Congress Poised to Let Autonomous Car Companies Run Wild in Cities

Photo:  Ed and Eddy/Flickr
Photo: Ed and Eddy/Flickr

As Zipcar found Robin Chase puts it, a world with autonomous cars can either be heaven or hell.

In the hellish scenario, cities are overrun by privately-owned robot cars circling all day, mostly without passengers, and our transportation system becomes even more isolating. In the heavenly one, fleets of shared vehicles reduce private car ownership, cut traffic, and make streets more conducive for walking and biking.

While these technologies are probably many years away from hitting the streets as mature, finished products, public policy decisions are already being made that will determine which path cities follow. Both houses of Congress are crafting regulations that will set the stage for autonomous vehicles, and they’re not off to good start.

Legislation introduced by South Dakota Republican John Thune and Michigan Democrat Gary Peters [PDF] would allow carmakers to deploy up to 80,000 autonomous vehicles on public roads to test and refine their technologies. The bill cleared the Senate Commerce Committee in a unanimous voice vote this week.

Watchdogs are especially troubled by the preemption of local governments in the bill, which says states and cities cannot impose “unreasonable restriction on the design, construction, or performance” of automated vehicles. Language like that is too sweeping and vague, and would give cities and states almost no power to regulate the testing of autonomous vehicles, warns Beth Osborne of Transportation for America.

For example, would basic safety precautions like mandating non-lethal vehicle speeds on streets where people walk or bike be construed as “reasonable” by the courts? As written, the bill leaves too much open to interpretation.

“It’s not clear if a city could even pass a law saying you have to follow our laws,” said T4A’s Russ Brooks.

Advocates say the disclosure measures in the bill are too weak. Automakers would only be required to submit annual safety reports to the feds, and the reports would only have to describe changes to the vehicles on the road and “any significant safety deviation from expected performance.” Basic data like crash rates per mile would not necessarily have to be spelled out, and local officials would not be entitled to access the data that is disclosed.

“City and state law enforcement will have zero information about how these vehicles are operating on these roads,” said Brooks. “It’s really quite terrifying.”

In a statement, Osborne said the bill would “creates a climate of secrecy around AV testing or deployment.”

Advocates have ample reason to be wary of ceding regulation and oversight to the automotive industry. Recent experience suggests that in the rush to get autonomous vehicles to market, companies are all too willing to cut corners.

California ordered the removal of Uber’s self-driving cars after one was caught on video speeding through a red light at a crowded intersection in San Francisco last year. In Pittsburgh, Uber damaged local political relationships by failing to provide data from its experiments with self-driving cars.

T4America says it will work to either amend the problematic sections or strip them out before the bill is brought up for a floor vote.

  • reasonableexplanation

    One interesting side effect of the coming self drive car era, is i feel like we’re going to see the death of most ‘fast cars.’

    Nobody cares how fast their taxi gets from 0-60, and that’s all self driving will be, a cheaper, nicer, personal taxi.

    I actually foresee the comeback of the large american land-boat style cars; think big roomy Cadillacs. Because if all you care about is comfort and lots of room; it doesn’t get any better. It might be a boom for American car manufacturers.

  • Vooch

    agreed – and self driving cars will never exceed the speed limit. another benefit

  • Joe R.

    They’ll still be “fast” by the true definition of the word even if they don’t accelerate rapidly. Most scenarios envision self-driving cars running at far greater speeds on highways. Of course, with vehicles cooperating at merges, it won’t much matter how long the car takes to go from 0 to 100 or 125 mph or whatever the cruising speed is.

    As fleet vehicles, we’ll also see designs focused mainly on efficiency, which would mean very aerodynamic bodies around those roomy interiors. I personally like aerodynamic vehicles but I suppose it’s not everyone’s cup of tea. With good aerodynamics you probably won’t need anything more than 50 or 60 HP to do 125 mph. Vehicles might have a little reserve for hills and merging, but you’re right, the days of quick acceleration will be over. Just from a personal standpoint, I’ve always been more excited by the top speeds of vehicles rather than how quickly they can reach some arbitrary speed like 60 mph. I’m happy with the acceleration rates on my bike, for example, but I wish it could go a heck of a lot faster.

  • Larry Littlefield

    We don’t know that. They could do speed limit plus 10, and decide to take the corner at that speed if they won’t roll.

    And what are the cops going to do, pull them over?

    What I fear is fewer injuries and deaths in crashes with other vehicles, and the same or more in crashes with pedestrians and bicycles. And the AV companies saying “we’re done, it works” and the suburban politicians saying “great, move to the suburbs or die.”

  • Vooch

    agreed on corners –

  • Joe R.

    Liability reasons will probably force them to go slow enough around vulnerable users so they can stop in time to avoid a collision.

    Also, I think once we’re 100% AV, the concept of a speed limit will no longer need to exist. The criteria will be is your speed low enough so you can stop in time to avoid any hazards detected by the sensors? As sensor technology becomes more advanced, it may be that you can detect pedestrians or cyclists 1/4 or 1/2 a mile away. If the road is clear up until that point, no reason you can’t drive 50 or 60 or 70 mph until you get within a block of the vulnerable users. Maybe you reduce speed so you pass them at 10 or 15 mph, then resume speed until the next hazard is detected.

  • twk

    Your assuming the current speed limits stay as they are. I would not be surprised that there will be a push to increase speed limits. You know, because the computer is a better driver and can react faster. Plus always going the speed limit will seem too slow for the customers. Speed over safety will re-establish itself.

  • Vooch

    dang – I think you are correct.

  • Larry Littlefield

    “Liability reasons will probably force them to go slow enough around vulnerable users so they can stop in time to avoid a collision.”

    So they’ll get Congress to exempt them from liability in the name of freedom. Is the manufacturer whose product was used to mow down concertgoers with a machine gun going to be held liable?

  • Joe R.

    The difference here is the machine gun was controlled by a person, and therefore the manufacturer couldn’t control how it was used. When you have a car driving itself with the automaker’s software, then the automaker is just as liable if someone dies as it would be if the software in an airbag malfunctioned. There is already plenty of legal precedent for this.

    Of course, they might be able to weasel their way out via legislation. I’m personally dubious if such legislation would have any hope of passing once the general public was aware of it.

  • Frank Kotter

    That we still look to technological improvements optimistically is a really baffling. Remember in the 90’s when the first generation of tech bros were speaking in lofty terms about the digital peace and the wonderful world of connectivity and the social cohesion it will bring? We are tearing apart at the seams exactly because of it!

    I see a massive danger that this coming revolution will only deepen the primacy a person in car has over all others – Not to mention putting ten million people out of work (directly, U.S. alone) who are least prepared to transition to other professions.

    We need to get over our ‘shiny things’ obsessions.

  • Larry Littlefield

    There are also precedents limiting airline liability in crashes, so people can get there a little faster, and limiting power company liability in nuclear accidents, so nuclear power could be developed.

    So there has been trading of lives, but with everyone involved on both sides of the tradeoff.

    How much easier when the lives being traded are those of a minority — urban, walking, or (snicker) biking?

  • I as a bicyclist like my chances of being noticed by a driverless car’s sensor better than my chances of being noticed by an incompetent and distracted driver.

    Also, the driverless car is not going to speed up to make a turn in front of a bicyclist, as sociopathic drivers do.

    So right now I think that driverless cars will be an improvement for bicyclists and pedestrians.

  • mckillio

    Still likely better than the human drivers running wild in cities.

  • Joshua Putnam

    “For example, would basic safety precautions like mandating non-lethal vehicle speeds on streets where people walk or bike be construed as “reasonable” by the courts? As written, the bill leaves too much open to interpretation.”

    Such a restriction applied *only* to self-driving vehicles clearly *is* unreasonable.

    Apply it to *all* vehicles on streets with vulnerable users, then you don’t need special regulations for automated vehicles, you just need competent traffic enforcement.

  • Joe R.

    OK, but look at the numbers. How many people die in plane crashes each year? At worst, it’s in the hundreds. And given the major investigations of every crash it’s apparent we really consider anything much above zero unacceptable.

    How many have died in nuclear power plant incidents in the US besides nuclear plant workers? The number is exceedingly small.

    Over 6K pedestrians and cyclists die by motor vehicle each year. The only reason we tolerate this number is because the numbers killed inside motor vehicles are even higher. I can’t envision any scenario where even more pedestrian/cyclists deaths would be politically palatable.

    Moreover, by most accounts AVs are going to reduce deaths inside motor vehicles by way more than 80%. Indeed, I suspect once the technology is mature motor vehicle deaths will be so rare they’ll make the national news. Given this, even if AVs speed in cities, it’s hard to see how there will be any increase in pedestrian/cyclist deaths unless the vehicles are intentionally programmed to not slow down or change direction when encountering people in the street. That can’t happen ( first law of robotics—”A robot may not injure a human being or, through inaction, allow a human being to come to harm.”).

    The only time pedestrians or cyclists might be killed is when the sensors failed to detect them and the vehicle couldn’t slow down enough or change paths. Even this is preventable if vehicles always slow to some safe speed when objects are blocking lines of sight to potential pedestrians. Even if they don’t, I’m not seeing mass casualties. Human drivers rarely slow down when something like a large parked truck is blocking their view of jaywalking pedestrians. Despite that, how many people are killed crossing midblock? Most pedestrian deaths are caused by failure to yield or signal violations. AVs in this area will be far better than humans. My educated guess is we’ll have at worst low tens of pedestrian/cyclist deaths annually once we’re 100% AV, even if those AVs don’t exercise maximum caution. If they do (at the expense of longer trip times), the number should be close to zero. So that’s the tradeoff we might make—a few tens of deaths for faster trips, not many thousands.

  • Joe R.

    To some extent we can counter any attempts to increase the primacy of motor vehicles by doing exactly what is already being done in Europe, namely making more and more of our cities off-limits to motor vehicles. Cities can legally ban all vehicles from certain areas even if they can’t regulate other aspects of AVs.

    I have mixed feelings on the job losses but in all honestly with the coming robotics revolution we’re going to see massive job losses across the board. The only question is what to do about it. We might need to have some system where people are entitled to some amount of goods and services produced by robotic labor. I think the idea of working will change. It will no longer be a means to provide for necessities, but rather a way to buy extra goods and service beyond the bare necessities. If you can’t find work, you’ll still have food and a roof over your head.

  • reasonableexplanation

    What is considered a non-lethal speed would be higher for an autonomous car: if you look up a chart of stopping distances, it’s usually broken down into two portions, perception/reaction distance, and braking distance. So at 30mph for example, the perception distance is 44ft, while the braking distance is 45ft.

    The braking distance won’t change in an autonomous car, but the perception distance will, significantly. Thus the same safety margin in terms of total braking distance will actually be for a somewhat higher initial speed than with a human driver.

  • reasonableexplanation

    It will be a safety improvement for everyone, no question. It might not be an improvement for urbanists, as driverelss technology will make living further from other people (sprawl) much more painless.

  • reasonableexplanation

    Speeds will likely greatly increase on the highways (this is a good thing). Speeds will be unlikely to change much in cities.

  • Johnny

    I agree with you. But the idea that autonomous cars will be able to go faster than our cars today is (mostly) bunk. Air resistance is proportional to velocity squared, meaning engine efficiency falls off steeply above 60-70 mph. Some of the more wild ideas have cars “platooning” on the highways, but this is extremely unlikely to happen simply from a safety and liability standpoint (what if one car gets a flat tire?), and besides, it only partially solves the aerodynamic problems. We could, of course, design super aerodynamic vehicles, but these would be impractical for city driving, so you’d have to switch cars at your destination. In the end, the only way to achieve high speeds with any efficiency is to move a large number of people within a single vehicle–in other words a train or a bus. And no matter how good Google’s algorithms are, if I’m moving at 100 mph+, I’d much rather be on a fixed track than on rubber tires.

  • John Murphy

    The whole speed limit plus 10 canard is not because it’s safe to go the speed limit plus 10, it’s because there is enough error in measurement that the cops don’t pull over people they will have to go to court with and get beat all the time.

    It will be very straightforward for the car companies to have the vehicles travel at the speed limit, modulated by lower speeds when the car detects a hazard.

    In the world of the self driving car, all the features we see in car commercials that push the envelope just aren’t relevant. If we can turn the car from “toy” to “basic transportation”, we win.

  • farazs

    > but the perception distance will, significantly
    So you say! The truth is none of them are anywhere close to having better perception than the best human driver … and that is the standard they will be held to. Talk when there is credible data to back that theory.

  • John Murphy

    “Over 6K pedestrians and cyclists die by motor vehicle each year. The
    only reason we tolerate this number is because the numbers killed inside
    motor vehicles are even higher.”

    No. We tolerate these numbers because we empathize with the driver. A driver hits a cyclist and a high percentage of the public cannot empathize with the cyclist but a high percentage does empathize with the driver. There is an instinct to empathize with the driver because the reader can visualize themselves in that situation.

    When the car self drives, the reader does not empathize with the car – they didn’t design the car, the automaker designed the car. Therefore there will be hell to pay if this entity causes a death.

  • farazs

    Hmm, you obviously haven’t seen the uber video or read about the Tesla crash. What makes you think that a system incapable of seeing a solid red light of a semi across its path will be better at detecting a cyclist? Some day … sure, but what until then?

  • John Murphy

    Ten Million?

    You misspelled Fifty Million probably.

  • John Murphy

    self driving cars are getting better. Human driven cars are getting worse.

  • farazs

    > Speed over safety will re-establish itself.
    Stupidest thing I’ve read in a while! If speeds are be increased only because an equivalent level of safety is guaranteed at the higher speed (because as you propose, computers become better drivers), then that is NOT ‘speed over safety’. Of course, this rests on all traffic being exclusively autonomous.

  • farazs

    AFAIK, most equipment has a tolerance of +- 2mph. The limit+10 canard is because cops tend to prioritize more egregious violators.

    Of course there is also the self-fulfilling effect. Since most people expect to get away with limit+10, more of them will contest citations issued within that range, which would overwhelm both law enforcement and judiciary. The people know this which is why they exercise limit+10 with more impunity.

  • John Murphy

    I still have an analog speedometer, even if it’s spot on accurate, the tick marks on it are such that a pretty solid deviation could be claimed just from visual inspection.

    Yes, more egregious violations get prioritized, but if the cops get dragged in to fight every 3 MPH over enforcement would grind to a halt

  • Joe R.

    I don’t think we need to use speed as a criteria here. Speed is used as a proxy for safety when human drivers are involved. Instead, AVs will be judged on how often they kill or injure people, not how fast they’re driving. It’ll be much like any other machine. We don’t care about the fine details of what a machine does so long as it doesn’t hurt people.

  • Joe R.

    You’re blaming the car for the crash but in truth the truck driver did something totally unpredictable. Also, sensor technology is improving by leaps and bounds. What we have now is only a preview to what we’ll have 5 years down the road.

    You should never pass judgement on a technology still in its infancy. I heard the same thing about LEDs 15 years ago when they were less efficient than incandescent bulbs. Now they’re more or less well on the way to replacing every existing light source.

  • Jeffrey Baker

    This is just totally false. Autonomous vehicles have a wider field of view, a higher vantage point, and can track a virtually unlimited number of fixed or moving objects from huge distances. “The best human driver” can’t see the walking person hidden behind a car but the AV can, because it saw the person from literally a mile away and has been tracking its movements.

    The truth is that today AVs are already better, way better, than even the best human drivers. Compared to an average meat-driven car it’s not even a contest.

  • Jeffrey Baker

    Tesla’s and Uber’s autonomous vehicle technology is garbage. Their systems should probably be outlawed.

  • Jeffrey Baker

    I agree completely. I think it’s ridiculous that every state regulates the power and top speed of electric bicycles but no state regulates those features of cars.

  • Joe R.

    Explain why aerodynamic vehicles would be impractical for city driving. I’m trying hard to figure out how this would be so.

    As for the higher highway speeds, that’s a carrot we’ll need to get people to give up driving the vehicle themselves. Also, airless tires are coming. That mitigates the issue of blowouts at high speeds. Or we could lay rails on the highways and the vehicles could have a separate set of rail wheels. This would help efficiency.

  • Joe R.

    That’s another good point. I think it’s a given that we’ll have a very low tolerance for any deaths caused by AVs. As a result, they will always err to the side of caution, even at the expense of trip time.

  • Joe R.

    Cruising speeds in cities may even be a bit lower but average speeds might be higher by avoiding the need to stop frequently for red lights.

  • I did read about the crash of the self-driving Tesla into a truck. I thought that the issue was that the truck was so broad and wide that the sensor did not detect it as an object, and crashed into its side. This would not be the case with an object the size of a bicyclist.

  • Joshua Putnam

    On streets with vulnerable users, “non-lethal speed” has nothing to do with avoiding collisions. Collisions are going to happen. A “non-lethal speed” is a speed at which the great majority of pedestrians will survive impact. That doesn’t change with driver types.

  • Joshua Putnam

    Case in point, our school zone speed cameras will ticket for 22 mph in a 20 mph zone. The radar is well-calibrated and certified, the readings are all video recorded as well as photographed, and the tickets hold up just fine unless you claim someone else was driving your car.

  • farazs

    Totally false, eh? The real world has weather conditions, animals, variable road conditions, debris etc. Perception is the sum effect of understanding the lay of the land, not just being able to receive and distinguish one minute signal from another. You know an AV model than can already predict that a cyclist moving at 12mph in a 25mph zone will need to take the lane soon because there is a tree branch or a truck parked in the bike lane? You know one already capable of driving an emergency services vehicle in traffic? One already capable of responding and yielding to an emergency services vehicle in the best possible way?

    Those who insist on living in the theoretical world will be held responsible for the gap with reality when things go wrong. If what you are claiming were true, AV would already be pervasive. Instead we have buggies operating at low speeds in ideal conditions and gathering data and learning and improving their detection mechanisms. Regulations are not the only thing holding them back from unleashing. There are plenty of real technical challenges to overcome.

  • Jeffrey Baker

    You are a bag of chemicals clinging to your own self-image. AVs are already clearly superior to meat drivers in adverse weather conditions because AVs reliably halt in situations where the meat people proceed. All the wailing on the Internet about how Google’s car can’t drive into a whiteout blizzard is quite hilariously missing the point. Nobody can drive in a whiteout blizzard.

  • You know an AV model than can already predict that a cyclist moving at 12mph in a 25mph zone will need to take the lane soon because there is a tree branch or a truck parked in the bike lane?

    The autonomous vehicles that exist today would see the cyclist moving over to take the lane and would immediately decelerate. Whereas a driver would not notice the cyclist until several seconds later (or maybe not at all if that driver were playing with a mobile phone or doing some other distracting ativity).

    Even a driver who did notice the cyclist would be unwilling to decelerate dramatically for fear of being hit from behind, whereas an autonomous vehicle would be free to make the appropriate adjustment in speed because behind it would be another autonomous vehicle which would also decelerate immediately upon detecting the lowered speed of the first car.

    Even in its current embryonic and primitive form, self-driving technology is far superior to humans in terms of reacting to sudden changes and avoiding collisions. And the gap is rapidly expanding.

  • Jason

    To play devil’s advocate, and speaking as someone extremely skeptical of this fantasy of autonomous cars being the savior freeing us from needing to develop mass transit: if fully autonomous cars are able to slam the brakes faster than a human driver could then it WILL change things because it’s going to be supremely less less for collisions to happen in the first place.

    Outside of heavily contrived “do you pull the railroad lever to save the baby or three adults?” moral dilemmas it’s hard to think of situations in which a proper autonomous car is going to perform worse than a human driver. I try to be extremely conscientious as a driver and even I have, much to my shame, had uncomfortably close calls over just not being able to see people because I’m driving a two-ton murder box that’s designed to have blind spots.

  • They say cars don’t have blind spots, but you do need to move your head, and you do need to vary your speed approaching potential points of conflict: (because saccades…)

  • com63

    The Tesla crash was caused because Tesla’s system was not designed to work on highways with cross traffic. It could not detect a truck turning across the road because it was not designed to do that. It only works on limited access highways. The driver was negligent for using it on a road type it was not designed for. All of the “white side of the tractor trailer against a brightly lit sky” statements by Tesla were just BS. Their system isn’t that good yet.

  • com63

    The Tesla driver was to blame for the crash the most. He was using autopilot on a type of road it was not designed for and he was not paying attention. The truck driver and Tesla should get partial blame.

  • That’s not just a problem “for urbanists,” it’s a problem for the environment. The more painless it is in the short term to damage the planet, the more damage we get.

  • I don’t see how that makes Tesla’s claim BS.

    You yourself say that Tesla’s system “could not detect a truck turning across the road because it was not designed to do that”. Is that not precisely what Tesla is pointing to with its explanation about the side of a truck?


Choose Your Own Utopia: What Will We Make of Driverless Cars?

Cross-posted from the Frontier Group.  A century ago, a new transportation technology burst onto the scene that threatened to disrupt everything: the car. Thinkers of the day, along with boosters of the new technology, dreamed grand dreams of the utopia it would bring. General Motors’ Futurama exhibit at the 1939 World’s Fair (shown in the amazing […]

How the Self-Driving Car Could Spell the End of Parking Craters

Here’s the rosy scenario of a future where cars drive themselves: Instead of owning cars, people will summon autonomous vehicles, hop in, and head to their destination. With fewer cars to be stored, parking lots and garages will give way to development, eventually bringing down the cost of housing in tight markets through increased supply. […]

High Stakes for Cities as Feds Start Regulating Self-Driving Cars

Last week as part of his State of the Union Address, President Obama announced a $4 billion investment over the next 10 years to test autonomous vehicles and get them ready for the market. Two days later at the Detroit Auto Show, Transportation Secretary Anthony Foxx announced that federal regulators would begin to develop coherent safety regulations for […]