this physically prevents the indigent and the paperless from entering the nice parts of town, well, that's just a sacrifice we'll reluctantly have to make.
posted by alexei at 8:33 PM on August 20, 2012 [2 favorites]


BUT NOT IN FLORIDA!!!
posted by schmod at 8:44 PM on August 20, 2012


No, the claim is that safety will increase significantly even with existing roads and non-driverless cars. If they have fewer accidents than humans (currently true, and there are many reasons to think it will continue to be true) that is all the benefit that is needed.

Efficiency would be nice but is hardly required for self-driving cars to be worth it.


Cars that can switch over to individual driving under same conditions is also what is being claimed. Worse though is that drunks, children, and the elderly may be placed in defensive conditions to switch over to manual driving in emergencies, or because they lack judgement in non-emergencies. Bad decisions in cars cause accidents, and we're entertaining the idea of increasing bad driver decision makers in cars. Additionally, traffic accidents must factor conditions, and when nobody is driving, but somebody forces the trip to occur, then we have a new avenue of trip failure. I think it is too soon to argue that these cars blend well by allowing poor drivers-passengers on the road in potentially worse conditions. At no time did I assume a system-wide failure, which is also a possibility in blended traffic.
posted by Brian B. at 8:55 PM on August 20, 2012


This article seemed fairly convincing until it decided that public transit couldn't possibly solve our transportation problems. But even before then, it didn't really go very far to explain how an automated fleet of self-driving electric cars would solve the problem of rush hour, when a substantial majority of people go to work or come back home. Having one car take on the ferrying duties of several people is useless if they all need the car at the same time, and if your fleet of vehicles needs to be large enough to handle rush hour loads, what problem have you solved, exactly?
posted by chrominance at 8:59 PM on August 20, 2012 [3 favorites]


A nice 200sqft. RV with a self-driving setup will probably run $100,000 or less in a few years. The same size apartment in a nice neighborhood in Manhattan can run $400,000. And the speed limit is a maximum, so why not just get an RV and cruise around a few blocks at 10mph or so, rather than dumping all of your money into an apartment?

(That might well be the straw that breaks the camel's back on road pricing.)
posted by akgerber at 9:32 PM on August 20, 2012 [4 favorites]


this entire thread and just one single mention of how the batteries even get their charge. it doesn't even matter that they can't come close to the energy density of what the sun and the earth have cooked up over the millenia. in fact it's worse, since that energy almost certainly comes from just a different carbon fuel. california, a place with an ideal moderate climate that won't mess with battery's ability to function, a population wealthy and progressive enough to adopt the tech, and even a healthy car culture with an appreciation of nice cars of any type, is at the very edge of being able to support the current power demands of the population. and with san onofre (nuclear) out, they're having to dial back capability for the whole region. so how would they support a fleet of these things big enough to transit the public about?

if you live in a cold climate, remember how terrible it is to go out and start your car in January? well if your entire car is a battery, how weak is it going to be on a day like that? what will your range be like? you know what killed the electric car? they suck is what killed it. but if, on every corner, there was a little building where you could go buy Doritos and refill your electricity in less than 10 minutes then an electric car becomes sort of viable. how ridiculous would the stress on the power grid be, though? and at 5pm when everyone who ISN'T driving home and charging up is already at home with the TV and the lights on making dinner? seriously? the existence of the electric car will not in any way blend with generations of people raised on the lifestyle that gasoline has given them. the younger generations are progressive enough to adapt to the changes, but the infrastructure is not.

I have $2,500. I'm not really a high school kid who probably cares about the environment, but I need a car to get to school/work. I'm gonna go on craigslist and find an 89 Honda Accord. I'll get 4 seats, hvac, and like 29mpg. what I won't find for 10x that money is an electric car. i know that's a little off topic, but this is not something that will change anytime soon.

so, in short, thorium nuclear thorium nuclear thorium nuclear. and then when that happens, i'll say in a thread about autonomous electric cars "Hey, that's a great idea that will also work right now."
posted by ninjew at 9:35 PM on August 20, 2012


i..e the traffic congestion, quadrupled.

Actually it would probably be reduced, much congestion is causes or exacerbated by bad driving - and not just accidents, simple slowing down and speeding up in an inefficient manner will do it.

Smart cars enable smart speed limits etc. Traffic management with a large proportion of self-running cars would almost certainly be better, even if trafic on the roads increased considerably.
posted by smoke at 9:45 PM on August 20, 2012 [4 favorites]


It always amazes me that people are more willing to entertain the idea of robot cars than of good public transit infrastructure.

Because robot cars are most people's ideal public transit infrastructure?
posted by ChurchHatesTucker at 9:50 PM on August 20, 2012 [7 favorites]


how ridiculous would the stress on the power grid be, though? and at 5pm when everyone who ISN'T driving home and charging up is already at home with the TV and the lights on making dinner? seriously?

You mean the power grid in a world where energy companies are no longer able to sell petroleum-based fuels (as everyone apparently has gone all-electric), and so have converted their billions upon billions of dollars in capital to producing electricity instead in an effort to remain in business and competitive?

I have $2,500. I'm not really a high school kid who probably cares about the environment, but I need a car to get to school/work. I'm gonna go on craigslist and find an 89 Honda Accord. I'll get 4 seats, hvac, and like 29mpg. what I won't find for 10x that money is an electric car. i know that's a little off topic, but this is not something that will change anytime soon.

Your '89 Accord is $2,500 because it is 23 years old. How much do you think a 2012 Nissan Leaf is going to cost in 2035? You can only buy an '89 Accord for $2,500 now because someone paid full price for a brand new one two+ decades ago.
posted by tylerkaraszewski at 9:55 PM on August 20, 2012 [1 favorite]


Jpfed: "The hard part is the software, which is kinda free to copy."

Free software from a corporation is usually ad-supported.

Car: We are approaching McDonald's in 45 seconds. Just say "Yes" to enjoy a Big Mac value meal.
posted by Bonzai at 10:01 PM on August 20, 2012 [1 favorite]


if I had a self driving car I'd take a whole lot of all night road trips to every city within a thousand miles of my house. Which would be awesome, but not good for the planet.

It probably wouldn't be any that much worse than flying the same distance, which you can do today if you wanted to, and the biggest cost component of a cheap seat on a jet is fuel. The biggest part of the marginal cost of taking a car trip is also fuel. Either way, you are paying for the energy that it takes to get you from one place to another, and if your desire for travel is limited by cost, you should end up doing about the same.

well if your entire car is a battery, how weak is it going to be on a [cold] day like that?

Gee, I don't know, I wonder if they could do something crazy like use the power cord that's plugging the car into the grid (which it's using to charge) to keep the batteries warm enough to function. Nah, I'm sure the engineers will just let the batteries get cold and fail to operate. Did you seriously think that's not something they've anticipated?

There are plug-in hybrids on the market today that solve a bunch of your problems. (And they'll even pre-heat or pre-cool themselves in preparation for a trip, so you never have to get into a hot or cold car, without having to idle it like you would with a gas engine.) The grid loading isn't that much of an issue either; if you want faster charging than your household electrical service supports, you could store energy in some form as part of the charging system, either in a separate stack of batteries or some other method. Real-time electricity pricing might even make that economical, by making nighttime electricity so much cheaper than daytime peak power that it overwhelms the efficiency loss in a local storage system. Systems like this would have the effect of making the power system much more resilient, and could potentially be tied into local generation, or use a hybrid's engine to backfeed the grid during times of exceptionally high power demand (and thus, pricing).

I have $2,500. I'm a high school kid who probably cares about the environment, but I need a car to get to school/work. I'm gonna go on craigslist and find an 89 Honda Accord. I'll get 4 seats, hvac, and like 29mpg.

That's nice. And you should do that; it's pretty much what I'd tell you to do if you asked it on the Green right now. (Although 29MPH is not really impressive; an old Civic will do a lot better than that.) But be aware that we're living in the last days of cheap gasoline. The reason that most people will eventually move to electric cars isn't because of environmental consciousness, it's because electric cars — more likely in the short term, some sort of CNG hybrids — will eventually be cheaper. Whether on their merits or because of subsidization (or just an end to subsidies on gas) doesn't really matter. But the current system is not sustainable.
posted by Kadin2048 at 10:04 PM on August 20, 2012 [1 favorite]


most cars don't have any kind of electronic control for steering, braking or shifting gears

Sure they do.


That's interesting, care to share any examples?
posted by teraflop at 10:05 PM on August 20, 2012


Hybrids have electronically-controlled braking and shifting. Steering does take extra work.
posted by GuyZero at 10:08 PM on August 20, 2012


ninjew: "this entire thread and just one single mention of how the batteries even get their charge. ... that energy almost certainly comes from just a different carbon fuel. california ... is at the very edge of being able to support the current power demands of the population. and with san onofre (nuclear) out, they're having to dial back capability for the whole region. so how would they support a fleet of these things big enough to transit the public about?"

Most places have excess capacity at night. This is because power plants don't power down for the night and they have little to no place to store excess. You could power a great number of cars on the current wasted electricity.
posted by Bonzai at 10:10 PM on August 20, 2012


billions of dollars cannot make more reserved of carbon energy available to turn into electricity.

the age of my accord isn't an issue. it is cheap to run, cheap to fix, and functions as viable transport.

yes, I understand the concept of battery warming. just use more energy..

many modern cars operate with drive-by-wire controls. instead of moving cables, gears, or levers, your controls operate the functions of the mechanical bits via servos and wiring. thus they can be remotely controlled.

we can hit peak coal and nat gas, too. if we are talking about a mass move over to electricity on the scale of how many billions of people use gasoline every day now, I hate to keep saying this but how we generate that power will become THE issue. because you're right, we cannot store it. after all, most of gasoline's benefit is that it is STORED energy.
posted by ninjew at 10:15 PM on August 20, 2012


All the arguments for these electric cars also work for electric busses as busses are currently used.

Indeed, electric buses were tried and failed in London over 100 years ago.

But who knows, perhaps today with better batteries they could work. But perhaps it's just not worth it against LPG and diesel busses.
posted by sien at 10:19 PM on August 20, 2012


Well, but the driverless thing could be a way around battery problems. If the cars have enough range to get themselves back to a station, a mechanic there could just swap out the drained battery for a charged one. Ideally suburbs would be on commuter train lines, but in America that's rarely the case - if this is what it would take to make it convenient and cost-effective enough for commuters to go electric, I'd certainly be for it.

Also:

In other parts of the country, taxis are scarce and expensive. In the town of Princeton, New Jersey, for example, taxis are found only at the train station, and the brief round trip from there to downtown Princeton costs approximately $40.

Hah! I feel totally vindicated that this is the specific example they chose. To add insult to injury, you can't split taxi fares in Princeton (even if you're going to exactly the same place -- I know), so even if you get home at 2:00 am with two friends (or can find two friends in the parking lot) that three mile trip is still going to cost each of you around $15.

In fairness, the article doesn't mention the shuttle train, which goes that same route and is much cheaper. I feel churlish complaining about it when it is so much better than most suburbs, and really it is a great option as long as the shuttle has started running for the day, hasn't left without you, and hasn't stopped running for the night or for lunch. So well-designed public transportation can help a lot.
posted by en forme de poire at 10:27 PM on August 20, 2012


It always amazes me that people are more willing to entertain the idea of robot cars than of good public transit infrastructure.

Google has built robot cars. You can watch the demos. It is easy to imagine how further development will produce a product for sale, which you could buy, for some price, and then own, and use. This sort of thing happens all the time.

Public transit infrastructure involves politics - worse, it involves local politics, a well-known morass of incomprehensible complexity and intractable, decades-long conflicts. Public transit is not a problem you, an individual person, have any hope of solving. At best you'll get to vote on some complex, flawed, frustrating, expensive, best-we-can-do-for-now initiative every five or ten years. And then public transit will still suck.

Robot cars? All you have to do is wait ten years, then swipe your card. That's easy to imagine.
posted by Mars Saxman at 10:36 PM on August 20, 2012 [3 favorites]


This would be such a great bargaining chip if only scientists and engineers would unionize. Although instead of negotiating with management they'd be bargaining with society at large: "Sure, you can have self-driving robot cars ... if America switches to the metric system."

I know, I know, that's not how science works.
posted by compartment at 11:02 PM on August 20, 2012 [1 favorite]


If automated cars are hired for short trips, like driverless taxis, how would the company owning the cars protect against things like vandalism and people spitting/vomiting in the cars? With a taxi driver in the car at least there is some oversight there. What if some drunk asshole maliciously vandalized a car and the next passenger was injured or killed in an accident as a result? People vandalize vending machines all the time for eating their 2 dollars and not spitting out a coke, what if a driverless car takes them to the wrong location or eats their credit card and they decide to start kicking in the paneling? Making these things safe for the general public also means making them safe from the general public. In my home town about 15 years ago there was a collectivized bicycle scheme where bicycles were left around the city in special "green bike" bike stands and anyone could take one and ride it for free to one of the other green bike stands around the town. It was really cool and handy but eventually it was abandoned because people (mostly schoolchildren) were vandalizing the bikes. Any driverless car scheme will probably require constant in-car camera surveillance, remotely monitored by humans with the ability to lock the passengers inside and redirect the car to a police station or private security post if there is any trouble. That has it's own problems too obviously.
posted by L.P. Hatecraft at 11:05 PM on August 20, 2012 [1 favorite]


Because robot cars are most people's ideal public transit infrastructure?

Which is exactly why I am completely opposed to this technology in every way.

Cars are toxic regardless of what their carbon footprint is. They have poisoned every design decision we make about our environments. We prioritize things like parking and congestion over the quality of life of the actual people who live and work in our cities and towns. Our geography is built on automobile scale - things must be spread out and further apart because traffic congestion demands it to be so. In most places, the environment is actively hostile to all non-motorized transportation (like the radical act of walking to the store in most American towns). Our communities are obstacles to be moved through as rapidly as possible, in our own little isolation bubbles, air tight and sealed away from everyone and everything around us.

The automobile is the ultimate technology of alienation. What does it say about us that we would rather be shut away in our little glass bubbles, all alone, rather than just take a bus or train?
posted by bradbane at 11:05 PM on August 20, 2012 [12 favorites]


I hate to keep saying this but how we generate that power will become THE issue.

Well, yeah. It's kind of the issue now in much of the world, in those places where they realize they are not going to be able to outbid the industrialized nations for the easy-to-use carbon sources.

But the advantage of having an all-electric transportation fleet is that it's relatively easy to switch or diversify between various input sources. Building a new powerplant and attaching it to the grid is a relatively straightforward process compared to rolling out a few hundred million cars and all the infrastructure they require to use a new fuel source.

However — and quite sadly, from an environmental (and long-term humanist) perspective — we'll probably go through another round of internal combustion engines powered by CNG before we actually get our shit together and eliminate the onboard carbon fuel source. It's a far less dense fuel than gasoline will have been, but it's comparable or better than most batteries and uses engineering that is one step above plumbing. That it will drive us to extract and release tremendous amounts of additional carbon is something that I doubt we'll think too hard about until it's too late.

But the electric technology will be there, on the off chance that we get our collective heads screwed on in time and discourage carbon fuel use, and of course the same technology that you need for an all-electric car can be easily paired with an carbon-fueled engine and generator to extend the vehicle's range. Plug-in hybrids get close to the best of both worlds, and as battery technology gets better over time, their range without starting the engine gets better and better. Realistically, they seem like the best and most easily implementable solution for weaning most drivers (and the communities that have grown up around the idea of cars) away from fossil fuels.
posted by Kadin2048 at 11:10 PM on August 20, 2012


Bradbane, while I don't entirely disagree with what you're saying, a lot of these urban-planning misfires have already been made and I think we also need ways to mitigate the problems. I think robocars shouldn't replace public transit, but they absolutely should supplement it, especially in cases where it is too impractical or difficult to connect, via trains/buses, existing suburban homes to the places where people work, eat, and socialize.
posted by en forme de poire at 11:13 PM on August 20, 2012


"But perhaps it's just not worth it against LPG and diesel busses." I know I shouldn't be, what with our non-stop love affair with classic cars and sports cars and cars that go vrooom!!!, but I'm always surprised when people stick up for the combustion engine. It's so noisy and it stinks so bad. It's like something out of an H. G. Wells novel. I'd like to think that we'd all like to be able to breath and talk easier than the combustion engine will allow, particularly considering how many of us want to drive these days.

"and anyone could take one and ride it for free" Well, there's your problem right there. Folks pay for something, they tend to take better care of it.
posted by fartknocker at 11:17 PM on August 20, 2012


Also, some good links here: http://www.reddit.com/r/SelfDrivingCars
posted by fartknocker at 11:19 PM on August 20, 2012 [2 favorites]


It always amazes me that people are more willing to entertain the idea of robot cars than of good public transit infrastructure.

People hate riding with other people.

Go low-tech for starters. Everywhere possible, give roads concrete dividers and curbs high enough that a car can't easily leap them, so a single failure doesn't turn into a head-on collision or a mowing down on the sidewalk. This would also help keep bad human drivers from killing people.

Use retractable bollards as needed to guarantee traffic stoppage at certain points.

Require that all cars (human-driven or automatic) include those devices that report the car's speed and behavior back to the insurance company. Now all the worst driving (constant speeding, etc.) is eliminated by prohibitively high insurance costs, and the rest of us are driving like robot cars because we can't afford not to. So why not just get a robot car and relax?

Put a common-sense computer in every car (automatic or manual), with independent location, speed, and direction sensors. Give it permission to shut down the car when the car does something out of bounds. This would work for automatic cars that fail and for manual cars being driven by people other than the car's owner (your kids, for example).

And always let a human with appropriate training and license override everything and take control, but automatically let the car owner and insurance company know that the nice safe computer system was manually taken offline at 11PM by the current driver (17-year-old son of the car owner), who then exceeded the speed limit by 30 mph, took a lot of corners at high speed, and did a lot of rapid starting and stopping before parking your car in the neighbor's lawn. If the son can convince his dad and the insurance company that it was an emergency, maybe the insurance rates won't go up.
posted by pracowity at 12:40 AM on August 21, 2012


What does it say about us that we would rather be shut away in our little glass bubbles, all alone, rather than just take a bus or train?

It says a variety of things but basically the same thing that people preferring to sleep in their own beds and own rooms instead of sleep communally in dormitories does.

I think you are seriously overestimating the benefits that would come from everyone in the world trying to arrange it so that whenever they travel they only need to go from somewhere that is serviced by public transportation to another place serviced by public transportation. Just things like having to travel ten times as far as the crow flies when your origin and destination aren't directly connected by a transportation network probably eats up much of any efficiency gains that result from the per-mile cost of traveling on a bus or train being better than the per-mile cost of a car, as well as things like buses and trains running when they're mostly empty.

I also don't think that land usage would be that different with more concentration of travel around public transportation. Things still have to be pretty spread out so that the public transportation itself and vehicles delivering cargo or ambulances, fire trucks, and other emergency vehicles can get in and out.

The notion of people sharing vehicles as the primary benefit which this article has is odd to me, since there would seem to be many more substantial ones, but self-driving vehicles will make the ideals you have for human behavior and land use more possible, not less.
posted by XMLicious at 12:41 AM on August 21, 2012


give roads concrete dividers and curbs high enough that a car can't easily leap them

Fuck that idea. Roads are public spaces and should be easily accessible by bicycles, mopeds and all sorts of vehicles other than cars. If you are concerned about safety there is is jut one simple solution. Reduce speeds.
posted by uandt at 1:20 AM on August 21, 2012


Public transit infrastructure involves politics - worse, it involves local politics, a well-known morass of incomprehensible complexity and intractable, decades-long conflicts. Public transit is not a problem you, an individual person, have any hope of solving. At best you'll get to vote on some complex, flawed, frustrating, expensive, best-we-can-do-for-now initiative every five or ten years. And then public transit will still suck.

Robot cars? All you have to do is wait ten years, then swipe your card. That's easy to imagine.


A night with Anne Hathaway is also easy to imagine. Monorails are easy to imagine. Easy to imagine is not a useful metric.

Robot cars involve liability. I totally agree they'll be better drivers than people. Instead of killing 32,000 Americans a year, maybe they only kill 3,200 Americans. Or they only kill 320. That's a 100x improvement, which would be remarkable -- sure computers don't get drunk and don't text, but people don't overheat and short out; people don't mismatch shapes; people don't give 404 errors.

What company is going to sign up to have 320 wrongful death lawsuits every year? And computers fail in different ways than people; look at face recognition failures -- many are baffling in ways we (or a jury) can't understand. The computers that kill people will often do things that no reasonable human could.

The only way to sort that out would involve changing liability law and auto insurance. And, just ask Obama, changing everybody's insurance is really easy to do with politics.

And it's not like we have self-driving cars ready to go. We have a few highly-funded prototypes that have only ever been driven while supervised by people. The most comparable recent high-tech automotive system I would use to compare is satellite navigation. I haven't been able to find any solid recent numbers on GPS use, but it seems like it might be ballpark 50%. About 10% of new cars were being equipped with GPS a few years ago, and of course many people either own a GPS device or a phone with navigation that they can take from car to car. So GPS penetration is probably faster than automated driving, which must be hardwired into the car.

And GPS has grown a lot in the last few years. But it was first made commercially available on the Oldsmobile 88, back in 1995. 17 years for the technology to become mid-market. And, again, that's from the first commercial application; robot cars are in early testing. The next car you buy will not drive itself. The car you buy to replace it will not drive itself. The one you buy to replace that one will not drive itself. The one you buy to replace that one may drive itself, especially if you're a bleeding-edge luxury car buyer.

I can't see anything that could possibly go wrong with ignoring our problems, on the assumption that techno-utopianism will solve them two decades from now.

Go low-tech for starters. Everywhere possible, give roads concrete dividers and curbs high enough that a car can't easily leap them, so a single failure doesn't turn into a head-on collision or a mowing down on the sidewalk. This would also help keep bad human drivers from killing people.

I live in a progressive (i.e. Canadian) city, with a much denser urban form than the typical US city, and transportation policies favouring pedestrians and transit riders. There was (I think it's been cancelled) a program in place to build sidewalks in all of our industrial areas. New suburbs only have sidewalks on one side of the street. And you're proposing concrete dividers and high curbs everywhere possible?

The industry standard for concrete road dividers is the Jersey barrier; it has a crossection of about 5 square feet. A mile of Jersey barriers is therefore 961 cubic yards of concrete; that's as much concrete as in 50 or 60 driveways. There are over 4 million miles of roadway in the United States. Putting Jersey barriers down the centre of even 1% of these roads would be enough concrete to pour a dozen Hoover Dams. That's on the order of 2-3 billion dollars, pure concrete cost alone. And that's for one percent. And just the barrier down the middle.
posted by Homeboy Trouble at 1:30 AM on August 21, 2012 [2 favorites]


What company is going to sign up to have 320 wrongful death lawsuits every year?

Is that actually a high number for a car company, much less for the entire auto industry, compared to the number of lawsuits already resulting from mechanical or other failures in cars?

Besides that, according to what you're laying out, every company that employs humans as drivers would face one hundred times more wrongful death lawsuits from its human drivers killing people than it would if it replaced those drivers with computers.

I don't know about ten years but we do already have driverless trains. It doesn't seem like driverless road vehicles would be that far off.
posted by XMLicious at 1:54 AM on August 21, 2012


People will simply travel much more when they don't have to drive.

I don't see this myself. The main constraint on how much most of us get to travel is all the other stuff we have to/want to do, like work, visit friends who live nearby, do the housework and shopping, and so on. We usually have a limited amount of time that we can spend on leisure trips out of town. Maybe more of those will end up as 300-mile round trips instead of 100-mile round trips - but maybe not, because another constraint is how long we want to sit in a car rather than be at our destination.
posted by rory at 2:26 AM on August 21, 2012


You really aren't going to need a bunch of concrete barricades. And Google is gearing up for self-driving on snowy roads. They've also dropped their requirement for two humans supervising the car (one "driver," one "navigator"). Almost makes me wish I still worked at Google!

I'd rather have a Google self-driving car watching out for me while I'm walking across the road or having to stop suddenly than I would a distracted texting or phone calling or drunk human driver. Sebastian Thrun's original motivation for working on self-driving cars was that his best friend was killed in a car accident when he was 17. We'll see how the liability works out (that is my largest area of concern) but Google has a pretty good in-house law firm.

The current cost of human driving errors is ENORMOUS.

A couple more advantages to self-driving cars:

- Safe platooning (or "road trains"). This has been possible for decades, but might actually be feasible with more self-aware cars. This can easily triple multi-lane road capacity.
- Platooning also saves fuel, estimated at 20% or more.
- Smaller intersections. (This one is slightly terrifying.)
- Your car could drop you off at the door, then go find its own parking.
- I, for one, would save money on speeding tickets, and I wouldn't care that I took a few minutes longer to get to my destination, because I would spend the trip doing something I enjoy.

Oh, and for the electric car skeptics who are concerned that we're just moving pollution sources from tailpipes to power plants, even in the areas of the U.S. where electric generation is the dirtiest (17% of the population), electric car-caused emissions are equivalent to gasoline cars getting 31 - 40 mpg. In the vast majority of the country, emissions caused by electric cars are far lower.

Re Vandalism: Zipcar and other subscription car sharing services deal with this just fine. You check in, you check out. Next subscriber reports any problems.

Finally, someone mentioned electric buses. Seattle has some pretty decent electric buses with no batteries. I've always wanted to set up an electric car to poach the power from the overhead wires, but I'm sure that wouldn't end well.
posted by Hello Dad, I'm in Jail at 2:34 AM on August 21, 2012


That's interesting, care to share any examples [of electronic control for steering, braking or shifting gears]?

Maybe I'm misunderstanding you, but all modern cars have power steering and electronically modulated brakes. Most cars have electronically controlled transmissions, too (at least in the US). Electronically controlled throttles are also very common. The only function that isn't usually electronically controlled or assisted in modern cars is the parking brake.
posted by ryanrs at 2:51 AM on August 21, 2012


It's curious that we're debating the safety of driverless cars when the technology to land passenger planes autonomously has existed for years and was first trialled successfully nearly 60 years ago.

What isn't being vocalised amid all the fake scaremongering about driverless cars, and I don't know if this is perceived as an issue, is the privacy, centralised control and hacking risks from a unified car and highway system.

Still, the real bar for driverless vehicle safety is set absurdly low judged against the number of actual accidents caused by drivers. In terms of acceptance of change are currently at the same point cars were at when a man with a flag was required to walk in front of moving cars.
posted by MuffinMan at 2:54 AM on August 21, 2012 [1 favorite]


A passenger plane needs to deal with drastically fewer uncontrollable variables on landing than a car doing a short urban commute. I am obviously not an expert but anyone who has programmed anything more complex than a bubblesort would not find it hard to be skeptical a car that drives itself across urban traffic, particularly in its ability to deal with unanticipated corner cases.
posted by Dr Dracator at 3:40 AM on August 21, 2012


I'm afraid the article fails basic logic : If you could build a robot electric car, then you could build a robot gasoline car. It follows that AI alone cannot improve the viability of electric cars. I suppose Google might improve their viability by only licensing their AI patents for electric vehicles, but still.

There is however one real massive advantage to electric cars, namely that they could acquire their electricity form the highway itself. If the French can power trams on old cobble stone streets using APS, then modern highways with carefully designed drainage could certainly support a ground-level power system designed for cars.

You need not upgrade all the roads simultaneously, market a couple electric and hybrid vehicles that support it, and offer discounted electricity on a few highways in eco-consious countries like Germany. As more roads are added, you eventually find that electric vehicles with smaller and smaller batteries are viable in specific markets, meaning you can build much lighter and cheaper vehicles with much less off-wire range but unlimited on-wire range.
posted by jeffburdges at 4:04 AM on August 21, 2012


Actually, self-driving cars are thus far safer than human drivers. Human drivers are actually really really bad at driving safely. The bar for "safer than humans" is extraordinarily low.

That's what fact says. But humans are also really really bad at making rational decisions.

As soon as one kid dies in an accident involving a robot car, the whole program gets shut down. And then we're back to 10 kids in that timespan dying from drunk drivers, idiots on their cellphones, or simple poor situational awareness.
posted by Foosnark at 4:51 AM on August 21, 2012


Wait til they learn how to hoon.

They're working on it, or at least giving it some relevant skills: Shelley at Thunderhill. Supposedly its lap times "rival those posted by professional drivers".

I'm looking forward to buying one of these auto-pilot cars some day, soon as they make one that can handle snow, and has a manual transmission for when I want to do the driving. I picture a little robot arm reaching out of the dashboard to do the shifting.
posted by sfenders at 4:57 AM on August 21, 2012


I'm afraid the article fails basic logic : If you could build a robot electric car, then you could build a robot gasoline car. It follows that AI alone cannot improve the viability of electric cars. I suppose Google might improve their viability by only licensing their AI patents for electric vehicles, but still

The article does a poor job of making its point. Electric cars have high capital costs but low operating costs, which means that increasing utilization gives them an overall cost advantage over gasoline vehicles. Due to the limited range, they can't be used for long distance driving but they can be used for many short trips if they operate as part of a fleet with a certain percentage of vehicles charging at any given time.
posted by atrazine at 5:42 AM on August 21, 2012


That's interesting, care to share any examples?

Yes!

Take anti-lock brakes. We have them because people suck at driving, even "good" drivers - it can modulate the brakes far more efficiently than a human foot. Then take Electronic Stability Control, because unless you're a trained and practiced racecar driver, you suck at tricky corners at speed, in bad weather conditions. It modulates the brakes and application of power to the wheels to break you out of a skid before you even realize you're in one. My mom's minivan has little alert arrows in the side mirrors to let me know if someone's in my blind spot, or coming up into my blindspot, and Cadillac has a system that uses IR to detect and point out potential road hazards ahead (like a deer in the road) on a heads-up display. Mercedes has a cruise control system that keeps you a safe distance from the car in front of you automatically.

And these are all OLD technologies. Time tested, and in the case of ABS and ESC, NHTSA mandated. The machines have already proven better drivers than people, and that's without them being able to see anything.

Apart from the whole "millions of lives saved" deal, there are a couple of other interesting effects:

- Traffic stops will no longer be routine for anything. If the car isn't registered, or the seatbelts aren't fastened, it won't run. If it isn't inspected, it won't go anywhere but the inspection station. It doesn't matter if the owner is drunk. The car, by definition, won't be breaking any traffic rules, unless the software is hacked (in which case the cops will come down on the road-racers for making their car unsafe. It will be pretty rare.) The car will drop you off in front of your destination, and trundle off to find legal parking, or to just loop the block if it's a short trip. This means the least progressive form of taxation - traffic and parking fines - will be reduced to near zero. This will cause a sharp increase in property taxes, and a reduction of the police force, which was essentially operating as a shakedown operation in terms of traffic enforcement. DWB(Driving While Black - race-based traffic stops) will either be a memory, or instantly identified as racist behavior.

- Traffic jams will be dramatically reduced, and in most instances they will merely be slow-downs, not true stop-and-go jams. Cars have no ego, they know the best way to merge; cars have no prurient interest, they won't rubberneck.

- Fuel consumption will drop through the floor, even if purely electric cars are a way off. With the exception of specialty vehicles, like heavy duty trucks, horsepower stops being a consideration. The driver will no longer be in any way responsible for power and steering, and the car won't be speeding or carving corners along twisty roads. The owner will likely not even be paying the slightest bit of attention to the road, but chatting with the other occupants, or watching a movie. A smooth, comfortable ride is suddenly of paramount consideration, and engine and handling characteristics become irrelevant. Luxury cars will boast of their comfort and efficiency - a full size Audi with only a hundred and fifty horsepower engine pushing 40mpg? Why not?

So, yeah, I'm excited.
posted by Slap*Happy at 6:06 AM on August 21, 2012 [7 favorites]


All I can say, as a visually impaired person is hell yes
posted by Ephelump Jockey at 6:19 AM on August 21, 2012 [2 favorites]


Oh, yeah - and a cabin that doesn't need any controls will be a godsend for the disabled.
posted by Slap*Happy at 6:40 AM on August 21, 2012 [1 favorite]


Take anti-lock brakes. We have them because people suck at driving, even "good" drivers - it can modulate the brakes far more efficiently than a human foot. [...]

All of these examples are assistive technologies - solving one small part of driving, under well-defined conditions. Major decisions are all taken by the driver: If you decide to brake, ABS will brake better than a human. If you decide to turn, ESC will turn better than a human. The major task of parsing the situation on the road, identifying everything around you and deciding what to do next under unconstrained conditions seems like a qualitatively harder problem.
posted by Dr Dracator at 6:44 AM on August 21, 2012


I've seen people in here citing the Google cars as a proof of concept. I just wanted to note the actual numbers we're dealing with, here. The Google cars have autonomously driven 300,000 miles without a (non-human caused) accident. Yes, a lot of those miles are on a closed course; however, a lot are also out in the wild.

There have been two accidents: one, when a human had taken manual control of the car; the other, when a human rear-ended the car.
posted by gilrain at 6:44 AM on August 21, 2012 [3 favorites]


Yeah, I was curious about the conditions of the test courses. I mean the car's only going to be as good as its gear for detecting obstacles. Can anyone point me to more info on the courses themselves?
posted by Ephelump Jockey at 7:11 AM on August 21, 2012


All of these examples are assistive technologies

Except the emergency auto-braking and radar cruise control system that new Mercedes Benzes have, which allows the car to make its own decisions on when to accelerate and when to brake.
posted by Slap*Happy at 7:54 AM on August 21, 2012


The major task of parsing the situation on the road, identifying everything around you and deciding what to do next under unconstrained conditions seems like a qualitatively harder problem.

Listen up. If you are at this point in the thread and have not seen video of Google's car, you do not understand how much of these problems have been solved in real-world situations. I was completely floored when I watched the videos. The cars already navigate live traffic situations at full speeds. They are no longer confined to test courses, and they are extremely good drivers. It is way beyond what you might imagine, if you haven't checked it out lately.
posted by odinsdream at 8:20 AM on August 21, 2012 [2 favorites]


Yeah, I was curious about the conditions of the test courses. I mean the car's only going to be as good as its gear for detecting obstacles. Can anyone point me to more info on the courses themselves?

This is three-part talk shows a bunch of video from real-world tests and discusses the technology in detail. These cars are analyzing surroundings in real-time and making dynamic decisions continuously. They're not following pre-determined paths or navigating predefined obstacle courses.
posted by odinsdream at 8:23 AM on August 21, 2012 [4 favorites]


Listen up. If you are at this point in the thread and have not seen video of Google's car, you do not understand how much of these problems have been solved in real-world situations.

Thank you for instructing me.
posted by Dr Dracator at 8:35 AM on August 21, 2012


I'm sure that after you explained to the rest of us what anyone who has programmed anything more complex than a bubblesort would think, he was just returning the favor.
posted by XMLicious at 9:09 AM on August 21, 2012 [1 favorite]


I truly didn't mean my comment as snark. Before I saw the videos I was in exactly the same place - thought "yeah well it'll be cool *some day* but autonomous vehicles surely can't do X Y Z things yet"... and there Google goes demonstrating things way beyond any of my expectations.
posted by odinsdream at 10:39 AM on August 21, 2012


Eh, guess I read it the wrong way - sorry about that.
posted by Dr Dracator at 12:02 PM on August 21, 2012


Drive autonomously via inductive power on freeways; use conventional battery on local streets. You could travel cross-country on grid power, sleeping/reading/fucking all the way, then pull off the freeway and get to your final destination on battery power, fully charged from the freeway.
posted by LordSludge at 3:40 PM on August 21, 2012


pla writes "One and only one advancement in technology will make electric cars practical (and make no mistake, I would sincerely love them to enter the realm of fiscally sound choices, rather than just another form of 'bling') - Cheaper, higher capacity batteries that can take a near-full charge in under five minutes."

Or gas that is really expensive relative to electricity. $12 a gallon gas would go a long ways to convincing people they only need a car that handles their typical day and that they could rent a car for those handful of days in a year where they don't have typical usage.

sonascope writes "Build me a computer that doesn't crash, even rarely, and I'll trust one to carry me around."

These computers are all around. When was the last time your Microwave crashed?

Homeboy Trouble writes "Self-driving cars have been coming soon, any day now, for decades. So has nuclear fusion. "

And so was LED lighting, DC power transmission, truely flat screens, video telephones, cell phones, computer controlled manual transmissions, satellite navigation, and countless other technologies that we now take for granted.

teraflop writes "most cars don't have any kind of electronic control for steering, braking or shifting gears, so that stuff has to be added too. And it has to be very reliable."

Many cars already have computer control of these functions; pretty well all modern cars can electrically apply brakes and all modern automatic transmissions are already computer controlled. Complete drive by wire is common place on heavy equipment; adapting it to automobiles and reimagining the joysticks as steering wheels, brake pedals and accelerators would be straight forward engineering.

Homeboy Trouble writes "sure computers don't get drunk and don't text, but people don't overheat and short out; people don't mismatch shapes; people don't give 404 errors."

A 404 "error" usually isn't a computer error. It's a notification that the human has asked for something that the computer can't provide.
posted by Mitheral at 8:01 PM on August 21, 2012 [2 favorites]


A 404 "error" usually isn't a computer error. It's a notification that the human has asked for something that the computer can't provide.

Or that a human has misconfigured something. Or that a human forgot to pay for their hosting. Or that a human has made some insidious, hard to detect error in the code of a CMS that only comes up 0.01% of the time. Or that a human didn't bother trying out a bad patch on a test server and has hosed the production machine. Or that some other human, completely oblivious to the first one, decided this web page isn't necessary any more. Or that a human is doing a bit of scheduled preventive maintenance and didn't bother with a temporary fallback. Or that a human with a backhoe has taken out a critical piece of cabling.
posted by Dr Dracator at 12:14 AM on August 22, 2012


LED lighting,

I've been a big ol' fan of LEDs for years now and have built a number of burning man-related art projects based on them. I have recently been retrofitting the local hackerspace with LEDs: not as a special effect, but as simple, ordinary illumination.

You know what's cool?

NOBODY HAS NOTICED.
posted by Mars Saxman at 11:14 PM on August 22, 2012 [5 favorites]


In less than a month, LED bulbs have halved in price... I know, I bought a Phillips 60W-equiv for $50 in late july, and found the smaller, more efficient next gen bulb for $25 when I went back for another. I didn't buy it... because the Home Depot house brand LED bulb was $14.

In other screw-the-power-company news, brand-name 40W dimmable ceiling-fan-size CFL bulbs, ones that don't need a warmup to reach full brightness, are now less than two bux a pop in packs of four.
posted by Slap*Happy at 9:20 PM on August 24, 2012


Makes you wonder if the phase out of incandescent bulbs was the push that was needed.

It's going that way in industrial applications too. Many of the standard exterior wall packs using Metal Halide or High Pressure Sodium bulbs have been delisted and the specified replacements are LED wall packs. Besides being wildly more efficient and longer lasting they are also about a 1/4 of the size. Looks weird when you end up with a mixture of old and new on the same wall.
posted by Mitheral at 2:44 PM on August 26, 2012


Dimmable CFLs with no warmup? I've been mixing CFLs with incandescents to get the instant-on lighting. Warmup-less CFLs sound wonderful.
posted by Pope Guilty at 3:18 PM on August 26, 2012 [1 favorite]


« Older Fan versions of the Game Of Thrones theme are not ...  |  Radio contact ceased. Temperat... Newer »


This thread has been archived and is closed to new comments