Will Our Cars Become Our Chauffeurs? 792
Roland Piquepaille writes "According to this long article from EE Times about the 'Self-Navigating Vehicle,' the answer is a resounding yes. Many car experts think that autonomous vehicles which avoid collisions and communicate wirelessly with other cars will be the norm in two to three decades. In the meantime, the enabling technologies for self-navigating cars are emerging, from sensors embedded in the brake or accelerator pedals to more powerful computers. Already, partial solutions exist for adaptive cruise control or for staying in a highway lane. One day, we'll be able to do something else than driving our cars through traffic jams, saving us about two hours per working day. This is the future that engineers are building, but will you accept to be driven by your car? So many people like driving that the concept of a completely autonomous car might be delayed for psychological reasons, not technical ones. This summary contains selected details of the original article."
I don't think I could ever trust it (Score:3, Insightful)
Amazing technological breakthrough (Score:5, Insightful)
Public transport, this is America.
Have a nice day.
Benefit Number One (Score:4, Insightful)
That would shave lots of time right there.
Not just nice, ESSENTIAL (Score:5, Insightful)
Having automated transport systems removing the human (idiot) factor will be essential to prevent utter gridlock in the future. The only other alternative is to stop immigrating people faster than we can expand the infrastructure they use. Yes this ultimately is the problem - highway construction cannot keep pace with US population growth.
No time soon, methinks. (Score:4, Insightful)
In the USA, the risk of lawsuits will surely delay this kind of thing for a long time to come.
Sadly, that will probably mean more people get hurt in the long run.
Mass Transit (Score:1, Insightful)
I drive my car to get away from being driven around.
What about the legalities? (Score:5, Insightful)
Furthermore, you need black boxes and monitoring/recording systems - how do you know who was driving in an accident, the autopilot or the human driver?
Sure, planes have "autopilots" but there's very little stuff in the air to avoid, and lots of air traffic controllers and rules to basically make flying in a straight line in your own empty area of airspace possible.
Technical and psychological issues aside (and those issues are still huge), unless the system was flawless and perfect (which it won't be) I see the legal morass here as nearly insurmountable.
Re:I don't think I could ever trust it (Score:3, Insightful)
I think if properly tested, computerized vehicles would make far better driving decisions than a lot of people I know.
Saving 2 hours? I don't think so. (Score:3, Insightful)
Re:But how deep? (Score:5, Insightful)
Although this argument never held much water with me. Consider all the tired drivers, drunk drivers, old people, teenagers, and in general crappy drivers on the roads. There's like, what, 60,000 deaths a year due to car crashes, and that's nearly all human error. Can't imagine computers doing worse job than we're doing already.
Re:I don't think I could ever trust it (Score:3, Insightful)
Re:Amazing technological breakthrough (Score:5, Insightful)
Re:I don't think I could ever trust it (Score:4, Insightful)
For much of the flight, a computer is controlling the aircraft with the pilot and copilot only monitoring it.
I'd think if computers were safe enough to work in three dimensions controlling vehicles with a multitude of control surfaces, in two dimensions with only gas, brake, and steering, they'd be at least safer than most drivers on the roads today.
Benefit Number Two (Score:2, Insightful)
Switchable (Score:5, Insightful)
I'd imagine that the first fully automated cars will be airport shuttles and similar vehicles which make a repeated circuit of stops. City buses and taxi cabs will come next, other commercial vehicles such as delivery vans and trucks, then finally personal automobiles. How much would a long haul semi-truck operation save if they could run their trucks 24/7 and didn't have to pay for drivers? That's a lot of profit to be had and profit drives innovation.
Re:it's called the bus (Score:2, Insightful)
And it limits your cargo-carrying capacity, ignores your schedule, subjects you to a bunch of wack jobs who can't afford any other kind of transportation, who may or may not be carrying a bunch of communicable diseases. Don't sit in the very front or the very back; the elderly sit in front and the mentally handicapped sit all over. Lots of those people have hepatitis and shit like that, because they are not equipped for the real world and they spend a lot of time going in and out of mental health organizations which are generally filled with very clean individuals. I say all this as a Santa Cruz native who used to work at County Health there, first as a MIS employee and later as a security guard.
Paranoid? Sure. But there's just a shitload of reasons why public transportation in most American cities is a joke. Santa Cruz's bus system is pretty good, there are a couple of buses that run until midnight and one (that goes to the university) that runs all night, and few buses run less often than once per hour, with the most popular lines running once per 15 minutes or per half hour. The sad truth is that buses are not cost-effective in most places and trains are useless without buses, so basically any non-major city will have a useless public transportation system, if any.
Re:I don't think I could ever trust it (Score:4, Insightful)
Re:Not just nice, ESSENTIAL (Score:3, Insightful)
And we won't even have to shoot illegal immigrants as they try to cross the border.
It will never happen (Score:4, Insightful)
Re:I don't think I could ever trust it (Score:2, Insightful)
Humans shouldn't control vehicles (Score:4, Insightful)
Re:But how deep are their pockets? (Score:5, Insightful)
You'll see it as HOV/Toll lanes first ... (Score:4, Insightful)
By breaking them out of the normal traffic situations the navigation computers will be able to avoid having to deal with the random actions of normal drivers and be easier to trust during the roll-out. Once you get into the city autopilot will go off and you'll be asked to start driving. Over time when the system is perfected and the market is more fully penetrated you'll see autopilot everywhere, but it will probably start on dedicated for pay lanes first.
My $0.02
Re:This would be great (Score:4, Insightful)
Person A is driving a safe distance from the car in front of him, person B is certianly more important that A so he pulls into the space in front of A causing A to slow down. CDE are all only 3-6 feet from A sothey JAM on their breaks because they can not simply slow down but must now PANIC stop in order to not hit the car in front of them.
THAT is the cause of traffic jams, espically the ones where there really is no visible cause.
In otherwords, very poor driving.
Re:Amazing technological breakthrough (Score:3, Insightful)
Why is the parent flamebait? Mods on pot? (Score:3, Insightful)
Being driven in my own car sounds like the pefect solution since most gridlock is actually caused by bad driving. Driving too close has been proven to cause traffic jams due to the wave effect (can't remember what its called in this situation) as people have to break to a stop rather than simply slowing down gradually. And the other big factor is the idiots who have to cut in too late or avoid moving out of closed lanes until the last minute.
Stick everyone in self driving cars which follow logical rules, drive the right distance apart can be updated of problems ahead and mostly aren't operated by the average selfish driver, and everything will flow much smoother. And then, like the parent said, we can all get drunk, smoke pot and still drive home.
The only potential problem I see is that once you take the boy racing syndrome out of driving, everyone would want gas-heavy RVs so they could lie back and have a snooze on the way home.
Re:I don't think I could ever trust it (Score:4, Insightful)
There are hundreds of activities you do to drive your car, it's a complex machine that has thousands of parts that have to work. Its maintenance is up to the user, and not carefully controlled and checked by the FAA.
Autonomy in software is EXTREMELY hard to test. Every combination of action, fault, and surroundings has to have an experiment to show the software works. This software will need to deal with every possible reality that can exist on the US freeways, city and town roads.
This software can't be fully tested in a lab, either, since in a lab you can only test what you can think of. Real life causes problems that nobody ever anticipates. If you can't anticipate it, you certainly can't expect a programmer to plan for that eventuality.
This problem is FAR more complex than people realize and will take time to solve. Even then, if a majority of people don't trust it, it will not come to be--since it will increase the price of vehicles, it will take legislation to make it happen, and that takes at least a majority vote.
Re:Saving 2 hours? I don't think so. (Score:3, Insightful)
Here in Europe a huge proportion of people commute on the train, often for more than 2 hours a day.
Usually a resonable arrangement is made with your employer. Many people really enjoy the quite time to get work done balanced with meetings in the middle of the day.
Fear (Score:2, Insightful)
However, _some_ accidents probably will happen with autonomous cars. We've all seen or had our systems crash every so often - glitches occur in the best designed systems.
The problem is that the media is likely be very vocal about these ('Robots Cause Twenty-Car Pileup, Many Dead' - or some such). And this will scare the heck out of people. People don't mind taking their life into their own hands - driving yourself you at least nominally have some control over the system. But putting your life into the hands of
It's not the fear of death so much as the fear of dying and not being able to do anything about it. That's scary.
The core problem (Score:5, Insightful)
This will spell the end of car ownership (Score:5, Insightful)
Deer in mid-air (Score:2, Insightful)
Re:But how deep? (Score:3, Insightful)
When the light turns green (Score:2, Insightful)
At speed zéro, it is OK to have your car at a very close distance from the one before you.
At 50km/h, it's dangerous.
At 100km/h, you must keep quite a big distance.
Then the queue of idle cars waiting for the light to turn green must be seen as a rubber band that is going to take expansion as speed increases.
AI in cars won't eliminate risks when cars are close to each other at high speed.
Re:But how deep? (Score:4, Insightful)
I saw an intersting Open University TV program about this issue a while back. Over 60% of the code was to deal with exceptions that happen less than 1% of the time.
Their major stumbling block? Anything their software couldn't cope with, there was no point handing control back to the human, because they wouldn't be able to react fast enough either.
The sight of 20 strech limos moving in absolute (down to the fraction of an inch) synch was very impressive... a bit un-nerving, but very impressive.
I think the problems facing robot cars are more to do with psychology than engineering. Look at how much fuss is raised over a train crash that kills people "not in control of the vehicle" therefore innocent compared to the number of people who die in car wrecks "in control" therefore less innocent.
I realise this issue is conflated with the number of deaths in an instant too, but i think one of the key "shock" factors is the helplessness of the passangers
Re:Amazing technological breakthrough (Score:3, Insightful)
One current problem is that people want to LIVE in rural areas even if their jobs are URBAN, and this is a selfish position. I live near where I work and feel my quality of life is better without the traffic congestion, and so have traded the bigger home for the ability to walk to a bus stop.
However, I am tired of the heavy traffic around my neighborhood as commuters race down our side streets trying to get to their suburban homes faster. They don't realize that their rural lifestyle rides on the backs of urban residents. And I doubt that automated vehicles will enforce traffic regulations any more than existing vehicles have speed governors.
Besides, if traffic improves as this post suggests, you'll simply find more people willing to live farther from where they work and congestion will increase: it's a self-regulating mechanism and urban sprawl will continue.
I trust it more than I trust you. (Score:4, Insightful)
Computer sensors could (in theory) operate in darkenss, fog, snow, or rain far better than a human could.
Considering that driving is usually a fairly mechanical activity, I think that this would be a good thing to automate. Plus, a coumputer could be programmed to drive in a more fuel efficient fashion. It could moniter traffic situatons and rout around them. Because it doesn't drive eratically, drive times become more predictable. As more cars become automated, driving becomes safer for everyone. This stupid weight escalation shit of buying an SUV becasue it is 'safer' can end.
There will always be some people that like driving a car. There are people that still enjoy knitting, even though there is no real need to make your own sweaters anymore. For most though, I think that a car is a source of freedom to go anywhere they want, and not so much a pleasure to drive. For those people, it wouldn't matter who drove, just that they got where they wanted to go.
Re:I don't think I could ever trust it (Score:2, Insightful)
This impression of the air traffic control system is incredibly oversimplified, and mostly backwards. The actual implementation involves a large number of parallel tracks at various altitude levels while air traffic control, which is run by very trained people in tandem with some very good computers, routes planes on these "roads". In fact, the margin of safety required in three dimensions is HIGHER than in two -- the likelihood of any two objects in random trajectories in 3-space colliding is tiny by comparison to the same situation in 2-space. If you've ever flown from LA to SF, you might notice the parallel sets of contrails where the last few planes have gone just a few hours previous. It's neat to see.
Anyway:
The airway system is unbelievably simple relative to the incredible variety of complicated situations on streets. There is only one kind of sky, and it's full of air. Excepting thunderstorms, you can drive a big jet straight through pretty much any part of the sky, and all the references needed to guide it can be computed internally (like pitch/roll/yaw) or from simple external means (like GPS). A vehicle must contend with terrain, markings and the lack thereof, pedestrians, ever-changing road geometries and new construction, and myriad other complications.
Also -- auto-pilot is not a complete solution. Yes, we've gotten good enough at control systems to keep a very complicated plane flying straight and level, and avionics assist in every aspect of commercial airline flight (can't steer a jet without hydraulics anyhow), but dynamic situational analysis is still the province of the mind rather than the computer.
All of this ignores the difference in volume, too. There are thousands of planes in the air most of the time, virtually all of which have fairly well-planned routes and destinations. There are tens of millions of cars, whose destinations often change, and which don't plan their schedules weeks or months in advance.
Computerized driving (esp. in a world where people are still allowed to drive, too) is a MUCH harder problem than air traffic control.
Re:Amazing technological breakthrough (Score:3, Insightful)
If you want services, more to a populated area that has them. And don't bitch about your taxes.
Re:I don't think I could ever trust it (Score:3, Insightful)
IMO, the solution is to make cars more like elevators. Instead of trying to put them on the current free-form road system. Put them on something more like a model railroad set.
I imagine cars about the size of golf carts that could run at high speeds on the constrained tracks, and at low speeds on regular roads to account for the fact that you can't build tracks to every single building.
This would have many advantages. First is the obvious time savings because the driver doesn't have to pay attention for most of the trip. Much more time is saved by central router pre-planning all traffic, totally eliminating slowdowns. (If there's too much demand for the system to handle, you'd be told to chill out for a while before the trip even starts.)
There would be huge safety gains, because the tracks could be built with no grade-level crossings whatsoever. In the constrained environment, I think that a 10X to 100X improvement in safety (similar to airplanes per mile) would be doable. Precise central scheduling would eliminate most needs for local traffic decision making, and the cars would only need to have local backup systems (based on radars, cameras and/or wireless P2P with other cars) that are designed only to avoid collisions due to errors.
The tracks could have mechanical features to support a "tail hook" kind of feature on the cars. In an emergency, the car could use the tail hook to stop at something like 10G deceleration, so it could come from 60 mph to a dead stop in about 12 feet without harming the (seatbelted) passengers.
Cargo capacity would be infinite by having robot cars that are programmed to follow you like a trailer. You could add any number of these to make your own train. You would summon them at places like a lumber yard, then when you're done, you send them back on their way to the next job.
Energy savings would be huge, because people wouldn't need to drive around with excess cargo capacity at all times. 99% of the time, the golf-cart passenger compartment would be sufficient, so the total mass being driven around would drop by a huge factor. Moreover, they would probably be electric, with small batteries for driving off-track. Centrally-generated electricity would power the cars and charge their batteries while on the tracks.
The tracks themselves could be prefabricated and put together like model railroad tracks, allowing huge flexibility in transportation. They might even be temporarilly set up for special one-time events, and then taken down again. I think that each track might be able to carry the equivalent of two of today's freeway lanes because of central scheduling. The footprint could be very small, like 6 feet wide by 5 feet tall per track, saving massive amounts of real estate. In cities, most tracks would be elevated or buried to reduce congestion.
I don't think the tracks would necessarily have actual rails; the cars might just automatically steer to stay in the center of the track. This would allow for flexibilty in design of track intersections. T and cross intersections could be used in low-traffic segments, and the central scheduling would eliminate the need for stop lights or similar local controls. (There would probably actually be stop lights, but they would cycle instantly on a per-car basis. They would only be actually used for emergency collision detection.)
A system like this seems pretty radical, but it could be initially tested out as a cargo-only system to replace trains and/or trucks in a limited area. Once the kinks are worked out there, it's use and scope could be expanded.
Re:Switchable (Score:3, Insightful)
Re:it's called the bus (Score:3, Insightful)
I say all this as a Santa Cruz native who used to work at County Health there, first as a MIS employee and later as a security guard.
Were you downsized or what?
You had me with this part, "And it limits your cargo-carrying capacity, ignores your schedule..." and then it all went psycho after that.
You're right about service in smaller areas being bad/nonexistent though.
Cheers!
It's not that I don't trust my car... (Score:4, Insightful)
On a rural road, I could easily imagine thugs with a computer emitting signals that fake a deer-sighting or accident-ahead event, causing you to pull over and slow down. You are then easy prey to carjacking or simple robbery.
This is similar to spam and envelope/header forgery. For a long time, email software trusted everything that was said in the SMTP transaction and the email header. We're still dealing with that today, slowly adding features to try to limit email's exploitability.
Since car navigation presumably affects the passengers' lives, you can't simply add wireless warning protocols to the navigation computer without thinking seriously about how much it should trust those signals.
Not First in the USA (Score:2, Insightful)
Re:Share the road with automated 18-wheelers? (Score:3, Insightful)
I love it when similar problem manifest in an sun-regularr "I" control system's sensors on humans. Sleeping and drinking at the wheel, talking on cell phones, badly misjudging the relative speed of convergin objects, administering punishment to 1/2 clones. Arrrrh! I freaks the shit out of me every time I look at a car. 2000 lbs POORLY guided bombs.
Re:I don't think I could ever trust it (Score:3, Insightful)
Planes and Trains which are massively computer controlled and require humans for monitor duty only, are monitored by professionals who have been trained for the task, and who don't have to deal with screaming kids in the back seat.
Your average driver? NOT a professional. NOT generally even qualifed to monitor a technical series of systems.
On the other hand, think of the benefits. Speeding becomes a thing of the past for most people (yes, someone will "hack the system") but that'll likely be beyond the purvey of most people. School Zones restrictions get obeyed. Highway Construction Zones get obeyed. I think it's a good thing if they can solve the "id10T" problem.
Re:I don't think I could ever trust it (Score:2, Insightful)
Modern planes do take-offs and landings, too. I don't think missing the runway by 30m is an acceptable error :)
That said it is much more difficult to make autonomous cars, though I still hope and think I'll live to see it. It's not the precision --- computer system are good at precision, and when Galleio goes up we can use that for dm scale precision. It's the road system that we humans have built for humans thatis going to be difficult. Marking the middle of the road with anything ranging from nothing over a single white stripe over complicated marking over a green area and a fence isn't precisely easy for a computer+sensor to navigate by. Then there are road signs, pedestrians and the odd cow.
Still, I'm told that that plane automatics have brought down the accident rate by a factor 100. Even if the real number is 10, that's 50 kills a year in little Denmark, and scaled to the world, it's 50000 people. Thats a lot of people --- enough to warrant forcibly introducing such a system if need be, IMHO.
Re:Benefit Number One (Score:3, Insightful)
All this is moot, since it would just spur more suburban development until the congestion rises back to some equilibrium level of annoyance. Build it and they will come.
Re:But how deep? (Score:3, Insightful)
I currently find it hard to believe that cars can drive themselves effectively on city streets. I don't see much of a problem (technically) on interstates though. In fact, if we could just get an automonous system running on the interstate, with human control for exiting & entering, I would be really happy.
But like I said, after a few years on the interstates I might not see a problem with autonomous driving everywhere.
Re:Switchable (Score:2, Insightful)
Re:Share the road with automated 18-wheelers? (Score:2, Insightful)
Re:But how deep? (Score:3, Insightful)
Actually the biggest problem right now is the cost of implementation. Highway markings and video detection are not good enough across enough of the country to reliably introduce a system right now. Non-video guidance, which is technically capable and is the basis for most of the technology demonstrations you see, is usable now, but the infrastructure installation costs are too high for large areas. What you will see over the next 10-30 years are HOV/Toll lanes that are installed and restricted to autonomous vehicles, once there are enough on the road using this, there will be a gradual re-balancing of the roadway, so in 50-80 years you will have multiple autonomous lanes and a single drive-it-yourself lane. The incrementalism isn't just psycology it is the only way to solve the chicken-egg problem.
car vs hospital (Score:3, Insightful)
And remember, when you're driving, there are other people besides yourself out there whose lives are on the line.
Obey the speed limit, keep right, and stay alive. It's a good thing.
Re:Amazing technological breakthrough (Score:3, Insightful)
The problem with your post is, you're comparing the convenience of a car to that of the current mass transit system, not the sort of mass transit system we could have if, say, one person in ten could give up their cars altogether and put that money into a serious system. For the purposes of this discussion, a "serious system" is one in which the buses run more often than every half hour, and don't stop running altogether after 7PM. In short, one not at all like the one in Salt Lake City.
I'm not imagining shiny new things like, say, an intercity monorail to replace the bus you take to Boston. I think the current systems, but faster, cleaner, and more efficient, would be more than adequate for the needs of most urban dwellers.
Regarding your trip to your parents' house, I think you make it sound worse than it is. Say you're spending 8 hours in the car rather than 11 hours on mass transit. But if you assume that half that time can be spent in productive ways that a car doesn't allow (reading, etc) then suddenly mass transit is competitive again. It could also be argued that you're ignoring the time and effort spent actually earning money to pay for, insure, maintain, gas up, wash, park, and store the car.
Finally, in a fit of anti-American, self-loathing pique, I would point out that we as a society would be spending a heck of a lot less on transportation if our suburbanized ancestors had stayed and dealt with urban problems instead of fleeing like rats off a sinking ship.
Re:Why does "Keep Right to Pass"... (Score:3, Insightful)
I'm not sure I agree with the sentiment that '"Speed Kills" is just propaganda?!' but I do think it is terribly incorrect.
The difference in speed is what kills.
There is a route to get to that end. (Score:3, Insightful)
So in the first stage of AI control, we make computers only do the simplest task: 'cruise control plus'. They stay at a specified speed or minimum distance from the car in front, so very little unless the vehicle in front slows down or someone cuts them up. They don't even stay in lane, the driver can continue to do that. This means the first task to the AI is simply object:location mapping in 2d in real time and I think we can already do this. It would be enough of an improvement not to have to keep braking and accelerating in heavy traffic that I suspect lots of long distance drivers would want pay for it as an add-on.
Next, the AIs take over lane following. I suspect they can already do this too, but it won't get into the mainstream for five years after the first section is considered normal. They'll need to be able to recognise a stationary object or lost pedestrian/cyclist and react sensibly enough till the driver can take over. Hopefully it would become legal to read a book or do your paperwork if your car is in the inside lane under AI control.
Then they get taught how to overtake. This is where it starts getting interesting, but it's still only clever object:location mapping.
For the first few years the driver takes over speed control and steering if there is any problem. Not that they will be able to do much!
Thereafter, we might improve the technology to A roads (main roads?) and eventually B roads (rural roads?). Howwever, these environments are so damn random that we won't see it for a long time, till AIs are much, much clever or roads are much better defined.
The legal remifications needn't be a problem. As the technology comes in piecemeal, we'll adapt. It's only if we went for a complete AI solution that there would be a legal nightmare.
I look forward to it myself.
J.
Re:Why does "Keep Right to Pass"... (Score:2, Insightful)
I think you've got that wrong there. The 'keep right except to pass' would still be valid if no one went above the speed limit. Because there's always going to be someone going slower than you, even if the maximum is 65. It's not to accommodate speeding, but to accommodate the flow of traffic regardless of speed.
The whole reason for the 'keep right except to pass' is to prevent traffic jams. If there is someone in the right lane going slow and the guy in the left lane is not passing him - guess what? There's going to be a traffic jam.
And let me explain my comment about tailgating a little for all those offended. I was using a bit of hyperbole. I don't actually tailgate very much. My strategy is to let the guy know that I desire to pass him. If he does not respond, I use other methods. I'm generally very patient even if there is a clear path in the middle lane to pass him. I think it's dangerous to pass on the right. Only as a last resort will I pass on the right. I find that most people are accommodating as long as they are alerted to the fact that you want to pass them. Only once in awhile do I find that the guy's a jerk and feels like he owns the lane.
I also feel like I'm doing a service for all the other people in the left lane that are behind me. I get people to move out of the way so traffic can be uninterrupted by a slow person in the left lane.