Slashdot Mirror


California To Join Nevada With Rules For Autonomous Cars

thecarchik writes "As of now, the only state where self-driving cars are legal on public roads is Nevada, thanks to its vast expanses of open space and lightly traveled byways. California, recognizing that autonomous cars are an inevitable progression of technology, is moving to establish its own rules for driverless vehicles. A bill proposed by California Senator Alex Padilla would set guidelines for the testing and operation of self-driving vehicles within the state. As California is home to Google, Stanford and Caltech, all of which have active autonomous vehicle programs, the state is positioned to be a leader in driverless car development. It stands to reason that self-driving cars will be allowed on California's roads, probably in the near future."

35 of 194 comments (clear)

  1. so it begins by Anonymous Coward · · Score: 4, Interesting

    This is just too awesome. It looks like we're solving the parking, traffic, and driving death (drunk driving and otherwise) issues in my lifetime. The microchip is the gift that keeps on giving.

    1. Re:so it begins by TFAFalcon · · Score: 4, Insightful

      Those things aren't going to go away. Parking and traffic have to do with the number of cars, not just their driver's skill. And driving death will still happen until EVERY car is driven by an infallable AI. Which won't happen for at least a few generations after the AI is developed, since people are much to attached to driving cars.

    2. Re:so it begins by icebraining · · Score: 4, Insightful

      Parking is a much smaller problem if when you're planning to stay put at a certain location for a few hours, the car can simply drive itself to another parking location.

      Also, self-driving cars are a major boon to car sharing services; which should reduce car ownership; for the user, there's a big difference between having to go to a parked car somewhere and then leave it there again than just have it parked outside his home and then leave it anywhere.

    3. Re:so it begins by CastrTroy · · Score: 4, Insightful

      Computerized cars could be a lot smarter than humans and reduce traffic. Take a simple traffic light. If there's 5 cars at the traffic light, it takes about 10 seconds for the 5th car to start moving (people really are this slow). If computers were driving, all the cars could start to move in unison. Also, take highway driving. People slow down to look at something interesting on the side of the road. Traffic piles up behind them. With a computer driving, this wouldn't happen. Automated cars will be able to make traffic much less of a problem. If you cut out accidents and stupid drivers, the amount of traffic will go down significantly.

      --

      Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
    4. Re:so it begins by Kleen13 · · Score: 2

      And with the invention of the "stack em and rack em" automated garages, this could really solve some downtown problems.

      --
      That sinking feeling deep in your gut when you KNOW you screwed up bad summed up with: {head desk} {head desk}
    5. Re:so it begins by mlts · · Score: 2

      What I see happening are some roads being auto controlled only, and cars being spaced apart by a central road computer. Easily twice as many cars can be packed on a road because there isn't the reaction time a human requires. To boot, cars can be spaced where longer distance commutes can take further left lanes.

      Heck, even four way intersections can be changed to not require any signals... just slow up one set of cars so another gets through without smashing.

      I'm all for this. It not just allows cars to be moved around autonomously, but it opens up the ability to borrow cars. Not just like Zipcars or Car2Gos, but for people to allow others to "rent" their vehicle at times they don't need them in use.

      This is arguably the most effective solution in the US, especially the suburban areas to handle the higher traffic densities where rail, even bus stops are just plain economically not doable. Of course, more urban areas are better served by subways, trams, light rail, and bicycles, but for areas where a commute may be 50+ miles, autonomous vehicles and roads that can space/maneuver vehicles would be ideal.

    6. Re:so it begins by perryizgr8 · · Score: 3, Informative

      slippery ice covered roads already are quite covered by the computer in your existing car.

      --
      Wealth is the gift that keeps on giving.
    7. Re:so it begins by eln · · Score: 3, Informative

      An automated vehicle would be able to detect the moose and apply the brakes far faster than any human possibly could. There already are rudimentary collision-avoidance systems in some cars, and they'll only get better over time.

    8. Re:so it begins by an+unsound+mind · · Score: 3, Informative

      Computers are perfectly able to see what is going to be on the road, all you need is more sensors and better shape recognition.

      Better than humans, in fact. Humans can't see infrared so well, and it's going to be a heck of a lot more useful in the Canadian wilderness than normal sight. Your concerns are a design problem.

    9. Re:so it begins by similar_name · · Score: 3, Interesting

      You forget the cost problem. I can not afford a half million dollar car and people in canada have to drive rediculous distances all the time so our cars do not last very long.

      Half million dollars? That's like saying Intel's next chip won't catch on because they spend hundreds of millions of dollars developing it and who could pay that much for a computer.

      You said you drive a 95 neon and an 06 Dodge 3500. What do you mean your cars don't last long?

      The driving conditions you describe actually seem an ideal place for AI to start to become feasible. Replacing a truck driver with an AI would save over $30,000 each year (I don't know about Canada but U.S. truck drivers start around $30,000).

    10. Re:so it begins by similar_name · · Score: 2

      A robotic car will never get distracted, will never tire, will never drive unsafely, will never get frustrated, will never get bored and so on.

      Don't forget the rest of it.

      It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear. And it absolutely will not stop, ever, until you are dead.

    11. Re:so it begins by Dusty101 · · Score: 2

      Wow. You really are a true multitasker.

      Just be careful how you hold that razor.

    12. Re:so it begins by Sarten-X · · Score: 2

      In an ideal AI-controlled automotive world, the road is monitored by multiple sensors, giving the vehicle complete knowledge of its environment.

      In the case of a moose standing in the road, the vehicle's camera and rangefinder will simple see that where the road was expected, there is instead a gap in the trees with a moose-shaped object in the middle. The AI, having detected the moose from ample distance away, will slow down the vehicle until it can safely avoid the obstacle by going to one side or the other. A more advanced AI may even be specially configured for avoiding a moose, and will be able to see which end is the head (and therefore the most likely way the moose is walking). While approaching the moose, the smarter AI could veer slightly toward the rear of the moose, likely allowing a larger window through which to pass.

      This does, of course, depend on having enough distance to detect the moose, recognize it, and stop. The AI can be programmed to be paranoid, slowing down to pass the top of hills or rounding a blind corner. With government help, those occasionally-useful "animal crossing" signs can be augmented by radio beacons, alerting the AI that its behavior should be even more paranoid, and assuming that stationary things on the side of the road are likely animals, not rocks (and be pleasantly surprised when they don't move).

      Then there's two more difficult cases to consider: The moving moose and the hidden ice. Both are very difficult for humans to deal with already, and can be managed with adding more sensors to a car. Decent infrared vision is cheap, on the order of a few hundred dollars for an infrared/light-amplifying night-vision device. With IR vision showing living things separate from the surrounding trees, most of the AI's algorithms can be directly applied to otherwise-invisible objects. A moving moose in infrared is as easy to see as another car (that doesn't respond to radio negotiation), and can be avoided as easily.

      The ice is actually more difficult, because it's easiest to see it (through basic sonar, laser, and stereo vision) as just plain road. To tell that it's actually ice, vehicles intended for cold conditions could be outfitted with additional sensors and AI to note the subtle differences, such as light reflection, snow thickness, color of tracks, or even density. All together, the extra cost would be about $1000, by my estimate. The vehicle would be best served by having independent traction, but that's not a necessity. The point of traction control is to apply brakes to a wheel that's spinning faster than needed, so it will slow down and grip the road again. That can be done with a whole axle, but it's just not as efficient.

      None of the AI algorithms are affected by the size or weight of the vehicle, or what it's pulling. Rather, the vehicle AI knows its own base design, and can estimate weight within the first few minutes of driving, just based on acceleration profiles. It knows the engine put out a certain amount of power, and the vehicle accelerated up to a certain speed. From that, mass can easily be deduced.

      Finally, all of this is accomplished with the car being fully independent. With the addition of radio communication with a nearby town, the vehicle could have known in advance about the ice or a moose in the area, but that's just helpful, not necessary.

      , pulling in more data than sight alone.

      --
      You do not have a moral or legal right to do absolutely anything you want.
  2. As California is home to... by HockeyPuck · · Score: 3, Insightful

    ...some of the worst drivers in the world.

    I've lived in Boston, New York and Chicago. And Northern California easily takes the cake for worst drivers. They hesitate when they should commit, they never use turn signals, roll through stop signs, drive until 7-8pm without their lights on (or just use their parking lights).

    So I would welcome driverless cars, because it can't get much worse than this.

    1. Re:As California is home to... by BasilBrush · · Score: 4, Informative

      Clearly you've never travelled. Try Italy or India for example.

    2. Re:As California is home to... by s2jcpete · · Score: 2

      Bah. I lived in Italy for a few years. Rome may be bad, but I found Boston to be worse

    3. Re:As California is home to... by Anonymous Coward · · Score: 5, Interesting

      Yes, I agree. Rome may be a mad house but, I loved driving in Italy. There are far fewer rules (and often a lack of lanes) but I interpreted it as "We trust you -- just don't crash into anyone." It was a breath of fresh air to not have a million signs like in the U.S. that you simply tune out.

      There are some experiments in Germany where they are getting rid of all but a couple signs and simplifying the rules to just a couple rules (like yield to the right). They (last I heard) have found it to be far more effective as people don't tune out the few signs they see.

    4. Re:As California is home to... by niftydude · · Score: 3, Funny

      Once when I was driving on the 101 between San Jose and mountainview, I saw a guy playing a flute while steering with his knees.

      Still not sure if that was an instance of poor driving, or an awesome display of physical and mental prowess. But what I can say is that I haven't seen anything like that anywhere other than california.

      --
      You can never know everything, and part of what you do know will always be wrong. Perhaps even the most important part.
  3. New ways by Anonymous Coward · · Score: 4, Funny

    Here in Nevada we are are at the forefront of gambling....

  4. Liability mitigation is the crucial rule by ShooterNeo · · Score: 4, Insightful

    What is REALLY needed is a law to mitigate liability risks for automated cars. Here's how a fair law might read :

    All operators of automated vehicles are required to buy additional insurance. If someone is harmed by an automated vehicle malfunction, a panel is empowered to compensate the individual with a FIXED amount of money depending on the severity of the injury and or death. This is how vaccine injuries are handled : if a vaccine harms someone, they get a certain amount of injury depending on the risk.

    Neutral, third party laboratories would be paid to examine the 'black boxes' from automated cars after a crash and present their findings to the panel.
    The panel would be required by law to make a decision within a certain amount of time (~180 days sounds about right)

    Advantages :
        1. Lawyers eat up a large chunk of the money when litigation is allowed. This way, most of the money goes to the victims.
        2. Everyone gets some compensation money instead of most getting nothing and a few hitting the jackpot
        3. Faster decisions instead of lawsuits that take 5-10 years.

    Disadvantages :
          1. Panel can be unfair or biased and little can be done
          2. The amounts of money seem low compared to jury awards for successful lawsuits. Lose a hand, it might be 100k not a million, etc.
          3. Legislators who are lawyers have to write the legislation for this.

    The reason to do this is the same reason we do vaccines, but it would save a LOT more lives. Automobiles kill far more people than the number who would die if we stopped most vaccinations. Automated cars will occasionally malfunction and kill someone. However, on the aggregate, the total deaths per passenger mile caused by automated vehicles will very likely be more than 10 times or more lower. Automated vehicles have short reaction delays, no need to take risks, ability to see in all directions they have sensors pointing at the same time, can predict a crash is about to occur and take mitigating actions (pre-firing the airbags, etc), activates the brakes quickly enough to avoid pileups, etc.

    The thing is, an automated car will have software bugs, and will occasionally make mistakes. Maybe a good model will be as good a driver as the average driver on their best day. EXCEPT, an automated car's systems cannot become distracted, board, drunk, or fall asleep. I suspect that this advantage over millions of miles will prove to be huge. Sure, the average human might be smarter, but we don't give our best effort during every minute of the many hours we drive.

    1. Re:Liability mitigation is the crucial rule by 0123456 · · Score: 2

      Oh, also, an automated car cannot become scared or panic.

      Meanwhile, back in the real world, the automated ABS systems in many cars will cut the brakes on fresh snow where locking the wheels would typically result in a shorter stopping distance.

      Who exactly is going to program the car to deal with every possible dangerous situation?

    2. Re:Liability mitigation is the crucial rule by fuzzyfuzzyfungus · · Score: 4, Interesting

      There is arguably one major gap in the analogy with vaccines(unless your plan includes it but simply didn't state it):

      The vaccine injury system is designed to deal, as efficiently as messy compromise allows, with the fact that vaccines(as with other drugs and procedures) tend to have risks that show up at the population level that couldn't have been detected in clinical trials of any feasible size and/or are substantially lower than their benefits. The logic is that these cases have victims deserving of compensation; but arise without culpable negligence or malice.

      It doesn't, and isn't intended to, cover other risk/liability issues arising in medicine that incidentally involve vaccines. If, say, your doctor stored a vaccine improperly and administered a contaminated or spoiled dose, that wouldn't be a vaccine injury, that'd be malpractice that happened to involve a vaccine rather than some other drug. In such a case, the damages would be partially to compensate you and partially to punish them; because there are both damages and culpable negligence or malice at play.

      In the case of an autonomous car, the 'vaccine analogous' set of risks/compensations would only cover the set of risks inherent to the system's operation(corner cases where physics simply doesn't allow for a safe solution on the navigational system's part, system defects sufficiently rare and esoteric to have escaped reasonable diligence on the manufacturer's part, and so forth). It wouldn't usefully cover negligence on the part of either the manufacturer(in, say, corner-cutting on testing or design of safety critical systems) or the operator(operating a vehicle despite sensor or system faults, defeating safety-critical systems in order to achieve faster trips, etc.)

      When dealing with small, essentially unavoidable, risks there is a strong logic in favor of efficient compensation purely on the basis of injury(assuming that those risks carry benefits sufficient to justify their broad imposition...); but one must be careful not to immunize negligence and malice in a system designed to handle mere accident...

      I suspect that there will be fewer impaired computers than there will be impaired drivers; but I suspect that operators running cars with the sensor equivalent of shot breaks and dead turn signals will hardly be unknown, and corner cutting by some manufacturer or other is just a matter of time.

    3. Re:Liability mitigation is the crucial rule by Vanderhoth · · Score: 3, Interesting

      I live in Nova Scotia, Canada. I'd trust the car over a human anyday. I've seen to many accidents where someone made a slight miscalculation that shouldn't have been a big deal. Then they end up over compensating and taking out someone in an on coming lane instead of vearing off into a parking lot, just ending up on the side of the road or even just staying on course and having nothing come of a small skid, swerve or bump.

      The only issue I see with and autonomous car is there are times here where a person has to guess where the road is. I'd like to know how the car would track the road when it's more or less just a blanket of white.

    4. Re:Liability mitigation is the crucial rule by BasilBrush · · Score: 4, Insightful

      Meanwhile, back in the real world, the automated ABS systems in many cars will cut the brakes on fresh snow where locking the wheels would typically result in a shorter stopping distance.

      Unlocked brakes means you still have some ability to steer. Locked brakes have no steerability. Locking the wheels will often result in the car going sideways down the road with no ability to steer into the skid. If there'a one thing worse than not being able to stop, it's the car travelling sideways whilst not being able to stop.

    5. Re:Liability mitigation is the crucial rule by jmichaelg · · Score: 4, Interesting

      What you're proposing is a No Fault liability scheme. Circa 1989-1992, the insurance companies attempted to get a proposition passed that would have established No Fault insurance. Their pitch was very similar to your list of advantages plus they said that since their costs would decline, our rates would have as well.

        Despite the idea making a lot of sense, the personal injury lawyers succeeded in killing it as they viewed the proposal a direct threat to their livelihood which of course, it was. The proposition was aimed at cutting their take out of the transaction.

      Your post makes a lot of sense but unfortunately, I think the political climate in California has gotten more bizarre over the intervening 20 years and what makes logical sense doesn't mean too much in California.

  5. Rosy future, but there will be system failures by damm0 · · Score: 5, Insightful

    I am wholeheartedly for the development of robot cars! I can hardly wait for the day when I can command my car to drive my drunken ass home, or tell it to go to the grocer and pick up my milk and cheese (which the grocer will load into my car for me) while I'm at work. Not to mention the possibilities for car sharing!

    However, there will be system failures. The cars will have to develop "reptile brain" like functions that can make the car pull over and stop in the case of byzantine failure of the controller. Think about car-worms and viruses that command cars to crash into each other, or remote car hijacking. It is going to be *very* interesting to watch all this develop. Consider the people who will drive recklessly in their "classic cars" expecting that most other cars are autonomous, which may make the road more dangerous for those who don't have one.

    That said, I'm looking forward to the robot-car only lanes on the freeway where we can have fuel-efficient car-trains and the social benefits of being able to hop out of your robot car in front of your destination and have the car valet itself.

  6. Re:I look forward to becoming rich by hot+soldering+iron · · Score: 2, Interesting

    You can try... Until the onboard video is played in court and shows you staged it, at which point the judge begins to legally sodomize you.

    --
    When you want something built, come see me. If you want correct grammar and spelling, get a F*ing liberal arts student.
  7. Several Points by DERoss · · Score: 4, Insightful

    Padilla's bill is SB 1298 at http://www.leginfo.ca.gov/cgi-bin/postquery?bill_number=sb_1298&sess=CUR. It has not yet had its first committee hearing.

    I was a software test engineer for over 30 years. There is no such thing as a computer system that is completely error-free. While SOME drivers are impaired or simply have poor judgement, other drivers are alert, coordinated, and generally safe. On the other hand, all autonomous cars from the same manufacturer will have the same software errors.

    The current leader in developing autonomous cars is Google. I would not drive one of Google's cars unless I knew that Google was not tracking where I went and what route I took to get there. I am concerned that, even if the car does not transmit its location and route in real-time, a mechanic might still be able to download the car's history while servicing the car. That information should be available only to law-enforcement agencies and even then only when a judge issues a warrant after being convinced there is probable cause that the history is relevant to an actual crime.

    1. Re:Several Points by ShanghaiBill · · Score: 5, Insightful

      There is no such thing as a computer system that is completely error-free.

      It doesn't have to be error free. It just needs to be better than humans. That is not a high bar.

      On the other hand, all autonomous cars from the same manufacturer will have the same software errors

      And when one of those errors causes an accident in ONE car, it will be fixed and patched in ALL the cars. So the number of bugs, and the number of accidents will decline quickly . Autonomous cars already have millions of miles of testing, and are probably already safer than the average human driver.

      Demanding absolute safety is foolish, and delaying the introduction of autonomous cars will cause many unnecessary deaths.

  8. Re:Unusual Road Conditions ???? by damm0 · · Score: 2

    More likely the car will be able to detect the dropping air temperature and reduce speed to adjust. It is shocking how many people blithely drive into dangerous situations unawares. People who don't come from cold places never think to tap their brakes and test traction from time to time. Heck, lots of people who should know better don't do it. And autonomous cars do have the advantage of being able to detect collision vectors and respond far more quickly than any human driver could hope to.

    The Robot 500 should be fun to watch too :) Everyone wants to see a crash in a car race, and they'll deliver!

  9. Crossing state lines by RyoShin · · Score: 2

    I'm a big states-rights kind of guy, and I applaud California and Nevada on taking initiative in a technology that will hopefully become widespread sooner than later, but this is one situation where the federal government should be involved (cars often cross state lines, after all) and at least form a committee/study (insert committee uselessness here) to set a ceiling on limitations for these vehicles. States can relax the limitations as they see fit (open-space areas like Nevada, Montana, and Wyoming might allow a higher auto-speed), but if each state is left to set its own devices you'll get a large amount of different standards that each automaker has to adhere to in order to sell the vehicle on a national or even regional level. The fed is going to step in at some point, but better sooner than later; not only will they create a nice standard for all states (/. likes open standards, right?), but it will make a lot of states that are on the fence about the whole thing (or not even caring) have an extra push to allow the vehicles (assuming the feds don't allow them nationally in addition to standards).

    I haven't read up on the various rules put in place (or recommended), but I sincerely hope there's a size limitation on the vehicle. No more than T tons, no more than XxYxZ dimensions to be allowed an autopilot. That way in the case of a catastrophic failure (we'll get at least a handful) the risk to other drivers is far less. Also, energy savings. Maybe even create a standard within the auto industry for censors that can be included in an "autopilot only" lane to enhance the cars capability in those environments. (Ooh! How about extra sensors within parking garages so that the car auto-drives itself to the closest available open spot?)

  10. Free Insurance by michael.j.boldischar · · Score: 2

    Does this mean we won't have to pay auto insurance soon? Computers should be able to drive better than humans, right? Imagine how much extra time we would have if computers could drive us to work each day.

  11. Re:Unusual Road Conditions ???? by Baloroth · · Score: 2

    I have to wonder how these autonomous vehicles will handle unusual road conditions such as snow patches on the road, black ice, a ball bouncing out into the road, etc. It may be that autonomous vehicles are by far safer than vehicles with drivers -- until the unexpected happens at which point they completely fail.

    I can just see it now: You are sleeping as your autonomous car is driving across country and then the horns and bells go off with a voice says: "Quick driver, take over, we are spinning out on black ice at 70mph and you have three seconds to recover or you and this car will die."

    Far better than a human who isn't programmed to deal with every possible situation a car can find itself in. The car can calculate up to the millisecond road and atmospheric conditions, as well as scan the road ahead for patches of unexpected alterations in the road that could be ice/snow/water/etc, and of course know local weather through radio transmission (as well as the position, velocity, and condition of every driverless car nearby). You can simulate and test every possible condition for the car, and train the algorithm to handle each situation in the ideal fashion. Humans, on the other hand, can do none of these things.

    So, most likely, the car wouldn't wake you up because it would never need to. It wouldn't travel 70mph in conditions that can result in black ice. It would see the ice well in advance and slow down, not overreact if it began the spin, and recover better once it entered it (humans tend to turn against the spin, which just makes it worse.)

    --
    "None can love freedom heartily, but good men; the rest love not freedom, but license." --John Milton
  12. Seems Nevada's driverless car rules by mysidia · · Score: 4, Informative

    Are actually rules... that autonomous cars can't be driverless

    2 passengers required; the human operator has to be able and ready to override the car; which means the human has to have a license, can't be drunk, etc. And the human operator (rather than the manufacturer) is responsible if there is an accident and the vehicle has fault because of improper decisions/failure.

    I guess the restrictions "sound good", but they eliminate some of the selling points for the concept of an autonomous vehicle. Probably without making it safer.

    You can't be relaxing, chatting on your cell phone, watching TV, or eating while the car drives you.

    Makes more sense to require that driverless cars be safe enough and have enough failsafes and instrumentation that a human operator will not ever be required to override; e.g. by ensuring that the safest reasonable response is always what the autonomous car will execute, and facilitated by multiple redundant highly robust systems.

    Such that the greatest remaining danger would be that the human erroneously overrides the computer and makes bad choices.

    1. Re:Seems Nevada's driverless car rules by Krneki · · Score: 2

      Yap, it's just a first step. Once we see it's safer if the driver doesn't touch anything we will move on.

      --
      Love many, trust a few, do harm to none.