Great confidence is being expressed about how robot cars are about to change the world. Robot cars, says a typical headline that Google (one of the prime movers in this new technology) has just today alerted me to, may be coming sooner than you think. But doubts are also being expressed:
A good technology demonstration so wows you with what the product can do that you might forget to ask about what it can’t.
Case in point: Google’s self-driving car. There is a surprisingly long list of the things the car can’t do, like avoid potholes or operate in heavy rain or snow.
Yet a consensus has emerged among many technologists, policymakers, and journalists that Google has essentially solved – or is on the verge of solving – all of the major issues involved with robotic driving.
“Essentially”. That’s a word that often means “not”. And “on the verge of” often signals a problem that turns out to be hideously intractable, as year after year passes with nobody any nearer to a definitive answer. I seem to recall an entire British high speed train project being abandoned because they just could not make the tilting of the carriages work perfectly. It worked okay, but okay wasn’t good enough. It had to be perfect, and perfection proved elusive. Here is what wikipedia says about that, for whatever wikipedia may be worth when reporting a story that remains controversial.
Even that constantly repeated refrain about how robot cars are coming “sooner than you think” is, if you think some more, an acknowledgement from robot car boosters that there are actually widespread doubts out there in the regular, non-techy world about how well these devices really will work, and how completely, above all how quickly, all the problems that they will face have really been and will really be solved. Yes, the techies will eventually get their robot cars working, probably. But for a few more years yet, there will surely be a nasty little clutch both of known unknowns and of unknown unknowns to deal with, all of which will have to be thoroughly dealt with. Crucially, such problems will all have to be solved. If robot cars get the go-ahead and work flawlessly for two months, followed by a lurid catastrophe like something out of a disaster movie, when a bunch of robot cars all follow each other into a swamp or over a cliff, or just run amuck and kill dozens or even hundreds in one catastrophe like in a plane crash, then their introduction will be judged a failure rather than given nine out of ten for technical accomplishment and an A plus for effort.
This strikes me as a lot more immediately promising:
I found that picture here. I first came across the story here:
The latest truck concept from Mercedes-Benz doesn’t look like anything crazy. Its design is a bit unusual, and it’s loaded up with LEDs instead of headlights and cameras instead of side mirrors. But those modest tweaks to conventional design hide the fact that this is a serious bid to revolutionize the trucking industry. That’s because the “Future Truck 2025” drives itself. And while it’s a prototype, Mercedes is serious about spending the next decade getting it – and us – ready for commercial use.
Robot passenger cars will eventually bring huge benefits. They will be epoch making, when the robot car epoch does finally arrive. I truly believe this. But in the shorter run, the problems of robot cars strike me as bigger than all the car and hi-tech companies are implying, and the benefits less immediate. Robot cars will presumably be good at finding their own parking spaces, and at making themselves useful to others if you aren’t using yours. Robot cars will presumably be less prone to error than humans, except when that turns out not to be the case. But what of those potholes?
In the meantime, making lorry (as we Brits call “trucks”) transport only somewhat more efficient will yield huge, very quantifiable, and fairly immediate benefits. Even if all they do to start with is robotise lorries on motorways, that would surely make a huge difference.
The motorway is the natural habitat of the big lorry, and is a place of far greater predictability than roads in general and hence more congenial for robots, especially robots in their early stages. Motorways are already highly controlled places, and are surely the right part of the road system to start introducing robots, not country lanes or city streets filled with complicated and unpredictable hazards.
Human lorry-drivers get tired, but robots don’t. A robot lorry could cross a continent with all the dogged, error-free serenity of a jumbo jet on autopilot crossing the same continent in the air.
At first, humans would need to sit in the lorries to check on their progress all the time, but pretty soon the human could be taking a nap and it wouldn’t matter. Not long after that, once everything has been shown to work, humans would not be needed to sit in lorries on motorways at all. Soon, all that the humans would need to do is collect the lorries (perhaps just the load bit) from their local off-motorway lorry parks, to which the robot lorries had driven themselves. Upon that solid technological foundation, lorries further into the future could then start travelling much faster. (I seem to recall a plan to concrete over railways and turn them into roads. Maybe that notion will be revived.) They could also make their way into the road system generally. The economic impact will surely be colossal, and more immediate than is the case for robot cars.
Before this Mercedes robot lorry story emerged, I had already asked fellow Samizdatista and techno-guru Michael Jennings what he thought of the notion of robot lorries. His main observation, if I remember what he said right, was that large-scale freight transportation in the USA is now a largely railway thing, hence the greater interest in the USA in robot cars rather than robot “trucks”. However, in Europe, where the railways are too full of humans to be able to carry much freight, manually driven lorries are already a very big deal, which means that robot lorries in Europe might make a much bigger difference. So it is no surprise that the robot lorry idea is getting its first big public boost from a famous European motor vehicle company.
But, what do I know? I look forward to reading the comments on this, both from Michael Jennings if he has more to say, and from anyone else who cares to join in. This is the kind of thing that the Samizdata commentariat is especially good at talking about.
Need it be restricted to lorries? If motorway driving is the low hanging fruit of vehicle automation, then surely might also benefit drivers of smaller vehicles?
By way of example; my working month involves multiple five or six hour drives, most of which is on one or more Interstates. If just the Interstate portion of those journeys could be automated, that would redeem many hours which I could use for more productive work, aided by the ubiquity of wireless Internet access along major corridors.
It seems to me that the key element here is the nature of the road, rather than the vehicle.
Don’t hold your breath.
How many of you would really get in a ‘robot car’ for a two hour trip and go to sleep ?
A career with computers and electronics has taught me that NOTHING electrical / electronic beyond the complexity of a light switch is 100% perfect forever ( even light switches wear out ! ).
And the way most vehicles are maintained ( or not ) ………………………………..
Sorry, not with my life.
I’ll believe my fellow programmers can write the code for self driving vehicles when I no longer have to re-start Windows regularly, when QuickBooks stops corrupting my accounting data files and when Adobe Flash stops freezing my browser.
How many downloads a week do we get from our software providers to fix idiot errors in their code? How many of the corrective downloads then need further fixes?
What will you do when your AutoDrive system freezes at 80 mph? Other than die?
I think some people grossly overestimate the risk of hardware or software failure. We know that commercial air travel depends very highly on hardware and software and is simultaneously the safest form of travel by a wide margin. And, when you compare the risk of hardware or software failure to the risk of human driver error that it will replace, it’s bizarre to me that anyone even tries to make an issue of it. It’s engineering-by-allegory: “Windows 95 used to blue screen, therefore all software is inherently unreliable.” That is the 21st century equivalent of: “If God had meant man to fly he would have given us wings.”
I think the last great roadblock to autonomous driving is liability. The simple fact is that even if robot driving is provably better than human driving, that isn’t all that great, and who gets held legally responsible when it goes horribly wrong? Is it the owner? The manufacturer? The programmer? Autonomous mechanicals on public roads make for some convoluted legal theory, and the risk costs associated will be too high until we come up with a common solution.
Yes, I agree that self-driving lorries are a better initial idea than self-driving cars. However, we don’t even have self-driving trains yet, with the exception of the DLR. I’d say it’s worth investigating why this is so, to anticipate the issues self-driving lorries may also raise.
I’m with Tedd in principle, but I’m afraid that air travel cannot be compared to a congested road full of other cars and many other objects. I do presume that when the self-driving cars are finally ready for prime time (which is not quite yet, as Brian said), the roads would have to be modified to better accommodate them.
What I don’t seem to understand is: what makes trucks more suitable for this than passenger cars?
San Francisco’s medium-speed train system, BART (which opened in 1971) was designed to be self-driving (though from a central computer, not one aboard each train), but it was never really attempted. At first this is because the operators’ union objected, but in the first few years of operation there were several serious system failures, which forced the contractor, Bechtel, to go back and overhaul the control system twice. The most spectacular was a train approaching the end of the line in Fremont, which sped up instead of slowing down, went flying off the end of the track at 45 mph and came to rest in the parking lot.
And this, mind you, was a completely grade-separated system, with no road crossings and the entire right-of-way fenced off.
Automated trucks on the roads are a great idea in principle, but even on a freeway they are going to have to be capable of dealing, as safely as a human, with all manner of unexpected hazards, from a deer or cow wandering onto the road, to fools on motorcycles who think it’s funny to swarm around a truck like bees, to coming on the scene of an accident, or group of pedestrians, blocking all or most of the road. Not to mention the traffic cop who is there, directing people safely past the wreck — is the robot going to understand his signals?
Then there’s the likelihood that teenagers or hooligans (or plain old impatient drivers) will think it’s fun to play games with the automated vehicle, by darting out in front of it so that it will be forced to slam on the brakes. I think if it were my vehicle, I would want the authority to write tickets to those people, since the camera on the vehicle will certainly have captured what they did.
Lorry driving on motorways is pretty much roboticised anyway.
Drive up on ramp; put right foot to floor and keep there; if anything ahead going slower than you pull out into next lane and keep foot to floor.
(Some advanced robots have an ‘indicate’ routine, but so far few that I’ve seen have a ‘check mirrors’ function.
You need humans in the car to avoid the lorries, (and to curse and swear).
Dave Walker there are actually many self driving trains. Look up ATO on Wikipedia. There are four on the London underground – all the “driver” does is open and close the doors. One on the Paris metro’ and many others round the world. Reason there are still “drivers” on many of them – unions. Nevertheless I agree much easier than lorries or cars.
Steve Sailer made the point about motorway driving recently on his blog. As “the other rob” (and Sailer) pointed out, it can be applied to cars and lorries. Sailer points out it could encourage suburbanization because people will be able to do email during the computer controlled part of their commute.
I work for a company that makes parts for automotive radar systems. The major car manufacturers think that things will go much slower than Google think. They think that lane change assist and cruise control Radar will only become standard in ten years time.
The software challenges are difficult. The sensor challenges are very difficult too because of reliability. Cars already have many duplicate sensors because of that problem (2 airbag sensors per airbag is standard, for example).
Regarding rail…. The Washington DC metro used automatic driving (and no drivers) for a long time. Two accidents arising from poor maintenance of the magnetic sensors on the tracks led to drivers being put back in the cabs. I think currently automatic operation is used but drivers are still present in case of problems. There are plans to move back to full automatic operation.
I sure would. The primary cause of accidents is human error, not equipment failure.
As a frequent user of the A1, (a road running from London to Edinburgh) doing 400 miles in a day quite often, it has not passed me by that I often pass by the carcasses of blown lorry tyres, sometimes one carcass every ten miles or so. All these ‘blowouts’ result in a lorry grinding to a halt in various states of controllability, and needing a wheel change. Now sorting that sort of thing out is the kind of job that robots like Red Dwarf’s Scutters were invented for, except that they don’t exist yet.
At the end of the day, as Tarrou says, who goes to jail, and who pays, if an automated lorry crashes and does harm? Do I see company directors, partners or sole traders running transport operations forming an orderly queue for any blame?
Those wow-dow fancy fly-by-wire (computer-flown) airliners all have actual human pilots glued to their seats for the duration of every flight, just in case. How silly is that! What could possibly go wrong.
(The Foot of All Knowledge
https://en.wikipedia.org/wiki/Fly-by-wire
allows as how “The main concern with fly-by-wire systems is reliability.” The whole article is both interesting and exciting. 🙂 )
I suppose it’s just anti-technological, Luddite Internet legend about airliners whose passengers were saved by its human pilots when this or that anomaly occurred aboard the aircraft, such as wings falling off. (“WARNING: The Boeing 747 is not certified for use unless at least one wing is present,” or whatever the quote was about whichever airliner it was that kept on going despite the loss of the skin of one wing for 2/3 of its length — if I remember the incident and subsequent in-joke vaguely correctly.) Or do people remember the crash of U.S. Airways Flight 1549 into the Hudson River when it hit a flock of Canada geese right after takeoff from LaGuardia, Jan. 15, 2009? See the fascinatingly detailed article on this at
https://en.wikipedia.org/wiki/US_Airways_Flight_1549
So-called “robotics” of all kinds are great helpers (and I would still like a Roomba); they enable us to do things that would otherwise be impossible; but systems that entirely delete the human control put the horse in charge of the wagon.
Not that there aren’t plenty of horses more sensible than their human masters, if you ask me.
I know a few things about the road haulage industry in the UK, so it may be informative for me to share a few things here.
In terms of routes, drivers and lorries; things are split roughly into 2 areas: ‘multi drop’ and ‘trunking’. ‘Multi drop’ is the delivery to or collection from customers, such as the lorry that delivers to the local supermarket, or pallets of things at a time to a particular premises on an industrial estate.
The large haulage companies operate on a hub-and-spoke system, with the largest having multiple hubs; so if you send a parcel or pallet to the other end of the country then 3 vehicles might be involved. One to collect the item and transport to the local hub, one to transport to the hub nearest the recipient, and one to finally deliver the item. The shuttles between hubs are ‘trunking’, so named as they form the trunks of a particular company’s network.
‘Trunking’ work is particularly coveted by lorry drivers, because it is easy, predictable and requires no manual labour, as unloading is carried out entirely by staff at the hubs. The same factors make it especially suited to automation.
The major hauliers already endeavour to build their hubs as close to a motorway as possible, often linked directly to motorway junctions (for example, here: https://www.google.co.uk/maps/@52.5979424,-1.6247547,5574m/data=!3m1!1e3?hl=en). Particular ‘trunking’ runs thus often spend 95%+ of their time and miles on the motorway network. It is conceivable that fully automated lorries might be approved not only for motorway use, but also on the short links between a particular company’s hub and the motorway. So at least some trunking would be completely automated and driverless.
If this happens, then another possibility is raised. Currently within the EU lorry drivers’ hours are restricted and enforced by tachograph. A driver is limited to 9 hours per day, with a 45 minute break, or 10 hours twice per week, with additional 45 minute breaks between the main 9 hours and the additional hour. So long-distance drivers are compelled to stop for 14-15 hour periods every day. Note that the hubs of any one company tend to be roughly 3-4 hours drive apart, to allow one driver to load, go to another hub, unload/reload and return to the first hub.
If ‘driverless’ lorries are approved, then what is to stop a driver hitching a ride? A single driver and lorry could do multi-drop around the London area (for example), automatically trunk to Glasgow or Rotterdam while the driver has a mandatory daily rest, then do multi-drop again at the other end. Rinse and repeat, ad infinitum. If that happens, then the prized hubs of the big players begin to look obsolete. So while big players would be the key to adopting driverless lorries, the same driverless lorries could be their downfall, without ever needing to work other than on the motorway.
My suspicion is that Google cars are already safer than their non-robot competitors.
I was surprised to hear a few months ago that the government is looking into robot cars. Surprised because governments are rarely keen on new technology and because I don’t think they realise the implications for their precious HS2 etc. With robot cars there would be no need for a speed limit and much more efficient use of road space. It would be curtains for inter-city rail.
One possible benefit of robot freight trucks is more intensive use of highways. I saw this brought up in a discussion of a new intermodal freight terminal for the Port of Los Angeles. The terminal would require an 8-lane highway for the truck side. But with robotized trucks, four lanes would suffice.
Why? Because robot trucks can follow one another much closer. With radio links between the controllers, all the trucks in a “convoy” can instantly copy any acceleration or deceleration by the front truck in unison. They become in effect a train. Sizable intervals for stopping distance aren’t needed.
Robot cars may work fine on the motorway, but in city driving? No way. It is inevitable that people will demand/regulators require that a manual override is possible. Would a robot car overtake a bus parked on the road, overtaking it by driving on the opposite lane? Seems unlikely. Unfortunately, the situations in which the human thinks ‘fuck this, I’m going!’ are the ones in which the override would get activated and also the ones in which human error is most likely to occur. All that said, I can see many areas in which technology could make travelling safer and less stressful, but it would be through empowering the human driver rather than demoting him to passenger.
another go at “centralized control”
the jihadis will have a field day hacking lorrys to use as bombs. allan ackbar!!11!!
Here in Australia, some of our biggest mining companies are already using robot trucks in their operations. In an age of FIFO (Fly In, Fly Out) miners, how soon before most mining equipment is operated remotely, from a central office, with ‘miners’ being drone operators, or the like? I don’t think there’ll be the same enthusiasm when they rescue 33 remote mining robots from a mine collapse, but that might be the news in the future.
> In an age of FIFO (Fly In, Fly Out) miners, how soon before most mining equipment is operated remotely, from a central office, with ‘miners’ being drone operators, or the like? <
probably soon. do you want that control over all of society?
How about robot politicians? Automated tax and spending cuts? Automated repeal of fiat laws? Automated sacking of layers of bureaucrats? Automated expenses claims and bribes?
Of COURSE Google “reports” driverless cars are just around the corner. They have a huge stake in the operations. Just wait till your car registration come with the navigation/monitoring fees attached!
“…then their introduction will be judged a failure rather than given nine out of ten for technical accomplishment and an A plus for effort.”
May I introduce the current US “executive administration” industrial complex, including the attached revolving door of “lateral promotion”?
General Motors (not exactly my favourite company – to put the matter mildly) has been promising these unmanned cars and truck since the 1930s.
Why not get rid of the regulations and the government road subsidies – and have railways back again.
Paul, the railways have declined in importance not due to any government (in)action, but simply because they’re horribly inefficient for the majority of freight movements, especially for distances less than say 300 miles.
The only thing they’re useful for is the trunking I discussed above, and even then only if you’ve got something approaching a trainload of stuff to transport at once, which rarely happens. When it does happen, then trains are actually used. Tesco for example runs a train from somewhere in the south midlands up to Glasgow and Inverness, carrying stock to their stores in that area.
The Tesla model SD which is coming next spring comes with a fully functional autopilot for use on the freeway. The robot car is here already.
To be fair you need to tell it when you want to change lanes, and it doesn’t work off the freeway, but even Arthur C Clarke envisioned robot cars needing to be driven manually away from freeways.
The technical issues are all solvable. The only reason Windows still crashes is economic: people are willing to put up with it, so no-one spends the resources to fix it.
Brian is onto something here, because lorry transportation companies will notice even the most marginal of benefit *immediately*. If the tech can be done, and the politics can be done, I expect this change to be overnight and take a lot of people by surprise.
I do like this idea and I say it should be developed.
but considering what junk we already put up with driving today, I can’t imagine trusting a self driving car. you think it’s bad when the muffler needs to be replaced? just wait until the connector pins rust in your self driving AI hardware.
that’s why I would actually prefer these be used in buses. you can expect some regular maintenance, and they might even put a driver in there for emergencies.
As with all difficult technological problems, we will continue to proceed with automated driving in incremental fashion. While I think robot trucking is inevitable, the conditions we are expecting robot trucks to deal with, such as city driving, or even highway usage, seems to me to be so difficult as to make them unfeasible. I think it is more likely that in order to have autonomous trucks in the mainstream, we will modify the roads themselves to make the problems with autonomy much more simple. ‘Robot only’ lanes, embedded guide wires (so weather becomes a non-issue), and centralized human monitoring will all contribute to moving stuff increasingly autonomously.
We have cruise ships that can drive themselves from a Florida coastal port to the Bahamas, while a human sits in the control room and reads a book. Granted that the ocean contains a whole lot of nothing to run into, but, it isnt completely empty, and imagine the liability issues that are being dealt with in that case, today.
The English Electric Wren was an aeroplane developed in the 1920s when it might not have been clear if cars or aircraft would be the transport of the future. There is one left, it has a top speed of around 50mph and provided that you have a small pilot, it can fly around at hedge height. It took some tuning up of the engine and a new propellor to get it that far, but with a bit of weight-reduction and/or a better engine, it might have made the motor car seem earthbound and rather pointless for all bar the shortest journeys.
It was only 3 decades before the same company had made the Canberra and the astounding Lightning.
It seems to me that Darryl is right. Dedicated “robot only” lanes will make highways more like railroads (i.e., extremely limited access), and embedded sensor systems will keep the lorries on track with minimal computer intelligence involved (so the software becomes a minor issue, and redundant systems cheap and ubiquitous). The occasional deer on the highway would be a problem, so they’ll probably have to put very high fences around those lanes. As Friday Night Smoke says, it will probably be limited to a “trunking” system, with human drivers handling the local non-highway portions at either end of the trip. Whether this can be made economically competitive with railroads, especially for very long trips (as in the US) remains to be seen. I suspect that if it can, it will only be because of the relative ease of changing main routes: converting one lane of an existing highway into “robot use” will be much cheaper and easier than building a new rail line.
I do not want to be driving down a two-lane road with my family and see one of those goddam things coming at me almost-headlong. The answer is no, and it will stay no.
Robot lorries? Then who’s gonna murder the hookers and hitch hikers?!
Robot tractors are the norm for broad acre farming now.
As (perhaps) the most conservative commenter on this here website, I need to make a couple of points against all this glowing talk of automated cars.
I don’t trust automation to do anything other than the most repetitive mechanical tasks — such as manufacturing. But throw in the slightest deviation, and automation starts cracking up quickly. If this were not the case, all space travel and exploration would be done exclusively by robotic craft and systems — but it isn’t. Now multiply that seemingly simple task by the billions of computations and permutations of vehicular road travel on a series of busy motorways… and I don’t trust computer systems to fulfill those activities seamlessly and a MTBF of infinity.
Here’s my take: automated cars answers a question which hasn’t been asked (other than “Can we do it? Cool!”) and it addresses a need which doesn’t require a solution. What this kind of automation does is eliminate all vestiges of personal choice and variation, and hands control thereof to an un-elected entity.
And here’s my final comment. I don’t trust Google. I don’t trust them with my data, I don’t trust them with my communications, and I don’t trust them to have my interests at heart any more than I would trust CitiBank to have my personal financial interests at heart when they make the same kind of corporate decisions. This is not a personal dig at Google; I would have the same reservations about a bank which controlled the global financial industry to the same extent that Google controls data and information. (And I think that enough has transpired that “Don’t be evil” is no more than a panacea for fools. Google’s not evil, per se; but in a choice between your interests and, say, advertiser Ford Motor Company’s, don’t expect them to spend much time thinking about you, for all the right business reasons.) I have the same reservations about Facebook and other social media, by the way, which is why I don’t have one of their accounts.
But (he said, wrenching the comment back on topic), the only reason I can think in favor of automated cars is laziness or, by a stretch, increased productivity (by working instead of driving), the drive for the latter being far stronger now than ever dreamed by, say, Henry Ford.
I don’t have a problem with people wanting to use automated cars. I do have a problem with the outcome of this technology, which is that at some point we will all be compelled to use automated cars, with which “remote control” is inevitably and irrevocably linked.
Screw that.
Oh, and by the way, we already have automated goods transportation. It’s called “railways” and has been around for a while.
Regarding railways, I refer you to Friday Night Smoke’s very useful comment.
Actually people are really superfluous unless they are there for fun or as payload (as opposed to pilot). People just get in the way. Not one of the moon landing actually *needed* to have the spam-in-a-can, even back then. And what do you think the crew of Cassini is?
Speaking as an aviation enthusiast, the main reason aircraft crash is because of human error. For every Chesley Sullenberger, there are a great many more CFIT incidents. Ideally the pilots should be there as a failsafe for the robot, not the other way around.
I think self-driving cars will be self-driving options in areas suited to them, as there is very little reason to have a driver on a motorway, rather more reason in a city, unless the city is purpose built for driver-less cars to keep pain in the arse pedestrians out of the way.
Kim:
No, it doesn’t. Automated systems are much better than human beings at detecting and responding to anomalies. In the past, one drawback of automated systems was responding to anomalies that had not been thought of and built into the programming in advance. But even that is no longer a limitation because programming methods have been developed that are based on achieving objectives, rather than simple input-output functions.
What’s important to note about that video is not that the autopilot recovered and safely landed the aircraft. A good human pilot could do that. What’s significant is that the autopilot was not programmed to fly the aircraft on one wing, it was programmed to achieve stability and mission-success parameters, and to create new flying techniques, as required, to do so.
It’s understandable that lay people are not aware of these kinds of capabilities. But they exist, and they are coming to a car you ride in very soon.
But it is. Because of the signal lag, all space missions outside of the Earth-Moon system are essentially completely automated. That’s where a lot of this technology started.
Automated cars would be nice because they would combine the ease of travel by train with the flexibility of travel by car. I would love to be able to read on a long road trip rather than having to drive.
Also I think it might be possible to save fuel once the system was sophisticated enough to allow the robots to tailgate each other and thus cut down on wind resistance. Likewise properly programmed cars should be able to maximize the flow of traffic and thus cut down on delays. However, it’s much easier to imagine all this working on a road with only robotic cars than with a mixture of human and computer drivers.
Tim, I am sure that robots can be programmed for these tasks! However, in the future, robots may also be the hitch-hikers, looking for seasonal work. The one aspect of ‘Real Steel’ which I found disappointing was that there weren’t more robots, even if only as chauffeurs.
I somehow got smited and can’t be bothered to reproduce my original post. But here are a few additional points.
Advanced control systems don’t work at all they way most people think they do. Many critiques of autonomous vehicles contain the implicit assumption that somebody has to think of every scenario the vehicle could encounter in advance (including systems failures) and plan for it. But that’s not how it works at all. Autonomous vehicles are programmed to achieve real-time control objectives and longer-term mission objectives, and are capable of innovating, as required, to do so. These systems are as capable at meeting unexpected challenges as highly skilled drivers or pilots are. But, unlike drivers or pilots, that skill can be perfectly reproduced in every system, and it can function at its best performance level 24/7.
Don’t get me wrong, there are still big challenges to overcome with autonomous vehicles. They just aren’t the challenges most people at first think of.
I agree with Tarrou that liability is a big issue. (Though not the biggest. Over regulation will almost certainly be the biggest limitation to successful implementation of autonomous vehicles.) And I’m a bit surprised that there’s even much to debate about liability, frankly. The passengers are not in control, so I can’t see any way that they can be deemed liable. It’s rare that the owner of the roadway is found liable, and I see little reason for that to change. So the vehicles manufacturers must, by default, assume essentially all the liability risk.
But, when you look at it from their point of view, that’s not such a big problem. It already costs auto manufactures millions in liability for things that are not even their fault (such as “unintended acceleration”). That just goes with the territory. With autonomous vehicles, at least they have control over the system they’re held legally responsible for.
Okay, okay. I’m sure it will all work out exactly as planned: Big Whatever is going to do a great job of managing your travel, accidents will be a thing of the past, and the efficiencies will be remarkable.
I’ll either be driving my old stick shift truck, or dead. But I won’t be in an automated vehicle, I promise you.
This is the only real worry I share with Kim. Other than that, what Tedd said.
Screw that.
Kim, the robots will have the last laugh. Your hearse will be fully automated.
Nick, the problem will be how to pry the still-smoking AK-47 from my fingers, in order to fit me in a coffin. Think they’ll be able to automate that process…?
Yes, they will, Kim, because robots will be doing everything that might have any danger in it. I envisage a future where robots forbid fallable humans from practicing surgery, or treating with infected people, or even being allowed to do any cooking, all in the cause of saving lives. The end result will be that we’ll all be cryogenickized for our own safety, and our still-living bodies preserved in deep, cold, space, forever.
Wanna shoot some robots sometime?
Kim, oh yes they will. Have you never heard of “mechanically-recovered meat”?
The liability issue should be straightforward, whoever tells the vehicle where to go has a duty of care to his ‘neighbours’ i.e. anyone in reasonably foreseeable harm’s way, which includes the occupants of the vehicle. This is distinct from discerning who told the vehicle where to go, presumably it would be the registered keeper unless he showed otherwise.
@ Mr Ed – ah, a man after my own heart. Indeed, the EE Wren won several prizes at the 1923 Daily Mail Motor-Glider contest at Lympne. Pretty amazing for the time.
But not realistic in any sense.
As you note, it can barely fly now, being grievously under-powered and suffering from certain undesirable handling characteristics.
It might be well to note that Shuttleworth also houses the remains of another design that did quite well in the 1923 contest, the de Havilland DH53 ‘Humming Bird’ G-EBHX named ‘L’Oiseau Mouche’. This aircraft crashed in 2012 and killed the pilot, a vastly-experienced commercial pilot and former Chief Pilot at Old Warden, who was also one of only 2 persons on earth certified on type. Despite having been up-engined after the 1923 competition and holding a current COA, this aircraft suffered some of the same problems as the EE Wren – under-powered, and subject to several challenging handling peculiarities. If it can trap and kill a pilot of that calibre, it is in no sense a realistic design for general use.
There have been a billion attempts since to create an aircraft comparable to a private car. None have been any more successful.
As it happens, I will be at Old Warden in 3 weeks and I will smile at the EE Wren for you. Unfortunately, they won’t let me visit the DH53, understandable, I suppose.
llater,
llamas
Kim:
I genuinely believe that you will ride in autonomous vehicles, you’ll do it regularly, and you’ll probably forget why you once thought you wouldn’t. Unless, of course, you don’t live another 20 years or so, in which case you might be right. It’s just not going to be that hard to accept. But it is going to be a huge revolution in transportation, every bit as significant as the invention of the automobile in the first place.
Tedd:
I don’t even care to shoot automatic (or semi-automatic) guns anymore. Nope, with the exception of my AK-47, all I have around the place are revolvers, bolt-action rifles and double-barrel shotguns. (The AK gets a pass because it’s the only weapon to have around when civilization crumbles, and I’ve been hearing the apocalyptic hoofbeats for some time now.)
True, I drive an SUV with an automatic gearbox, but that’s because The Mrs. can’t drive a stick shift anymore and we only have the one car; but I prefer to drive The Daughter’s manual Fiat 500 whenever I can. With a stick shift, it’s driving; with an automatic transmission, it’s steering.
So given all that: what makes you think that I’d ever choose, or acquiesce to give up the steering too, and be driven around in a little computer car controlled by Google?
I find it highly ironic that I, a (very-) small government conservative, find myself wanting to preserve a teeny element of personal freedom; whereas all these libertarians around here seem only too willing to give it up because automation is cool, or something.
The problem with automation is that as cool and efficient it may be, at some point it ends up being controlled by a large entity — inevitably and always — and that large entity will use it to control individuals in some way. Feel free to come up with all sorts of clever rebuttals; but in the end, I’ll be proved right, even though I don’t want to be.
Kim, you are probably correct about some “large entity” ultimately controlling the technology. But that won’t stop it. I think Tedd is right; autonomous vehicles are inevitable, at least to some extent (i.e., highway-only driving). For most people the convenience will far outweigh whatever minimal loss of personal autonomy that may entail. (YMMV, as the saying goes!)
I am perfectly capable of driving a stick shift, too, but generally choose not to do so. Most of the time I simply want to get somewhere, not “drive”; mere “steering” is perfectly fine with me. And best of all is leaving that “steering” to someone else entirely, when possible. At present that’s my wife. If, in the future, it’s a self-driving car we’ll probably both be happier.
Robot drivers will be continually improved. Human drivers will not. Ergo, sooner or later robot drivers will be safer than human drivers, as a general rule.
Occasional catastrophes caused by robot drivers, and occasional examples of massive lifesaving caused by the exceptional skill and heroism of particular human drivers/pilots do not change the validity of the fact that, as a general rule, robot drivers will be safer one day. Perhaps this day will come very soon (next 5-10 years), perhaps only moderately soon (in my lifetime), but the day is coming.
However, occasional spectacular mishaps with robots drivers could seriously inhibit take-up not only in the market, but due to the conservatism of regulators, licensing, resulting in robot drivers being unavailable on safety grounds even long past the point when they are safer on average. This would be sub-optimal.
Having read this story, it appears that robot lorries offer a potential solution to this conundrum, and if so that may be the more significant thing about it in the long run. Robot lorries claim to offer not just better consumer convenience (the major claim of robot cars) but large cash savings in the logistics market (not just fewer drivers; trucks on the road 24/7 instead of 9 hrs/day implies much greater efficiency of capital); given that the logistics market is core economic infrastructure such efficiency savings will drive broader economic growth. Even conservative regulators may be willing to look past occasional mishaps to enable those economic gains, once they can produce evidence that safety is statistically improved as well. This would get robot trucks out on the road – and that will lead to the real-world experience that will end teething troubles in software. A combination of ever-safer robot trucks and a sense that robots are no longer new and scary will eventually make it possible to introduce robot cars. “well there have been robot trucks on the roads for years; why worry about a robot hatchback if you’re not scared of a robot HGV bearing down you?”.
I don’t even have a car any more, I just use trains and buses, but I might buy one again if I don’t have to drive the damn thing (I view driving as a complete waste of my time, I prefer to be a passenger so I can do something I find interesting or useful, like watch the scenery or leave rude comments on the Guardian or phone a friend or whatever). A self driving car is just a cheap chauffeur as far as I am concerned. A great many trends worry me, but this ain’t one of them 😀
I don’t see the future of driverless cars coming any time soon, mainly for reasons of cost. What it comes down to is the fact that reliability is expensive, and the levels of reliability required for driverless cars would be along the lines of those seen in the airline industry. Expensively reliable systems are cost effective for mass transport machines like aeroplanes because the same system can transport several hundred people at the same time, but to put such a system into a machine for 4 people? Something would need to change.
I know quite a lot about reliability. In the design of an oil and gas installation there are many, many automated actions controlling the process and shutting down the plant in the event of an emergency. The devices which govern the actions (which includes the entire “loop” from initiating the action to executing it) are rated according to what is known as a Safety Integrity Level, or SIL. A SIL-1 system will fail every 100 times it is asked to do something; SIL-2 every 1,000 times; SIL-3 every 10,000 times, SIL-4 every 100,000 times. For non-critical actions you use a SIL-1 or -2, but for anything which will shut down your plant in the event of a gas leak or fire you will need a SIL-4. The reason we don’t install SIL-4 on everything is because a SIL-3 or -4 system is staggeringly expensive compared to a SIL-1 or -2. As I said, the whole “loop” needs to be rated: your gas detectors, your Process Logic Controller (PLC), your valve actuators all need to be of the same rating.
Then you have redundancy: anything which can shut your plant down should do so on the basis of 2 out of 3 voting. So you put up 3 gas detectors (all of which are SIL-4) and the PLC will only initiate the action if no less than 2 gas detectors say the same thing. This prevents false positives, or spurious trips as they are known in the industry. I can tell you from experience that the SIL levels and redundancy levels required to achieve the reliability seen in aircraft, the nuclear industry and (snigger) the oil industry (stop laughing at the back!) is seriously expensive, and it is not simply a case of the price of electronics falling and the Chinese will take care of it all. This kit is high-end stuff, and rigorously tested, plus it needs to be maintained properly.
I’m not saying that SIL-4 levels of reliability will never be cheap, just that we are one hell of a long way from it being so now. And I would like to see the cost of a production car with SIL-4 levels of reliability if it were made now. I’d be guessing somewhere between $500-$1m.
$500k-$1m, obviously.
Stick-shift vs automatic is a great example of where well-done automation will lead us.
I, too, love the delicate ballet of clutch and gas, nowhere more so than on the mighty Ducati. Yes, indeed, llamas got laid off from his previous employment, and spent his severance check on what he’s been promising himself since he was a cria – the biggest Desmo Ducati that money can buy. If you like a manual transmission, no place better to like it than on a very powerful motorcycle.
Same in an old-school semi with totally-manual, multi-range transmission, where you have to double-declutch every gear, up and down. There’s a particular joy in running all the way up and all the way down all 12 gears and never missing a shift. I totally grok this.
And yet . . . . the ultimate expression of the manual transmission is the electronically-shifted manual transmission, where automation will get you a gear shift faster than any human with a mere hand lever can possibly attain, with gear speeds matched down to the tooth level, no clutch wear, and never a missed shift. Ever. There’s no denying how much more effective this kind of technology is than any human can ever possibly be – never mind the average human.
Concerns about the possible failures of automated vehicles overlook the basic fact that the vast majority of vehicle accidents are caused by human error. Eliminate those errors and the accidents due to failures of automation that remain will appear miniscule by comparison. The computer may not enjoy the bark of a well-made down-shift, this is true – but the computer will also not succumb to the human failings which cause the great majority of accidents. If I have to watch someone putting on her mascara at 70 mph on US-23, I’d much sooner she be in a Google driverless car while she’s doing it.
llater,
llamas
If driverless trucks become the norm, I would guess new jobs would be available for security guards. I wonder what defenses they would have to prevent high jacking and subsequent looting. There are some very remote places on the interstate highways in the US.
I’m surprised that Google hasn’t solved this one, given that the mining equipment companies have solved it for autonomous dump trucks.